Hydrogeological Processes in

Karst Terranes
 SOME GROUNDWATER AND KARST TITLES PUBLISHED BY
The International Association of Hydrological Sciences (IAHS)
Improvement of Methods of Long Term Prediction of
Variations in Groundwater Resources and Regimes Due
to Human Activities. Proceedings of a symposium held
during the First IAHS Scientific Assembly, Exeter, UK, July
1982
edited by G. P. Jones
Publ.no.l36(1982), price $30
Challenges in African Hydrology and Water Resources.
Proceedings of the Harare Symposium, July 1984
edited by D. E. Walling, S. S. D. Foster & P. Wurzel
Publ.no. 144 (1984; reprinted 1986), price $48
ISBN 0-947571-05-1
Relation of Groundwater Quantity and Quality.
Proceedings of the Hamburg Symposium, August 1983
edited by F. X. Dunin, G. Matthess & R. A. Gras
Publ.no.l46(1985), price $30
ISBN 0-947571-15-9
Land Subsidence. Proceedings of the Venice Symposium,
March 1984
edited by A. L Johnson, Laura Carbognin & L. Ubertini
Publ.no.151 (1986). price $45
ISBN 0-947571-40-X
Hydrogeology in the Service of Man, volumes 1-4.
Proceedings of the IAH/IAHS Cambridge Symposium,
September 1985
volumes 1-4
Publ.no.l54(1985). price $40 the set
ISBN 0-947571-55-8
Conjunctive Water Use. Proceedings of a symposium held
during the Second Scientific Assembly of IAHS, Budapest,
Hungary, July 1986
edited by S. M. Gorelick
Publ.no. 156 (1986), price $48
ISBN 0-947571-65-5
Karst Water Resources. Proceedings of the Ankara
Symposium, July 1985
edited by Gultekin Gunay & A. Ivan Johnson
Publ.no. 161 (1986), price $45
ISBN 0-947571-90-6
Developments in the Analysis of Groundwater Flow
Systems. Report prepared by a Working Group of die IAHS
International Commission on Groundwater
edited by G. B. Engelen & G. P. Jones
Publ.no. 163 (1986), price $35
ISBN 0-947571-01-9
Groundwater Monitoring and Management. Proceedings
of die Dresden Symposium, March 1987
edited by G. P. Jones
Publ.no. 173 (1990), price $55
ISBN 0-947571-51-5
Please send orders and enquiries to:
IAHS Press, Institute of Hydrology
Wallingford, Oxfordshire OX10 8BB, UK
Telephone: +44491 38800; telex: 849365 hydrol g
Fax: +44 49132256
Please send credit card orders (VISA, ACCESS, MASTERCARD, EUROCARD) and IAHS membership orders to
the Wallingford address only. A catalogue of publications may be obtained free of charge from either o f the above
addresses.
Consequences of Spatial Variability in Aquifer Properties
and Data Limitations for Groundwater Modelling
Practice. Report prepared by a Working Group of the
International Commission on Groundwater
by Adrian Peck, Steven Gorelick, Ghislain de Marsily, Steven
Foster & Vladimir Kovalevsky
Publ.no. 175 (1988), price $45
ISBN 0-947571-61-2
Karst Hydrogeology and Karst Environment Protection.
Proceedings of die IAH/IAHS Guilin Symposium, October
1988
volumea 1 and 2
Publ.no. 176 (1988), price $5
ISBN 0-947571-66-3
Groundwater Contamination. Proceedings of a symposium
held during die Third IAHS Scientific Assembly, Baltimore,
Maryland, May 1989
edited by Linda M. Abriola
Publ.no.185 (1989), price $40
ISBN 0-947571-17-5
Groundwater Management: Quantity and Quality.
Proceedings of die Benidorm Symposium, October 1989
edited by A. Sahuquillo, J. Andreu & T. O'Donnell
Publ.no. 188 (1989), price $60
ISBN 0-947571-32-9
ModelCARE 90: Calibration and Reliability in
Groundwater Modelling. Proceedings of the symposium
held in The Hague, September 1990
edited by Karel Kovar
Publ.no.l95(1990), price $55
ISBN 0-947571-67-1
Groundwater Contamination Risk Assessment: A Guide
to Understanding and Managing Uncertainties. Prepared
by a Working Group of the International Commission on
Groundwater
by Eric Reichard, Carl Cranor, Robert Rancher & Giovanni
Zapponi
Publ.no.l96(1990), price $40
ISBN 0-947571-72-8
Land Subsidence. Proceedings of the Fourth Land
Subsidence Symposium held in Houston, May 1991
edited by A. I. Johnson
Publ.no.200(1991), price $5
ISBN 0-947571-92-2
Hydrological Basis of Ecologically Sound Management of
Soil and Groundwater. Proceedings of a symposium held
during the JUGG Assembly, Vienna, August 1991
edited by H. P. Nachtnebel & K. Kovar
Publ.no.202(1991), price $55
ISBN 0-947571-03-5
Office of the Treasurer IAHS, 2000 Florida
Avenue NW, Washington, DC 20009, USA
Telephone: +1 202 4626900: telex: 7108229300;
Fax: +1 202 3280566
I Hydrogeological
Processes in
Karst Terranes

Edited by

GULTEKIN GUNAY
International Research and Application Center for Karst Water
Resources, Hacettepe University, Beytepe 06532, Ankara, Turkey

A. IVAN JOHNSON
A. Ivan Johnson Inc., 7474 Upham Court, Arvada, Colorado 80003,
USA

WILLIAM BACK
US Geological Survey, MS 431, National Center, Reston, Virginia
22092, USA

Assistant Editors:
HAYATÏ KOYUNCU & HATIM ELKHATÎB
International Research and Application Center for Karst Water Resources,
Hacettepe University, Beytepe 06532, Ankara, Turkey

Proceedings of the International Symposium and

Field Seminar held at Antalya, Turkey,
7-17 October 1990.

IAHS Publication No. 207

Published by the International Association of
Hydrological Sciences.
IAHS Press, Institute of Hydrology, Wallingford, Oxfordshire OX 10
8BB, UK.
IAHS Publication No. 207.
ISBN 0-947571-28-0

The designations employed and the presentation of material

throughout the publication do not imply the expression of any
opinion whatsoever on the part of IAHS concerning the legal
status of any country, territory, city or area or of its
authorities, or concerning the delimitation of its frontiers or
boundaries.

The use of trade, firm, or corporate names in the publication is

for the information and convenience of the reader. Such use
does not constitute an official endorsement or approval by
IAHS of any product or service to the exclusion of others that
may be suitable.

The editors would like to acknowledge with gratitude the

assistance of colleagues at the International Research and
Application Center for Karst Water Resources, Hacettepe
University, Beytepe, Ankara, Turkey, who helped in the
preparation and publication of this volume.

The help provided by Dr Rodier in reviewing the papers in

French was also appreciated.

The camera-ready copy for the papers was assembled and finished at
IAHS Press, Institute of Hydrology, Wallingford, Oxfordshire, UK.

Printed in Great Britain by Galliard (Printers) Ltd, Great Yarmouth.

 V

PREFACE

Limestone and dolomite are common sedimentary rocks that are moderately soluble (50-
400 mg/1) in meteoric waters. These rocks outcrop over approximately 12% of the dry
land area of the earth, being found in every morphostructural and climatic region
(Fig. 1). Calcareous soils formed on them have been preferred for cultivation for
thousands of years. This has led to a disproportionate concentration of people upon the
carbonate rocks, so that today it is estimated that 25% of the world's population depends
largely or entirely upon karst aquifers for their water supplies.

Most outcrops of carbonate rocks display karst dissolutional enlargements to some extent;
many are intensively karstified. The sedimentologic and diagenetic environments of
limestone and dolomite are so varied that the number of distinctly different types of these
rocks that occur is greater than the number of all other types of consolidated sedimentary
rocks. When the fracturing effects of differing tectonic histories in differing regions, and
the episodes of paleokarstification that are often found, are superimposed upon this initial
lithologie variety, it will be appreciated that the range of hydrophysical conditions
encountered in modern karst aquifers can be very large indeed. Some are amenable to
treatment as classical Darcy-flow granular aquifers, other as fracture aquifers or as
double porosity (fracture-granular or wide fracture-narrow fracture) aquifers, while yet
others are purely conduit systems (cavernous) that can be analysed only as complicated
pieces of piped plumbing. Many karst aquifers range between these differing conditions
over distances as little as a few tens or hundreds of metres. The problems of predicting
their behaviour, of managing them and conserving their valuable properties are
correspondingly great.

This complexity was considered at the International Symposium and Field Seminar on
Hydrogeological Processes in Karst Terranes held at Antalya, Turkey, in October 1990.
The papers were divided into sessions that considered karst geochemistry, geomorphology,
(input) hydrology, hydrogeology, quantitative and analytic modelling, groundwater tracing,
remote sensing, geotechnical installations, and regional studies. All of these themes are
represented in the 37 papers from 14 nations that are included in these Proceedings.

Paper sessions at the attractive coastal resort of Kemer were interspersed with one-day
excursions to the Hellenic hill town of Termessus with its cisterns (enlarged caverns) for
the dry season caulked with yolk of eggs, and to the great travertine terraces and springs
above Antalya; then to an astonishing calcareous formation, the Koprucay Conglomerate,
which forms half of a mountain and regional springs with a discharge that never fall
below 30 m3/s even during severe droughts. Following the papers there was a five-day
field seminar across the Taurus Mountains west of Antalya to the shores of the Ionian
Sea. In the interior we were shown major closed basins such as Elmali Polje (karstic
capture of a river valley, with the downstream polje newly replaced by an artificial
tunnel to regulate flow) and the hot springs and travertines of the Pamukkale geothermal
basin. The coast was reached at Ephesus and Koycegiz, a spring-fed lake on the shore.
Two days were spent visiting Gokcesu, Orenkoy and other large springs in karstic
canyons that discharge the Lycian nappes of the western Taurus. The seminar concluded
with an unforgettable voyage along the seashore, spotting mixing zone springs and
visiting the sunken Hellenic city of Kekova, whose fate impressed upon us the tectonic
vigour of this beautiful karstland. All foreign participants will join me in thanking
Professor Gunay and his impressive team of young hydrogeologists at the International
Research Center for Karst Water Resources, Hacettepe University, for their hard work
on our behalf. The arrangements for the meeting were excellent.

The work and the Proceedings of this Symposium should be viewed within two broader
contexts:
VI Preface
 Preface vu

— The neglect of karst in the general practice of consulting hydrogeology. In the

English-speaking world, at least, too many hydrogeology professionals have trained
only in conventional granular and fracture aquifer procedures. They are profoundly
ignorant of the distinct nature of many karst aquifers, despite the latter's widespread
occurrence and economic importance. For example, a recent textbook published in
the United States is advanced, well ordered and comprehensive in most subjects that
it considers, but discusses karst in only four or five pages scattered among seven
hundred pages; yet more than 40% of the USA east of the Mississippi River is karstic
to a significant extent. The erroneous practice, incorrect recommendations and waste
of money that result from this neglect are considerable. Quinlan & Ewers (1985)
have emphasized the need for special training and procedures when dealing with
carbonate terranes. The contributors and readers of these Proceedings are urged to
stress them at all meetings with other hydrogeologists, as well.

— The entire karst system and its fragility. The circulation of groundwater through
pores, fractures and/or caverns is the core, but the whole system is larger than this.
At the groundwater input end are the distinctive landforms created by water sinking
underground, such as karen, corridor karst, dolines and cockpits, dry valleys and
gorges, marginal poljes and internal basins; we saw representatives of most of these
in the Taurus Mountains. At the outlets are springs, springhead gorges, and
travertines. In between, the living aquifer may display river caves with large volumes
of clastic sediment in transit through them. Above the active galleries every mature
karst displays higher levels of older caves with relict sediments and precipitated
secondary minerals. The beauty of such caves has been appreciated by all peoples,
and they have preserved the majority of artefacts and other evidence of early human
culture. In the face of demands from a growing world population with an increasing
ability to alter terrain by engineering works at all scales, every component of the
karst system is threatened. Karen are ripped up for ornamental stone; everywhere,
dolines are treated as pre-dug garbage pits. A polje can appear to be a prefabricated
tailings pond for a bauxite mine, and a dry gorge an ideal site for a hydroelectric
dam. Caves are filled by sediments washed from carelessly developed roads, or
destroyed entirely by quarries. All of this damage diminishes our natural and cultural
heritage and, to return to the theme of the Symposium, it pollutes the aquifers which
we study and wish to manage and conserve. Contributors and readers are urged to
teach their colleagues, farmers, engineers, politicians, etc. and the general public
about the fragility of karst, the dangers of polluting its aquifers, and the difficulties
and costs of restoring them.

Derek Ford
Department of Geography, McMaster University
Hamilton, Ontario, Canada L8S 4K1

REFERENCES
Ford, D. C. & Williams, P. W., 1989, Karst Geomorphology and Hydrology. Unwin-Hyman (now
Chapman & Hall), London.
Quinlan, J. F. & Ewers, R. O., 1985, Groundwater flow in limestone terranes: strategy, rationale and
procedure for reliable, efficient monitoring of groundwater quality in karst areas. In: Proceedings
of the National Symposium on Aquifer Restoration and Groundwater Monitoring, 197-234. National
Water Well Association.
 IX

CONTENTS

Preface by Derek Ford v

1 Environment
Impact of quarries on karst groundwater systems Mehmet Ekmekgi 3

Environmental hydrogeology of a karst system with thermal and normal

groundwaters: examples from the Bursa region (Turkey) Thomas Imbach 7
Effects of hydrogeological development in selective karst regions of China
Lu Yaoru 15

2 Geochemistry, Geomorphology, Geotechnics and Remote Sensing

Karstic hydrogeological investigations for the construction of the Rules

Reservoir (Granada, Spain) J. Benavente, J. Cardenal, J. Cruz-Sanjulian,
S. Garcia-Lopez & A. Nevot 25
A model of karstification in the vicinity of hydraulic structures
W. Dreybrodt 33
Potential of remote sensing techniques in karst areas: southern Turkey
Hatim Elkhatib & Gultekin Giinay 47
Travertine formations in the Antalya area as correlated sediments of
karstic erosional phases in the surrounding Taurus Mountains
Oguz Erol 53
Karst phenomena and dam construction in Greece P. Pantzartzis,
G. Emmanuelidis, L. Krapp & P. Milanovic 65
The influence of geostructural setting upon water thermomineralization in
certain areas of Apulia (southern Italy) Tiziano Tadolini & Giovanni Bruno 75
Storage watertightness problems and technical solutions in the highly
developed karst area of the Niksic Polje, Yugoslavia Dragoljub Zogovic 85

3 Hydrogeology
Origin and catchment area of the Olukkôprû karst springs Mustafa
Degirmenci & Gultekin GUnay 97
First results from the monitoring system of the karstic complex "Grotte di
Frasassi-Grotta Grande del Vento" (central Apennines, Italy)
W. (V. U.) Dragoni & A. Verdacchi 107
Absorption massive d'eau de mer par des aquifères karstiques côtiers
C. Drogue 119
Analysis of sea water intrusion associated with karstic channels beneath
Ovacik plain, southern Turkey Hatim Elkhatib & Gultekin Giinay 129
The hydrogeology of some coastal paleodunes in an equatorial area and
their karst related morphologies: the case of Gesira (Somalia) P. Forti &
F. Francavilla 133
X Contents

Geohydrological behaviour of the Cerritos well system, northeastern

Mexico Jaime A. Tinajero Gonzalez & L. Velazquez Aguirre 139
Alimentation artificielle des systèmes aquifères dans les régions karstiques
Borivoje F. Mijatovic 153
Hydrogeological characteristics and groundwater mismanagement of
Kazerun karstic aquifer, Zagros, Iran Petar Milanovic & Bijan Aghili 163
Karstic hot water aquifers in Turkey Sakir Simsek 173

4 Hydrology
Flood frequency analysis in karst river basins Ertugrul Benzeden,
M. Sibel Olcay & K. Bagali 187
Spatial distribution of precipitation in the western Taurids Lèvent Tezcan &
Alparslan Arikan 203
Effect of karst springs on flow duration curves of rivers Ûnal Ôzis,
Ahmet Alkan & Ergun Unal Tatlioglu 209

5 Modelling
A rainfall-runoff model for large karstic areas Alparslan Arikan &
Lèvent Tezcan 225
Assessment of mathematical models with exponential functions describing
karstic spring discharges Turkay Baran & Nilgttn B. Harmancioglu 231
A conceptual model for the Bey§ehir Lake karst system Mehmet Ekmekçi 245

Numerical versus stochastic modelling of water balance and minimum

discharge of a karst hydrogeological system Neven Kresic, Nebojsa Kukuric
6 Milena Zlokolica 253
Double porosity models in karstified limestone aquifers: field validation
and data provision Martin Sauter 261
An extended double-porosity concept as a practical modelling approach for
a karstified terrain Georg Teutsch 281

6 Regional Systems
Results from studies in several karst formations in southern Catalonia
(Spain) Emilio Custodio, Alfons Bayo, Manel Pascual & Xavier Bosch 295
Use of the physico-chemical variations of karst waters to understand how
an aquifer works: example from the Fontaine de Vaucluse system
(southeastern France) Bernard Blavoux & Jacques Mudry 327
International cooperation in research on karst environments
Brian Finlayson & Ugo Sauro 335
The karst aquifers of Tuscany (Italy) Paolo Forti, Luigi Micheli, Leonardo
Piccini & Giovanni Pranzini 341
Historical development of concepts of regional groundwater flow in the
Floridan aquifer system, southeastern United States Richard H. Johnston 351
 Contents

7 Tracing
Parameters of carbonate rock aquifers from tracer methods A. Plata

An example of the multitracing approach in studies of karstic and coastal

aquifers Luigi Tulipano, Vincenzo Cotecchia & Maria Dolores Fidelibus

Relations entre réponses impulsionnelles et conditions hydrodynamiques

dans le cadre de traçages artificiels des aquifères karstiques: applications
sur colonne de laboratoire et sur un système karstique à double entrée
dans le Causse de Gramat (Lot, France) M. Dzikowski, E. Carlier,
N. Crampon & G. de Marsily

New results on Cretaceous limestone karst in Westphalia, Germany

Eckehard P. Leohnert
1 Environment
Hydrogeological Processes in Karst Terrenes (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHS Publ. no. 207, 1993. 3

IMPACT OF QUARRIES ON KARST GROUNDWATER SYSTEMS

MEHMET EKMEKÇI
International Research and Application Center for Karst Water Resources, Hacettepe University, 06532 Beytepe,
Ankara, Turkey

ABSTRACT

Quarrying of carbonate rocks for various purposes dates back to early times. The impact
of quarries on the environment and particularly on the groundwater system in terms of
quality and quantity has often been ignored until the last few decades. Regardless of the
small amount of quarried rock compared to the volume of an outcrop, the result of
removing the protective cover of an aquifer may cause severe pollution of the
groundwater. Another impact is that groundwater flow patterns may change due to
manmade effects, such as blasting. Sites of quarries, therefore, should be selected by
considering the hydrogeological, environmental and economic factors.

INTRODUCTION
Quarries provide blocks and limestone chippings for the cement industry and the
construction industry. About 30% of Turkey is underlain by carbonate rocks that are
lithologically suitable for such purposes. The site of for a quarry should be studied in
detail for hydrogeological, economic, geological, and technological aspects, which, until
the last decade, have often been ignored.

Carbonate rocks cover productive aquifers in many places particularly where

karstification is well developed. Removal of this cover through quarrying may result in
dramatic changes not only in the groundwater regime but also in the quality of the karst
water. This problem directly affects human health. Quarried landscapes should be
restored for aesthetic reasons.

PARAMETERS TO BE CONSIDERED IN THE SELECTION OF SITES FOR

QUARRIES
Quarrying is one of the oldest methods for utilization of limestone. Until recent times,
the operational economy was the most important factor in the selection of quarry sites.
Generally sites with smooth relief and easy accessability are preferred. Therefore, sites
not far from cities are favoured in order to minimize transportation costs. In Turkey,
carbonate rocks that are lithologically suitable for use in the cement industry can be
found over a vast area. Since remote areas generally have a steep and rugged topography,
these are not suitable. Instead, sites are selected in areas with smooth topography, a
parameter which also affects the selection of sites for the settlement of man.

Some landforms including hills, valleys, caves and closed depressions which are of great
importance in a karst groundwater system are often destroyed by quarries. Geological
factors such as structural elements, the strike and dip of the strata, the underlying and
overlying lithology, and landslide problems are generally less important in the selection
of quarry sites.

Landforms of quarried rock slopes are generally analogous to natural limestone features
like limestone towers, collapse dolines, rock debris chutes, cones and flows which have
resulted from a combination of solutional and mechanical processes (Gagen & Gunn,
4 Mehmet Ekmekçi

1987). The hydrogeological and environmental parameters should also be considered in

the site selection study for a quarry.

IMPACT OF LIMESTONE QUARRYING ON A KARST GROUNDWATER SYSTEM

Regardless of the small amount of quarried rock compared to the volume of an outcrop,
removing the protective rock cover of an aquifer may cause some undesirable results. In
many areas of quarries, the limestone bears a significant amount of groundwater
resources. These potential resources are available in most places for domestic use. In the
case of quarrying a limestone outcrop which acts as a protective cover for the underlying
aquifer, two major changes may occur in the hydrogeological system related to water
quality and the flow system. Where the groundwater flow is in conduit karst aquifers, or
where the water table of a flooded fractured/fissured aquifer is near the surface, removal
of the limestone outcrop leads to contamination. The scar created by a quarry may easily
act as a sinkhole which conveys surface water to the groundwater system rapidly (Gunn
et al., 1985). Quarries surrounded by villages may become a source of pollution for
springs that emerge downstream (Fig. 1).

Fig. 1 - Limestone quarries as sources of water pollution.

The other impact of a quarry is that quarry blasting may result in the destruction or
disruption of groundwater flow paths, changes in the pattern of groundwater movement
and changes in the quantity of water flowing through the karst system (Fig. 2). The flow
path may change direction and contribute to another karst subsystem or spring. Thus the
amount of water abstracted from boreholes fed by the system may decrease significantly.

CASE STUDY OF THE IMPACT OF LIMESTONE QUARRYING AT BEYTEPE,

ANKARA, ON THE QUALITY OF GROUNDWATER
A limestone outcrop of about 30 km2 was selected as a site for quarrying at Beytepe near
Hacettepe University. The limestone unit of Permo-Triassic age is underlain by
impervious Paleozoic greywacke. The limestone is the aquifer utilized by wells and
springs on the Campus and by local people.

There are four active quarries on this outcrop removing the relatively thin cover of the
aquifer. Besides studying the impact on the flow regime of the groundwater, a systematic
sampling programme was implemented particularly for microbiological analyses.
Springs that are downstream of the quarries were found to be microbiologically polluted
while those far from the quarries were still clean. Karst hydrogeological studies showed
that this might be due to the removal of the protective cover. This removal caused a rapid
 Impact of quarries on karst groundwater systems

BEFORE BLASTING

QUARRY

KARST SPRING

DRY OR ALMOST
DRY

Fig. 2 - Impact of quarries on a karst groundwater flow system: (a) before

blasting, and (b) after blasting.

infiltration of surface runoff from the adjacent villages that carried debris containing
animal wastes and other contaminants. The relation between the distribution of quarries
and the polluted springs is shown in Fig. 3.

SCALE!1/25000
O 5Km

LEGEND

Pliocene Clay

Permian-Triassic
Limestone

Paleozoic Greywake

A£ Quarry

1-1
- r Polluted Spring

-^- Borehole
NW Kirkgoz S:

1
'5 ' J ' *

Borehole SW
Quarry
•f f-Polluted spring

Fig. 3 - Hydrogeological map of the study area.

6 Mehmet Ekmekçi

CONCLUSIONS

The karst groundwater system can be altered by limestone quarrying in greater magnitude
and extent than by any other activity of man. This change may be concentrated within a
small area compared to the whole limestone outcrop, but its negative effects on the
quality and quantity of the groundwater resources potential can extend further.

REFERENCES
Gunn, J., Hardman, D. &Lindesay, W., 1985, Problems of limestone quarrying in and
adjacent to the Peak District National Park. Ann. de la Société Géologique de
Belgique, 108, 59-63.
Gagen & Gunn, J., 1987, A geomorphological approach to restoration blasting in
limestone quarries. In: Proc. 2nd Multidisciplinary Conference on Sinkholes and
Environmental Impacts of Karst (edited by B. F. Beck & W. L. Wilson) (Florida,
USA).
Hydrogeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHS Publ. no. 207, 1993. 7

ENVIRONMENTAL HYDROGEOLOGY OF A KARST SYSTEM WITH

THERMAL AND NORMAL GROUNDWATERS: EXAMPLES FROM THE
BURSA REGION (TURKEY)

THOMAS IMBACH
Dr. Werner Balderer Engineering Geology Section, ETH Hbnggerberg, 8093 Zurich, Switzerland

ABSTRACT

Several earthquakes in and around Bursa, NW Anatolia, are known since historical time
and are indications for present tectonic activity. Also hot thermal springs, which have
been used since Roman and Byzantine times flow out on the northern slope of the
Uludag massif within a lateral zone just above the Bursa plain. These springs can be
distinguished by the geographical and geological situation of the outflow zone and also
by the chemical composition of the outflowing water. Based on the hydrochemical data,
a conceptual model was developed which shows there is a common origin for all the
thermal springs. Evidence from isotopic analyses which indicates that there is no
connection between two thermal water districts is explained by a mixing with cold young
groundwater.

INTRODUCTION
Within the framework of a scientific collaboration between the Mining Geology
Department of the Technical University of Istanbul and the Engineering Geology Section
of ETH Zurich (Swiss Federal Institute of Technology), hydrogeological, geological and
geophysical studies were undertaken in the Bursa region as part of the Ph.D. thesis of
the author of this paper.

Several earthquakes in and around Bursa are known since historical time and are
indications for the present tectonic activity. Also hot thermal springs, which have been
used since the Roman and Byzantine epoch flow out at the northern slope of the Uludag
massif within a lateral zone just above the Bursa plain. One aim of this work was to
study the thermal water, the mixing processes with cold groundwater from the Uludag
massif and their relation to recent tectonic activity.

GEOGRAPHICAL INTRODUCTION
Bursa is situated south of the Marmara Sea at the northern slope of the Uludag massif.
The city is located on a huge alluvial fan and on travertine terraces which has made
Bursa famous both in the past and also in recent times.

GEOLOGICAL INTRODUCTION
The northern slope of the Uludag massif, which rises from 300 to 1700 m consists
mainly of marbles, amphibolites and gneisses. This part of the massif is morphologically
characterized by steep, deeply eroded narrow valleys. The central high plateau (1800
m) is formed by plutonic rocks (granodiorite). The highest peaks on the Uludag massif
(2500 m) consists of white marble. The western part of the Uludag massif consists of
deeply altered, slight metamorphic schists on which relics of sedimentary Neogene
deposits can be observed (freshwater limestone, volcanic sediments and tuffs). There are
also two important travertine complexes, which lie discordant on Neogene deposits (Figs
1 and 2).
 Thomas Imbach

Fig. 1 - Geological units on the Uludag massif with the most important
springs and their electrical conductivity (/iS/cm).

Fig. 2 - Location of hot and cold springs in the Çekirge-Bursa area. For
location names see Table 1.

Further information about the geology of the Uludag massif may be found in Ketin
(1985).
 Environmental hydrogeology of a karst system 9

Table 1. Chemical analyses from hot and cold springs in the Çekirge-Bursa area (their
locations are shown in Fig. 2)
m* rTCS9 8s*@ LI t el U k I £3 g^ c& Si k ta Mdt % Cl 8r I M SB m £83 tl i SI
i t t t
O/a (C) S/I * • • • • » • *
6.53 « .13 11.8 <9.2 8.8M 284 0.59 9.128
1 T«l E«Ua
2 K.I
858983 USS
a.i
891686 1391 £3.6 .«8
228
223
22.8
22.8
9.882
8 88!
8 «9
7 «3
0.561
8.»
9.885
8.888
8.855 9.518 3.8
5.5 11.4 M
<2i
8.3 8.59
531
S29 9.125
1.697
1.778
45.37
48.71
ÏS.I S2WH3 1W 63.7 6.31 1 M 223 23.9 9 112 7 « 0.574 9.857 9.857 8.277 5.5 12.3 <9 2 0.869 277 8.5 9.58 529 9.094 1.775 «8.37
stB.in 85t98S 778 29.5 6.SS « .12 69 9.3
e 882 23 M 0.53 8.8St 8.873 1 599 1.3 U.7 <9 2 8 830 77
<0 2 0 859 254
<a 8.31 433 9.856 9.107 16.04
I I U mm 1171 73.7 «.S3 0 .ti 2M 22.8 9 M2 8 «9 0.545 8.815 8.813 9 824 5.2 12.S <E. 9.88 516 9.148 1.841 43.29
SHI emus 1469 76.8 «.SI 0 .«7 218 22.8 0
m 7 93 9.555 8.689 9.M 8 893 5.3 12.9 «J 2 8 859 274 8.2 8.73 S3! 9.688 1.725 45.57
«OtikjUra 898917 U«2 59.1 7.68 9 .55 is* 28.9 8 m U 64 8.581 8.853 0.825 8 839 4.3 13.» <9 2 8 849 232 5.8 8.49
m 9.108 1.487 39.38
7 UT328SU{& 899968 UM 61.5 6.77 « .54
6.C3 « .81
us7 19.8 8 882 9 77 8.4S2 8.154
t.m m 8 4.5 U.3 <8 2 0 858 224
<0 2 0 048 15
1.6 8.38
6.27
449 0.074
0.832
1.419
9.621
39.86
3.78
Ifimtel S9998S Sit 15.1 1.9
• M2 15 53 8.112 8.8U
0.614
e.«M 8 N7 8.1 7.8 2.2 215
9 St^iea 890917 SSI 15.5 7.U 6 .«2 12 t.3 9 8*2 918S 8.135 9.1 9.4 <• 2 0 858 21 8.3 6.48 349 6.620 6.652 5.55
unitb^it 816997 S73 I t . ) 7 . » « .82 li 3.7 9 892 18183 e.us 0.82i 9.1 11.2 <9 2 0 040 25 18.9 8.(2 331 0.020 6.654 5.67
6.7» 0 8.464 8.184 5.4 <0 2 0 659 59 0.2 8.38 0.035
11 Trii&lk» 1
12 fejvg&el
mm HI «5.7
8*9905 S3S «3.1
.es
7.(5 I .Ci
35
32
S.t
4.6
9 882
8 882
21 «5
21 66 8.(87 6.116
8.939 8.234 9.9
8.823 0.114 9.9 5.4 <9 2 0 045 57 0.2 0.25
296
293 0.651
6.676
6.671
15.35
18.94
8.444 19.9 <0 2 0 848 61 5.8 9.28 6.638 9.859 13.63
13 Banx txpscl
mm 537 3S.I .83 39 i.9 9 •82 29 75 6.6JS 9.7 296
14 k m «rill.
mm 563 34.6
15 Sig. sio£ sa 8S0S87 S31 39.8
7.61 ( .83
7.7» 0 .84
28
29
«.8
i.l
9 882
9 882
29 77
22 82
8.449
0.452
9.9»
«.191
9.8
8.6
8.6
9.2
<0 2 0 049 (3
<9 2 0 045 89
0.2
7.6
0.34
0.48
381
273
6.655
6.078
9.053
9.162
13.15
15.62
It Hlli it h>(aU86912 Ut 15.3 7.40 B. [ i . i 1.8 9 882 22 66 9.227 9.914 0 . » 9.844 8.1 4.1 <8 2 0 040 35 <H 8.19 244 6.821 6.064 5.24
0.643 0.213 8.1 <0 2 0 8» 35 5.8 0.604 5.35
ni*m
16 S1|.M11.
8J6SU
m U.l
858963 711 14.8
7.310. Ëb.
(.82 « .81
5
18
1.2
1.3
9 «82
0 «82
29 76
25 122
9.235
9.357
9.911
9.852 8.1
5.8
17.6 <8 2 0 049 43 19.6
0.25
6.31
284
4M
6.049
0.831 0.004 6.73

la&sitary:
Isseiie&gnlesit, IBHisgg&&»g, C3-t3S93 2iMt&
E&. PwwtessassUlt filr laid, Scteas ®£ Isalscèaft, G§-gSâ3 HissssÉff
EsaalGoxirkfca-, fcsicsttrtr. U, ©-£355 Eritis

HYDROGEOLOGICAL INTRODUCTION
The Bursa plain groundwater within the alluvial plain aquifer is recharged mainly by
water from the Uludag massif. Fractures are the most important flow paths in
metamorphic and igneous rocks. Huge karst springs on the border of the plain, the
Karapnar and Pnarba springs, show the importance of the karstified marble series as a
groundwater aquifer above the valley on the Uludag massif. River water from Uludag
is artificially infiltrated into this karstified marble series to intensify the groundwater
resources in the plain. Bursa itself is built on a third important aquifer.

The natural terraced morphology of this complex was used as a fortification by the
founder of Bursa. But only the highest part of this karstified travertine complex is
visible. Drill holes in the plain at Kûltûrpark and Stadion, prove the extent of 100 m
thick travertine deposits, which act as an aquifer with artesian water.

On another travertine terrace not far to the west, also on the border of the plain,
thermal water with a temperature of up to 82°C flows out. These thermal springs within
and around the historical bath of Yeni Kapl'a (Kukiirtlii district) all orginate from a
travertine complex lying on metamorphic rocks. This travertine complex, which is
karstified, can be distinguished from the Bursa travertine complex by a lamination
structure.

The hydrogeological situation of the thermal springs within the second thermal water
district of Çekirge is not well defined because the spring locations are situated in the
urban area. The outcropping rock within the Çekirge district consists of Neogene
sediments and limestone.

The only information about the hydrogeological situation is given by the existing spring
captures and boreholes from DSL The hot thermal water seems to originate from these
rocks, as can be deduced from the old capture gallery of Byzantine or Roman times at
the Vakfbah'e spring.

FIELD PROGRAMME

Geological mapping of the Uludag massif is the basic work of the present studies. The
10 Thomas Imbach

most important springs from all over the Uludag massif were sampled for hydrochemical
analyses and some for isotopic investigations (D, T, <5180) in the summer of 1989
(Fig. 1).
This hydrological information on a large scale was the basis for the geographical
limitation of further detailed field programmes on a smaller scale (Fig. 2). The isotopic
analyses were done at GSF Neuherberg.

RESULTS
The interpretation of the data from the hydrochemical analyses, isotopic investigations,
ideal gas measurements and periodical water investigations in the Bursa-Çekirge area
(Tables 1 and 2) permits the postulation of a conceptual model about origin of the
thermal waters and the mixing processes.

Thermal springs in the Kukiirtlii district (Nr. 17) are characterized by a Na-Ca-
HC0 3 S0 4 water composition. Their total mineralization is about 1 g/l and the maximum
temperature is 82°C (Yeni Kapl'a).
Based on observations from the established monitoring network (through sampling on
the thermal springs and boreholes by DSI) the flow path of the thermal water within a
hillslope aquifer of Kukiirtlii towards the plain can be distinguished from the cold
groundwater within the main Bursa travertine complex, which exits also in confining
layers within the plain (Nr. 910).

The thermal water in the Çekirge district (Nr. 1115) is characterized by a Ca-Mg-Na-
HC0 3 -S0 4 water type with a total mineralization of about 0.5 g/l and a maximum
temperature of 46°C (Vakfbah'e). From the present observations the thermal water in
the Çekirge district also behaves as ordinary groundwater from a hillslope aquifer.

Model I (based on hydrochemical analyses)

The hydrochemical analyses and physical parameters alone are misleading and postulate
a common origin for the two thermal water types which have evolved along different
flow paths. The Yeni Kapl'a thermal water represents the most original thermal water
type and the thermal water of the Çekirge district seems to be a result of mixing of the
original thermal water endmember and a cold groundwater endmember with the chemical
composition of the Inkaya karst springs (Nr. 1617).
J
Dissolved silica-enthalpy graph

SiOj — //
1
[mg/U
//4
:

100

-
50

CGSft] 1 1 1 1 1 1 1 1 I I I , I i n i

30 ISO rq
Fig. 3 - A silica-enthalpy graph (Truesdell & Fournier, 1977) for the
Kukiirtlii thermal water within the hillslope aquifer.
 Environmental hydrogeology of a karst system 11

Table 2. Tritium analysis

Nr. Probe Datua TritiBB
[TBI
1 Yeni Raplica 890905 0.9 +-0,7
2 KOk.I 891004
Kdk.I 890905 1.1 +-0,7
3 KOk.III 890905 3.1 +-0,7
4 KOk.II 890905 2.2 +-0,7
5 Dil 890905 1.6 +-0,7
6 Oto Kaplica 890907 4.5 +-0,7
7 Karaaustafa 890906 4.1 +-0,7
8 Pinarbasi 890908 18.1 +-1,4
9 Stadion 890907 20.0 +-1,4
10 KtUtdrpark 890907 22.4 t-1,6
11 Vakifbabce 1 890906 1.2 +-0,7
12 Boyugdzel 890906 1.8 +-0,7
13 Havuz kaynagi 891004 5.0 +-0,7
14 Havuz drill. 890907 3.8 +-0,7
15 Sig. sicak su 890907 5.6 +-0,7
16 Villa et Hangal 890912 10.9 +-0,9
17 Inkaya 890908 12.1 +-0,9
18 Sig.drill. 890908 18.1 +-1,3

Laboratory:
GeseiUsahaft £0? StiraMea.- <mi ItamltforsclMBg HBBCBCD

Model IT (based on a combination of hydrochemical data and the results from isotopic
analyses)

The low tritium concentration in the Çekirge thermal water does not confirm the single
hot water origin model. Figures 4 and 5 prove that the Çekirge thermal water did not
develop from the hotter Kukurtlii area as it behaves as an ordinary groundwater from
a hillslope aquifer.

The linear trend of the Kiikiirtlii and Çekirge data towards the Inkaya karst water data
indicates there is a mixing with cold water within the hot water spring aquifer
(travertine). For the hot unknown endmember of both thermal water provinces a tritium
content at or below the detection limit has to be admitted.

For the interpretation of the cold water endmember data we need the tritium input
function. The tritium concentration in present meteoric water is not known but the
Inkaya karst water probably represent very young rainfall water. The evidence for a
connection between thermal water and rainfall water is shown in Fig. 6.

cold Bursa gw.

a
a a

Inkaya

O D Kiikiirtki
- Ç«Wro» R °

t.» %A

Fig. 4 - Electrical conductivity-tritium plot for sample number 1-18.

12 Thomas Imbach

D pf@clf*itatlfKi hi Efe^s© ây@© 4. Boyugyz@l ©Lcon*£uctlvity [MS/CRI]

Fig. 6 - Precipitation measurements from DSI in the Bursa area, doganci

barraj, compared with electrical conductivity from thermal water in the
Çekirge district.

Figure 6 shows a positive correlation between rainfall in the Bursa area and electrical
conductivity of the thermal water. The positive correlation without any delay proves an
that the residual hot water with higher electrical conductivity (summer water) is forced
out by a rapid rising of the hydrostatic pressure. The conductivity measurements were
done on 18 October 1989, one day after a heavy rainfall (180 mm in 24 hours) in the
Bursa area. This behaviour provides evidence for a karst or fracture aquifer.

CONCLUSION

New isotope data do not support the first model developed only on hydrochemical data.
This application demonstrates the importance of environmental tritium as a tracer,
especially in the case when old and young groundwaters are mixed.

ACKNOWLEDGEMENTS

We are grateful for the accommodation provided by DSI Bursa. Special gratitude is
expressed to Mr E. Saker, Mr Yuksel and to Mr H. Tezel from DSI Bursa for their
 Environmental hydrogeology of a karst system 13

periodical measurements. Thanks also to WSL Birmensdorf and GSF Munchen for their
collaboration and to Prof. Yiizer and Prof. Shindler for organizing the project.

REFERENCES
Ketin, I., 1985, New developments in overthrust nappe tectonics of Turkey with the
example of Uludag. In: Ketin Simpozyumu 20-21 February 1984, pp. 19-36T.
Truesdell, R. O. & Fournier, R. O., 1977, Procedure for estimating the temperature of
a hot water component in a mixed water by using a plot of dissolved silica versus
enthalpy. J. Research US Geol. Survey, vol. 5, no. 1, p. 4952.
Hydrogeological Processes in Karst Terrants (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHS Publ. no. 207, 1993. 15

EFFECTS OF HYDROGEOLOGICAL DEVELOPMENT IN SELECTIVE KÀRST

REGIONS OF CHINA

LU YAORU
Institute of Hydrology and Engineering Geology, Chinese Academy of Geological Sciences, Zhengding, Hebei,
China

ABSTRACT

In nature hydrogeological conditions are continuously evolving within the hydrosphere-

aerosphere, lithosphère, and biosphere. Karst provides the basis for discussing natural
and artificial effects caused by the actions influencing hydrogeological conditions.
Particularly, pollution, deforestation, desertification and karst collapse disasters are
discussed. Four kinds of geological-ecological environments are classified according to
the environmental features affecting hydrogeological conditions.

INTRODUCTION
Development of hydrogeological conditions is controlled mainly by geological structure and
climate. Commonly their development involves: 1. aquifer formation and destruction;
2. hydrogeological basins and characteristics ; 3. hydrogeological systems and sub-systems ;
4. changes of flow such as velocity, gradient, direction, quantity and pressure; 5. storage;
6. recharge; 7. water quality; 8. water circulation; 9. changes of energy between surface
water and subsurface water and between different aquifers; 10. effects of water processes
related to erosion corrosion and deposition; 11. phase changes between liquid, solid and gas;
12. changes in temperature; 13. changes in chemical features; 14. changesinhydrodynamics.

Human actions have influenced the above factors which are the main features of
hydrogeological systems. In other words, environmental effects result from
hydrogeological conditions that have been generated by artificial influences upon their
natural features.

SIMPLIFIED DISCUSSION ON NATURAL EVOLUTION

Natural geological-ecological environments are within the global system and their
evolution follows the whole global change. It is important to emphasize that natural
evolution of hydrogeological conditions usually have a rapid response to related
phenomena. Natural evolution and development of karst hydrogeological conditions are
complex but their basic features and processes can be simply expressed on maps (Lu
Yaoru, 1986). The essential factors for evolution of hydrogeological conditions include:

- types of karst hydrogeological processes: corrosion, erosion and deposition;

- features of karst hydrogeological processes: chemical, physical and biogenic;

- ranges of karst hydrogeological processes: surface, subsurface, shallow; and
- physical-chemical regime: open, closed systems, normal, high pressure, normal and
high temperature.
Natural karst processes can generate many unfavourable phenomena and related disasters
which can be separated into geological, compound, climatic and biogenic disasters. Most
of these natural disasters have close relationship with different karst processes.
16 Lu Yaoru

DUAL NATURE OF HUMAN-INDUCED ENVIRONMENTAL EFFECTS

Geological-ecological environments have a dual nature, part of which provides favourable

and beneficial conditions for mankind and part provides unfavourable conditions. The
Yunnan, Guizhou, Sichuan, Guangxi, Hunan, and Hubei provinces are typical karst
regions, and may be used as examples (Fig. 1). This large area, within complex
landscapes ranging from high mountains at altitudes of more than 5000 m to a low plain
of only several tens of metres, makes up a sub-system. The dual nature there is
exemplified by: 1. favourable and beneficial conditions, which include energy, mineral,

Fig. 1 Regions of main geological disasters in central-south China and

southwest China:
I. West Sichuan-west Yunnan region activating deep and shallow geological disasters:
I. - Daxue Shan-Cao Yuan strong earthquake, freezing and melting erosion sub-region;
I 2 - Hengduan Shan-San Jiang strong earthquake, strong erosion sub-region; \ - East
Yunnan earthquake, karst disaster sub-region; \ - Jinsha Jiang-Yalong Jiang strong
erosion sub-region; I5 - Min Shan-Qionglai Shan strong earthquake, erosion sub-
region.
II. Yunnan-Guiihcu Plateau-Daba Shan Mountain region activating shallow-surface
disasters: II, - South Guizhou-west Guangxi strong karst, erosion sub-region; 1^ -
Middle Guizhou-Wu Jiang karst-erosion sub-region; II, - Jiuwanda Shan-Xuefeng Shan
karst erosion sub-region; II4 - West Hunan-west Hubei karst erosion sub-region; II5 -
Qinling-Daba Shan karst, strong erosion sub-region; 1^ - Sichuan basin and
surrounding belts less karst, strong erosion sub-region;
III. Guangxi basin-Jianghan Plain region activating shallow-surface special geological
disasters: III, - Guangxi basin strong karst diluvial sub-region; III2 - south Hunan
karst, diluvial sub-region; III3 - Jianghan Plain less karst, diluvial sub-region;
IV. Southeast Guangxi-east Hunan region activating shallow surface geological
disasters (including soil erosion, landslides, and debris flows).
 Effects of hydrogeological development in selective karst regions of China 17

water, and biogenic resources; 2. unfavourable which include the four kinds of natural
disasters. Geological disasters can be separated into: those caused by deep earth crust
movement; those caused by shallow-surface endogenetic and exogenetic dynamic
processes; and those caused by special geological processes such as karst processes.
Considering the different circumstances of the main geological disasters, this subsystem
is divided into several regions and sub-regions shown in Fig. 1.

A dangerous situation is where natural karst processes combine with artificial influences
to deteriorate karst environments in two ways: to harm valuable resources and to
aggravate disasters. Environmental damage mainly originates from artificial effects on
karst hydrogeological processes.

POLLUTION
Because of industrial and agricultural development with little attention to environmental
protection, the environmental quality in many areas is decreasing day by day. Generally,
pollution includes water, atmospheric, and land pollution. Among them, water is most
important and the others are influenced by water movement to cause related
environmental problems. For example, in the belt north of Guiyang, acid rain, with a pH
of only 3-4, increases the solutional capability of surface and subsurface water, decreases
vegetation, promotes soil erosion and causes landslides. The ratios of waste water to the
total water resources in Sichuan and Guizhou provinces are presented in Table 1.

Table 1 Ratios of waste water to total water resources in Sichuan and Guizhou.
Province Annual water resources Annual industrial waste Ratio
(million m 3 /year) water (million m 3 /year)
Sichuan 3 212 000 > 20 470 >6.56% 0
Guizhou 1 022 000 > 5 000 >4.89%o

The comprehensive polluted coefficient P c is:

P c = £ C pi /C ai /n
i=l

where C pi = content of i-th element in polluted water; C ai = content of i-th element in

standard water.
Usually the five elements - phenol, mercury (Hg), lead (Pb), arsenic (As), and cadmium
(Cd) - are used as the standard constituents for studying water pollution. For example,
the polluted coefficient of phenol in a river mouth in southwest China will reach 222 and
the comprehensive polluted coefficients in many belts will be several tens to over one
hundred.

DECREASE OF VEGETATION
As a result of the denudation and construction, the vegetation cover in many belts is
decreasing. For example, the reduced amount of forest in some belts within the Daba
Shan Mountains is listed in Table 2.

The influence of decreasing vegetation upon environment will involve many processes
which are shown in Table 3.
18 Lu Yaoru

Table 2 Reduction of forest cover in percent in some belts of the Daba Shan Mountains.
Belt Wexi Jiange Wanchuan Chuanxi Chengkou

Forest cover rate in 37 45 27 35 40

1957 (%)
Forest cover rate in 10.5 28 14.5 14 15
1985 (%)
Average annual 0.94 0.60 0.44 0.78 0.89
reduction of forest
cover (%)

Table 3 Analysis of trends related to decreasing vegetation in some belts of south China.
DENUDATION
By developing industry and agriculture and other artificial action
DIRECT RESULTS
Annual loss jf water stored in the soil and in the upper vadose zone of rock mass is
about 300 000 nvVkm2
(in karstified mountains in south China)
DIRECTLY CAUSING
drought, flooding, waterlogging
DIRECTLY AND INDIRECTLY PROMOTING
weathering, karst collapse, landslides, soil erosion, debris flows, land subsidence devolution
etc.
DANGEROUS EVOLUTION
TREND TO ROCKY DESERT

KARST COLLAPSES

Karst collapses occur by natural processes and may be caused by artificial factors.
Origins of natural karst collapse are:

— cave collapse caused by erosion, gravitation, or earthquake to form windows or

natural bridges;
— collapse caused by erosion, gravitation, or earthquake to form collapse depressions
or collapsed funnels;
— compound karst collapses caused by erosion, expansion, gravitation, or earthquake
to form collapsed columns;
— earth cave collapses caused by underground erosion, gravitation, or earthquake to
forms collapses in covered soil.
Origins of artificial karst collapse include:
— underground erosion or heavy loading against negative pressured air mass in karst
caves when pumping karst water;
— ground erosion from water leaking from under a reservoir storing water;
 Effects ofhydrogeological development in selective karst regions of China 19

— gravitation and percolation under a storage reservoir;

— bursting high pressured air mass in a karst cave under a storage reservoir;

— explosions.

The most important causes of karst collapse over large areas are decreasing and
increasing karst water level, pumping, draining or storing karst water (Table 4.)

Table 4 Analysis of karst collapses upon the environment.

DECREASING OR INCREASING KARST WATER LEVELS
DIRECT RESULT
Directly causing natural or artificial karst collapses
EVOLUTION OF HYDROGEOLOGICAL CONDITIONS CAUSING OR PROMOTING
Soil erosion, decreasing vegetation, debris flow, pollution, drought, earthquake, landslide, land
subsidence, devolution, waterlogging etc.
MAIN HARM
Deteriorated geological-ecological environment

Table 5 Classification of geological-ecological environments.

Class Evolution trend Main features Regions
Natural geological- geological-ecological original forest and some belts in West
ecological environment good vegetation, only Sichuan and West
environment characterized by natural disasters, and Yunnan of natural
natural evolution good environmental ecology
quality

Contaminated natural environmental less vegetation, industrial zones,

geological-ecological evolution deteriorated serious pollution, many big mining
environment by artificial exploitation without regions, larger urban
influences, which good protection of belts and many
include pollution and environment prevent industrial belts with a
other harmful effects many kinds of bad environment
disasters

Protection of prevention of surface presenting surface- larger cities and

geological-ecological and shallow shallow geological developing industrial
environment geological disasters, disasters, such as zones based on
sustainable landslide and karst reasonable plan
exploitation, and collapse, debris flow, related to
managing less pollution and environmental
environment good environmental protection
quality

Controllable control of deep and preventing surface no typical regional

geological-ecological shallow disasters and and shallow example, need to
environment sustainable geological disasters monitor for
exploitation to and forecasting earthquakes but
improve environment earthquake disasters, forecast with a
good vegetation, and problem of security
excellent
environmental quality
20 Lu Yaoru

CLASSIFICATION OF GEOLOGICAL-ECOLOGICAL ENVIRONMENTS

It is clear that the quality of the geological-ecological «nvironment is influenced by the

dual nature of the factors discussed earlier. The fuzzy mathematics method may be used

Fig. 2 Kinds of evolutions of geological-ecological environments and

related artificial effects.
 Effects of hydrogeological development in selective karst regions of China 21

for evolution of karstified environments.

The evolution of natural environment and artificial effects, particularly the changes of
hydrogeological conditions and considering the main sources of major disasters
percentage of forest cover, rate of soil erosion, and polluted areas. The geological-
ecological environment can be classified into four classes: natural geological-ecological
environment; contaminated geological-ecological environment; protected geological-
ecological environment; and controllable geological-ecological environment. Their
features and characteristics are summarized in Table 5 and expressed in Fig. 2.

For protecting environmental quality, it is necessary to use environmental engineering.

It is important to make a reasonable plan promoting what is beneficial and abolishing

what is harmful based on scientific research. Secondly, according to the natural
conditions and the needs of economic development, environmental engineering includes
three programmes: (a) initiate biogenic projects; (b) building environmental projects to
protect hydrogeological conditions and (c) implementing a waste disposal project.

CONCLUSION
The whole globe as a system in space is naturally influenced by the evolution of the
universe and of the solar system. At present, the artificial effects show the early natural
geological-ecological environments are minimal in the Antarctic Continent, North Pole,
oceans, snow capped mountains, protected natural ecological zones; but most of the lands
in the world contain areas of contamination.

It is possible to protect, monitor and ameliorate contaminated environments as indication

in the above third class. The fourth class will long remain an ideal goal.

REFERENCES
Lu Yaoru, 1986, Karst in China - Landscapes, Types, Rules. Geological Publishing
House, China.
Lu Yaoru, 1987, Karst geomorphological mechanisms and types in China. International
Geomorphology, part II. John Wiley.
Greeley, R., 1985, Aeolian geomorphology from the global perspective: Global mega-
geomorphology. Proceedings of a workshop held at Sunspace Ranch Oracle,
NASA.
Peck, D. L., Troester, J. W. & Moore, J. E., 1988, Karst hydrogeology in the United
States of America. US Geol. Surv. Open-File Report 88-476.
Filson, J. R., 1987, Geological hazards: programs and research in USA. Episodes, 10
(4), 292-295. International Union of Geological Sciences.
2 Geochemistry, Geomorphology,
Geotechnics and Remote Sensing
Hyâroseological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHS Publ. no. 207, 1993. 25

KARSTIC HYDROGEOLOGICAL INVESTIGATIONS FOR THE

CONSTRUCTION OF THE RULES RESERVOIR (GRANADA, SPAIN)

J. BENAVENTE, J. CARDENAL, J. CRUZ-SANJULIAN

Institute) del Agua, Vniversidad de Granada, Rector Lopez Argueta sin, 18071 Granada, Spain
S. GARCIA-LOPEZ
Departamento Geodinamica, Vniversidad de Granada, Campus de Fuentenueva s/n, 18071 Granada, Spain

A. NEVOT
Confederation Hidrografica del Sur de Espana-M.O.P.U., Avenida de Madrid n"' 7, 18012 Granada, Spain

ABSTRACT

The Guadalfeo and Adra Rivers (southeast Andalusia, Spain) drain the crystalline massif
of the Sierra Nevada and, before reaching the sea, cross a mountainous region occupied
by permeable Triassic carbonate materials. The Beninar reservoir on the Adra River is
situated in this mountainous area and is affected by considerable seepage through a series
of sinkholes in the carbonate outcrops. A new reservoir (Rules on the Guadalfeo River)
is planned in a similar, but more complex hydrogeological context. The results of
preliminary research, presented here, raise doubts about the impermeability of such a
dam-site, particularly in relation to the main aquifer (the Los Pelaos Unit) which presents
within the sector three discharge zones of very similar altitude, although their physical-
chemical, isotopic and hydrodynamic characteristics are quite dissimilar. Two of these
discharge zones would be submerged by the reservoir and the possibility exists of leakage
taking place towards the third discharge zone or adjacent areas downstream from the
dam. In any case, the regulation objectives allow for this seepage up to a maximum value
of 2 m 3 /s.

THE REGIONAL HYDROGEOLOGICAL CONTEXT

In the southeast of Andalusia where the climate is semiarid (mean annual precipitation
less than 350 mm), there is a great demand for water, particularly for the irrigation of
highly profitable agricultural products. The permeable materials of most hydrogeological
interest in this region are the Triassic carbonate materials of the Lujar-Gador nappe
(Alpujarride Complex, Internal Zones of the Betic Cordillera).

The outcrops of these materials extend parallel to the coast in the form of a mountain
chain of approximately 1000 km2, with peaks of over 1800 m (Sierra de Lujar) and
2200 m (Sierra de Gador), all of which represents favourable conditions for the recharge
of these aquifers, since the precipitation on these peaks (frequently in the form of snow)
is normally over 600 mm per year, a figure considerably greater than the regional
average mentioned above.

The sequence of the Alpujarride units is made up of a lower set of Palaeozoic and lower
Triassic metapelitic materials (schists, phyllites and quartzites with occasional gypsum
intercalations), and a higher carbonate member of middle and upper Triassic age
(Delgado et al., 1981). The general structure is characterized by the superposition of
thrust nappes. The degree of metamorphism is greater in those nappes of upper tectonic
position.
The Lujar-Gador nappe occupies the lowest tectonic position. In this case, the thickness
of the carbonate formation is more than 1000 m, although it presents some considerable
marly and evaporitic intercalations. The internal structure is one of recumbent folds and
26 J. JBenavente et al.

important inverse faults (Estevez et al., 1985; Campos & Simancas, 1989). Moreover,
the neotectonic activity of the area has produced important tensional fractures (Sanz de
Galdeano et al., 1985). From a hydrogeological point of view, all of the foregoing
implies the existence of highly partitioned aquifers with large reserves, in which different
flow systems can be identified and whose waters have very diverse physical-chemical
characteristics.

The mountain chain is crossed by the Guadalfeo and Adra Rivers (Fig. 1(a)). These rivers
suffer loss to the aquifers in certain sections of their courses, as will be described later.

Fig. 1 - Geological sketchs: (a) general and (b) the studied area.
 Karstic hydrogeological investigations for the Rules Reservoir (Granada, Spain) 27

The high basins of both rivers are situated in the meiapelitic massif of Sierra Nevada,
where the highest altitudes of the Iberian Peninsula are to be found (over 3300 m). The
interest in the regulation of these rivers stems from the irregularity of their discharge,
which is greater at the beginning of summer due to the rapid thaw, and also in early
autumn because of occasionally frequent catastrophic floods.

In the carbonate materials of the Lujar-Gador nappe the karstic forms are not well
developed either on the surface or at the level of speleological networks. The discharge
of the springs is very regular, with great inertia in their response to pluviométrie
recharges. In general, the behaviour of the aquifers is more similar to that of a porous
medium than that of a typically karstic one. The predominantly dolomitic nature of the
carbonate formation with a microfractured texture probably contributes to this behaviour,
as too does the regulating effect due to the important thicknesses (500 m or more in most
of the massifs) attained in the non-saturated zone.

T H E H Y D R O G E O L O G I C A L CONTEXT O F T H E BENINAR R E S E R V O I R

In the Beninar Reservoir on the Adra River (Fig. 1(a)), used for irrigation and domestic
water-supply, water is lost through a series of previously not recognized sinkholes as
they were covered, in the main, by the alluvial deposits of the river. This gives rise to
important, as yet uncontrolled losses, with the result that the reservoir functions only at
a fraction of its initial capacity.

This water drains out some 5 km downstream from the dam through large springs in the
river bed whose normal discharge has tripled as a result of the construction of the dam
(Carrasco et al., 1981). As various attempts at impermeabilization have not produced
totally satisfactory results, it is planned to use pumping wells to recover part of the
losses and return them to the channel fed by the reservoir.

Another dam (the Rules dam) is at present being planned on the Guadalfeo River, in a
similar, although perhaps somewhat more complex geological context to that of the
Beninar Reservoir. There follows a description of the main results of the preliminary
hydrogeological investigation undertaken between February 1989 and May 1990. The
primary objective of this investigation was to analyse the conditions of impermeability
of the projected reservoir in order to avoid the problems encountered with Beninar.

H Y D R O G E O L O G I C A L FEATURES O F T H E RULES DAM AREA

A study was made of a sector of some 40 km 2 around the Guadalfeo River upstream from
Vêlez de Benaudalla (Fig. 1(b)). In this sector the Lujar nappe comprises two tectonically
superposed units (Estevez et al., 1985): Los Pelaos, whose carbonate formation
constitutes the Sierra de Lujar, and the Escalate Unit, superposed on the former and
which occupies a narrow band on the western border of this Sierra. The Murtas nappe,
which in this sector is represented only by its metapelitic formation, appears overthrust
on the Escalate Unit. From the hydrogeological point of view, the most important post-
orogenic materials in this sector are the travertines of Vêlez de Benaudalla and the
alluvial deposits of the Guadalfeo River. The hydrogeological context of the studied area
is sketched in the cross sections of Fig. 2.

Difficulties were encountered for the analysis of the impermeability of the reservoir due
to a series of factors, the most important of which were: (a) the great geological
complexity of the area, which means that the extrapolation in depth of field observations
cannot be relied on; (b) the scarcity of information provided by the electrical geophysical
investigations undertaken, partially as a result of the geological complexity; (c) the
relative failure of the drillings meant to reach the main permeable formation (Los Pelaos)
28 J. Benavente et al.

due to the thicknesses of overlying materials and to the subsequent technical problems
of drilling through them; and (d) the presence of permeable surface formations which
mask the characteristics of the carbonate aquifers and their relations with the Guadalfeo
River.

The main data obtained from the most important springs is collected in Table 1. With the
exception of spring No. 6, the others correspond to the drainage of the Los Pelaos
aquifer, although Nos 2 and 3 in fact appear in material belonging to the Escalate Unit
as a result of a local connection. Considerable differences can be observed in the values
of the discharge and in the physical-chemical and isotopic characteristics of the water.

As a result of the tectonic and stratigraphie features of the region the Los Pelaos aquifer
is partitioned into two sectors of different piezometric levels, so that discharge takes
place at the two following altitude intervals:
(a) Between 300 and 310 m a.s.l. This discharge corresponds to springs Nos 5 and 2.
The first is situated in contact with the alluvial deposits, towards which a

Table 1. Main data of the most important springs.

Springs and Height Discharge (1/s) TDS (mg/1) T 0-18(0/00) Deut.(0/00) Hydrochemical F a d e s
Ref.Number m (a.s.l) mean/range mean value (°C> mean value mean value

1 Vêlez de Benaudalla 182-194 230/130-460 60 18-19 -7.95 -51.0 Calcium bicarbonate

2 EÎ Algarrobo 305 100/0-160 430 18-18.5 Calcium bicarbonate
3 La Colora 185 9/8-11 3200 25-26 -7.81 -50.5 Calcium sulfate
4 Rules 187 140/120-180 1120 18-20 -9.09 -63.0 Calcium sulfate-chloride

5 El Castillejo 303 23/22-26 390 17-17.5 -8.01 -52.0 Calcium bicarbonate

6 Fuente Rafael 335 1.5/0.5-10 390 18-19.5 -6.40 -41.2 Calcium bicarbonate
 Karstic hydrogeological investigations for the Rules Reservoir (Granada, Spain) 29

subterranean discharge must also take place as has been shown by the use of
piezometric data and differential gauging. The discharge of spring No. 2 is highly
variable: it was nil during most of the control period and only began to reach values
of up to 160 1/s after the heavy rains of the winter of 1989.

(b) Between 180 and 200 m a.s.l. The three Velez springs (No. 1), the two at Rules
(No. 4) and that at La Colora (No. 3) all appear in this interval. The Velez springs
rise in a travertine outcrop and are assigned to the Los Pelaos aquifer on the basis
of the size of their discharge since the drillings made cannot confirm the presence
of the aquifer under the Quaternary materials.

The Rules springs appear in the contact with the alluvial deposits of the Guadalfeo River.
Upstream from this point the river loses part of its discharge towards the Los Pelaos
aquifer (Carrasco, 1981), which is why the springs present a mixture of surface water
and water from the aquifer. This is most clearly noticeable as a result of the isotopic
data, since the river water gives the most negative values of the area due to its origin on
the high slopes of Sierra Nevada. This fact even allows for an initial estimate of the
surface water contribution as being 75% of the total volume of the mixture (Benavente
et al., 1990a,b).

The La Colora spring appears in material of the Escalate Unit in a sector with abundant
presence of intercalated gypsum, which explains its hydrochemical characteristics. It
presents a slight thermal anomaly. The Los Pelaos aquifer is here confined beneath
metapelitic materials of the Escalate Unit and as a consequence the water must rise
through fractures parallel to the course of the river. The isotopic content is similar to
that of the Velez springs. See Table 1 for the main data from the most important springs.

The comparison of the discharges recorded with those obtained from previous
investigations (Benavente, 1981) indicates a progressive reduction in the discharge of the
Velez springs since 1981 ( the mean value for the 1972-1980 period was greater than 500
1/s). This has also occurred with the Algarrobo spring, whose activity is increasingly
sporadic and with decreased values of discharge. These circumstances are independent
of the climatic conditions and appear to be related to the increase of pumping of the
alluvial aquifer downstream from Velez de Benaudalla, which would indicate a
connection, for which there is as yet no direct evidence, between both permeable
formations. The piezometric levels indicate a close relationship between the Guadalfeo
River and the alluvial aquifer; downstream from Velez de Benaudalla the river may be
influent (losing water), due to the effect of pumping in the alluvial aquifer.

Differential gauging in the river reveals the existence of both seepage of river water
towards the aquifer and discharge from the aquifer into the river at different points
(Fig. 3). These differences may be attributed to variations in cross section or in the
hydraulic parameters of the alluvial aquifer.

Recharge from the permeable carbonate material to the river-alluvial aquifer system is
also identified. Such recharges bear a relation to the Castillejo, Rules and La Colora
springs; moreover, they suggest variations in the physical-chemical characteristics of the
river water, which are particularly noticeable in the dry season. Thus, the electrical
conductivity of the river water rises from 850 to 1250 /nS/cm on passing through the
Rules springs sector and again to 1350 fiSlcm after the La Colora sector.

REMARKS ON THE IMPERMEABILITY OF THE PLANNED RESERVOIR

The planned reservoir would submerge outcrops of three aquifers: the alluvial, and the
carbonate materials of the Escalate and Los Pelaos Units. In the alluvial aquifer, pumping
 J. Benavente et al.

Fig. 3 - Schematic cross section along the Guadalfeo River showing (a)
the river discharge in summer period; the piezometric levels in (b) the Los
Pelaos aquifer, and (c) the alluvial aquifer. The location of the most
significant springs in this area is also shown ( 1 : Velez; 3: La Colora; 4:
Rules).

tests have shown very high values of permeability (more than 1000 m/day: Benavente et
al., 1990a,b), which, together with the thickness of around 30 m, would require sealing
by means of an impermeable screen penetrating sufficiently far into the substratum.

In the area of the dam, the Escalate aquifer has a thickness of several tens of metres.
Injection tests reveal not very high values of permeability, but the tectonic structure
makes doubtful the continuity of this Unit under the river. Any seepage which might
appear would not, therefore, be important, although impermeabilizing treatment should
be undertaken in order to minimize such a possibility. The gypsum levels which appear
in the proximity of the La Colora spring also appear to be only local in extent.

As regards the Los Pelaos aquifer, a broad sector would be submerged upstream from the
Rules and La Colora springs to an altitude of approximately 240 m (Fig. 2). The results
of injection experiments indicate permeability values of up to 1 m/day in the Rules sector
(Euroestudios, 1983). The system whose drainage is situated between 300 and 310 m does
not, therefore, pose permeability problems. Although outcrops of this aquifer have not
been recognized downstream from the dam, there are springs which drain its resources
(the Velez springs).

The main questions about impermeability of the reservoir concern the drainage system
at altitudes of 180-200 m. Three possibilities should be considered given the present state
of knowledge:

(a) The flow system drained by the Velez springs may be independent from that drained
by the Rules-La Colora springs, there being a mere coincidence in their altitudes.

(b) Some type of hydraulic barrier or hydrogeological groundwater divide may exist
between the two groups of springs, which could be altered as a consequence of the
rise in water level due to the reservoir.
 Karstic hydrogeological investigations for the Rules Reservoir (Granada, Spain) 31

(c) The two groups of springs may correspond to drainage points of one single flow
system.

In the two latter cases a loss of reservoir water would be produced towards the Velez
springs or adjacent sectors (Fig. 2), the amount of which cannot be estimated with the
information at present available.

Apart from the action being considered to solve the remaining problems, it must be
pointed out that, according to the variations in water requirements in the region (winter
irrigation) and taking into account the pumping needs in the alluvial aquifer downstream
from the dam, losses of up to 2 m3/s in the reservoir would be acceptable.

REFERENCES
Benavente, J., 1981, Sobre el drenaje de la Sierra de Lujar (Granada). / Simposio sobre
élagua en Andalucia (Granada, Spain), vol. II, 589-598.
Benavente, J., Cardenal, J., Cruz-Sanjulian, J., Garcia-Lopez, S., Araguas, L. & Lopez-
Vera, F., 1990a, Content analysis of stable isotopes in aquifers from the coastal
mountainous chain of Gador-Lujar (southeast Andalusia, Spain). International
Conference on Water Resources in Mountainous Regions, Lausanne, Switzerland.
Benavente, J., Cruz-Sanjulian, J., Cardenal, J. & Fernandez-Gutirrez, R., 1990b,
Caracterizacion hidrogeologica y efectos de la explotacion del acuifero aluvial del
rio Guadalfeo en el sector de Escalate (Granada). In: XXIII International Congress
IAH: Aquifer Overexploitation (Puerto de la Cruz, Tenerife, Canary Islands).
Campos, J. & Simancas, F., 1989, Los pliegues tumbados del manto de Lujar y su
modificacion por el sistema de cabalgamientos fragiles (Complejo Alpujarride,
Cordillera Betica). Geogaceta vol. 6, 50-53.
Carrasco, F., 1981, Caracteristicas hidrogeologicas del borde Norte del acuifero calizo-
dolomitico de la Sierra de Lujar: Su relacion con el rio Guadalfeo. / Simposio sobre
el Agua en Andalucia (Granada, Spain), vol. I, 233-248.
Carrasco, A., Carrasco, F., Ruiz-Tagle, M. & Thauvin, J.P., 1981, Caracteristicas
hidrogeologicas del acuifero calizo-dolomitico de la Unidad de Gador-Lujar, en el
sector de Berja-Beninar (Almeria) y su relacion con las Fuentes de Marbella. /
Simposio sobre el Agua en Andalucia (Granada, Spain), vol. I, 249-268.
Delgado, F., Estevez, A., Martin, J. M. & Martin-Algarra, A., 1981, Observaciones
sobre la estratigrafia de la formacion carbonatada de los mantos alpujarrides
(Cordillera Betica). Estudios Geologicos vol. 37, 45-57.
Estevez, A., Delgado, F., Sanz de Galdeano, C. & Martin-Algarra, A., 1985, Los
Alpujarrides al Sur de Sierra Nevada: Una revision de su estructura. Mediterranea
vol. 4, 5-32.
Euroestudios, S. A., 1983, Anteproyectode la presa de Rules: Anejo No. 5, Estudio
geologico y geotecnico. Technical Report (unpublished). Confederacion Hidrografica
del Sur de Espana.
Sanz de Galdeano, C , Rodriguez-Fernandez, J. & Lopez-Garrido, A. C , 1985, A strike-
slipfault corridor within the Alpujarra Mountains (Betic Cordillera, Spain).
Geologische Rundschau, vol. 74, 641-655.
Hydroseological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHS Publ. no. 207, 1993. 33

A MODEL OF KARSTIFICATION IN THE VICINITY OF HYDRAULIC

STRUCTURES

W. DREYBRODT
Institut fur Experimentelle Physik, University of Bremen, Pf 330440, D-2800 Bremen 33, Germany

ABSTRACT
To model the development of karst channels from primary fissures in limestone a
computer simulation of solutional widening of a fracture by calcite aggressive water is
suggested. The parameters defining the problem are the initial width a 0 of the fracture,
its length 1 and the hydraulic gradient i driving water through it. The dissolution rates
of limestone determine how fast enlargement of the fractures proceeds. At a calcite
concentration c far from equilibrium they follow a first order rate law F ( 1 ) =
a c
p( eq ~ c ) ; close to the equilibrium concentration c a slow fourth order rate law
pw = /3(c — c)4 is valid. The results show that at the time of initiation the water flow
through theTcarst channels increases slowly in time until an abrupt increase occurs. After
this moment of breakthrough the channel enlarges rapidly and evenly over its entire
length by first order kinetics. Breakthrough times have been calculated for karstification
under natural conditions for low hydraulic gradients as functions of a 0 , 1 and i. Special
attention is given to karstification in the vicinity of hydraulic structures where hydraulic
gradients are high ( > 0.5) and channel lengths are below 200 m. We find that the
breakthrough event will occur after less than 100 years, if (i/1) > 1.1 • 10"7 a0"263
P Co2 " 0 ' 55 , 1 is in m and a0 in cm; (i/1) is given in m . P Co2 [atm] is the C0 2 pressure of
the water entering the fracture. After this event the channels will widen to a width of
about 1 cm within only 10 years, which can cause considerable leakage. Finally, we
discuss critical values of the parameters i, 1, a 0 , which give the conditions of failure in
various types of hydraulic structures.

INTRODUCTION
Primary permeability of carbonate rocks results from an interconnected network of small
fissures and cracks comprising joints, bedding plane partings and eventually also faults.
Initial aperture widths are about 2 • 10"2 cm (Motyka & Wilk, 1984). Once by some
geological event an input-output configuration for surface water is given, which supplies
a hydraulic gradient i, driving surface water through the percolating network of fractures
to an output, solutional attack of the C0 2 containing water starts to widen the fractures
and secondary permeability originates. Figure 1 shows some of the many small fractures
constituting a percolating pathway from a point input to a point output. By solutional
widening a conduit arises which is also shown. This process generally known as
karstification is intimately connected to the chemical kinetics of the system H 2 0-C0 2 -
CaC0 3 (Dreybrodt, 1987,1988).

If dissolution proceeds very fast, the calcite aggressive H 2 0-C0 2 solution very quickly
attains equilibrium and thus cannot penetrate deep into the rock without losing its
solutional power. Therefore only surface denudation should be expected. On the other
hand, if dissolution rates are extremely small, water can penetrate deeply into the rock
but solutional widening is then extremely slow and it may take geologically unrealistic
times until an aquifer, typical for carbonate rocks develops. It is known (Ford, 1988;
Mylroie & Carew, 1986; Bakalowicz, 1982) that minimum times of about 10,000 years
are required until such an aquifer develops to maturity. It is of special interest to model
the development of carbonate aquifers since new land-use programmes of limestone
terrains provide human impact onto these systems and a deeper understanding of their
34 W. Dreybrodt

Fig. 1 - System of fractures with percolating pathway. By solutional

widening a karst channel is created along this pathway from a point input
to the output.

development may prevent damage.

In principle karstification processes under natural conditions and those resulting from
man's impact on karst regions are governed by the same physical and chemical laws and
each model of karstification should describe both. The parameters which determine both
types of karstification and consequently the time scales are quite different. In natural
systems hydraulic gradients at the initiation of karstification are low, usually smaller than
0.1 and the length of percolating pathways subject to solutional widening is in the order
of several hundred metres to several kilometres.
Output-input configurations resulting from hydraulic structures in karst regions, such as
dams or artificial underground reservoirs by plugging karst channels (Milanovic, 1981,
1988), exhibit extremely large hydraulic gradients (>1) and comparatively short
percolating pathways with lengths less than 200 m or so. Generally, it is assumed that
leakage of hydraulic structures results from pre-existing karstic permeability of the rock,
which may have been blocked by sediments and eventually is opened by removal of these
under high water pressure. The question, however, arises whether under extreme
hydraulic conditions solutional enlargement of primary permeability arises in a period
which is short compared to the lifetime of the artificial structure. In this case one would
expect leakage problems which are not apparent when the reservoir is filled, but which
will increase slowly and finally when they are detected might endanger the construction.
It is of interest therefore to model the development of permeability under the conditions
which exist in the vicinity of hydraulic structures.

Recently, the author developed a model from which the times for development of karst
systems can be computed. So far this model has been applied mostly to karstification of
natural systems and has been quite successful in estimating realistic times in accordance
to those inferred from geological observations. A few calculations for artificial conditions
with high hydraulic gradients have also been performed, which show that a considerable
amount of karstification might occur in the order of several ten to several hundred years
(Dreybrodt, 1988, 1989, 1990). Palmer (1988) by a similar approach independently
reached the conclusion that significant leakage can be caused by solutional enlargement
during the lifetime of a hydraulic structure.

This paper describes the model and its physical and chemical parameters. After applying
the model to a case of natural karstification, a study is made of karstification subject to
a variety of conditions as they may by created by man's activity.
 A model of karstification in the vicinity of hydraulic structures 35

THE MODEL

The basic questions to answer are: Provided the hydraulic gradient i between input and
output stays independent of time, what is the time dependence of the flow rate Q(t) of
water driven through the percolating pathway? How does this rate depend on the
geometrical parameters such as the length 1 of the pathway, its average aperture a and
the magnitude of the hydraulic head h between input and output? And furthermore, how
does this system depend on chemical parameters such as saturation concentration c e q and
chemical rate laws describing dissolution.

To give answers to these questions the percolating pathway is simplified to an initially

uniform fracture of rectangular or ellipsoidal shape with an aperture width a 0 and breadth
b 0 . By solutional activity these initial parameters will not stay constant along the length
of the fracture but will become functions of x, the coordinate along the length. Figure 2
shows a part of the fracture, between x and x + dx. The concentration of dissolved
cal cite, given as c = [Ca 2 + ] at the inflow x is c(x) and at x + dx it is c(x + dx) =
c + dc. The flow velocity is v(x). Finally dissolution proceeds by a rate law F(c). The
mass balance now requires:

F(c) P(x) dx = v(x) A(x) dc (1)

stating simply that the amount of calcite dissolved from the wall with perimeter P(x) is
equal to the amount of calcite transported away by flow through the cross-sectional area
A(x). Now from the equation of continuity we have a constant flow rate along the
conduit:

v(x) • A(x) = Q (2)

Integrating equation (1) yields:

( - * L = f P(x) dx (3)
J F(c) J0

Thus one obtains the concentration c as a function of x along the pathway. To calculate
Q one uses the equation of Hagen-Poiseuille for laminar flow (Beek & Mutzall, 1975).
From this a flow resistance R can be defined:

I.
12 n dx
(x, t) (4)
a (x,t)b(x,t)M 0 (x,t)

V(x*dx)

C+dC =
C(x«dx)

Fig. 2 - Fracture along the pathway between positions x and x + dx. The
concentration of the inflowing solution at x is c(x) = c. At x + dx it has
increased by dc and is c(x + dx) = c + dc.
36 W. Dreybrodt

where 17 is the dynamic viscosity, p is the density of water, and g the gravitational
acceleration. M0 is a geometrical factor depending on the ratio b/a with values between
1 and 0.3 for rectangular and elliptical cross sections. The resulting flow rate Q is:
Q = h/R = il/R ; i = h/1 (5)

where h is the hydraulic head and i the hydraulic gradient along the flow path.
To calculate Q(t) as a function of time, one starts with an initially uniform fracture from
which R0 = R(0) can be easily calculated. Then with this value equation (3) is integrated
numerically and the dissolution rates are obtained as a function F(x) = F(c(x)). Using
this function one obtains for the fracture aperture width a(x,t) and the fracture breadth
b(x,t) at later time t:
a(x,t + At) = a(x,t) + -yF(x) At
(6)
b(x,t + At) = b(x,t) + YF(X) At

where 7 is a factor converting dissolution rates from mol cm"2 s"1 into retreat of bedrock
in cm/year and At is a suitable time increment. Thus the time dependence of the flow rate
and of the aperture width along the fracture can be calculated by iteration, if the
dissolution kinetics are known.

DISSOLUTION KINETICS OF CaC0 3

Dissolution rates of CaC0 3 by a calcite aggressive H 2 0-C0 2 solution flowing laminarly
under conditions of a closed system in a completely water-filled fracture of aperture a 0
are given by Buhmann & Dreybrodt (1985):

F») (c) = a(x) (ceq - c); a(x) = a 0 -11 + a ° a ( x ) (7)

6D
The kinetic constant a 0 depends on the initial C02-content of the solution and on the
aperture a. It is almost constant for 5 • 10"3 cm a0 10"1 cm. The effective constant a
considers the influence of mass transport by diffusion for large apertures a. D is the
coefficient of diffusion and its value is 1 • 10"5 cm2 s"1. Depending on temperature and
initial C02-concentration of the solution prior to dissolution of calcite, values of a 0 are
in the range of 1 • 10"5 to 1 • 10"4 cm s . For pure calcite free of magnesium, this rate
law holds up to saturation (Dreybrodt & Svensson, 1989), for natural marble and
Icelandspar containing Mg, however, close to saturation a change in the chemical kinetics
at the surface occurs and dissolution rates follow a higher order rate law (Plummer &
Wigley, 1976; Palmer, 1991):

F(4)(c) = I3(ceq - c) 4 ; B = 1.2 • 1013 cm10 mol"3 s"1 (8)

In our calculations we therefore apply the first order kinetics of equation (7) for
concentrations c < 0.9 c eq and use fourth order kinetics for c > 0.9 c . This implies
an abrupt drop of dissolution rates by two orders of magnitude at c = 0.9 c e q , which in
nature will be smoothed. We will keep it, however, in our model as it is convenient in
the interpretation of the results.

KARSTIFICATION UNDER NATURAL CONDITIONS

Figure 3 shows the profile of a fracture a with 1 = 200 m, b 0 = 100 cm, a0 = 0.01 cm,
subject to a hydraulic gradient of i = 0.1, as it develops with time; a0 = 2.5 • 10"5
cm s"1, iff = 1.2 - 1013 cm10 mol"3 s-1 and ceq = 2 • 10'6 mol cm"3. Note that the scale
of a is logarithmic. In the early stage of evolution, i.e. for times below 2000 years, c =
 A model of karstification in the vicinity of hydraulic structures 37

10%

100 150 200

Length ( m )
Fig. 3 - Evolution of the width a of the fracture along its length. The
widths are plotted logarithmically. Numbers on the curves give the time
in 1000 years needed for the evolution of the profile. After breakthrough
at 5000 years, the profile widens evenly by first order kinetics over its
entire length.

0.9 ceq is attained so quickly that practically fourth order kinetics determine dissolution
along the entire length of the fracture, which increases slowly in width. As a consequence
Q(t) increases in time also, as shown by Fig. 4. The penetration length L s , i.e. the length
water can travel until under the regime of first order kinetics the concentration has
reached c = 0.9 ceq is given (Dreybrodt, 1988) by:
3
a
Q(t) o i r„„i (9)
L,= 1.15' = 9.4- 10
a0. b a0

10 *T
10'I

,—, 1 0 * 1
m :
10*1
*
o 10 4
i

*> 1 0 * 1
aJ
1 0 * 1:
*
10 i

1 1

10" lll|IIIIIMII|lllllllllllllllllll|IIIIIIIH|lllllllll
100 200 300 400 500 6O0
Time (*10 Years)
Fig. 4 - Flow rate through the fracture of Fig. 3 as a function of time.
The steep increase marks the event of breakthrough at breakthrough time
T = 5000 years.
38 IV. Dreybrodt

Thus Ls(t) increases in time and after about 2000 years, where high dissolution rates
determine the further widening of the aperture, starts to penetrate into the fracture, as
can be seen by the two profiles which have developed after 3000 and 4000 years. The
region of fourth order dissolution correspondingly becomes smaller but dissolution rates
are now more evenly distributed along its length. Finally the region of first order
dissolution reaches the output of the fracture and dissolution now quickly widens the end
of the funnel with a dramatic increase in flow rate as shown in Fig. 4. We call this event
breakthrough and the corresponding time is the breakthrough time T. After breakthrough
has occurred, dissolution proceeds with first order kinetics exclusively and solutional
widening is in the order of several 10 2 cm/year and fairly uniform over the entire length.
If finally flow becomes turbulent at conduits widths of several mm due to the influence
of eddy diffusion, a further increase in solutional widening occurs up to about 10 -1
cm/year (Buhmann & Dreybrodt, 1985). This behaviour of time development of the
profile is valid as long as Ls(0) < 1. For the parameters used initially L s = 37.6 cm.
Note that if one assumes first order kinetics to prevail up to c = c eq the dissolution rates
drop by a factor of 10 after each further travel over the length L s , such that after 376 cm,
dissolution rates have dropped to 10"10 of the initial rate. Therefore, if first order kinetics
predominated up to saturation secondary permeability in spatially extended fracture
aquifers would not develop at all. For natural conditions of karstification, calculations
of breakthrough time in dependence on the hydraulic parameters a, 1, and h as well as
on the chemical parameters a0, 15 and ceq are given by Dreybrodt (1990). They are in the
range between 104 and 107 years.
These results can be summarized by an expression for breakthrough times given by:

T = 3.33 • 10"2 a0"2-8 (i/1)"1-3 (years) (10)

which is valid for T > 5 • 103 years and only slightly dependent on the values of a0 and
15 in the region covering natural conditions.

KARSTIFICATION IN THE VICINITY OF HYDRAULIC STRUCTURES

Now we will focus our interest on the calculation of karstification times under the
conditions that exist in the vicinity of hydraulic structures. We have calculated
breakthrough times for a realistic range of the geometric constants i, 1, and a 0 .

The length of the percolating pathway was varied in the range 10 m <. 1 < 200 m.
Hydraulic heads range in the same region. Therefore hydraulic gradients between 0.5 and
10 were considered. The fracture apertures were chosen between 5 • 10"3 cm and 1.5 •
10"2 cm. Unless otherwise specified we used a0 = 2.5 • 10"5 cm s"1 and R = 1.2 • 1013
cm10 mol"3 s"1. We furthermore assume that the initial calcite concentration of the
solution entering the fracture is c ; = 0.

One first very important result of all these many calculations is that the breakthrough
times depend only on the ratio (i/1) of the hydraulic gradient i and the length 1 of the
percolating pathway. This allows us to plot the breakthrough times as a function of (i/1),
which we term the hydraulic ratio.

Figure 5 shows breakthrough times for various values of a 0 , which are given in cm on
the corresponding curves. The equilibrium calcite saturation concentrations cover the
conditions found in lake and river water in karst areas.

It is interesting to note that breakthrough times depend critically on the width of the
aperture. From the curves one can read the critical values of (i/1), above which
breakthrough will occur during the lifetime (100 years) of a dam. At a value of a 0 =
0.01 cm the critical value of the hydraulic ratio is (i/1) = 6 • 10"2 m"1 for the curve with
 A model of karstification in the vicinity of hydraulic structures 39
\ V
\ x \A N \ \ \\
1000i
'A
\\\ \ M \\\ \ \\

100; AH ANA A*\ v

M
v\ A \M
i
10-5
u il
I
0.015 0.01 0.0075 0.005
1- 1 1 1 |
"! r- I -1 1 I1TT i i 1 I 1 ITT
0.01 0.1
i/l (1/m)
Fig. 5 - Breakthrough times T as a function of hydraulic ratio (i/l) for
various fracture widths (numbers on curves in cm) and the following
equilibrium calcite saturation concentrations:
c
eq = l 10'6 mol cm"3;
Ceq = 5
ccq = 2
2.5 . 10"5 cm s'1, I5 = 1.2 1013 cm10 mol"

ceq = 1 • 10"6 mol cm"3. This value increases to 5 • 10"1 irr1, if a0 is lower by a factor
of two, i.e. a0 = 0.005 cm. If one plots the corresponding points (i/l, a 0 ) for a fixed
breakthrough time logarithmically one finds straight lines.

If one assumes the breakthrough time to be 100 years one obtains from the corresponding
line:

(a 0 ) cr = 3.5 • 10"3 (i/l)cr"°-37 (cm) (11)

The numerical value of (i/l) cr is measured in m"1, (a 0 ) cr in cm.

As the next step we compute the dependence of (i/l) cr on the kinetic constant a0. Figure
6 shows the breakthrough time as a function of (i/l) for values of 1 • 10"5 cm s"1 < dn
< 1 . 10 cm s and for various values of ceq as given in the figure caption. a0 = 0.01

x "•
>><V »Vv
x W
1000i \ \\ N ^ \
;
\ \ \-
_^ \ VVV^\\\
*V\
vv s
e 100-i
U VA \ \
0)
>\
I- 10 =
-
\\ Av»\ \\\V\
v\\ \ \ \\ x v

V! \\\A \\\
,M
W \\ \\
1e - 5 2.5e-5 5e-5 1e-4
1- "T~ 1 T 1 ! ! , • • , ,

0.01 0.1
i/l (1/m)
Fig. 6 - Breakthrough time T as a function of hydraulic ratio (i/l) for
various values of <x0 given as numbers on the curves in cm s"1 and
equilibrium concentrations. c „ , labelled as in Fig. 3. a0 = 0.01 cm,
15 = 1.2
40 W. Dreybrodt

cm is kept constant. As in Fig. 5 only a slight dependence of the critical hydraulic ratio
on ceq is seen. The change in (i/l) or upon variation of cc0 however is significant. By
plotting (i/l) cr for T = 100 years and <x0 logarithmically one finds the relation:

ex0 = (i/Ocr1'35 • 8.2 • 10-4 cm s"1 (12)

for ceq = 2 • IO"7 mol/cm3 and a0 = 0.01 cm.

For a closed system Buhmann & Dreybrodt (1985) have calculated the dependence of a 0
on the initial P' Co2 of the solution and its temperature, when it enters a fracture with an
initial Ca-concentration c ; = 0. From these data we find by plotting them logarithmically:

a 0 = 3 • 10-6 P c / J 5
cm s"1 (13)
where P'Co2 i s measured in atm and the temperature is 10°C. Inserting equations (12) and
(13) into equation (11) we find:

(i/l)„ = 7.3 • lu"8 (a0)cr-2-63 F Co2 -°' 55 m"1 (14)

1
where (i/1) is given in m" , a0 in cm and P Co2 in atm. Palmer (1988), by using dissolution
kinetics, derived from experimental data of Plummer et al. (1978) found a similar result:
(i/l) cr p = 8.2 • 10-7 (a0)cr-2-2 F Co2 - 0 - 71 m 1 (15)

Figure 9 shows the critical values from our and from Palmer's expression for P C o 2 =
0.01 atm and P Co2 = 0.002 atm, respectively. For all points above the corresponding
lines breakthrough will occur within the lifetime (100 years) of the hydraulic structure.
If, for example, the percolating pathway below a dam has a value (i/1) = 0.1 m"1, then
one can read from the plot that at P Co2 = 0.002 atm the critical width of the percolating
pathway above which breakthrough will occur is a 0 = 1.4 • 10"2 cm in our work,
whereas from Palmer's data we find a0 = 3.6 • 10"2 cm. Correspondingly, if a0 of the
percolating pathway is given we may read the critical value of (i/l) cr above which
breakthrough will occur. For a 0 = 0.01 cm and P Co2 = 0.002 atm we find from our data
(i/1) = 0 . 3 m"1 but 1.7 m"1 from Palmer's work.

10 -" o.oe

-| 1 1 1 I I I M| 'I I |i F F l'l'U|l I I

ion io™ m *•
betei
Fig. 7 - Breakthrough time T as a function of a in the range between 1012
to IO14 cm10 mol"3 s"1 for various equilibrium concentrations (labelled as
in Figs 3 and 4) and hydraulic ratios given on the curves.
a 0 = 1 • IO"4 cm s"1 for the curves with ceq = 2 • IO"7 mol cm"3 and
a 0 = 2.5 • IO"5 cm s"1 otherwise.
 A model of karstification in the vicinity of hydraulic structures 41

The difference in the two data sets may result from the fact, that the dissolution kinetics
used in this work considers slow conversion of C 0 2 into H 2 C0 3 , which results in smaller
dissolution rates than those used by Palmer (1991). In any way, these are first data for
the development of breakthrough events and a more detailed study will be necessary in
the future.

After breakthrough has occurred widening of the channel is given by:

c c 9
da/dt = a*0»eq
0( eq ~ ) ' 1-2 * 10 (cm/year) (16)
6 3
For values of a o and "eq 1 • 10" mol cm" we obtain widening of the
fracture of about 0.1 cm/year. Therefore it takes only 10 years after breakthrough until
fractures of 1 cm width have developed. This can cause considerable leakage of 3 • 104
cm3 s"1 at i = 1 from a fracture of 100 cm extension b 0 , if turbulent flow is assumed.
In all our calculations we have so far kept the fourth order kinetic constant a at a value
of B = 1.2 • 1013 cm10 mol"3 s"1. It is mandatory, however, to know to which extent the
breakthrough times will change in dependence on a, since this parameter depends on the
chemistry and lithology of the dissolving rock. Figure 7 shows the breakthrough times
in dependence of /3 varied over a range of two orders of magnitude. The values at the /3
curves are (i/1) given in m"1. The curves with ceq = 1 • 10"s mol cm"3 and 5 • 10"7
mol cm"3 are calculated for a0 = 2.5 • 10"5 cm s"1. The curve with ce„ = 2 • 10"7
mol cm"3 is computed by using a 0 = 1 • 10"4 cm s"1. For all curves a0 = 0.01 cm. The
results show clearly that for breakthrough times below 100 years no significant
dependence on a is found.

So far we have assumed that the initial concentration c ; of the water entering the rock is
zero. This, however, is unlikely in hydraulic structures. Therefore as a next step we
explore the dependence of breakthrough time on the initial concentration C; of the

4E-007 6E-007 8E-007 2E-007 .. OEVOÔÔ 2É-007

(Ceq-Ci) (Ceq-Ci) (Ceq-Ci)
Fig. 8 - Breakthrough time T as a function of undersaturation (c eq - C;).
c; is the concentration of the solution entering the fracture. The upper
graph is for c = 1 • 10"6 mol cm"3. The lower left-hand graph is for
C = 5 • 10"7 mol cm"3, and the lower right-hand graph for 2 • 10
mol cm"3. The numbers on the curves give the hydraulic ratios.
Note that in all cases breakthrough time increases drastically with
increasing c;.
If breakthrough time for c ; = 0 is T = 100 years at a given hydraulic
ratio, this ratio must be increased by about 50% to obtain breakthrough
at T = 100 years in the case of c ; = 0.5 c eq .
42 W. Dreybrodt

aggressive solution. Figure 8 shows the breakthrough times as a function of

undersaturation (c eq - c;) for three equilibrium concentrations 10"s, 5 • 10"7 and 2 • 10 "7
mol cm"3. The numbers on the curves give the values of the hydraulic ratio; a 0 =
0.01 cm, a0 = 2.5 • 10"5 cm s"\ fl = 1.2 • 1013 cm10 mol"3 s"1. All cases show a steep
increase with decreasing undersaturation. This implies that with increasing values of c ;
the critical hydraulic ratio for breakthrough to occur at T = 100 years also increases.'
For c ; = 0.5 c the data in Fig. 9 show that (i/1) is to be increased by about 50% in
comparison to the critical values for a situation, where c ; = 0. Therefore to estimate
critical values, instead of equation (14), a better approximation to account for c ; = 0.5

(i/l)„ = 1.1 • 10-7(a0O-'cr

),( 2.63 rp i
Co2
-0.55 _m, - l
(17)
The resulting threshold curves are also shown in Fig. 9. A more detailed investigation
is necessary at this point. In reality, water will attain a certain value of c ; under
conditions of the open system at a given value P C o 2 and then will dissolve limestone
under closed conditions in the fracture. First investigations on the value of a 0 in this case
have been performed by Baumann et al. (1985) and show that a0 depends on both c ; and
p
Co2- I n comparison to the case of c; = 0, a0 is larger by a factor of about 3 in the case
c; = 0.5 c eq . This will also increase breakthrough times.

O.01 -
JZ

O.001

0.001

Fig. 9 - Threshold curves for failure of a hydraulic structure. Solid lines

represent equation (14) (upper curve P' Co2 = 2 • 10"3 atm, lower curve
P'co2 = 1 ' 10 2 at m )- Short dashes represent equation (17) which
considers c; = 0.5 ceq. Long dashes give the result of Palmer's work,
equation (15). In all cases the upper curve of a pair refers to P' C o 2 =
1 • 10"2 atm. Range A gives the region of the hydraulic ratios of grouted
dams; range B refers to zones of high permeability or caves below the
reservoir floor, range C refers to ungrouted arch and buttress dams, and
range D to ungrouted gravity dams. Range E gives the region for
underground reservoirs by plugging cave conduits. If the point (a0> i/1),
where a0 gives the width of the percolating pathway and (i/1) the hydraulic
ratio of the structure, is located above the corresponding threshold curve,
failure may occur.
 A model of karstification in the vicinity of hydraulic structures 43

DISCUSSION
Most important hydraulic structures are dams. Many of them have been built in karst
areas and often failures have occurred by leakage beneath the dam or abutments and also
from the floor of the reservoirs. Palmer (1988) has estimated hydraulic ratios from
existing dams by assuming the length of the percolating pathway to be twice the basal
dam thickness. For gravity dams values between 0.005 to 0.05 m"1 are obtained. Arch
and buttress dams have higher values between 0.1 and 1 m"1. From Fig. 9 one reads
threshold values of a0 in the range from 2 • 10"2 to 5 • 10~2 cm for gravity dams (range
D in Fig. 9), and from 5 • 10"3 to 2 • 10"2 cm for arch and buttress dams (range C).

From the work of Motyka & Wilk (1984), who observed fracture widths in Triassic
limestone in mines, one finds that fractures with widths of about 0.02 cm are most
frequent. Percolating pathways in the critical region of a0 are therefore highly probable
for arch dams and cause a high risk of failure without grouting. If a grout curtain is
constructed with a depth of one third of the maximum water depth H (Milanovic, 1981),
then the length of the percolating pathway will be approximately 1.5 H and (i/1) ~
(0.5/H). This reduces the hydraulic ratio of arch and buttress dams to a region between
0.03 down to 0.002 m"1 (range A in Fig. 9) with critical values of a 0 > 2 • 10"2 cm.
Therefore proper grouting to at least one third of maximum water level is mandatory to
avoid the risk of breakthrough.

If caves or regions of extreme high permeability are located below the reservoir (range
B), high values of the hydraulic ratio will result. At a water depth of 100 m with a cave
50 m below ground (i/1) = 0.06 m"1. For a cave only 20 m below ground (i/1) raises to
a value of 0.25 m"1. Leakage will then occur in very short times. The critical values of
a0 can be read from range B in Fig. 9. Recently Milanovic (1988) discussed the
construction of concrete plugs in cave conduits to use cave systems as underground
reservoirs. Extremely high hydraulic ratios can result (range E), rendering these
structures to a high risk of failure, if no grouting is provided.

CONCLUSION

Hydraulic structures with hydraulic ratios above 0.2 m'1 exhibit a high risk of
unacceptable leakage during their lifetime. By solutional widening of initially narrow
fissures connecting an input point in the reservoir to an output point outside by a
percolating pathway, a breakthrough event can occur within several tens of years. After
this, the fissure widens evenly along its entire length at a rate of up to 0.1 cm/year.
Within 50 years, leakage rates of 0.5 m3 s"1 for a fissure breadth b 0 of 100 cm can result.
Proper geological inspection for distribution of fracture widths at the construction site
and correspondingly grouting is necessary to avoid problems. The increase in leakage
rates after breakthrough depends linearly on the undersaturation (c eq - c;). It is high
where allogenic waters are the main supply to a reservoir.

The results of the model depend on the kinetics of calcite dissolution. In our work slow
conversion of C0 2 into H + and HC0 3 " has been taken into account for the determination
of dissolution rates in narrow fractures. This has not been regarded by Palmer, who uses
higher rates and consequently finds for a given fracture width critical values of the
hydraulic ratio which are higher by a factor between 3 and 5 compared to our results.

More work on dissolution kinetics for chemical reaction paths as they occur in hydraulic
structures is therefore necessary. It is of high importance to model the dissolution
kinetics for the following path: calcite is dissolved at low P Co2-pressures in an open
system. After its concentration has attained the value c ; it enters into narrow fractures
and reacts towards equilibrium. The important result from these investigations will be
44 W. Dreybrodt

the dependence of a 0 on c ; , P' Co2 and the fracture width. Using these results, more
reliable data on critical values can be given. It might also be of importance to conduct
experiments on the reaction rates of the specific limestone from the vicinity of the
hydraulic structures. Investigations into that direction are under way in our laboratory.

The purpose of this work is to raise questions rather than to give answers. From the
model the important parameters which define the development of the secondary
permeability can be found. Field values are needed for proper modelling. Special interest
should be paid to the water chemistry at the floor of reservoirs with respect to P Co2 ,
hardness, and temperature, to the distribution of fractures of given aperture width and
to the parameters determining the dissolution kinetics of limestone.

REFERENCES
Bakalowicz, M., 1982, La genèse de l'aquifer karstique vue par un géochemiste. Reun.
Monogr. Sobre Karst-Larra, vol. 82, 159-174.
Baumann, J., Buhmann, D., Dreybrodt, W. & Schulz, H. D., 1985, Calcite dissolution
kinetics in porous media. Chem. Geol. vol. 53, 219-228.
Beek, W. J., & Mutzall, K. M. K., 1975, Transport Phenomena. Wiley, New York.
Buhmann, D. & Dreybrodt, W., 1985, The kinetics of calcite dissolution and
precipitation in geologically relevant situations of karst areas: 2. Closed system.
Chem. Geol. vol. 53, 109-124.
Dreybrodt, W., 1987, The kinetics of calcite dissolution and its consequences to karst
evolution from the initial to the mature state. Nat. Speleol. Soc. Bull. vol. 49, 31-
49.
Dreybrodt, W., 1988, Processes in karst systems. Physics, Chemistry and Geology.
Springer Series in Physical Environments 4, Springer, Berlin-New York.
Dreybrodt, W., 1989, Karst development in its initial state: A model of speleogenesis.
Proc. 10th International Congress on Speleology (Budapest, 1989) (ed. by A. Kosa),
vol. 1, 174-176.
Dreybrodt, W., 1990, The role of dissolution kinetics in the development of karstification
in limestone: A model simulation of karst evolution. J. Geol., vol. 98, 639-655.
Dreybrodt, W. & Svensson, U., 1989, The effect of MgC0 3 on the solubility of calcite:
enhanced aggressiveness. Proc. 10th International Congress on Speleology (Budapest,
1989) (ed. by A. Kosa), vol. 1, 177-178.
Ford, D. C , 1988, Characteristics of dissolutional cave systems in carbonate rocks. In
Paleokarst (ed. by N. P. James & P. W. Choquette). Springer, New York.
Milanovic, P. T., 1981, Karst Hydrogeology. Water Resources Publications, Littleton,
Colorado, USA.
Milanovic, P. T., 1988, Artificial underground reservoirs in the karst experimental and
project examples. In: Karst Hydrogeology and Karst Environment Protection (Proc.
21st Congress of IAH, Guilin, China, 1988), vol. 1, 76-87. IAHS Publ. no. 176.
Motyka I. & Wilk, Z., 1984, Hydraulic structure of karst-fissured Triassic rocks in the
vicinity of Olkusz (Poland). Kras i Speleologia vol. 14(5), 11-24.
Mylroie, J. E. & Carew, J. L., 1986, Minimum duration of speleogenesis. 9th Congr.
Int. de Espeleol. (Barcelona), vol. 1, 249-251.
Palmer, A. N., 1988, Solutional enlargement of openings in the vicinity of hydraulic
structures in karst regions. Proc. 2nd Conference on Environmental Problems in
Karst Terranes and their Solutions, Assoc, of Ground Water Scientists and Engineers,
3-15.
Palmer, A. N., 1991, The origin and morphology of limestone caves. Geol. Soc. Am.
Bull., vol. 103, 1-21.
Plummer, L. N. & Wigley, T. L. M. (1976) The dissolution of calcite in C02-saturated
solutions at 25°C and 1 atmosphere total pressure. Geochim. Cosmochim. Acta vol.
40, 191-202.
 A model of karstification in the vicinity of hydraulic structures 45

Plummer, L. N . , Wigley, T. L. M. & Parkhurst, D. L., 1978, The kinetics of calcite

dissolution in C0 2 -systems at 5°C to 60°C and 0.0 to 1.0 atm C 0 2 . Am. J. Sci. vol.
278, 179-216.
Hydrogeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHS Publ. no. 207, 1993. 47

P O T E N T I A L O F R E M O T E SENSING TECHNIQUES IN KARST AREAS:

S O U T H E R N TURKEY

HATIM ELKHATIB & GULTEKÎN GÛNAY

International Research and Application Center for Karst Water Resources,
Hacettepe University, Beytepe 06532, Ankara, Turkey

ABSTRACT

Remote sensing is the science and art of obtaining information about an object, area or
phenomena through the analysis of data acquired by a device that is not in contact with
the object or the area under investigation. In modern hydrological and hydrogeological
explorations of karst regions, aerospace technology including remote sensing techniques
is used more and more. The application of remote sensing technology may cover many
fields of studies, especially in geology, hydrology, hydrogeology and mineral
explorations. Generally, these applications may not provide requisite information needed
for investigating a karstic area, but, it can be a tool that appreciably assists in various
planning, development, and management problems which are related to karst regions.
Therefore, remote sensing can provide an extremely valuable data base in any field, as
well as continuing data collection capability that will allow the planners, scientists, or
engineers to perform their functions or works more quickly, accurately, and more
effectively. The present paper indicates the physical bases of remote sensing, describes
remote sensing sensors and platforms, and discusses the role and potential of remote
sensing techniques applied to karst hydrogeological investigations in southern Turkey.
Examples of these applications are concerned with the karst regions of the Kôprûçay
River basin (Antalya) and Ovacik on the Mediterranean coast of Turkey.

INTRODUCTION

Karst regions appear to be more sensitive to environmental changes than other rock
terranes, especially near coastal areas. The application of remote sensing technology can
be a tool that appreciably assists in investigating the hydrogeological processes which
may take place in such coastal karst regions.

This aerospace technique is mainly used for interpreting the distribution of different karst
features and studying the main geological and structural features that affect and/or
control the distribution of groundwater resources in such coastal regions.

The present paper gives an indication of the physical basis of remote sensing, data forms
and selection, and the role and potential of remote sensing techniques applied to the
hydrogeology of karst regions on the Mediterranean coast of Turkey. Examples of
applications concerned with karst hydrogeology are given for the Kôprûçay River basin
and the Ovacik submarine springs on the Mediterranean coast of Turkey (DSÎ-UNDP,
1983; Eroskay, 1968).

P H Y S I C A L BASIS OF R E M O T E SENSING

Remote sensing devices have been developed to exploit a variety of physical phenomena,
including acoustic energy (sound waves), electromagnetic energy (light waves), and the
"force field" of gravity and magnetism. For hydrological and hydrogeological
applications, those designed to exploit electromagnetic energy are considered the most
useful and important.
48 Hatim Elkhatib & Giihekin Giinay

Electromagnetic radiation is of great significance for remote sensing, because it is an

energy from which can be transmitted through all bodies and media which are not opaque
to it. This radiation is emitted from all bodies with temperatures above absolute zero.
Each radiator emits a signal whose configuration is determined by the physical
characteristics of its source. This signal can be described in terms of wave length,
frequency or photon energy which aids in the selection of the most appropriate wave
bands for specified remote sensing objectives. Moreover, the emitted radiation from any
source or target depends on its colour and temperature.

REMOTE SENSING DATA FORMS AND SELECTION

Remote sensing data may be provided in the following forms:

* Image data, in the form of "hard copy" films or paper prints (black-white and
colour).
* Analogue data, continuously variable electrical signals recorded on magnetic tape,
and representing radiation variations along scan lines or selected traverses.
* Digital data, involving radiation values for small individual areas (pixels), and
recorded in map form or on magnetic or punched paper tape.

Generally, Earth Resource Satellites, such as the Landsat Multispectral Scanner (Landsat-
MSS), Thematic Mapper (Landsat-TM), Return Beam Vidicon (RBV), and SPOT, provide
digital and photographic images (Table 1).

Table 1. Characteristics of Landsat images

Image Landsat-MSS Landsat-TM

type
Band no. Wave length Spatial Wave length Spatial
resolution resolution
1, 2, 3 - - 0.45-0.69 30
green-red
4 0.5-0.6 green 80 0.76-0.90 30
infrared
5 0.6-0.7 red 80 1.55-1.75 30
infrared
6 0.7-0.8 near 80 10.14-12.5 120
infrared thermal infrared
7 0.8-1.18 80 2.08-2.35 30
infrared infrared

Before applying remote sensing techniques in karst hydrogeology, the main objectives and
purposes of a study should be known. Fortunately, selection, processing, analysing, and
interpretation stages could aim in achieving these objectives.

During the application, it is important to identify the most suitable time coverage of data
and the acceptable maximum cloud cover percentage expected to be observed in the
selected imagery, especially for the extensive use of Landsat-MSS and TM. These data
 Remote sensing potential in karst areas in southern Turkey 49

may represent dry and wet period conditions in karst areas. This may help in providing
a good overview of the geological structures, karst features, and surface temperature of
water bodies and shallow groundwater in such regions.

ROLE AND POTENTIAL OF REMOTE SENSING TECHNIQUES

In many modern hydrogeological explorations of karst regions, remote sensing techniques

are applied more and more. In the majority of cases, application of remote sensing
techniques cannot eliminate the necessity for detailed hydrogeological mapping, test
drilling, or aquifer testing. But it can expedite such ground activities, improve their
accuracy, and increase the confidence of the field investigators in achieving the main
objectives of their studies.

Fig. 1 - Location map of the study area.

In this paper, examples of the application of remote sensing techniques are concerned
with the karst hydrogeology of the Kôprûçay River basin and the Ovacik submarine
springs on the Mediterranean coast of Turkey (Fig. 1).

CASE STUDY
Kôprûçay basin
Hydrogeological investigation of the Kôprûçay River basin using remote sensing
techniques involves other relevant studies for evaluating the geological and
hydrogeological characteristics of various lithological units, and identifying the main
sources of groundwater recharge of the most important karst springs in the study area.
These springs (Olukkopru and Dumanli) are located in the Kôpriiçay River basin (south
Beysehir-Egridir lakes), and northeast of Manavgat county seat respectively.

Analysis of remote sensing images from Landsat-MSS (band-7) and Landsat-TM (band-5)
covering the research area shows that the main geological and structural features (normal,
reverse, and thrust faults) are extended in the north and northeast-southwest directions,
and affect different lithological units in the study area. These lineament features are
extended along Kirkkavak fault that joints the Beysehir lake and the Olukkopru springs,
where most of the karst features (sinkholes, caves, and springs) are located (Fig. 2).

A comparison of data from Landsat images, geological and structural maps, and
50 Hatim Elkhatib & Giiltekin Gunay

Fig. 2 - Tectonic lineament map of the Kdprûçay basin area.

hydrogeological investigations shows there is a direct relation between the discharges of

the Olukkôprù springs and the groundwater recharge from the Bey§ehir lake. While the
main source of groundwater recharge of the Dumanli spring was related to deep karst
channels extended to Sugla lake at the northeast of Manavgat county seat.

Ovacik submarine springs

In Ovacik, where there is a great demand for irrigation water, groundwater discharges
from karst springs into the Mediterranean sea before its interception for beneficial uses.
Hydrogeological investigations of the Ovacik submarine springs using remote sensing
techniques are considered a tool for studying the main geological boundary conditions,
structural features that may control the different lithological units or water-bearing
formations in a karst area, and its hydraulic connections with these submarine springs.

Black and white prints from Landsat-MSS (band-4) and infrared images of Landsat-TM
(band-5) at 1/100 000 scale, are used for identifying the main geological, structural, and
karst features that control the groundwater distribution and the water resources of the
Ovacik submarine springs (Fig. 3).

Comparison of the field observations, hydrogeological investigations, and lineament maps

derived from Landsat photographs, have shown that most of the intersection images
which are located along the northeastern boundaries of the Ovacik plain, might be a
reflection of the northeast-southwest extended thrust fault and developed fractures in the
underlying beds of the coastal water-bearing formation. The extension and frequent
occurrence of karst features (natural springs, sinkholes, and caves) along the mainland
 Remote sensing potential in karst areas in southern Turkey 51

Fig. 3 - Lineament map of Ovacik (data from Landsat-TM, band-5).

fault line (75°N-80°E) may reflect the main direction of the groundwater flow throughout
the coastal porous units.

The application of remote sensing techniques with the other relevant hydrogeological
studies in this area, has shown that the discharge of the Ovacik submarine springs is
mainly related to underground, highly karstified zones directly connected to the sea
through two main parallel linear features. These features reach the coastline after
crossing the mainland dolomitic limestone unit in the 75°N-80°E direction.

CONCLUSION

Remote sensing capabilities obtainable from spacecraft and high altitude aircraft offer
valuable data for improved development and management planning of different water
resources and more comprehensive regional, interregional, and global hydrogeological
studies.

This capability can increase confidence in the operational procedures of many studies.
Remote sensing technology will be used more and more with the passage of time, and
eventually become an integral part of operational procedures.

REFERENCES

DSI-UNDP, 1983, United nations strengthening DSi groundwater investigative

capability. Phase II. DP/UN/TUR-77-015/1, Tech. Report, Karst Waters of Southern
Turkey. DSi, Ankara.
Eroskay, S. O., 1968, Geological investigation of Kôprûçay-Be§konak reservoir. ElE
Report no. 11-06-5, Ankara.
Hydrogeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHS Publ. no. 207, 1993. 53

TRAVERTINE FORMATIONS IN THE ANTALYA AREA AS CORRELATED

SEDIMENTS OF KARSTIC EROSIONAL PHASES IN THE SURROUNDING
TAURUS MOUNTAINS

OGUZ EROL
University of Istanbul, Institute of Marine Sciences and Geography, Division of Geomorphology, Istanbul,
Turkey

ABSTRACT
The Pleistocene travertine terraces of Antalya are well known karstic features and have
been investigated by several authors. They are principally the correlated sediments of the
karstic erosion in the Korkuteli-Kestel-Bucak area to the north. Some parts of these
travertines have already been attributed to Pliocene, but no travertines older than
Pliocene have been indicated in the area. According to paleogeomorphological studies,
karstic erosion must have started during the late middle Miocene as a part of the
formation of the denudational surface of Erol's D.I system in the Taurus Mountains. This
karstic activity had started as a surficial process under the influence of a warm-humid
climate and penetrated as horizontal extending karstic systems beneath this denudational
surface. These karstic systems have gradually developed in vertical extent, as the base
level lowered and the step-like geomorphological upper Miocene (D.II) and Pliocene
(D.III) denudational surface systems developed. Thus, the intramountain tectonic basins
solutionally deepened and enlarged in limestones and as correlated sediments of this
karstic phenomena, the Asartepe, Tasyayla and Kocaaliler travertine terraces developed
at an altitude of 1200-800 m in the Upper Aksu basin to the north during the late upper
Miocene, and eventually in the Messinian. As for the correlated sediments of karstic
solution in the Tahtali Dag area in the south, the Saylica fan-delta-travertine complex at
an altitude of 700-300 m developed during the Pliocene. The travertine terraces of
Antalya City at - 5 5 to 300 m are the correlated sediments of the karstic solutional
activity in the Kestel-Bucak polje depressions during the Quaternary.

INTRODUCTION
Recent geological and tectonic studies of the Taurus mountains concluded that this area
has been subjected to four episodes of compression, probably in the Upper Eocene-Lower
Oligocène, Langhian, Upper Tortonian and Upper Pliocene-Quaternary periods (Poisson,
1977; Akay et al., 1984; Akay & Uysal, 1988). Each of these tectonic episodes caused
uplift in the Taurus mountains and southward regression of the sea, resulting in a
denudational phase starting on the newly emerged land surface. The next episode then
started with a new transgression, but the coastline of this new phase could not reach up
to its former altitude and remained at a lower position in the south. According to the
author, following each of these tectonic events, denudational surfaces developed in the
Taurus mountains, and in other parts of Anatolia, during late Oligocène, Serravalian,
Messinian and late Upper Pliocene times (Erol, 1981, 1982, 1989, 1990). As the result
of these geological, tectonic and geomorphological phases, relief and denudational
surfaces of different types have developed and are called D.I, D.II, D.III, D.IV phases
of the Middle Miocene to lowest Pleistocene ages and terraces of the Quaternary (Erol,
1981, 1982). The denudational surfaces which developed during these phases, were
uplifted and dissected due to later tectonic activity that caused a lowering of the base-
levels in the continental or marine basins surrounding the denudational uplands. Thus,
step-like denudational surfaces developed on uprising mountain blocks (Erol, 1982,
Fig. 1), and the same geomorphological model can be observed along all the Taurus chain
(Erol, 1989, 1990). Therefore, the travertine formations in the Antalya area, have also
developed as a part of these relief generation phases and they are obviously correlated
54 Oguz Erol

chemical sediments of the karstic erosional phases in the surrounding mountains.

Additionally, by correlating these travertine sediments to the denudational systems in the
mountains one can indicate their relative sequence in time and space.

THE TRAVERTINE TERRACES IN THE ANTALYA AREA

The city of Antalya is situated on the large and well-known travertine terrace of Antalya.
This terrace covers more than 600 km2 (Co§kun, 1985), its maximum thickness is 275 m,
and it consists of nine steps of terraces between —55 and 300 m (Burger, 1985; Nossin,
1988, 1989).

According to different authors, these travertine steps have developed under the influence
of several factors such as travertine sedimentation (Burger, 1985; Co§kun, 1985; Kayan,
1990), karstic underground water levels and lateral karstic corrosion (Burger, 1985),
which may be related to sea-level changes, direct influence of some faultlines (Tietze,
1885; Penck, 1918), or indirect control of some buried northwest-southeast and northeast-
southwest directed lineaments that controlled karstic underground water circulation
(Sesôren, 1985; DSI, 1977, 1985; Guidebook, 1990; Giinay, 1977; ITC, 1971; Yiicel,
1981). According to Aydar & Dumont (1979) these lineaments are younger than
Messinian. The travertines lie on Pliocene sediments in the east (Darkot & Erinç, 1951;
Chaput & Darkot, 1953; Akay et al., 1985; Inan, 1985) and the main body of the
travertines is generally accepted as of Quaternary age.

However, some deeper parts of the travertines may have started to develop in the
Pliocene. The travertine sedimentation must have accelerated during the cooler pluvial
phases of the Pleistocene (Depape & Arènes, 1952; De Planhol, 1952, 1956a,b; De
Vaumas, 1968; Vita-Finzi, 1969). The submarine —55 m travertine terrace developed
during a glacial regression of the Mediterranean sea. These travertine-deposits are
obviously the correlated sediments of the karstic erosion in the Bademagaci-Bucak Polje
depressions, mainly during the Quaternary (Erinç, 1970; Gûldah, 1979).

In spite of the above explained several studies on the Antalya travertines of mainly
Quaternary age, no travertines older than upper Pliocene have been indicated in the
Antalya area. Only Akay & Uysal (1988) and Akay et al. (1984) have described some
travertine layers within the sediments of upper Miocene and Pliocene formations.

During field studies, however, the author observed fairly large travertine outcrops north
of Antalya near the villages of Çamhk, Ta§yayla and Kocaaliler-Hiiseyinler (Fig. 1). The
Asar Tepe travertines near Çamhk village are at an altitude of about 1200 m and lie over
the sediments of the Aksu formation of Tortonian age (see the geological map of Akay
et al., 1984), and they are in direct contact with the Duzagaç conglomerates (Fig. 1).

Akay et al. (1984) have attributed the Duzagaç conglomerates and the Asar Tepe
travertines to the Quaternary, because they are dominant over all the Tortonian marine
sediments and the depression which has developed on them (Fig. 2). After
geomorphological mapping, the author believes that the Duzagaç conglomerates and Asar
Tepe travertines are the marginal sediments of the D.II denudational surfaces of Upper
Miocene (Messinian) age (Figs 1, 3 and 4). They were obviously covering the inclined
surface which had developed during the Messinian after the regression of the Tortonian
sea from the Upper Aksu basin.
The travertines north of Tasjayla village are also a continuation of the same continental
deposits, only they are slightly lower than Asar Tepe due to the original inclination of
the pediment-like surface. These two travertine outcrops have been separated from each
other because of deep erosion of the Tasjyayla valley during the Pliocene and Quaternary.
The third part of the same continental conglomerates and travertines are found near the
 Travertine formations in the Antalya area 55

H* H»

17» » '

B .

» « 1S

-.—IT

Fig. 1 - Geomorphological map of the Antalya area, southwest Turkey.

1 - D.I late middle Miocene (Serravalian) relief; highest karstified limestone plateau
surfaces. 2 - D.II late upper Miocene (Messinian) relief; inclined denudational
pediment-like surfaces at the foot of the D.I plateaux. 3 - D.II late upper Miocene
(Messinian) Dûzagaç formation and related Asar Tepe travertines. 4 - D.III travertine
formations correlated to the pre-Quaternary karstic landform generations. 5 -
Quaternary Antalya travertine terrace. 6 - Quaternary fluviokarstic sediments of the
polje plains. 7 - Holocene valley bottom alluvium. 8 - Sharp crested hills. 9 - Slopes
(full lines: bluffs; dashed lines: foot of slopes). 10 - Altitudes. 11 - Isobath's. 12 -
Approximate position of the coastline of the Messinian sea. 13 - Approximate position
of the coastline of the upper Pliocene sea. 14 - Main roads. 15 - Towns. 16 - Aksu
River. 17 - Smaller rivers.
56 Oguz Erol

Tyfr .«î,^---

Fig. 2 - Asar Tepe travertines of Messinian age. Distant view, looking

from the upper Aksu basin in the northeast.

tftf

Fig. 3 - Asar Tepe travertines, nearer view.

villages of Kocaaliler and Hùseyinler in the south (Figs 1 and 5). Because of the general
inclination of the D.II surface towards the south the travertines are found there at about
800 m altitude.

All these outcrops of the Upper Aksu basin are probably the relics of an extensive
continental cover of conglomerates and travertines of Messinian age and probably were
deposited on the D.II surfaces, which had developed on the Tortonian marine sediments
after the regression of the sea from the former Upper Aksu gulf. In this sense, the
Messinian continental conglomerates and travertines of the Upper Aksu basin are the
correlated sediments of the D.II denudational surfaces, which were surrounding this
basin, and the travertines indicate that karstification was already extensive in the western
mountainous area during the upper Miocene.
 Travertine formations in the Antalya area 57

Fig. 4 - Upper part of the Tasyayla valley, looking from Asar Tepe
towards the north. D.II plateau surface, Dûzagaç conglomerates and part
of the Ta§yayla travertines are seen.

i* r

Fig. 5 - Hiiseyinler travertine terrace, 2 km to the east of Kocaaliler.

If the recent 1500-1700 m altitude of middle Miocene highly-karstified D.I denudational

surfaces and the 1100-1300 m altitude upper Tortonian-Messinian D.II denudational
surfaces around the Bucak-Korkuteli area are studied (see Fig. 1), one concludes that the
karstification must have started under the influence of a warm humid climate on the
limestone outcrops of the middle Miocene D.I denudational surfaces. During the upper
Miocene, because of tectonic movements, some intramountain basins started to develop
and they were deepened by strong weathering under the influence of a warm-humid
climate (for example, see Biidel, 1977). As with the correlated sediments of this karstic
erosion in the west, the Asartepe, Tasyayla and Kocaaliler travertines were deposited in
the upper Aksu basin.
58 Oguz Erol

North of Antalya, following the regression of the sea from the Upper Aksu basin and
relative lowering of the base-level, the travertine formations also shifted towards the
south and were at a lower position during the Pliocene and Quaternary (Fig. 1).
Therefore the highest altitude of the Antalya travertines of Quaternary age is only 300
m. This rapid lowering of the base level also influenced the groundwater level in the
Bucak Korkuteli polje areas; they were enlarged again and deepened at least 200 m, from
1000-1100 m to 800-850 m during the Pliocene and Quaternary.

It is interesting to note that the absence of some higher Pliocene travertine terraces might
be expected between the 1200-800 m altitude upper Tortonian-Messinian travertine
terraces of Asar Tepe-Kocaaliler and the —55 to 300 m altitude Quaternary travertine
terraces of Antalya city. As explained above, the travertine deposits of Pliocene age may
be found in a deeper position at the bottom of the Antalya travertine deposits. This is
probably as a result of the latest subsidence of the recent Antalya Gulf during the
Pliocene and Quaternary (Aydar & Dumont, 1979). This point will be discussed below.
Then a question arises: is there a travertine generation of Pliocene age, between 800 and
300 m? The answer is yes, it is at Sayhca Sirti south of Kemer.

TRAVERTINES OF SAYLICA IN THE KEMER AND TAHTALI DAG AREA

The travertines of Sayhca Sirti and Kepez are about 10 km south of the city of Kemer in
the southwest of the study area (Figs 1 and 6). They are actually seen as a huge coalesced
fan-delta complex consisting of large limestone pebbles and blocks that were cemented
by thick travertine deposits. The surface of these deposits is a fairly steeply inclined
depositional plane, as may be expected from fan-deltas. This surface begins at an altitude
of 700 m at the foot of the Pliocene faultline escarpment of Tahtali Dag in the west, and
lowering through the Sayhca Sirti it reaches the recent coastline about 10 km to the east
(Fig. 7). Accordingly, the inclination of this surface is 70/1000 and it continues
eastwards under the sea. This observation may indicate that this surface had been
developed according to a former base-level that was lower than the recent one. If one
follows the continuation of this surface towards the west, beyond the Pliocene faultline
escarpment of the Tahtali Dag, it can be seen that the erosion surface has extended into
the former mountain valleys.

Fig. 6 - Sayhca travertines south of Kemer, looking from Keçili village

in the north.
Travertine formations in the Antalya area 59

ppe:
CO

Mes inian
CO

o 3
»l
- t,

largei
adien
co
o O
CO CO

icen
u

idd
<u
fi 4-4 O
<u u
o
o ft
S °-i
I
as

chan;
depr
pper
is H
<u
ft 3 bô o
ft 03
3 CO x •-x
Q U
tU 03 >
fi w
03
03
1
^
"« O &
)~~i
X
o3 co~0O
J__J' « H t±1 . - cô

ins; 5 Blul

lected oint
in dia eter
Q Q O

plateau;
ft
CO CO , E

înt ea;
urface
key.
CO

H
5 3 •"!
Q a
ë is s s
ftO fi
ca

car
CO CO * - >T) O
03" 03 X CO
0<N u
CO

S "ft «2 S "S
CO
<+-. CO
Xo
S CO
o 13 X . 2

cène faultline
ars ified imestoni
Dag are, , southwi st of An

ions mailer t
s; 4 Incline
blés; B - Bea

ines; 11 - A l t
rfa es at the foot
!\1 l (Z)

o <i CS
ru O ~
"•4-1

.2 -° B" co
E ft O u! <-<
3
3 co ft P
03
^ T3

ic de
cont
CO tu

ntle
CO
<u CL, X
X! C w
- Rive
«S'a
atio

SO
ces.

X O >- fi
o3 o3
.2 C+-- ' ^H CO
03 C fi U tn i u
3
XI '3 CO co
<* • ~t> 3
avalia n) relief; m

M "O co J3
. ^.
strike ; 10 - Alti
e foot of slopes:
Mode;rate slope
of the Kemer-T

Valle y bottoms
tic se>dimentary
, pedi ment-like

tH T3 CO l i X^ _
ft u. .2 t-i +i 73
03 «J 03 (N *-* fi
Ë C/D c£ M > 03 03

CO O CO
aj
o
CO
c
v-i
'>
3
CO
o s
ft.ÏÏ !•-
Q
^« C
r; 9 - 1
ky slo

o3
'f grad

'5b u
fluvi 1 ter
artly
holo

iddl Mio
pau

ft ft o o u
CO
•—; °*
JOUI

OUI

T3~ o *-. *-*

£0
T3 C C
o E c C.O~CO a «^ 3«
co 03X
O co o c fi
03 _c CO
S* g.s
o oo CO >
r~ g
lief;

1
t-H
co co s « s
£P CO !5 '5 t/2 O c
E Q ES —< U (Ni
60 Oguz Erol

As seen in the detailed geomorphological map of Fig. 7, the Sayhca fan-delta-travertine

complex has been deeply dissected by rivers coming from the western mountains (Fig. 6).
Therefore under the recent conditions it is undergoing erosion. On the other hand, by
considering the Pleistocene terraces lower than 100 m that have developed in the valleys
cutting the Sayhca surface, one concludes that these deposits should be principally of
Pliocene age. The same Pleistocene terraces have also developed superficially on the
submerging easternmost part of the Sayhca surface (Fig. 7).
The steps of secondary importance on the Sayhca surface may be the result partly of the
headward erosion of Quaternary rivers, but some of these steps on the Sayhca surface
may be relics of the original but smaller travertine terraces of Pliocene age. But since
they are fairly well destroyed, it is difficult to show them on a geomorphological map
of today. To consider the origin and the age of the detritic and travertine material
forming the Sayhca fan-delta complex, the study was extended to the west into the higher
parts of the Tahtah and Teke Dag area. The results of this study are shown on Fig. 7 and
are summarized below.

The mountains in the west are up to 2366 m high and consist of mainly Beydagi
autochthonous sedimentary, mainly limestone, formations of Jurassic-Cretaceous age and
allochthonous ophiolitic and limestone thrust bodies of Antalya and Elmah nappes (Brunn
etal., 1970, 1971; Akay etal., 1984; Akay & Uysal, 1988; Erakman, 1987; Erol, 1989,
1990; Giinay, 1990; Inan, 1980, 1985; Kocaefe & Ataman, 1976; Kalafatcioglu, 1973;
Koçyigit, 1985; Poisson, 1977). Because of this complex structure, karstic features have
developed extensively on and in the limestone, while the rivers have eroded the
formations other than limestones, bringing huge bodies of sediments down the slopes.
Therefore, the Sayhca fan-delta-travertine complex developed as a mixture of soluble and
insoluble material.

At the summit region of the Tahtah Dag, Teke Dag and Dazkir Tepe limestone area one
can observe an extensive denudational surface with a well developed doline and uvula
terrain. This karstic area is between 1500 and 2000 m and has an undulating relief. The
dolines and uvulas are principally elongated on northwest-southeast and southwest-
northeast lines, which reflects the control of lineaments of tectonic origin on the study
area (Akay et al., 1984 map; Aydar & Dumont, 1979; Kocaefe & Ataman, 1976,
Sesôren, 19865).

The directions of dry valleys and some slopes also correspond to these lineaments.
Compared with several other similar karstified high plateau surfaces in the summit region
of the Taurus Mountains (Erinç, 1970; Erol, 1990; Louis, 1985), this karstified surface
should correspond to the D.I denudational surface system that developed under the
influence of a warm-humid climate during the late phases of the Middle Miocene, before
the tectonic collision in the south of the Taurus chain in the Serravalian. The extent of
this D.I surface can be seen in Fig. 1.

The primary karstification should have started on the limestone outcrops during the
Serravalian, following an extensive surficial denudational phase on the soluble and
insoluble rocks of this D.I surface. But since the altitude of the D.I surface was not very
high relative to the base level, deep karstic features could not develop and maybe
remained as large dolines or uvulas. During the following stages (D.II and D.III) of
geomorphological evolution, the karstic features extended vertically. Thus, the erosional
landform generations of upper Miocene (D.II) age developed at the foot of the rising
Tahtah Mountains (Fig. 7). Obviously because of the structure of the area, the karstic
solutional features and their correlated travertine sediments were not well developed
during this phase. But north of the study area — for example in the surroundings of
Bucak, the intramountain tectonic-structural basins of early Tortonian age had been
deepened and enlarged easily in limestone formations during the upper Miocene. As
 Travertine formations in the Antalya area 61

correlated sediments of these solutional processes the Asar Tepe, Ta§yayla and KocaaJiler
travertine terraces developed in the eastern neighbouring Upper Aksu basin.

During the Pliocene, as the karstic base-level gradually lowered, the vertical extent of
the karstic solution process also increased. This increasing activity is principally the
result of the increasing rate of tectonic subsidence in the Antalya Gulf southeast of the
study area. The influence of this event is the formation of a high faultline escarpment
east of the Tahtah Dag during the transition between Upper Miocene and Pliocene times.
As a result of this, the Yashca fan-delta-travertine complex started to develop and
gradually enlarged as correlated sediments of the fluviokarstic activity of the
mountainside in the west. Therefore, the dolines and uvulas on the D.I summit plateaux
of Tahtali Dag-Dazkir Tepe may be subjected to some tectonic deformation and the
karstic solutional forms may be deepened again.

The fairly steep gradient of the Yashca surface reflects the active relation between the
uplifting mountains in the west and the subsiding basin of the Antalya Gulf in the east.
If the development of the Pleistocene terraces in the Kemer area and the large travertine
terraces of Antalya city is considered, one concludes that the tectonic subsidence of the
Gulf was more active at the beginning of the Pliocene, gradually decreased and ended
during the Quaternary.

GEOMORPHOLOGIC DEVELOPMENT OF THE ANTALYA GULF

The Antalya Gulf in the southeast of the study area seems to be a subsiding counterpart
of the uprising mountains in the north. As explained above and shown in Fig. 1, the sea
had invaded the Upper Aksu basin during the Middle and Upper Miocene. It had receded
at the end of the Upper Miocene, eventually during the late Messinian from this former
Gulf and extensive denudational surfaces have developed.

According to recent geophysical studies in the Gulf of Antalya (Ôzhan, 1983), the
Messinian denudational surfaces extend as a clear unconformity at the base of the Plio-
Quaternary sediments of the Antalya Gulf. According to this observation and the
geological map of Akay et al. (1984), the Pliocene sea invaded the gulf up to the north
of Aksu town and the coastline receded to the present position. This recession must have
been accompanied by some tectonic movements, because the sediments of Pliocene
formations are found today above sea level in the Lower Aksu plain. But the same
sediments are much deeper in the south of the recent Antalya Gulf (Ôzhan, 1983) and
there are some traces of synsedimentary tectonic distortion in the Plio-Quaternary
sediments of the gulf.

Additionally, traces of a submarine canyon filled with Plio-Quaternary sediments are also
observed in geophysical profiles (Ôzhan, 1983). This may indicate that following the
tectonic distortion of the Messinian denudational surface, the Antalya Gulf had deepened
considerably. Today this surface is found up to an altitude of 1300-1200 m in the
mountains, its continuation is seen at 250 m south of Gebiz (Fig. 1), and it is found
deeper than —1200 m in the Antalya Gulf (Ôzhan, 1983, bathymétrie map and profiles).
The Messinian denudational surface, because of its continental pediment-like character,
has an original gradient, but it is believed that it had been also subject to considerable
tectonic distortion (Erol, 1990).
The inclined Sayhca travertine plain must have developed according to the lower sea
level of earlier Pliocene age and it may have been submerged later as the sea level rose
during the late Pliocene and early Pleistocene. Then it was dissected by rivers during the
later stages of the Pleistocene. Additionally, the submarine (—55 m) travertine terrace
62 Oguz Erol

at Antalya (Fig. 1) may be an indicator of some Pleistocene low levels of the sea surface.
In contrast to the tectonic distortion of the Messinian surface in the Antalya Gulf and its
north, the relics of D.II and even D.I surfaces are not very distorted on the mountain
blocks (Fig. 1). This may be proof of block faulting of the area after Messinian during
the Pliocene and Quaternary.

CONCLUSION
Based on paleogeomorphological studies in the Antalya area, it is concluded that the
karstic features are part of the general geomorphological landform system, and the karstic
activity probably started during the late Middle Miocene as surficial and horizontal
karstic phenomena on the D.I surfaces. These karstic systems gradually developed
vertically, as the base-level lowered and the step-like geomorphological upper Miocene
(D.II) and Pliocene (D.III) denudational surface systems developed. Thus, the
intramountain tectonic basins were solutionally deepened and enlarged in limestones.
Additionally, correlated sediments of these karstic phenomena, the Asartepe, Tagyayla
and Kocaaliler travertine terraces, developed at 1200-800 m in the Upper Aksu basin in
the north during the late upper Miocene. As correlated sediments of karstic solution in
the Tahtah Dag area, the Sayhca fan-delta-travertine complex at 700-300 m developed
during the Pliocene. The travertine terraces of Antalya city at —55 to 300 m are the
correlated sediments of the karstic solutional activity in the Kestel-Bucak polje
depressions during the Quaternary.

REFERENCES
Aydar, C. & Dumont, J. F., 1979, Observations sur images Landsat d'alignements dans
les travertines d'Antalya: discussion des relations probables entre neotectonique et
hydrogéologie. Bull. M.T.A., 92, 81-84.
Akay, E., Uysal, S., Poisson, A., Cravatte, J. & Mûller, C , 1984, Stratigraphy of the
Antalya Neogene basin. Tttrkiye Jeoloji Kur. Bait., 28.2, 105-120.
Akay, E. & Uysal, Û., 1988, Post-Eocene tectonics of the Central Taurus Mountains.
Bull. M.T.A., 108, 23-34.
Brunn, J. H., Graciansky, P.-C. de, Gutnic, M., Juteau, T., Lefevre, R., Marcoux, J.,
Monod, O. & Poisson, A., 1970, Structures majeures et correlations stratigraphiques
dans les Taurides Occidentales. Bull. Soc. Géol. France, 7.3, 515-556.
Brunn, J. H., Dumont, J. E., Graciansky, P.-C. de, Gutnic, M., Marcoux, J., Monod,
O. & Poisson, A., 1971, Outline of the geology of the Western Taurids. In: Geology
and History of Turkey (ed. by A. S. Campbell), 225-256. Tripoli.
Burger, D., 1985, Der travertin komplex von Antalya. Sudwestturkei. Ann. de la Soc.
Geol. Belgique, 108, 197-202. Comptes Rendus du Colloque International de
Karstologie Applique, Liège, mai-juin 1984.
Bûdel, J., 1977, Klima-Geomorphologie. Stuttgart.
Chaput, G. & Darkot, B., 1953, Sur la Pliocene des environs d'Antalya, Turquie. C.R.
des Seances de l'Acad. des Sci.Paris, 2259-2260.
Co§kun, N., 1985, Kirkgôz springs and Antalya travertine area. In: Karst Water
Resources (Proc. Antalya-Ankara Symp., July 1985), Abstract publication, p. 48.
Darkot, B. & Erinç, S., 1951, Terrasses de Travertin d'Antalya à l'ouest d'Aksu. 1st.
Univ. Cografya Enst. Derg., 1.2, 56-65.
Depape, G. & Arènes, J., 1952, Contribution à l'étude des flores fossiles d'Asie
Mineure. Bull, du Museum Paris., 2 e série, 24.3, 328-332.
De Planhol, X., 1952, Les terrasses de la plaine pamphylienne et le rajeunissement post-
glacière du Taurus. In: Volume jubilaire 50e Anniversaire Laboratoire de Géogr.
Univ. Rennes, 420-428.
De Planhol, X., 1956a, Contribution à la étude géomorphologique du Taurus occidental
 Travertine formations in the Antalya area 63

et de ses plaines bordières. Rev. Géogr. Alpine, 44, 609-685. Grenoble.

De Planhol, X., 1956b, Le Pliocène et la Quaternaire de la plaine d'Antalya. Resumenes
20 Congr. Geol. Int., 70-71. Mexico.
De Vaumas, E., 1968, Phénomènes karstiques en Méditerranée Orientale. Mémoires et
Documents 1967, 4, 193-281. Centre de Rech. Doc. Cartogr. et Geogr. Paris.
DSI, 1977, Antalya travertine area. Dye test preliminary Report. Ankara.
DSI, 1985, Antalya - Kirkgôz springs and travertine plateau karst area investigation
Report. Devlet Su îsleri Yayini, Ankara.
Erinç, S., 1970, Tûrkiye Kuvaterneri ve jeomorfolojinin katkisi. Jeom. Derg. 2, 2-35.
Ankara.
Erinç, S., 1977, Geomorphological evidence of neotectonics in Turkey. Review of the
Geogr. Inst of Istanbul Univ., 15, 37-42.
Erol, O., 1981, Neotectonic and géomorphologie evolution of Turkey. In: Neotectonics
(ed. by R. W. Fairbridge). Z. / Geomorph., Suppl. Bd 40, 193-211.
Erol, O., 1982, Bati Anadolu genç tektonigimnjeomorfolojik sonuçlan. In: Ban Anadolu
genç tektonigi ve volkanizmasi paneli (ed. by O. Erol & V. Oygûr). Tûrkiye Jeoloji
Kurumu Yayini, Ankara.
Erol, O., 1986. The relationship between the phases of the development of the Konya-
Karapinar obruks and the Pleistocene Tuzgôlû and Konya Pluvial lakes. In: Karst
Water Resources (Proc. Ankara-Antalya Symp., July 1985) (ed. by. G. Gûnay & A.
I. Johnson), 207-213. IAHS Publ. no. 161.
Erol, O., 1989, Geomorphological evolution of the Taurus Mountains, Turkey. Second
International Conference on Geomorphology, September 1989, Frankfurt. Abstracts
of Papers and Posters, 85. Geoôkoverlag Bensheim.
Erol, O., 1990, Messinian palaeogeomorphology and neotectonics of the Western Taurus
Mountains. 8th Petroleum Congress of Turkey, April 1990. Extended Abstracts, 91.
Guidebook for Field Excursions and Field Trip, 1990, Int. Symp. and Field Seminar on
Hydrogeological Processes in Karst Terranes, October 1990, 14-22. Kemer-Antalya,
Turkey.
Gûldali, N., 1979, Géomorphologie der Tiirkei. Erlaeuterungen zur geomorphologischen
Ubersichtskarte der Tiirkei. 1:2 000 000 Beihefte zum Ttibinger Atlas des Vorderen
Orients. Reihe A. Nr 4, Wiesbaden.
Gûnay, G., Altug, A., Dogan, L. & Coskun, N., 1977, Antalya Plain Biyikh sinkhole
system dye test. DSI-UNDP Project (TUR/77/015) no. 18.
Gûnay, G., 1990, The role of orogeny and related tectonics on karstification process in
Taurus karst belt. In: Hydrogeological Processes in Karst Terranes (Proc. Antalya
Symp. and Field Seminar, October 1990). Abstract publication.
Inan, N., 1980, Antalya travertenleri. MTA Rapor no. 7175.
Inan, N., 1985, Antalya travertenlerinin olusumu ve ôzellikleri (The features and genesis
of Antalya travertines (abstract). Jeoloji Mtihendisligi, 24, 31-39. Ankara.
ITC-UNESCO Centre, 1971, An Integrated Survey for the Land and Water Development
in the Aksu and KirkgOzler-Duden River Basins: Antalya Region, Turkey. ITC
Enschede.
Kocaefe, S. & Ataman, G., 1976, Seismotectonic events of the Anatolia. 1, Investigation
of the region in the Antalya-Fethiye-Denizli Triangle (summary). Yerbilimleri
Hacettepe Univ., 2.1, 55-70. Ankara.
Koçyigit, A., 1984, Gûneybati Tûrkiye ve yakin dolayinda levha ici yeni tektonik geli sim
(Intra-plate neotectonic development in southwestern Turkey and adjacent areas
(abstract). Ttirkiye Jeol. Kur. Bttlt., 27.1, 1-16. Ankara.
Kalaftacioglu, A., 1973, Geology of the western part of Antalya Bay. Bull. MTA, 81,31-
84. Ankara.
Kayan, I., 1990, Tarihôncesi yerlesme yerleri olarak Antalya magaralannmjeomorfolojik
ôzellikleri. Ege Cografya Dergisi, 5, 10-32. Izmir.
Louis, H., 1985, Landeskunde der Turkei: Vornehmlich Aufgrund eigener Reisen.
Geographische Zeitschrift Beihefte, 73. Franz Steiner Verlag, Stuttgart.
Nossin, J. J., Sesôren, A. & Kuran, I. H., 1988, PEPS-92: Karst and submarine
64 Oguz Erol

freshwater discharge in the Antalya area, southern Turkey. In: SPOT-1 Image
Utilization, Assessment, Results, 719-730. CNES, Cepadues, Toulouse.
Nossin, J. J., 1989, SPOT stereo interpretation in karst terrain, southern Turkey. ITC
J., 2, 79-91. Enschede.
Ôzhan, G., 1983, Kiyiôtesi jeofizik verilerin i§iginda Antalya Kôrfgezinin jeolojik
yorumu (Geological interpretation of the Gulf of Antalya in the light of the marine
geophysical data (abstract). Tttrk. Jeol. Kong. Bait. 4, 47-50.
Penck, W., 1918, Die Tektonischen Grundziige Westkleiasiens. Engelhorn, Stuttgart.
Poisson, A., 1977, Recherches géologiques dans les Taurides Occidentales, Turquie.
These Docteur des Sciences, l'Univ. de Paris Sud (Centre d'Orsay), Paris.
Sesôren, A., 1986, Potential of remote sensing use in a karstic area, Antalya Region in
the south of Turkey. In: Karst Water Resources (Proc. Ankara-Antalya Symp., July
1985) (ed. by G. Gunay & A. I. Johnson), 271-280. IAHS Publ. no. 161.
Tietze, E., 1885, Beitrage zur Géologie von Lykien. Jahrb. der Kaiserl. Kônigl. Geolog.
Reichsanstalt Wien, 35, 283-386.
Vita-Finzi, C , 1969, Late Quaternary continental deposits of central and western
Turkey. Man 4.4, 605-619. London.
Yûcel, A. M., Engin, T., Coskun, N., 1981, Antalya denizi kiyi sulanmn diçkisal
kirlenme yônûnden araçtinlmasi. Tiirkiye Bilimsel ve Teknik Ara§t. Kur. Proje no.
TAG 367.
Hydrogeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHS Publ. no. 207, 1993. 65

KARST PHENOMENA AND DAM CONSTRUCTION IN GREECE

P. PANTZARTZIS, G. EMMANUELIDIS
Hydro-Project Development Department, Public Power Corporation, 4 Korai Street, 10564 Athens, Greece

L. KRAPP
RWTH, Aachen, Germany

P. MILANOVIC
Energoprojekt, Lenjinov Bulevar 12, 11070 Belgrade, Yugoslavia

ABSTRACT
The report gives information about three dam projects in Greece with major karst
problems and the applied sealing measures.

INTRODUCTION
Limestone formations are found in many parts of Greece, mostly with deep reaching karst
features. Typical of the underground regimes of such massifs are small hydraulic
gradients, depending on the degree of karstification. This is why groundwater flow is
often observed over great distances into neighbouring catchment areas or even into the
sea. This hydrogeological situation had a decisive impact on the planning, performance
and costs of several large dams in Greece.
The dams of Polyphyton, Perdika and Mornos (Fig. 1) are given as examples of the
typical problems encountered and of the measures applied. The problems cover
submersion of karstic springs, occurrence of karst features in reservoirs, development
of swallow holes and the complete failure of a project.

POLYPHYTON HYDROELECTRIC PROJECT

The Polyphyton Hydroelectric Project in Macedonia is part of the Aliakmon River
Development Project which will finally consist of several dams, power stations and
coastal irrigation areas. The Polyphyton reservoir has a storage capacity of 1900 million
m3 and is primarily a peak power generating plant. Secondly, it provides ample storage
for flood control and irrigation.

General hydrological and geological conditions

The total catchment area is 5800 km2. Precipitation is restricted to the wet season from
October to March and amounts on average to 1050 mm/a. Mean evaporation is 700 mm/a
and runoff 350 mm/a or 2030 million m 3 /a. The specific capacity (mean annual inflow
storage capacity) is 2030 m3/1939 m3 = 1.05.
Due to the annual fluctuations of precipitation and runoff (1953 to 1961 ~ 1:2) and the
specific capacity of the reservoir, the maximum storage level can only be reached in very
wet seasons. The project is situated in the Pelagonian Massif, which comprises of
crystalline rocks (gneisses, diorites, metaschists, amphibolites and marbles).

The dam and power station have been built in a gneissic gorge section, while most of the
reservoir extends into the basin of Servia. This is a young tectonic graben structure,
66 P. Pantzartzis et al.

Fig. 1 - Project locations.

filled with Tertiary, lacustrine (mainly marls) and Quaternary lake and valley sediments
of up to 1000 m thickness. The basin sediments are in contact partly with gneissic rocks
and partly with karstified marbles and limestones (Fig. 2).

Main karst problem

According to geological and hydrogeological reconnaissance, one major problem area
with possible water losses from the reservoir was suggested at the contact between the
basin sediments and the Skopos marbles on the northern bank. Here the large natural
springs of Neraidha discharge about 10 m 3 /s of karst water into the basin, and it was
considered questionable whether the impoundment of up to 40 m in this area could divert
the flow gradient in the opposite direction, i.e. from the reservoir via Skopos marbles,
Vermion marbles and the Vardar zone to the coastal plain of Thessaloniki.

Approach and control

Hydrogeological field investigations and mapping was undertaken in the Skopos and
Vermion mountains. This survey revealed the geological conditions for the possible
groundwater flow and has proved the coherency of the two above-mentioned mountain
areas which are only surficially divided by the Sari Giol, an eastern extension of the
Ptolemais basin.

Further more a tectonic dome structure of gneisses and amphibolites was confirmed in
the southern part of the Vermion mountains which dips towards the east, north and west
 Karst phenomena and dam construction in Greece 61

Fig. 2 - Regional geology and hydrogeology (Heitfeld et al., 1969).

under the karstified marbles, leaving a potential groundwater passage for reservoir
leakages in the Verraion marbles of about 6 km width.

An initial drilling programme of four deep piezometric boreholes (Gx-2a, Gx-5, Gx-9 and
G-lOa) showed the water level of the Ptolemais/Sari-Giol basin to be perched, while the
main karst water level was met in all four piezometers below the maximum storage level
of the Polyphyton reservoir (291 m) with a gradient of about \%o from the Neraidha
springs in the north (Fig. 3).
Subsequently four other piezometric boreholes were drilled in the Vermion mountains
(Gx-8, Gx-11, Gx-12a and Gx-13), from which the same gradient and deep karst level
below the maximum reservoir storage were established reaching far into the marble
massif. In most of the piezometers two or three groundwater levels were detected during
the drilling of the holes (Fig. 4).

Finally a distinctly higher level was found in piezometer Gx-13 at about 40 km from the
Polyphyton reservoir, and situated close to the Vardar thrust contact. Indirectly
confirmed by a number of large karst springs along this contact and in the Vardar Zone
an effective and continuous groundwater division in the Vermion mountains could be
 P. Pantzartzis et al.

VEBHIOB MIS

V A D Â R Z.- -PELAGOH IAN ZONE-

P**- 40km
Fig. 3 - Schematic profile from Neraidha to Tripotamos.

5 N K n a a x u i s s i s sa-TMElétr*]—<=-

\ s \-m !«
1200

GW-LEVELS OURIHG
LOWERING OF BOREHOLES
GX-10a/GX-12a/GX-13 [19691 1000

GX-13

\ NO
! ji
\\"./--- ,
•HCl-l ljectim
GX-10( iY~~ \ n GX-1 1 600
^
K il £ \ \ \\ 1*831
V /
/ 1
1 un
POLYPHVT0H GX-1 a // | , 6 ) -12a
291 RESERVOIR
\ s.a"m il—
"1265
200

N 11
Fig. 4 - Groundwater levels during the drilling of boreholes GX-lOa, GX-
12a and GX-13.

assumed. Hydraulic calculations, neglecting the positive result of the last borehole
(Gx-13), and assuming a continuously low karst water level throughout the Vardar Zone,
a flow section of 5000 x 100 m2 and a rather high K-value of 10"2 m/s, resulted in
possible water losses from the reservoir of 6 m 3 /s, i.e. only 10% of the mean river
runoff. The remaining risk was considered acceptable, and in the meantime a full scale
test after the filling of the reservoir in 1974 indicated that no losses occur.

PERDIKA
The Perdika project in northern Greece was completed in 1961. The main project data
are:
storage volume 10 million m3 (for industrial water supply)
earth fill dam 40 m high
catchment area 164 km2
annual runoff 10-30 million m3
 Karst phenomena and dam construction in Greece 69

The first test impoundment was performed in 1962/1963 and the maximum storage level
(617.3 m a.s.l.) was reached for a short period despite leakage losses already at low
storage levels (604.2 m). The leakage increased rapidly and after drawdown a great
number of sinkholes, dolines, cracks and subsidence areas was observed in the reservoir.

Geology
The geological investigations of 1960 to 1962 were supplemented in 1966 to 1970 by an
additional drilling programme. Young alluvial sands and gravels were found in the river
beds. They are incised into the Plio-Pleistocene sediments of the Ptolemais basin which
consist of detritus of the surrounding mountains (Fig. 5). This formation can be
differentiated in an older fine-grained series (clayey silts and silty sands with sand and
gravel interbeds) and a younger coarse-grained series (coarse sand and gravel). The
thickness of the Plio-Pleistocene is highly variable between zero (small rock outcrops
at the right-hand side of the valley) and >90 m (boring at the foot of the dam). The
bedrock is built-up of karstic limestones which are bisected by faults.
m m in

«11 J&
WS Ë^b
^T—7 ^=-ÇppC

smF*f, XcocwnàK<358<
^LÂKE'7Jj$OÇ X

Ë7f $A A
ÊËÈÎJL
slga P t o I e'fflra i s I ÀÊ/$/yy$fflr7$™mci:'U40S

^-~J5XMADITIS,S S / / \ $T

JJEJCT) |MH '"[/i-niiy J

IÊLÛW^/ % F
\_"-~+f /^i(/~~^ «OLEUMS / /

^/f§
'A/^^z^xwCCCCC

~ ^.^^fe 'ffv / ,K/v&.

«M=^^SS£ °a\ J J ^ ^
*• -. m\ ~ ^ / g y X N . SW* S LiPÎOL V <!_-/

^=^Êcœœ$$$ s k i on»«?
^^^P^M'tîfs?^ ?& \ - ILIIOS
0 S «ft»
i i—.—i—i—1—.—i i i 1

Fig. 5 - Location of the Perdika Dam and groundwater levels.

The rock surface is very irregular and in some parts of the reservoir the rock was found
only a few metres below the surface. The limestones are slightly metamorphic and
folded. Their age is probably Upper Cretaceous.

Karst phenomena and groundwater

The limestones are intensively karstified with solution joints and cavities along bedding
planes, joints and faults. Several cavities with loam fillings were encountered by the
borings. The sinkholes dolines, cracks and subsidence zones in the reservoir follow two
lines (WSW-ENE) which converge northwest of the dam (Fig. 6). Both lines compare
70 P- Pantzartzis et al.

Fig. 6 - Sinkhole and subsidence zones and thickness of overburden in the

Perdika Reservoir.

very well with the general joint and fault pattern of the area. This observation indicates
a dominating influence of tectonic features on karstification.

On the other side it is quite obvious that no or little relation exists between the
development of sinkholes and the thickness of overburden, respectively the depth of the
limestones below the surface.
The above-mentioned sinkhole and subsidence lines are outside the areas of thin
overburden but close to them. It seems that the differentiation of low and high limestone
complexes (Fig. 7) is of tectonic origin and that the bisecting faults facilitate the
karstification.
According to the geological structure there is a perched groundwater level in the
Quaternary and Tertiary formations at about 580-600 m a.s.l. The groundwater level in
the Cretaceous limestone was found at about 530 m a.s.l. The upper level compares with
the general perched level of the Ptolemais basin and drains via tributaries and the Soulou
River to the Vegoritis lake, while the karst groundwater level drains probably to the
Tripotamos springs in the west (Fig. 2).

Rehabilitation measures
Following the first storage test loam blankets were applied in all leakage areas in 1962.
They proved to be ineffective. Alternative measures (full lining/grouting) suggested by
experts, were not accepted because of questionable success. From 1966 to 1970 a new
study was performed by the Federal Geological Survey of Germany with 30 borings,
hydrological assessment and large-scale geotechnical tests.

Storage tests were made with low water levels of 7 to 9 m. No relation between storage
height and daily leakage was found. Colmation processes were active at storage levels
 Karst phenomena and dam construction in Greece 71

Fig. 7 - Isolines of the top of the limestone formation (PRAKLA).

between 599 and 603 m a.s.l., whereas new sinkholes developed at higher storage levels:

Test Storage height (m a.s.l.) New defects

1967 603 1
1968 605 4
1969 609.12 (short) 74
Large-scale hydraulic tests were made to support the colmation processes, i.e suspended
silt streams were directed from the reservoir banks into the leakage areas. The
resumption of individual holes was up to 135 m 3 of solids.
Partly, the treated areas failed again during the storage test of 1969, and additional
defects appeared. The present operation scheme of the reservoir allows storage up to 603
m a.s.l. ( ~ 7 m maximum storage height). A final rehabilitation of the project would
require full scale grouting of the two sinkhole subsidence zones in the reservoir, the costs
being much higher than the benefit.

MORNOS
The Mornos reservoir was built for the water supply of Athens which has about 3 million
inhabitants and some of the major industries of Greece (Fig. 8). Since 1981 it has been
the main water resource for the growing city. Extensions are already under discussion.
The main project data are:
catchment area 550 km3
precipitation 720-2000 mm/a
mean runoff 11 m3/s
storage capacity 640 million m3
dam height
(rockfill dam with grout curtain) 135 m
water conduit 185 km (60 km tunnels; 5 km siphon; 120 km
canals)
72 P. Pantzartzis et al.

Fig. 8 - Location of the Mornos Dam and conduit.

General hydrological and geological conditions

The runoff of the Mornos River varies from less than 1 to about 4000 m 3 /s. The location
of the reservoir within a high-mountain area is favourable as regards morphology, high
rainfalls (up to 3000 mm/a) and low evaporation.
Geologically the reservoir is situated in the Pindos zone (Fig. 9) which consists of
limestone, siltstone, sandstone and conglomerate formations of Cretaceous to lower
Tertiary age.

Fig. 9 - Regional geological map of the Mornos Project (Syrmbos &

Wafiadis, 1973).
 Karst phenomena and dam construction in Greece 73

Both the Parnass and the Pindos zones are thrust over the Gavrovo zone in the west. This
is the reason for the intense folding and faulting of the Pindos flysch. The intercalated
limestones form an irregular massif with partly undetermined contacts. Generally the
karstification of the limestones in the Pindos appears less developed than in the Parnass.

Main karst system

The Mornos reservoir is crossed by smaller limestone interbeds which are probably
isolated within the flysch sediments. Greater limestone massifs occur in the north
(Vardusia) and in the south (Pyrnos), the latter is laterally inundated by the reservoir
along the tributary valley of the Belesitsa River (Fig. 10).

!3 |u [is [» [IT i |H 20

Fig. 10 - Mornos reservoir.

This area was found most critical for the water-tightness of the reservoir. The Pyrnos has
a low groundwater level of about 200 m below storage level, which dips in the southern
direction via the Tolophon limestone massif to the Gulf of Corinthos (springs of Eratini).
A close hydraulic connection between the Pyrnos and Tolophon massif must be assumed.
The distance from the reservoir to the sea is about 20 km and the hydraulic gradient
about 1.2%.

In order to investigate the critical area, 10 boreholes were drilled. By this programme
a relatively favourable geological situation in the reservoir was confirmed with flysch and
transition sediments covering the karstified Pyrnos limestones up to an elevation of
approximately 415 a.s.l. The maximum storage level is 443 m a.s.l.

Sealing measures
The measures applied consisted of surficial cleaning, an equalizing rock-fill dam along
the Pyrnos flank (2 km) with an inclination of 1:2 (27°) reaching from the valley floor
at elevation 380 m to the crest at 445 m, and an asphalt lining (Fig. 11).

This type of sealing was selected because of its effectiveness, long life-time and
flexibility. In order to ensure a reliable contact between the asphalt blanket and the flysch
74 P. Pantzartzis et al.

Fig. 11 - Pyrnos sealing (Heitfeld & Krapp, 1972).

sediments of low permeability a grouting and drainage gallery was built along the foot
of the dam with a drainage tunnel penetrating the flysch and the transition sediments as
well as about 20 m into the karstic limestones.

The drainage capacity of the tunnel is increased by drainage holes. Grouting was
performed all along the dam foot. The success of the grouting measures is not fully
satisfactory. The rock fill amounts to 5 million m 3 , the asphalt lining covers about 300
000 m2, the grout curtain has an extent of 100 000 m2 and the costs for the reservoir
sealing totalled 150 million DM (1980).

REFERENCES
Heitfeld, K. H., Kohlhaas, W., Krapp, L. & Kunert, N., 1969, Polyphyton Eng. Geol.
Reports (unpublished).
Heitfeld, K. H. & Krapp, L., 1972, Consulting Report (unpublished).
Syrmbos, S. & Wafiadis, N., 1973, Geologische und hydrogeologische Untersuchungen
im 57-Teil des staubeckens des Mornos. Projektes/Griechenlands Dipl. Thesis,
RWTH Aachen, unpublished.
Hydrogeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHS Publ. no. 207, 1993. 75

THE INFLUENCE OF GEOSTRUCTURAL SETTING UPON WATER THERMO-

MINERALIZATION IN CERTAIN AREAS OF APULIA (SOUTHERN ITALY)

TIZIANO TADOLINI & GIOVANNI BRUNO

Istituto di Geologia Applicata e Geotecnica, Facolih d'Ingegneria, Université di Bari, 70125 Bari, Italy

ABSTRACT
Despite a general homogeneity, waters circulating in the Mesozoic carbonate formations
of Apulia are characterized by a broad variation in temperature and salt content, mainly
resulting from the modes of replenishment, outflow and contamination by sea water
ingression. The aquifer is embedded in a complex structural-tectonic setting which
occasionally has a significant impact upon the processes that determine the underground
water flow and the chemico-physical characteristics of the groundwater. Some peculiar
examples are discussed.

INTRODUCTION
Apulia is a region of southern Italy mostly characterized by a relative simple
stratigraphie sequence with elements dating from the Mesozoic to the Quaternary. A
prevailing portion shows outcrops of a platform of carbonate rocks. The stratigraphie
sequence is more complex in the northwestern section which marks the beginning of the
characteristic structure of the Apennine Range.

Locally, the carbonate formation in the southernmost portion of the study area is covered
by a stratigraphie sequence of elements that can be ascribed to several different sedimen-
tation cycles. During the last orogenesis, the study area played a major role as a foreland
in the region's structural setting and, consequently in the lithostratigraphic evolution of
the subsequently formed sedimentation basins.
The region's geostructural setting is closely connected with the events that have marked
its history and, specifically, to the relationship existing between the Apulian-Gargano
foreland and the junction area between the Fossa Bradanica-Apennine Range. The most
apparent structural features in the foreland area are the result of tensional Plio-
Pleistocene tectonics whose main lines are oriented northwest-southeast (the Apennine
system) and northeast-southwest (the anti-Apennine system) (Fig. 1).

In contrast to this area of tensional tectonics, the region's western and northern sections
show a compressional type of tectonics, clearly seen in the Gargano Promontory, along
the Fossa Bradanica and at the margin bordering the Apennine Range. In the latter area,
outcrops and data taken from wells several hundred metres deep reveal the presence of
the Apennine allochthonous front, characterized by intensive phenomena of plication and
deformation of predominantly flysch rocks, locally affecting also the more recent Plio-
Pleistocene soils.

The platform too presents different paleogeographic domains, depending on whether one
considers the Murgia or the Salento. The hydrogeological traits of the region have been
affected by such different domains in a significant manner: the transition zone between
the two domains, known as the "Messapian threshold" is tectonic in origin and shows the
characteristics of a transition area particularly in connection with permeability traits. As
a matter of fact, in the Salentine Peninsula proper there are some outcrops that contain
a clay component which has resulted in the formation of some water-bearing horizons
with a resulting rather complex hydrogeological system.
76 Tiziano Tadolini & Giovanni Bruno

(3

ID o
o • -
O s>
e
o

0,0.
o •
HA
:% *°«g. >••%

•*2 S
M o o.
u S> c O « " "-»

g 2 §
O

•a °
cd o .s

o o"£ ° 'S
e O" i
il
C T3

a, •3JSS
ai
C rt

M •b •• « 8 . 1 <
C
2 »
w
w ti Ë o
2«»S
o >
q o: «'
3.23* ' & h
S
H 2
u

ma
11! ÏH
o E
PU o .-S
"o
o Ss

o
u , 60 - ^ M
S.SS
•7! "O S' - ^ - o 199!
£ 2 *Ï3:
O

if]! ln.§ï
o IH u
•a ™ o o
o «X~ •—< t ; t-i »-«
3 c ?-
S H E E J S lll O O
OSCQCQ
o O

inn-^in^Jt^ooa'
 Geostructure and water thermomineralization in Apulia, Italy 11

The lithology in these systems is of primary importance in respect of the underground

hydrology since it may force the groundwater occurring in the Mesozoic calcareous
substratum to flow under confined conditions. Confined flow is also observed in the
Murgia domain where it can be ascribed, however, to the fact that the permeability of
rocks may be extremely low locally due to faciès variations. Despite its apparent relative
homogeneity, a hydrogeological environment of this kind houses significant chemico-
physical anomalies in its deep groundwater.

Underground water flow has been fully discussed both in general terms and in detail
(Calo' et al., 1983; Cotecchia, 1977; Cotecchia & Magri, 1966; Grassi & Tulipano,
1983) with special reference to the various hydrogeological units making up the
underground water system in the study area. Also the relationships between supply waters
and water in the aquifer and between the latter and both lateral and deep sea-water
ingression have been thoroughly demonstrated (Cotecchia, 1977). The above studies also
describe the rather simple patterns of groundwater behaviour observed at different scales,
and stress the fact that they can be mutually correlated. Consequently, the chemical and
physical features of underground water have been clearly outlined.

Salt content distribution appears to be different in the different hydrogeological Apulian

units: at the centre of the region - in the Murgia - this phenomenon is observed as far as
several kilometres inland. However, it is well known that the occurrence of water of
marine origin underlying fresh water is also related to the groundwater head which here
may be as much as several tens of metres. In the southernmost portion, the Salentine
Peninsula, the whole of the groundwater body rests on top of sea water which penetrates
into the Peninsula joining, as it were, the two seas (Fig. 2).

100
LS. _-—«52EE» *-ij±r •"=" "rr=r y, „^_ A.S.
m 0
-100
W
-200 0 5 10 kre
I l I

Fig. 2 - Schematic section showing the distribution of salt content (g/1) in

waters from the aquifers and from sea-water intrusion.

The mean water temperature in the aquifer is rather variable as it is affected by the
temperature of the inflowing fresh water and of the sea water below. The temperature of
the inflowing fresh water depends on the elevation and is therefore conditioned by
microclimates: generally it is around 14-16°C (Figs 3 and 4). Previous studies (Carlin
et al., 1973; Cotecchia et al., 1978; Grassi & Tulipano, 1983) have established the
replenishment areas, the main lines of groundwater flow, and the influence of brackish
water encroaching into the mainland with isotherms of about 18-19°C. The chemistry of
the groundwater (Cotecchia, 1977) falls within the range of values observed in the
inflowing water and those of the sea-water continental ingressions.

Underground water resources are basically supplied by direct precipitation onto the area,
especially in the presence of large calcareous outcrops which make deep seepage possible
due to their high rate of permeability. In smaller areas due to the presence of impervious
covers, there has developed a surface hydrology which conveys water either directly into
the sea or towards areas where it can percolate into the subsoil system.

In particular, clearly separated " shallow" aquifers occur in comparatively large areas:
occasionally these are connected to the "deep" groundwater body flowing in the
calcareous and calcareous-dolomitic Cretaceous formation which constitutes the region's
basement.
78 Tiziano Tadolini & Giovanni Bruno

Fig. 3 - Horizontal section drawn at sea level to show the temperatures

distribution (°C) in the deep groundwater (from Cotecchia et al., 1978).

100 -.
m0 - kS. _______ — «waa»——; « a y a >-' ""«Bam ,_ ... A.S.
^ = g » — — * g Z f f _ _ _ j y ^ w.t. » - . . - _ _ . | 6 _ , 7 _ - — • ^^^Hg^j"
— __ ______ |g __ — — ___ _——
"2ao 0 5 t» km
i l i

Fig. 4 - Vertical schematic section showing the temperature distribution

(°C) in groundwater.

HYDROGEOLOGICAL PECULIARITIES OF THE REGION

Highly refined studies are now being carried out to arrive at a precise definition of
certain situations that occur in the very articulated and complex geological-structural
setting of the area even though they are in fact somewhat different from the
hydrogeological patterns typical of the same area. It has been the purpose of this study
to focus on such specific examples. The thermal and chemical features of groundwaters
as seen in relation to the lithostratigraphic and geostructural problems are examined in
the light of considerations of underground waters naturally flowing into the sea in the
form of coastal subaerial springs, and on groundwater tested from boreholes several
hundred metres deep.

Water point nos 1 and 2 (Fig. 1) (Molino Caldoli and San Nazario springs)
These springs are spaced some 500 m apart. The average discharge is 2001/s and 2501/s,
respectively, and the water temperature is 17.9°C and 26.7°C respectively. The springs
are replenished by water flowing in Mesozoic limestones. Spring no. 1 flows out in the
midst of the coastal sandy-conglomerate deposits covering the limestone, while spring
no. 2 flows directly out of the calcareous rock (Fig. 1). The chemistry of the water from
the two springs (Fig. 5) is similar, while temperature levels, as mentioned above, differ
markedly (Cotecchia & Magri, 1966). Both springs are located in areas where tensional
tectonic phenomena are characterized by normal faults that form the headland in the
Gargano calcareous massif. More precisely, spring no. 2 is part of the Apennine system
of faults running along Torrent Candelaro; conversely, spring no. 1 is affected by a
normal fault running in the direction of the anti-Apennine system.
The high temperatures recorded at spring no. 2 are due to the geological characteristics
of the area. A few kilometres away from the point at which spring no. 2 emerges, along
the above described tectonic alignment at Torrent Candelaro, some magma tic (Punta delle
 Geostructure and water thermoirdneralization in Apulia, Italy 79

9.10
£a , J Mg'2 NaVK* CI" SO;2 HCq

Fig. 5 - Chemistry of water obtained from the investigated water points.

Piètre Nere and San Giovanni in Pane) and Triassic evaporite outcropping rocks can be
found (Punta délie Piètre Nere). The magma tic rocks were almost certainly originated
from a late, possibly Pliocenic, dyke-forming activity (Cotecchia & Canitano, 1954; De
Fino et al., 1981; Mongelli & Ricchetti, 1970a). Again, the high water temperatures
found at spring no. 2 are very likely related to deep water circulation characterized by
a high residual thermal gradient which can be correlated with such magma tic activity.
Nevertheless, the waters that give rise to the two springs join each at some point and also
mix with the shallower groundwater or even with sea-water ingressions. That there have
been some mixing together is revealed by the silica content found in the waters sampled
at the spring outlet. Figure 6 shows the correlation between the temperature and the silica
80 Tiziano Tadolini & Giovanni Bruno

concentration of water (silica geothermometer) also stresses the fact that there is a
general degree of silica subsaturation in flowing water, especially that at spring no. 2.
This is further proof that water from two clearly differentiated circulations mix together:
one is the deep circulation influenced by the high residual thermal gradient and the other
is a shallower circulation which also receives water from sea-water ingressions.

Water points nos 3, 4, 5 and 6

Water points nos 3,4,5 and 6 (Fig. 1) are represented by boreholes, some of which about
500 m deep and reaching down into the aquifer by as much as 200 m. These water points
are all distributed along the border between the Apennine Range and the foreland area,
more particularly in the so-called Fossa Bradanica. The latter is a tectonic depression
which was filled by predominantly clay and clay-sandy Plio-Pleistocene soils following
the formation of a terraced series of normal faults that lowered the marginal portion of
the Murgia platform. More precisely, the boreholes are scattered in an area close to the
Bradano River the course of which is tectonically controlled by an extensive normal fault.

E
S êO -

/
/
40
/
/
S

20
^>

-i 1 r-
20 40 60 80 °C 100
Fig. 6 - Silica geothermometer showing data obtained at water points nos
1 and 2.

Chemical and physical analyses of water samples obtained from the boreholes have
revealed the typical chemistry of water flowing inside carbonate rocks unaffected by sea-
water intrusion (Fig. 5); temperatures of about 27-30°C are the highest water
temperatures observed up to now (Fig. 7).
The water points that were available and have been considered in this study are located
both in the Mesozoic carbonate formation alone (water points nos 3, 4 and 5) and in the
Plio-Pleistocene clay-sandy formation which covers the limestones over some hundred
metres at water point no. 6. This circumstance suggests that water thermalization is due
not only to the effect of the geothermal gradient (Mongelli & Ricchetti, 1970b), which
is rather high in that particular structural area, but also to the depth of the circuits that
groundwater is bound to travel starting from the sites of replenishment and which have
been identified in the most elevated portion of the Murgian system by ongoing, as well
as by previous studies.

Recent investigations (Bethke, 1985; Deming et al., 1990), however, indicate that in a
"mountain Range" environment there is often found a marked residual thermalization of
groundwater which becomes all the more apparent when the water circuit develops
vertically alongside tectonic discontinuities. Under such circumstances, the water
 Geostructure and water thermomineralization in Apulia, Italy 81

circulation in the aquifer is faster due to the high degree of permeability through
fractures, so that temperatures stay high up to the top of the aquifer.

<D ©
i «m s 2 lj/11 1
~A .-J
21 CCI 2»
n fa a
1
«J. ._^.^_
• •J

-s-
•
\ 3 •
—1-
1

-is-
1
(
' \
-B5 .
\ \
-ITS .

-2K-
>
"*l

X î
1
1
A
....

(D
2 fe/fl I ©
" M~""rcr"~2» M M n/ii ».o i ,«2.o Hi

«.» 22.0 2i.e f»C) 30 0

. »,7t B.I.

_»A

/ I; vAïL: :
r ...Scr
I
•
:
|
\"

\
•
\

- ~i a. vl
Fig. 7 - TDS and temperature data: their distribution along the water
column at water points nos 3, 4, 5, and 6.

Water points no. 7, 8, 9 and 10

As mentioned earlier, water flowing in the carbonate aquifer extending in the

southernmost portion of Apulia has fairly homogeneous chemical-physical characteristics:
any variability is essentially due to variations in the salt content resulting from the
mixing with sea water intruding both laterally and from below.

Heavy volumes of water are pumped from the aquifer for different uses, essentially for
82 Tiziano Tadolini & Giovanni Bruno

agriculture. Subordinately, drinking water supplies are integrated by significant amounts

of groundwater pumped from those parts of the aquifer where the TDS content is low
enough to allow the water to be used for drinking purposes. It is here that waters
currently labelled "mineral table waters" are drawn from the boreholes: though their TDS
contents are 0.51 g/1 and 0.32 g/1 (nos 8 and 9 respectively; Fig. 1), these waters are
nothing but ordinary groundwater (Fig. 5). Water from the deep aquifer is also used in
certain medical treatments. Water points nos 7 and 10 (Fig. 1) are in fact subaerial
springs that yield water with a heavy salt concentration (no. 7: 9.4 g/1; no. 10: 29.9 g/1)
due to strong groundwater contamination by encroaching sea water (Fig. 5). In particular,
point no. 10 was drilled into tectonic discontinuities with intensive rock fracturing that
favours the evolution of a karstic process, very active in the coastal sections.

The temperatures of these waters are similar to the mean values recorded in the region.
More detail, temperatures of 17.8 and 17.4°C, that are peculiar to groundwater in the
most inland portion of the Salentine Peninsula, are observed at water points nos 8 and
9; point no.7, with a mean temperature of 17.7 C C, provides a typical example of
groundwater found along the coastal belt which clearly reflects the thermal effects of the
sea. Lastly, point no. 10, with a temperature of around 23.0°C, owes its higher mean
values to the exothermic processes mainly from the sulphur reduction phenomena that are
triggered inside the Miocene cover rocks that occur locally throughout the supply basin
(Calo' et al., 1983; Zezza, 1980).

FINAL CONSIDERATIONS
Regional-scale hydrogeological investigations may reveal peculiar situations in some
places. On this basis, we were able to better develop and understand the geo-structural
events that have had a significant impact upon the hydraulics in the aquifers as well as
upon the chemico-physical properties of the groundwater. The examples discussed in this
paper explain how the two problems are mutually related.

Thus, ongoing studies are meant to evaluate how these phenomena are distributed and to
what extent they can affect the characteristics of the waters. Once the interdependence
between the geostructural features of specific areas in the region and their
hydrogeological characteristics are better understood, it will be possible, inter alia to
better plan the utilization of the groundwater resources.

ACKNOWLEDGEMENTS
The authors thank Mrs Maria Rosaria Paiano for her care in the drawings.

The study was carried out with MURST 40% and 60% funds.

REFERENCES
Bethke, C. M., 1985, A numerical of compaction-driven groundwater flow and heat
transfer and its application to the paleohydrology of intracratonic basins. / . Geophys.
Res., vol. 90, 6817-6828.
Calo', G. C , Sizzico, M., Tinelli, R. & Zezza, F., 1983, Hydrogeological investigation
on the area surrounding Santa Cesarea Terme springs (southern Apulia). Geol. Appl.
e Idrogeol., vol. XVIII, parte II, Bari.
Carlin, F., Magri, G. & Mongelli, F., 1973, Temperatura delle acque sotterranee della
Penisola Salentina. Geol. Appl. e Idrogeol., vol. VIII, parte II, Bari.
Cotecchia, V., 1977, Studi e ricerche sulle acque sotterranee e sull'intrusione marina in
 Geostructure and water ihermotnineralization in Apulia, Italy 83

Puglia (Penisola Salentina). Quaderni dell'1st. di Ric. sulle Acque del CNR, no. 20.
Cotecchia, V. & Canitano, A., 1954, SuH'affioramento délie Piètre nere al lago di
Lésina. Boll. Soc. Geol. It.
Cotecchia, V. & Magri, G., 1966, Idrogeologia del Gargano. Geol. Appl. e Idrogeol.,
vol. I, Bari.
Cotecchia, V., Tadolini, T. & Tulipano, L., 1978, Groundwater temperature in the
Murgia karst aquifer (Puglia - southern Italy). International Symposium on karst
Hydrology, Budapest.
De Fino, M., La Volpe, L. & Piccareta, G., 1981, Geochemistry and petrogenesis of the
Paleocene platform magmatism at Punta délie Piètre Nere (southeastern Italy). N. Jb.
Miner. Abh.
Deming, D., Nunn, J. A. & Evans, D. G., 1990, Thermal effects of compaction driven
groundwater flow from overthrust belts. J. Geophys. Res., vol. 95.
Grassi, D. & Tulipano, L., 1983, Connessioni tra assetto morfo-strutturale della Murgia
(Puglia) e caratteri idrogeologici della falda profonda verificati anche mediante
l'analisi della temperatura delle acque sotterranee. Geol. Appl. e Idrogeol., vol.
XVIII, parte I, Bari.
Mongelli, F. & Ricchetti, G., 1970, Heat flow along the Candelaro fault Gargano
headland (Italy). Geothermics, special issue 2.
Mongelli, F. & Ricchetti, G., 1970, The Earth's crust and heat flow in the Fossa
Bradanica, southern Italy. Tectonophysics, 10.
Zezza, F., 1980, Le sorgenti ipotermali di Santa Cesarea Terme. Estr. Riv. Sallentum,
anno III, nos 1-2.
Hydrogeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar. October 1990).
IAHS Publ. no. 207, 1993. 85

STORAGE WATERTIGHTNESS PROBLEMS AND TECHNICAL SOLUTIONS

IN THE HIGHLY DEVELOPED KARST AREA OF THE NIKSIC POL JE,
YUGOSLAVIA

DRAGOLJUB ZOGOVIC
Energoprojekt, Biro za Geologiju i Geofiziku, Lenjinov Bulevar 12, Belgrade 11070, Yugoslavia

ABSTRACT

The Niksic Polje appears to be a typical closed karstic depression, consisting of a number
of distinct units. The Niksic Polje has developed in an area of Mesozoic carbonate rocks,
including two Tertiary flysch narrow zones. Its bottom is covered by Quaternary,
limniglacial deposits. There are typical very complex karst phenomena, such as uvulas,
sinkholes, jamas, caves, underground channels, caverns, high yielding springs, swallow-
holes and estavelles. Numerous high yielding springs occurring mainly along the northern
periphery of the Polje feed a confluent network of the Upper Zeta River. Under natural
conditions, all surface water and groundwater discharging into the Niksic Polje
disappeared through the swallow-holes and estavelles situated along its southern
perimeter. After a relatively short underground flow the Upper Zeta River water occurred
again as major springs forming the Lower Zeta River, at a level about 550 m lower than
the Niksic Polje level. In order to use this huge hydropower potential, three dams have
been built, thus creating surface reservoirs under high developed karstic conditions.
Consequently, serious problems regarding the reservoirs watertightness have occurred.
The paper describes the measures used to prevent and limit large water losses from each
of these storage reservoirs. Surface and underground sealing of the reservoirs has been
applied. The plugging of the main swallow-holes or closing them with small cylindric
dams appeared to be unsuccessful. In order to seal some of the large karstic channels
more than 1.00 x 106 kg of grout mix was used.

INTRODUCTION
The construction of dams in highly developed karst areas began in Yugoslavia shortly
after the Second World War. At that time, there was no experience concerning the
technical solution of the storage watertightness problems in karst areas, particularly as
numerous swallow-holes could transmit several tens or even hundreds of cubic metres of
water per second, as in the case of the Niksic Polje.

Moreover, the idea of constructing dams and storing surface water within the Niksic
Polje was deemed to be an illusion by karst experts among both geologists and civil
engineers.

An irregular manifestation of complex karstic phenomena was commonly attributed as the

crucial feature to the karst terrains. Therefore, the widespread opinion was that any
attempt to achieve satisfactory watertightness by sealing the reservoirs was unlikely to
succeed. Despite this, a considerable number of dams have been built in the Dinaric karst
region. Most of the karstic storage reservoirs have been made properly watertight, but
in some cases the problem of large water losses still exists.

The three dams and related lakes situated in the Niksic Polje provide examples of the
particular watertightness problems encountered in such areas, and the technical solutions
necessary to solve these problems.

Very good sealing of the Krupac storage reservoir was attained by quite limited grouting
86 Dragoljub Zogovic

works. However, a long and deeper grouting curtain was indispensable in order to limit
losses from the Slano Lake.
Due to relatively deep karstification and groundwater circulation, the first trial to
improve the watertightness of the Vrtac storage reservoir was by installing surface
structures to close off the main swallow-holes. Unfortunately, this was unsuccessful as
new swallow-holes that swallowed water at a rate of about 20 m 3 /s appeared through the
Quaternary deposits.

GENERAL HYDROGEOLOGICAL SITUATION

From the hydrogeological point of view, the following formations are distinguished in
the area (Fig. 1):
— Limniglacial deposits, underlying the Niksic Polje floor,
— Flysch Formation,
— Complex of Cretaceous, Jurassic and Triassic limestones and dolomites, sporadically
interbedded with marly, siliceous and clastic sediments.
Apart from the limniglacial deposits, these rock formations are tectonically strongly
folded, ruptured, faulted and overthrust. In fact, the Niksic Polje is tectonically
predisposed.
Two large faults going along the southern and western rims of the Polje, respectively,
as well as a reverse fault, stretching along the Niksic Polje gave rise to an original
tectonic trough being afterwards shaped by fluvio-limniglacial erosion and karstification
into the karstic Polje. Though very heterogeneous in composition and very changeable
in hydraulic conductivity, the Quaternary mantle of fluvioglacial deposits does not
provide any watertightness for the reservoirs within the Niksic Polje.

Many swallow-holes have developed where the groundwater level occurs in the
underlying karst aquifers. Two narrow flysch zones extending along the Niksic Polje act
as hydrogeological barriers, impeding further groundwater flow downstream and limiting
the depth of karstification.

Consequently, water losses from the Niksic Polje are limited to its southern periphery.
The Cretaceous, Jurassic and Triassic rocks underlie the fluvioglacial deposits within the
Polje floor and constitute the mountain ridges surrounding the karstic depression. These
rock systems were the framework for the karst development.

Catchment area
The catchment area of the Niksic Polje is, in fact, the Upper Zeta River basin. Taking
into consideration an assumed groundwater divide, the catchment area of the Niksic Polje
amounts to approximately 890 km2 (Vlahovié, 1975). It is mainly a karst watershed with
groundwater occurring on the surface inside the Polje. The sole river coming from
outside is the Gracanica.

Surface flow
Perennial surface streams are exclusive to the northern portion of the Niksic Polje. The
Upper Zeta River is formed by two confluents within the Upper Polje (extreme north).
The two confluents (the Susica River and the Rastovac River) are fed by high yielding
karst springs. In addition, there is a network of Upper Zeta River tributaries.
 Storage watertightness problems and solutions in Yugoslavia 87

EXPLANATION
| a ) FLUWOGUCIAL KPOSITS

I n . w o i raauATiOH

J ceotocfCAL BOUMOART oaszRvss « t s « s u a t o

c 3 M U L T - 08SCTW0 AND ASSUMED

i * * * * * * * * * * ^ REVERSE FAULT- OBSERVED J

| CAVE
[ CAVE WITH WATER

^ ] SPRING a > N ' i o o i i u c

J SPRING te 1 0 3 - I O O O I / M S -

| ^> 1 SPUING Q , * WOO l/aee-

I g I ESTAVELLE
SWALLOW HOLE

BcrenuiTTCNT SWALLOW note

ESS] SWALLOW HOLES ZOHC

• J PERCKMAL SURFACE FLOW»

ma
INTERMITTENT SURFACE FLOWS

GROUNDWATER CONTOUR U K £
{ M i l 1161)
GROUTING OBTAIN COMPLETES

^ 3 AND UNDER EXECUTION

' ^

il

Fig. 1 - Hydrogeological map of the Niksic Polje (Vlahovié, 1975).

88 Dragoljub Zogovic

Under natural conditions, before the construction of the dams, the Upper Zeta River and
its tributaries dried up in their downstream courses during the dry season. All surface
water was lost through the existing swallow-holes. During wet periods and flooding of
the Niksic Polje, the Upper Zeta River would disappear into the Slivlje swallow-hole in
the extreme southeast of the Polje.

The Niksic Polje water occurred in several karst springs, forming the Lower Zeta River,
at an elevation 550 m below the Niksic Polje floor (Fig. 1).

Groundwater occurrence and flow

Groundwater occurs in the karst aquifers underlying the Niksic Polje and its mountain
surroundings, and also locally or intermittently in the overlying limniglacial deposits.
Before construction of the dams, the limniglacial deposits constituted an aquifer all over
the Niksic Polje only during the wet seasons when the water table of the underlying karst
aquifers was high.

Under natural conditions there was a specific intercommunication and alternation of

surface and groundwater flows. Groundwater and surface water substituted for each other
from north to south over short distances. But these flow conditions were considerably
changed by the construction of the Upper Zeta electric power system.

Water from the karst springs situated at the farthest upstream unit, the Upper Polje, flows
partly on the surface as the Upper Zeta River and partly underground through the channels
discharging downstream into the nearby Krupac Polje (now the Krupac Lake). Before the
Krupac dam was built, this polje was drained by the Mostanica River and the swallow-hole
which was called the Krupac hole. After flowing a short distance underground this water
emerged at springs on the northern perimeter of the Slano Polje. Upon the construction of
the Krupac dam, this underground communication of the two karst polje units was cut by a
grout curtain.

The southern belt of the Niksic Polje which comprises the Slano Polje, the Vrtac and Slivlje
represented a surface water sinking zone. The water from the river network sunk into
estavelles (swallow-holes acting intermittently as karst springs) along the southern periphery
of the Slano Polje, and swallow-holes farther to the east (Vrtac and Slivlje).
By the construction of the Slano dam and a continual grout curtain along the entire
southern bank of the reservoir, water losses were limited to an acceptable rate of 1-2
m 3 /s. Most of the estavelles were transformed into intermittent springs occurring outside
and overflowing the grout curtain.

Thus, the general groundwater flow within the Niksic Polje is from the north to the south.
Only local exceptions can be noticed. This general groundwater flow is indicated not only
by the distribution of springs, estavelles and swallow-holes but also by the piezometry of
the karst aquifers (Fig. 1).

The contour lines of the water levels show that the groundwater flows through the
southern mountain ridge, and farther away the direction is changed towards the east, i.e.
the karst springs forming the Lower Zeta River. The depth to the regional groundwater
level increases from the northern to the southern parts of the Niksic Polje, as well as from
the Slano Polje to the east, i.e. through the Vrtac and the Slivlje Poljes.

General water balance

The water balance of the Zeta River under natural conditions was estimated by Vlahovié,
(1975) as follows:
 Storage watertightness problems and solutions in Yugoslavia 89

Precipitation within the catchment area of

the Lower Zeta River springs 1.785 X 109 m3/year (56.60 m3/s)
Surface and groundwater flow 1.25 X 109 m3/year (39.64 m3/s)
Evapotranspiration (28%) 0.535 X 109 nrVyear (16.98 m3/s)
Surface flow rate 1.032 X 109 nrVyear (32.74 m3/s)
Groundwater flow exclusively 0.217 X 109 m3/year (6.88 m3/s)
Discharge rate of the Lower Zeta springs 1.5 x 109 m3/year (47.56 m3/s)
These figures show that most of the precipitation water recharging into the karst aquifers
came to the ground surface within the Niksic Polje, but after a short surface flow it
disappeared underground again and occurred through the Lower Zeta springs. By
successfully impounding water in the Krupac and Slano reservoirs the natural water
balance and flow were substantially changed. Most of the Niksic Polje water has been
captured and used for electric power generation. The watertightness of the Vrtac reservoir
still needs to be improved for an optimum utilization of the hydropower potential.
Eventually, groundwater from the eastern part of the Niksic Polje, outside the Vrtac dam,
could be considered for development.

SOLUTIONS FOR SEALING THE RESERVOIRS

The initial approach to prevent water losses from the reservoirs was the surface closing
of the large swallow-holes. The first step appeared to be the construction of a cylindric
dam around the Upper Zeta River swallow-hole, called the Slivlje swallow-hole, as a
feasibility test. This dam was constructed in 1953 and 1954. It is founded on the limestone
bedrock underlying the Quaternary deposits.

However, though the Upper Zeta River could no longer disappear into this swallow-hole
after the completion of the dam, there was not a considerable increase in the amount of
stored water. The discharge from the Lower Zeta River spring Perucica, connected with
the Slivlje swallow-hole, was practically unchanged. To be precise, by closing the
swallow-hole itself, water losses were decreased from about 118 m 3 /s to about 106 m3/s
or by approximately 10%.
After this unsuccessful trial, it was decided to seal the underground karst channels by
grouting the Krupac reservoir. Evidence obtained from the hydrogeological investigations
showed there was a relatively shallow groundwater flow through a rather narrow karst
zone extending from the Krupac swallow-hole. The grouting was executed during 1958
and 1959. As a result, all the karst springs situated downstream in the Slano Polje dried
up and the Krupac reservoir filled up. The reservoir watertightness is very good, without
any known water losses.

This success encouraged the experts to proceed with grouting to seal the Slano reservoir.
The scale of grouting was much larger, but a satisfactory watertightness of the reservoir
was achieved. The grouting work was carried out from 1961 through 1968.
Regardless of the unsuccessful efforts of closing the Slivlje swallow-hole, the advocates
of the surface method did not give up. In order to improve the watertightness of the Vrtac
reservoir, steps were taken to seal the swallow-holes at the surface. The following
methods were applied:

— encompassing the large swallow-holes by cylindrical dykes;

— closing the swallow-holes by concrete slabs equipped with non-return valves;

- covering karstified limestone banks by concrete blankets, furnished also with check
valves.
90 Dragoljub Zogovic

In total, more than 20 swallow-holes were treated in one of the above ways during the
summer of 1962.
As a result, after completion of the work, the storage reservoir filled up and the normal
reservoir level was established. However, during the following eight months more than
100 new small swallow-holes appeared in the limniglacial deposits around the structures,
and all around the southern periphery of the reservoir. None of the structures was
destroyed, but the beneficial effects were eliminated after a short period.

TECHNICAL DATA ON THE GROUT CURTAINS

The main technical data on the grout curtains for the three reservoirs are presented in this
section.

Krupac Dam Reservoir

The main water losses took place through the Krupac swallow-hole. A considerable water
quantity escaped also around the dam on both the left and right banks. Therefore, the
reservoir could not be filled up. Consequently, three grout curtains were installed after
the construction of the dam (Fig. 1). The karst channels from the Krupac swallow-hole
were sealed by a grout curtain 247.5 m long with an average depth of 41.83 m. The
distance between the grouting boreholes along a single-row curtain was 2.4 m. A total of
4308.63 m of boreholes were grouted. The average amount of grout used was 1282.29
kg/m.

The main problem was to seal the karst channels close to the minimum groundwater level.
5436 x 106 kg of dry grout and 2.28 x 103 m3 of crushed stone were used.
The grout curtain on the left bank is 504 m long, but the average depth is only 18.41 m.
The grout curtain on the right dam bank is shorter, 159 m, but somewhat deeper, on
average 26.73 m. The grout used for these two curtains was relatively small, some
hundreds of kg/m.

Slano Dam Reservoir

To achieve a satisfactory watertightness of the Slano reservoir was much more difficult.
It was necessary to install a 3660 m long grout curtain along the entire southern bank of
the reservoir. It is also a one-row curtain with the average distance between grout holes
of 2.83 m.

As for the representative grouting fields, covering 777 m out of the whole curtain, 18 178
m of boreholes were grouted, 10 446 110 kg of dry grout was used (574.9 kg/m). The
average depth of grouting was 66.42 m.

The main problem was again to grout and seal the karst channels corresponding to the
minimum groundwater level. For example, 1 072 805 kg of dry grout was used to seal just
one of the karst channels which had intense water circulation. This quantity corresponds
to about 10% of the total amount pertaining to the curtain interval of 777 m.

Thus, the grouting was adapted to the requirements for sealing the main karst channels,
neglecting possible leakage through the fissure network (Zogovic, 1980).
 Storage watertightness problems and solutions in Yugoslavia 91

Vrtac Retention
Evidence obtained from the hydrogeological investigations showed that sealing of the
reservoir should be beneath the dam up to the flysch zone and along the southern bank to
the Slano dam. It requires a large amount of drilling and grouting, particularly because
the necessary depth of grouting is considerably greater than in the cases of the Krupac and
Slano reservoirs.

According to the final plans (Energoprojekt, 1978), the total length of the grouting curtain
was 1500 m beneath the dam and 4500 m along the southern bank, making a total of 6 km.

The principal criteria used for the grout curtain design, as far as its contour is concerned,
were as follows:
— to extend the grout to a depth of up to 20 m below the minimum groundwater level,

— to grout the subsurface zone with permeability exceeding 10 Lu (Lu = Lugeon is a

permeability pressure test unit; one Lugeon unit corresponds to a water loss of
1 1/min along 1 m of a borehole, under 10 bars of pressure),

— to seal karst channels and caverns encountered whether or not they were open or
filled with material derived from fluvioglacial deposits or by terra rossa.

In the case of the Niksic Polje, by sealing the zone of groundwater level fluctuations,
down to a depth of 10-20 m below the minimum groundwater level, all the criteria were
satisfied. Higher permeability and karst underground forms are limited to this zone.
Consequently, the Vrtac reservoir curtain reaches from 30 to 130 m in depth. A
representative interval of the grout curtain contour has been shown in the hydrogeological
cross-section (Fig. 2).

The grouting is just under way. Two grouting fields, each 200 m in length, have been
completed so far (Fig. 1). Grouting is being carried out in one row, with boreholes spaced
at 2 m. The total drilling length is 26 617 m (194 boreholes), whereas 25 147 m were
grouted. The average consumption of dry grouting mass ranged from 206 to 720 kg/m,
or 505 kg/m in grand total.
In the process of grouting, special care was taken to seal the highly permeable zones. The
grout used for the water-bearing karst channels exceeded the 4000 kg/m and included
cement, clay, sand, bentonite and water-glass.
Along 400 m of the grouting front, 32 caverns were discovered, corresponding to 83.4 m
of boreholes. Out of this length, 37 m represented open caverns, but 46.4 m were filled
by sandy-clayey material.

Naturally, these first steps to seal the Vrtac reservoir do not enable a positive assessment
to be made. Control permeability pressure tests are not sufficient to evaluate the degree
of sealing while the piezometry is not indicative with such a limited portion of the curtain
installed.

However, judging from the experience gained in the case of the Krupac and Slano grout
curtains, proper sealing of the Vrtac reservoir will certainly be feasible.

CONCLUSION
Evidence obtained from comprehensive hydrogeological investigations and from the
92 Dragoljub Zogovic

III

Hill II IS of &s
.1 I 11 ïl ï Î
i | j ! 5 HI
(Hi! f V! =1 II II
•:;/

:q Mh-
ââ ^ \ f \' '
^•" '
-1
 Storage watertightness problems and solutions in Yugoslavia 93

surface and underground sealing of the storages shows:

(1) The Niksic Polje presents an exceptionally complex and highly developed karst area.
Despite this, the regional karstic water level is at approachable depths over most of
the Polje. Therefore it is possible to improve the watertightness of the surface
reservoirs considerably.

(2) The minimum karstic water levels indicate an approximate depth of an intensive
karstification and groundwater circulation. By sealing this zone, the problem of the
karstic reservoir watertightness is practically solved.

(3) The presence of numerous swallow-holes and estavelles makes grouting the only
adequate method of sealing the reservoirs. Surface closing of the swallow-holes would
not be effective.

(4) Thereby, the grouting procedure should enable proper sealing of the water-bearing
karst channels and privileged zones of the karst water circulation. Special remedial
measures have to be applied. Relatively low permeable rocks which usually prevail
in the curtain section may however be treated by the standard grouting procedure, or
even excluded from grouting.

REFERENCES
Vlahovié, V., 1975, Karst and hydrogeology of Niksic Polje. Special Issue, Science
Academy of Montenegro, Titograd.
Energoprojekt, 1978, Final Design of Vrtac Storage Grouting Curtain. Documentation
Centre of Energoprojekt, Belgrade.
Zogovié, D., 1980, Some methodological aspects in studying hydrogeological conditions
for dam construction in karst regions. Symposium on Hydrogeology and Engineering
Geology, Portorof.
3 Hydrogeology
Hydrogeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990). _
IAHS Publ. no. 207, 1993. 97

ORIGIN AND CATCHMENT AREA OF THE OLUKKÔPRÛ KARST SPRINGS

MUSTAFA DEGÎRMENCI & GULTEKÎN GÛNAY

International Research and Application Center for Karst Water Resources, Hacettepe University, Beytepe, 06532
Ankara, Turkey

ABSTRACT
The Olukkôprû springs are a significant karst groundwater discharge point in the western
Taurids, the most important karstic region of Turkey. It is well known that these springs,
comprising the major water source for the Beçkonak dam, are recharged from basins
other than the Kôprûçay basin. However, the relationship between the springs and the
adjacent basins has not been precisely explained. Due to this relationship, the Olukkôprû
springs are of great importance not only for the Beçkonak dam project, but also for other
karst water resources projects in adjacent basins. Basinwide rainfall and surface runoff
analyses, remote sensing techniques, caving, hydrogeochemical resolution and
environmental isotope analyses have been applied in order to determine the recharge-
discharge systems and the relationship between the springs and the adjacent basins. The
Olukkôprû springs discharge from Miocene conglomerates at the northern edge of the
Beçkonak reservoir and have an average dry-period discharge of 30 m 3 /s whereas, in wet
periods, the average discharge can be up to three times higher. Based on measurements
carried out at the Be§konak flow gauging station 4 km downstream from the springs, the
long-term average flow of the Kôprûçay River has been estimated to be 86.4 m 3 /s.
According to the basinwide hydrological budget calculations based on the flow data from
the Begkonak flow gauging station, a karst groundwater contribution of some 20 m 3 /s
from the adjacent basins has been evaluated. An overall analysis of the available data has
shown that some 10 m3/s of this groundwater recharge come from the Bûyûk Çandir basin
(a sub-basin of the Aksu basin); the remainder might come from northeast of the
Kôprûçay basin.

INTRODUCTION
The Olukkôprû karst springs, one of the most important karst discharges of Turkey, are
located in the Kôprûçay Canyon National Park, 40 km west of Antalya. These springs are
the major water source for the Begkonak dam planned to be constructed on the Kôprûçay
River. It is known that a significant part of the recharge to the springs take place through
adjacent basins. Therefore, special importance is attached to the springs regarding the
water resources projects in the adjacent basins as well as for the Beskonak Dam project
(Fig. 1). In this study, the recharge and discharge mechanisms of these springs are
determined and their relationships with adjacent basins have been investigated.

GEOLOGICAL UNITS AND THEIR HYDROGEOLOGICAL FUNCTIONS

The stratigraphical units in the region can be divided into allochthonous and
autochthonous units, in terms of their lithological and structural features. Paleozoic,
Mesozoic and Cenozoic rocks prevail in the region.

Autochthonous units
Paleozoic: The Paleozoic rocks consist predominantly of schists and outcrop at the north
of the basin (Fig. 2, Pz). These units have an approximate thickness of 2000 m and serve
as the impermeable base in the region.
98 Mustafa Degirmenci & Gultekin Gunay

Fig. 1 - Location map of the water resources projects in adjacent basins.

Mesozoic: Triassic shales known as the Kasimlar and Kirkkavak formations (Tr) have
a thickness of over 2000 m and are considered the most important impermeable units in
terms of their hydrogeological functions. The hydrogeological connections between the
Kôprûçay basin and the basins to the north and south are closely related with the extent
and structure of these units which serve as natural barriers for karst groundwater
movement; a divide between the Anamas mountains limestone aquifer to the north and
the permeable units to the south. The discharge from the Anamas mountains limestone
aquifer towards the Kôprûçay basin takes place through numerous karst springs (such as
Ba§pinar and Karapinar) located at the contact between the limestone and these
impermeable units. The impermeable units also outcrop between Kepez and Dumanh
mountains to the east of the Kirkkavak fault and impede the groundwater movement
between the polje systems and Kôprûçay basin.
 Origin and catchment area of the Olukkopru karst springs 99

Fig. 2 - Hydrogeology of the Kôprûçay River basin and vicinity.

1 - Alluvium; 2 - travertines; 3 - sandstone-claystone; 4 - conglomerates; 5 - flysch
(sandstone, shale, limestone); 6 - limestone; 7 - shale, sandstone, conglomerate; 8 -
limestone, dolomite; 9 - schist, quartzite; 10 - impermeable beds; 11 - limestone; 12 -
limestone; 13 - impermeable beds; 14 - main surface water divide; 15 - sub-basin
surface water divide; 16 - underground water divide; 17 - probable direction of
groundwater flow; 18 - groundwater flow direction defined by dye tests; 19 - spring;
20 - group of springs; 21 - river; 22 - discharge measurement (A: annual discharge
(m 3 /s); B: catchment area (km2)); 23 - sinkhole; 24 - poljes; 25 - cave; 26 -
hydroelectric power station; 27 - dam site; 28 - lithological boundary; 29 - thrust
fault; 30 - reverse fault; 31 - dip and strike of the bed; 32 - fault; 33 - rose-diagram
of joint systems; 34 - sampling station; 35 - residential area.
100 Mustafa Degirmenci & Gùltekin Gunay

In the study area, Jurassic-Cretaceous units are represented by widespread limestone

deposits (J-K) known as the comprehensive sequence. This limestone sequence, with a
thickness well over 1000 m, is extensively karstified and reveals striking examples of
most karst features such as karens, dolins, uvulas, sinkholes, poljes, caves and estavellas.
In the Kôprùçay and Aksu basins, Jurassic-Cretaceous limestones are overlain by
members of the Antalya nappes whereas there exists Tersiyer (T) (Lutecien) flysch unit
between these limestones and the Bey§ehir-Hoyran-Hadim nappes in the east. The
impermeable flysch is of special hydrogeological importance for it acts as an
impermeable barrier between the permeable nappe units and the underlying limestones.

Cenozoic: All the older rocks in the area and units of nappes are overlain by Miocene
deposits which are over 1000 m thick. They cover a large area of the Kôprùçay basin and
comprise gradually alternating conglomerates (Tk, Kôprùçay conglomerates) with coarse
components and alternating shale-sandstone sequences (Tb, Beskonak formation). The
conglomerates are karstified and permeable whereas the Beskonak formation is
impermeable.

Due to its dominant calcareous components and carbonate cement the Kôprùçay
conglomerate is karstified along fractures, joints and fault zones and thus permeable.
Numerous dolines, ponors and caves have been found in the basin. The Kurukôprû cave
with a total length of 530 m in the Be§konak reservoir area is a typical example of the
karstic features in the Kôprùçay conglomerates.

Allochthonous units
Units of Antalya nappes occur between the Kôprùçay basin and Aksu basin to the west
of the region. Antalya nappes cover large areas of the Aksu and Kôprùçay basins, from
Egirdir lake in the north to the Serik-Aspendos area in the south. Due to their lithological
and structural properties, permeable and impermeable nappe units in this area are of great
importance from the geotechnical and hydrogeological points of view.

In the Aksu and Kôprùçay basins, Antalya nappes are in contact with the underlying
Jurassic-Cretaceous limestones. In the study area the stratigraphie sequence from the
bottom to the top is as follows: Ophiolitic unit/Ispartaçay unit (Tr-K), Dulup unit/
Tahtahdag Group (P-K).

The impervious Ophiolitic unit comprising detritic rocks, ophiolite and radiolarite has
an average thickness of 500 m, whereas the Dulup unit (limestone) with a thickness of
100 m is extensively karstified and pervious. Similarly, the Ophiolitic unit represents an
impermeable barrier between two identical karst systems, the Jurassic-Cretaceous
autochthonous limestones (J-K) and the Dulup limestones (P-K) of the upper nappes (Fig. 2).
The Dulup limestones of the nappes are encountered as separate blocks within the
Ophiolitic unit. However, they are also widespread on the Ophiolitic unit (south of
Egirdir lake). The impermeable Ophiolitic unit of the nappes can be considered a natural
barrier which controls and limits the development of karst processes in the area. They
represent the groundwater divide between the Kôprùçay and Aksu basins, due to their
south-north extension and structural position (Fig. 2).

ORIGIN AND LOCATION OF THE OLUKKÔPRÙ SPRINGS

The Olukkôprû springs are named after the Roman stone bridge constructed in the narrow
canyon through which the Kôprùçay River flows. Along the right bank of the river,
numerous springs occur slightly above the river level. The occurrence of paleo outlets
 Origin and catchment area of the Olukkôprû karst springs 101

at the higher levels clearly indicates the evolution of the karst erosion base. The springs
discharge through bedding planes dipping 5-10° towards the valley and through joints and
fractures.

The Olukkôprû springs together with the Bôgrûmkôpriï, Alabalik havuzu and Oglanuctugu
(Fig. 3) springs in the south are discharges from the same aquifer. This conglomerate
aquifer is surrounded by the impermeable Be§konak formation (Tb) from the east and
south. Thus, all the groundwater discharges in the area take place between the Kikgeçit
creek in the north and the Oglanuctugu fault in the south.

Based on drilling data obtained from borehole 21551 to the south of the Oglanuctugu
fault, the thickness of the Beskonak formation (sandstone-shale), which acts as an
impermeable barrier in the region, is over 240 m.

Fig. 3 - Hydrogeological map of the Olukkôprii springs.

102 Mustafa Degirmenci & Gultekin Gtinay

SPRING DISCHARGE RATES

Due to the fact that the Olukkôprû springs are located between the Bulasan and Be§konak
flow gauging stations (Fig. 3) and that there is a negligible contribution from the
tributaries especially in the dry season, a precise analysis of the springs' discharge rates
is possible.

QBE$=3l,+1.8l8(lBUL
•.30*2.vuqmi
THE YEARS OF MAXIMUM
FLOWS (19661 THE YEARS OF MINIMUM
(daily flows) FLOWS I196U
(daily flaws)

20 30 40 50 60 30 «0 80 WO 00 10 20 30 40 50 60 70 00
9 Q
BULASAN, m/s BULASAN, n?/s

QgE?= 32.6* 1.96 QBUL

THE MEAN MONTHLY FLOWS

FOR THE PERIOD BETWEEN
1963-1971

20 M 4 0 5 0 6 0 T O K> !» 5 T

Q BULASAN, n?/s
Fig. 4 - Flow records from the Bulasan and Be§konak flow gauging
stations.
 Origin and catchment area of the Olukkôprtt karst springs 103

Comparison of Bulasan and Beskonak flow records

In comparing the flow gauging records at these two stations, two approaches have been
attempted: (a) for the years of minimum and maximum flows between 1963 and 1971, a
similar period in records at both stations, daily flow readings of the recession period
(April-September) were compared (Fig. 4); (b) based on the mean monthly flow rates for
the period 1963-1971, a comparison of monthly flows was made.

As seen from Fig. 4, in the year of minimum flows (1964) a significant flow rate of
30 m3/s was observed at the Be§konak flow gauging station even though the river was
completely dry at the Bulasan flow gauging station. This flow can be regarded as the
minimum contribution of the Olukkôprû, Bôgrumkôprû, Alabahk havuzu and Oglanuçtugu
springs. The contribution of these springs to the Kôprûçay River was evaluated from the
monthly flow records between June and September 1982. Accordingly, the average flow
rates for these springs are:

Olukkôprû springs: 31 m3/s

Bôgrumkôprû springs: 2.5 m3/s
Alabahk havuzu + Oglanuçtugu springs: 6 m3/s
Based on the same flow gauging records it can be concluded that in dry periods, about
4/5 of the total flow (48 m3/s) recorded at the Beskonak flow gauging station is supplied
by the Olukkôprû and other springs in the region, whereas 1/5 of the flow comes from
the catchment area to the north of the Bulasan flow gauging station.

The long-term (1941-1987) annual average flow of the Kôprûçay River is 86.4 m 3 /s.
About 68 % of this amount is met by baseflows. For a year of minimum flows, this ratio
can increase to as much as 75 %. The abundance of the baseflow in comparison with the
surface flow can be explained by the discharge of the Olukkôprû karst springs. This also
gives rise to the existence of a regular flow regime, which is an important factor where
the construction of dams, especially river-type power plants is concerned.

RECHARGE AREAS OF THE SPRINGS

Groundwater flow contribution from adjacent basins

In order to determine the recharge areas of the springs, water budget calculations based
on the rainfall-evaporation-surface runoff have been made (Table 1) for the sub-basins
of Kôprûçay and adjacent basins, in which flow data have been recorded (Fig. 5). The
evaluations gave a groundwater contribution of 20 m 3 /s from the adjacent basins to the
Kôprûçay basin. Based on the budget calculations and hydrogeological evaluations, the
following conclusions were drawn:

— About 10 m3/s of this groundwater comes from the Bûyûk Çandir sub-basin of the
Aksu basin (Fig. 2).
— The remaining 10 m3/s is the groundwater inflow to the Kôprûçay basin, that may
be supplied by the karst terranes loeated between the Dumanh mountain and Kartoz
(Fig. 2).

CONCLUSIONS
In this study an attempt was made to determine the hydrogeological connections between
the Kôprûçay basin and adjacent basins and, thus, the regional karst groundwater flow
pattern. For this purpose, a multidisciplinary study was conducted (Degirmenci, 1989;
104 Mustafa Degirmenci & Giiltekin Gttnay

Table 1. Results of water balances for the basin and sub-basins in the study area.
1965-1987: 1965-1971: 1986-1987:

Basin Method Method Method Method Method Method

(A) (B) (A) (B) (A) (B)

Begkonak +38.4 + 20.5 + 45.6 +35.2 +33.4 + 15.3*

Bulasan -4.4 -16.6 -10.2 -27.4
Olukkôprû** + 50.3 + 50.3 + 40.6 + 40.6
Aksu (Anamas) + 0.4 -5.6 -1.9 -8.9
Sorgun -0.68 -2.42
Bûyûk Çandir -5.3 -9.7
Degirmenlik -0.91 -2.7
Method A: Water budget based on mean annual precipitation estimated by the polygon method.
Method B: Water budget based on mean annual precipitation calculated from the precipitation-
elevation relationship.
* The values are recharge/discharge amounts (in irf/s) calculated from water balance studies of
related basins. The + sign indicates recharge through underground karst channels from
adjacent basins; the - sign indicates outflow to adjacent basins
** The sub-basin located between the Begkonak and Bulasan basins (Fig. 5).

Fig. 5 - Sub-basins of the Kôprûçay River basin.

 Origin and catchment area of the Olukkoprtt karst springs 105

Degirmenci & Giinay, 1990) and a hydrogeological map of the study area was prepared
(Fig- 2).

REFERENCES
Degirmenci, M., 1989, Kôprûçay ve dolaymin (Antalya) karst hidrojeoloji incelemesi
(Karst hydrogeological investigation of the Kôprûçay basin and its vicinity, Antalya.
Unpublished PhD Thesis, Hacettepe University.
Degirmenci, M. & Giinay, G., 1990, Analysis of hydrologie relations between Egirdir
Beysehir-Sugla Lakes systems and adjacent basins by means of remote sensing
techniques; (S. Turkey). Environmental Geology and Water Sciences, USA, (in
press).
Hydrogeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990). _
IAHS Publ. no. 207, 1993. 107

FIRST RESULTS FROM THE MONITORING SYSTEM OF THE KARSTIC

COMPLEX "GROTTE DI FRASASSI-GROTTA GRANDE DEL VENTO"
(CENTRAL APENNINES, ITALY)

W. (V. U.) DRAGONI

Earth Sciences Department, Perugia University, Piazza dell'Université, 06100 Perugia, Italy

A. VERDACCHI
Consorzio Frasassi, 60040 Genga (Ancona), Italy

ABSTRACT
The karst complex of "Grotte di Frasassi-Grotta Grande del Vento" is located in the
gorge cut by the River Sentino in the anticline of Mt. Valmontagnana, about 50 km from
the town of Ancona (central Italy). In order to manage the caves in a rational way, and
to get new information about the karstic processes at the cave "Grotta Grande del Vento",
a computerized monitoring system was installed for temperature, humidity, rain,
percolation and air velocity, inside and outside the cave complex. The first data collected
suggest the following preliminary results: (a) Normally the flow of groundwater is
towards the Sentino River. During floods this flow is reversed. The effect of the waters
mixing can increase karstic dissolution. This hypothesis seems to be confirmed by the
greater dimensions of the cavities close to the Sentino River, (b) As expected there is a
close correlation between air flow through the cave and the temperature difference inside
and outside the cave. However the data seem to show that in some zones of the caves the
air flow is mainly controlled by the processes of condensation-evaporation, (c) The
condensation phenomena probably play an important role in the karstic evolution of the
system, (d) An initial estimation of the groundwater draining into the Sentino River along
the Frasassi Gorge has been made (about 50 1/s); according to the Maillet equation the
depletion constant of the river is 2.8 X 10"2 day"1, that of the aquifer is around
8.4 x 10"2 day"1.

INTRODUCTION
In central Italy, about 50 km from Ancona, in the Frasassi Gorge, one of the most
spectacular karst complexes in Europe has developed. The largest underground system
in the area is that known as "Grotta Grande del Vento" (Great Cave of the Wind), the
explored part of which extends for over 15 km.

The Grotta Grande del Vento system is economically quite important to the area: since
being opened to tourists in 1974, approximately 400 000 have visited the cave each year.
The part open to tourists is run by a public agency, the "Consorzio Frasassi", which has
set up a Scientific Committee with the task of studying local karst processes and how
they might be altered by the steady flow of tourists in the cave. For example the visitors
alter the C0 2 content in the air of the cave and algae tend to grow in artificially lit areas,
spoiling the best speleothems. The Committee is composed of small teams of scientists
from various fields and is divided into three different sections: the first one deals with
the chemical, physical and geological aspects of the problem (Physical Section), the
second one is the Bio-Speleological Section; the third one is the Dating Section, with the
task of dating the speleothems using isotopic methods.

In early November 1989 a continuous real-time monitoring system was installed for
monitoring the physical data most relevant to the aims of the Scientific Committee. Up
to now priority has been given to temperature, rainfall, percolation inside the caves,
108 W. (V. U.) Dragoni & A. Verdacchi

atmospheric pressure, humidity, air velocity and carbon dioxide content in the air.
We expect these data will make an important contribution towards the understanding of
some of the phenomena taking place in a karst environment. In this paper the monitoring
system will be briefly described, and some of the preliminary results inferred from the
data obtained will be discussed.

GEOLOGICAL AND GEOMORPHIC FEATURES

Figure 1 gives a simplified geological sketch of the area where the Frasassi Gorge is
located. Here the geological situation has been synthesized. The reference list includes
many entries which give information on the Gorge.

The Grotta Grande del Vento cave system has developed entirely within the Mt.
Valmontagnana anticline core, which is situated on the western side of the Umbria-
Marche ridge. This anticline has an Apennine trend (northwest-southeast) with an
Adriatic virgence (northwest). Towards the west it joins with the Camerino syncline,
whereas towards the east it crosses the southern edge of the Pergola syncline.

, tssl 5 D o • 7 H o • '9 HlO

11 12 — 1 — 13 0 14 O 15 • 16
Fig. 1 - Geological sketch of the Frasassi area. 1: Calcare Massiccio;
2: Bugarone and heteropic formations.; 3: Maiolica ; 4: Marne a Fucoidi;
5: Scaglia Bianca e Rosata; 6: Scaglia Cinerea e Variegata; 7: Bisciaro;
8: Schlier; 9: alluvial deposits; 10: debris and landslides; 11: fault;
12: uncertain fault; 13: anticline axis; 14: syncline axis; 15: major
sulphatic spring; 16: beginning and end of the Frasassi Gorge.
 The monitoring system of the "Grotte di Frasassi-Grotta Grande del Vento" 109

The thickness of the sediments making up the Mt. Valmontagnana anticline is less than
that of the adjacent synclines, due to settling in low depth conditions over a "structurally
lifted block" existing during the entire Jurassic period. This brought about two different
sequences in the sedimentation area: the Mt. Valmontagnana sequence (condensed
sequence), and that of the adjacent synclines (complete sequence).

Triassic evaporitic formations ("Formazione di Burano"?), which do not outcrop in the

Frasassi area, should be present below both sequences. Both sequences begin with the
limestone formation of "Calcare Massiccio" (Hettangian-Sinemurian) which is made up
of beds ranging in thickness from about 1 m to 10 m. This formation has a total thickness
of between 500 and 700 m. The uppermost portion of the Calcare Massiccio has about
80 m of typical oolitic facies (Colacicchi & Pialli, 1974).
On a large scale the Calcare Massiccio is the most permeable formation in the area.
Below the degree of permeability of the geological formations is given with respect to
the permeability of the Calcare Massiccio.
In the condensed sequence, from Lias to Tithonian, the deposition of the "Bugarone"
formation took place, after the Calcare Massiccio, in conditions of a "structurally lifted
block". This formation is composed of thin beds of limestone nodules, reaches a
maximum thickness of 50 m (Centamore et al., 1975), and, on the whole, can be
considered as having medium permeability. The other sequence, deposited in pelagic
basins, is heteropic with the Bugarone, about 450 m thick, and composed of limestone,
marly-limestone, and cherty-limestone (from Sinemurian p.p. to Lower Tithonian), with
an overall medium-low permeability.

From the Tithonian on, both sequences continue with the typical Umbria-Marche
sequence:
— "Maiolica" (Lower Tithonian p.p.-Aptian), made up of bedded limestone, with
medium-high permeability;
— "Marne a Fucoidi" (Aptian p.p.-Cenomanian), consisting of marls, with very low
permeability;
— "Scaglia bianca e rosata" (Middle Cenomanian-Middle Eocene p.p.), made up of
limestone and marly-limestone in the upper portion, with medium-high permeability;
— "Scaglia variegata e Cinerea" (Middle Eocene p.p.- Oligocène), made up of marls,
marly-limestone, with low permeability;
— "Bisciaro" (Lower Miocene), made up of marls and silty-clays, with very low
permeability;
— "Schlier" (Lower Miocene), made up of marls, silty-marls and clays, with very low
permeability.
Although there is a certain amount of tectonic activity in the Frasassi area at the present
time (Forti & Postpischl, 1979; Centamore et al., 1978), the features of the zone are
essentially due to past tectonic stages; in particular, conjugate strike-slip systems were
most likely formed in the late Pliocene, the dextral faults striking NNW-SSE and the
sinistral faults northeast-southwest. The latter are definitely prevalent (Menichetti, 1988;
Cocchioni et al., 1988).

The north-south system of inverse faults existed before the strike-slip systems. The
Sentino stream flows through the northern part of the Mt. Valmontagnana anticline,
cutting deeply into the Calcare Massiccio core (Fig. 1). The latter is the formation which
is most affected by karst phenomena, although they also occur in the Maiolica and in the
Bugarone (Coltorti & Galdenzi, 1982). On the basis of the morphology of Frasassi Gorge,
it appears that the deepening of the Sentino River into the Mesozoic formations
underwent a progressive migration towards the north.
110 W. (V. U. ) Dragoni & A. Verdacchi

Fig. 2 - Map of the karstic system "Grotta Grande del Vento-Grotta del
Fiume". The dimensions of the cavities increase close to the river; the
main directions are parallel to the main fault systems.

The karst phenomena, enhanced by some primary porosity in the Calcare Massiccio,
developed accordingly to pre-existing tectonic features, such as faults and joints (Figs 2
and 3). At the Frasassi Gorge at least four main karst plains can be recognized (Cattuto
& Passeri, 1972; Cattuto, 1976); according to a more recent study there seem to be eight
plains (Antinori, 1979). These plains are often matched by particular external
morphologic features such as terraces, orographic benches, etc. This shows that the
formation of the gorge was characterized by stasis periods.

Two types of groundwater are present in the area. One has a bicarbonate chemical
composition, originating essentially from in loco infiltration; it floats in places on a
water with sulphate-chloride composition. This second water, which acquires its chemical
composition from contact with the Triassic evaporitic formations, belongs to a confined

Fig. 3 - Map of the karstic system "Grotta di Frasassi-Mezzogiorno". The

cave is located in the Frasassi Gorge, at the same altitude the upper levels
of Grotta Grande del Vento. Note the parallelism of the main directions
of the cavities and how the dimensions of the cavities increase close to the
paleo rivers.
 The monitoring system of the "Grotte di Frasassi-Grotta Grande del Vento" 111

aquifer. This aquifer, along the axis of the Mt. Valmontagnana anticline and in the
presence of trans-tensive faults, becomes unconfined. Both waters are drained by the
Sentino, and some sulphate-chloride springs are present along the Frasassi Gorge. The
karst phenomena of the Frasassi Gorge are largely controlled by the interaction between
the two waters and by the presence of H2S (Cucchi & Forti, 1988).

CLIMATIC AND HYDROLOGICAL CHARACTERISTICS

The Sentino basin, at the end of the Frasassi Gorge (Fig. 1), has an area of 285 km2,
with an average altitude of 495 m a.s.l. The average precipitation in the area is about
1200 mm/year, and the average temperature is approximately 13.5°C. Real
évapotranspiration according to the Turc formula averages 650 mm/year. Precipitation
reaches peaks in November and April, while July is the driest month. The greatest
infiltration, strictly tied to water excess, takes place during November, December and
January.

At the present time, continuous measurements are not being taken of the flow of the
Sentino River. Daily measurements taken between 1926 and 1929 at the end of the gorge,
give an average flow of approximately 7 m 3 /s, corresponding to about 775 mm/year over
the entire basin, and to a runoff coefficient of 0.67. This latter value is considerably
higher than that obtained using the Turc formula, on the hypothesis that the
hydrogeological basin coincides with the catchments. It should be emphasized, however,
that the average rainfall in the 1926-1929 period was about 300 mm higher than the
average precipitation, therefore only further investigations may make clear how much this
high runoff coefficient depends on underground contributions from neighbouring basins
and not on abnormal pluviométrie conditions.

As regards the sulphate waters appearing in the gorge, Cigna & Giogelli (1989) proposed
two possible underground permanence times, 5 or 35 years, based on isotopic methods.

Measurements taken using both a current meter and chemical methods show that the
overall flow of the sulphate waters in the Frasassi Gorge is about 50 1/s, which is much
higher than the 27 1/s estimated by Perrone (1911). Our measurements were taken during
a dry period in the winter of 1990 when the average flow of the Sentino River at the end
of Frasassi Gorge was approximately 1 m 3 /s.
According to our measurements the depletion coefficient of the Sentino River, according
to the Maillet equation, should be about 2.8 X 10 2 day"1, that of the sulphate aquifer 8.4
x 10"3 day"1.

THE MONITORING SYSTEM

Attempts were made in the past to take continuous measurements of some parameters in
the Frasassi caves, such as temperature, humidity, and C 0 2 level in the air. However,
until 1988 the instrumentation consisted of rather delicate mechanical systems (recorded
on paper) that made it difficult to obtain reliable and continuous data. In 1989 a new,
entirely computerized, system was set up for continuous, real-time monitoring of the
main underground physical parameters, with magnetic data storage. Figures 4 and 5 show
the planned system.

The completed system will consist of 33 sensors run by two TECA concentrator stations
(small computers which organize and record sensor data a first time). In turn the TECA
concentrators are connected to a computer which automatically processes and records the
data. The use of the TECA stations was necessary by the distances, in some cases more
112 W. (V. U.) Dragoni & A. Verdacchi

Fig. 4 - Sketch of the monitoring system. 1: TECA concentrator;

2: hygrometer; 3: water thermometer; 4: pluviometer; 5: barometer;
6: psychrometer; 7: carbacidometer; 8: thermohygrometer;
9: tacheoanemometer; 10: air thermometer.

than 1000 m, between the sensors and the computer. The voltage losses deriving from
this would have compromised the reliability of the measurements.

All sensors were appropriately chosen for the environment in which they would be
operating: underground climatic conditions are a considerable handicap for electronic
systems, especially the humidity, which is almost always close to 100%.

The fact that each TECA concentrator can handle around 60 sensors makes expansion of
the entire system easily possible. At the present time 23 of the 33 planned sensors are
in operation. The sensors at present activated are:

six air temperature thermometers (sensitivity 0.01 degrees);

two water temperature thermometers (sensitivity 0.01 degrees);
six hygrometers (sensitivity 0.01%);
two raingauges (sensitivity 0.25 mm);
five tacheoanemometers (sensitivity 0.1 mm/s);
two barometers (0.1 millibar).

Fig. 5 - Plan of the monitored portion of the Grotta Grande del Vento
system (for explanation see Fig. 4).
 The monitoring system of the "Grotte di Frasassi-Grotta Grande del Vento" 113

In the next few months another 10 sensors should be put into operation for measuring
C0 2 in the air of the cave. In the future it is hoped that gauges will be installed for
measuring the water table level and the flow of the Sentino.

All the instruments have to be calibrated periodically to prevent problems deriving from
instrumental inaccuracy. Up to now it was not possible to calibrate continuously all the
sensors; however the data on which the following results are based have been recorded
by calibrated instruments and are reliable.

INITIAL DATA AND IMPLICATIONS

Internal meteorology and karst phenomena

The data on temperature, air velocity and direction have provided interesting indications
as to the meteorology of the system and the development of karst phenomena. As usual
in caves fairly near the surface, the average internal temperature is more or less constant
and close to the average external temperature. For caves with openings at different
heights there are two types of air circulation. Using T ; for internal temperature and T e
for external temperature, we have:

(a) if T; > T e , air enters from the lower openings and is warmed. This warming
causes a decrease in density, therefore triggering a flow of air from the lower
openings towards the upper openings. This situation is common in winter, and
thus is conventionally called "winter circulation" (Fig. 6A).
(b) if T; < Te , air enters from the upper openings and is cooled. This cooling
causes an increase in density, therefore triggering a flow of air from the upper
openings towards the lower openings. This situation is common in summer, and
thus is conventionally called "summer circulation".
The two situations outlined above may take place within a 24-h span, when the day and
night temperatures are alternately higher and lower than the internal temperature; this
generally takes place in spring and autumn (Fig. 7).

B i D ! H a ^ 4
Fig. 6 - Air circulation in the Grotta Grande del Vento cave. A: Winter-
type circulation; B: summer-type circulation. 1: oversaturated area;
2: condensation zone; 3: underground water; 4: direction of air
circulation; T ; , Te: temperature measurement points (see Fig. 7); P t , P 2 :
air speed measurement points.
114 W. (V. U.) Dragoni & A. Verdacchi

Fig. 7 - Air flow and temperature of the Grotta Grande del Vento cave
system. Data were recorded at minute intervals between 18 and 21 March
1990. T; = internal temperature (°C), T e = external temperature (°C), V^
= air speed (cm/s) at point Fl of Fig. 6, V2 = air speed at point P 2 of
Fig. 6. Up and down arrows show the air flow direction.

This outline does not take into consideration other factors influencing the flow of air in
a cave, such as the velocity and direction of the external wind with respect to the
openings, the variations in barometric pressure outside, and the effect of evaporation and
condensation phenomena inside the cave. As regards the last two factors in particular,
it should be kept in mind that, with other conditions being the same, humid air is less
dense than dry air. This implies that, with other conditions being the same, a mass of
humid air tends to rise when immersed in a mass of dry air. Thus, in a cave with two
openings at different heights, a liquid water surface inside, the same temperature inside
and outside and undersaturated air outside, a rising current is set off (Castellani &
Dragoni, 1986).

Evaporation-condensation factors inside caves have been studied by various authors, who,
although having little available data, have hypothesized as to the importance of such
factors in the development of karst phenomena (for example, cf. Castellani & Dragoni,
1982, 1987; Cigna & Forti, 1986). The data and observations currently available at
Frasassi seem to confirm these hypotheses, adding new elements.

Figure 6 is a conceptual sketch of the Frasassi system, topographically too complex to

be shown here exactly. Figure 7 gives the internal and external temperatures and the air
velocity registered in the period between 18 and 21 March 1990. Figure 7 gives the air
temperatures; the air velocity through an upper opening, corresponding to passage P t in
Fig. 6; and the air velocity recorded in an internal shaft (Falconara shaft) at about 450
m from the lower opening, corresponding to point P 2 in Fig. 6, and also shown in Figs
2 and 4.

Figure 7 shows that the Grotta Grande del Vento cave system, the behaviour of the air
flow in the external zones of the caves is different from that in the internal zones. As
 The monitoring system of the "Grotte di Frasassi-Grotta Grande del Vento" 115

expected in the external zones the air flow is controlled mainly by the value and sign of
the difference (T; - T e ).
In the internal-lower zones the air flow is controlled mainly by the humidity-evaporation
effect. The air flow is always ascending and the thermal effect only slows down the air
velocity, without inversion. Probably only when the values (T e — T;) are bigger than a
threshold value greater than 8°-9°C "summer circulation" reaches the sensor P 2 at the
Falconara shaft.

It is interesting to underline that in any of the cases there can be condensation on the
cave walls. During "winter circulation" (Fig. 6A) the air rises, expands and cools.
Condensation also takes place, obviously, in the upper openings near the outside, where
the temperature is lower than the internal temperature. During "summer circulation"
condensation could happen due to the internal ascending circulation and because the flow
coming from outside could cool below dew point (Fig. 6B). The presence of condensation
phenomena are suggested by the value of the relative humidity, which is always close to
100%, and by the frequent formation of fog in the upper part of the cave.

The evaporation of the internal lakes of the cave is indicated by the fact that the surface
temperature of the internal lakes is generally lower than the air temperature by almost
a tenth of a degree. Furthermore the air temperature in different points of the cave
system shows that the innermost lower zone is constantly colder than the outer zones
(about 1°C). This can be explained by admitting that the innermost lower zone is affected
essentially by evaporation, which is an endothermic process; from this point of view, the
peripheral upper-middle areas, where usually condensation should take place, are warmer
due to the liberation of latent condensation heat.

The condensed water is aggressive against carbonates given the presence of C 0 2 and H2S,
so that it is difficult to deny the importance of condensation in post-phreatic karst
processes.

Flood events and karst

Data on the temperature of the aquifer have brought to light an aspect of the processes
present in the Frasassi system, which until now has been overlooked. It was said
previously that in ordinary conditions the Sentino drains the groundwater from the Mt.
Valmontagnana limestone massif. During floods, however, water from the river
temporarily feeds the aquifer.
Figure 8 gives data on the temperature of the water inside the cave, at a linear distance
of about 450 m from the Sentino River, during a flood which took place in November
1989; the same figure gives the recorded rainfall. During the flood the external water,
colder than the internal water, was able to lower the temperature of the groundwater by
up to 10°C in six days. As it is known, the mixture of two waters with different C 0 2
contents creates an undersaturated mix, even if both are saturated in carbonates; the
effect is increased when the two waters are at different temperatures (Bogli, 1960a,b).
It should be pointed out that the available data indicate a stratification of the C 0 2 at
Frasassi in the lowest levels of the cave, with concentrations in the air up to 1000 ppm
(Castellani, 1989); a cooling of the water implies an increase in the C 0 2 content of the
water with a subsequent increase in the aggressiveness against carbonates. All of this
suggests that flooding contributes considerably to the development of karst phenomena.
This appears to be confirmed by the fact that, going towards the river, the dimensions
of the caves get larger and larger (Figs 2 and 3).
116 W. (V. V.) Dragoni & A. Verdacchi

Fig. 8 - Temperature of groundwater in the cave at a point 450 m from the

Sentino River (November 1989). 1: Rainfall in millimetres; 2:
Temperature of the water.

FINAL REMARKS

The data collected so far have allowed us to confirm some hypotheses made in the past,
and have also offered the possibility of pointing out some aspects of karst phenomena
which would not have been observable with less sophisticated equipment. In the first
place, we were able to better define the mechanisms and the characteristics of the
circulation of air in the "Grotta grande del Vento" cave system. In this context, it was
possible to indicate the areas where condensation may take place, according to the
different types of circulation. The result of this is important as the condensation waters
are aggressive. In this light, one could ascribe to their action the formation of particular
karst features such as inverse gutters, mount milk, etc. often encountered at the Frasassi
karst systems.

The data suggest the existence of a particular karst factor due to the action of the Sentino
River flood waters. The development of the cave dimensions within the Frasassi karst
complex, with respect to the distance from the river, seems to prove this hypothesis.

The results reached up to this point are qualitative and preliminary. The completion and
final setting up of the monitoring system should permit more definitive conclusions and
the making of quantitative models, possibly with results of general interest.

REFERENCES
Antinori, A., 1979, Studio geologico-strutturale dell'area carsica di Frasassi. Tesi di
Laurea, Univ.di Camerino.
 The monitoring system of the "Grotte di Frasassi-Grotta Grande del Vento" 117

Bogli, A., 1960a, Kalklosung und karrenbildung. Z. Geomorphol., suppl. 2, 4-21.

Bogli, A., 1960b, Les phases de dissolution du calcaire et leur importance pour les
problèmes karstiques. Rass. Speleol. It., 12, 167-180.
Castellani, V. & Dragoni, W. (V. U), 1982, About the genesis of karstic cavities. 2°
International Symposium on Utilization on Karstic Areas (Bari, May 1982).
Castellani, V. & Dragoni, W. (V. U.), 1986, Evidence for karstic mechanism involved
in the evolution of Moroccan hamadas. /. Speleol., 15, 57-61.
Castellani, V. & Dragoni, W. (V. U.), 1987, Some consideration regarding karstic
evolution of desert limestone plateaux. International Geomorphology 1986, pt. II,
1199-1205.
Castellani, V., 1989, Frasassi e speleomonitoraggio. Speleologia, 18, 33-35.
Cattuto, C , 1976, Correlazione tra piani carsici ipogei e terrazzi fluviali nella valle del
fiume Esino (Marche). Boll. Soc. Geol. It., 95, 147-160.
Cattuto, C. & Passeri, L., 1972, Relazioni tra idrologia carsica e litologia nell'area
umbro-marchigiana. Acta Cong. Naz. Spel., XI, vol. 1, 227-238.
Centamore, E., Catenacci, V., Chiocchini, M., Chicchini, V., Jacobacci, A., Martelli,
G., Micarelli, A. & Valletta, M., 1975, Carta geologica d'ltalia: 1:50.000, Note
illustrative del foglio 291 Pergola, 1-40.
Centamore, E., Deiana, G., Dramis, F., Micarelli, A. Carloni, G. C , Francavilla, F.,
Nesci, O. & Moretti, A., 1978, Dati preliminari sulla neotettonica dei fogli 116
(Gubbio): 123 (Assisi), etc. Contributi preliminari alia realizzazione della carta
neotettonica d' Italia, 113-148. CNR.
Cigna, A. & Forti, P., 1986, The speleogenetic role of air flow caused by convection.
Int. J. Speleol., 15, 41-52.
Cigna, A. & Giorgelli, F., 1989, Underground water dating by tritium measurement. Ac.
Int. Spel. Cong. (Budapest).
Colacicchi, R. & Pialli, G., 1974, Significato paleogeografico di alcuni depositi nella
parte sommitale del Calcare Massiccio (Nota preliminare). Boll. Soc. Geol. It., 92
(suppl), 173-187.
Coltorti, M. & Galdenzi, S., 1982, Geomorfologia del complesso carsico: "Grotta del
Mezzogiorno"(4 MA-AN) con riferimento ai motivi neotettonici, etc. Studi Geologici
Camerti, VII, 123-132.
Cocchioni, M., Coltorti, M., Dramis, F., Mariani, F. & Tazioli, G. F., 1988,
Circolazione idrica e chimismo delle acque sotterranee dell'area carsica di Frasassi
nelle Marche. Ac. Nat. Cong: "Carsismo nella gola di Frasassi", preprint.
Cucchi, F. & Forti, P., 1988, Evoluzione speleogenetica del complesso carsico: "G. G.
del Vento- G. del Fiume", S. Vittore, Marche. Ac. Nat. Cong. "Carsismo nella gola
di Frasassi", preprint.
Forti, P. & Postpischl, D., 1979, Determinazioni di dati neotettonici da analisi di
concrezioni alabastrine. I e II contrib. alia Carta Neotettonica d'ltalia C.N.R., pp.
1399-1409, 634-644. CNR.
Menichetti, M., 1988, Influenze strutturali nello sviluppo del carsismo della gola di
Frasassi, (Appennino marchigiano). Ac. Nat. Cong. "Carsismo nella gola di
Frasassi", preprint.
Perrone, A., 1911, Memorie illustrative della carta idrografica d'ltalia: N.35, p. 203.
Hydrogeol&gical Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHS PÛbl. no. 207, 1993. 119

ABSORPTION MASSIVE D'EAU DE MER PAR DES AQUIFERES

KARSTIQUES COTIERS

C. DROGUE
Laboratoire d'Hydrogéologie, Unité Associée au CNRS, XJSTL, Montpellier 11, Place E. Bataillon, F-34095
Montpellier Cedex 5, France

RESUME
Des phénomènes d'absorption massive d'eau de mer par des cavités d'aquifères karstiques
littoraux ou sous-marins, sont connus en diverses régions du monde. Deux exemples
étudiés dans cette note, sur des karsts, méditerranéens permettent d'expliquer certaines
caractéristiques hydrodynamiques de ces écoulements très particuliers. Dans le Sud de
la France, une source absorbe en certaines périodes jusqu'à 150 1 s*1 d'eau de mer,
simultanément à un débit sortant d'eau douce par le même griffon (jusqu'à 200 1 s"1). Ces
écoulements juxtaposés et de sens inverses, de deux veines liquides de densités
différentes, sont dûs aux surélévations du niveau de la lagune côtière sous l'effet de la
marée et des vents venant du large. En Grèce, dans l'île de Céphalonia, l'eau de mer
absorbée de façon permanente (jusqu'à 300 1 s"1) réapparaît sur la côte opposée de l'île
par des sources d'eau saumâtre. La nouvelle explication proposée, repose sur la mise en
évidence d'une différence d'altitude des niveaux de la mer de part et d'autre de l'île, qui
peut atteindre 0.3 m, due à un courant marin SSE-NNW. L'absorption est entretenue par
des écoulements densitaires. Ces deux phénomènes sont rendus possibles par une
karstification profonde héritée vraisemblablement des régressions de la Méditerranée, qui
se sont produits au Messinien et au Wùrm.

ABSTRACT
High rates of flow of sea water into littoral or submarine karstic caves have been
described in several areas through the world. Some of the hydrodynamic features of such
flow can be explained, based on two examples from Mediterranean karst areas. In
southern France, in certain periods a spring absorbs up to 0.15 m 3 s"1 of sea water and,
simultaneously, discharges up to 0.2 m* s"1 of fresh water. These two superposed streams
of fluids with distinct densities and inverse flow directions, are caused by uprisings in
the level of the coastal lagoon, under the effects of the tide and of winds blowing from
the sea. In Greece, on Cephalonia island, sea water that is permanently absorbed (up to
0.3 m3 s"1) re-emerges at briny springs on the opposite side of the island. The new
explanation relies on the observed difference in the elevation of the sea level between
both sides of the island, up to 0.3 m, due to a SSE-NNW marine current. Buoyancy
effects keep the flow on. Both studied phenomena are possible in view of a deep
karstification that originates from a lowering of the Mediterranean sea level during the
Cenozoic period.

INTRODUCTION
L'absorption d'eau de mer par des cavités karstiques littorales ou sous-marines est un
aspect tout à fait original de l'interaction entre hydrodynamique souterraine et
hydraulique océanique. C'est un phénomène très étonnant si l'on considère que le niveau
des océans représente pour les eaux, dans le champ de gravitation terrestre, la côte de
référence à potentiel minimal à laquelle elles ne peuvent fournir aucun travail.
120 C. Drogue

Un potentiel moindre ne peut donc être atteint qu'au prix d'un apport d'énergie. C'est
le cas lorsqu'un courant d'eau de mer pénètre dans un sous-sol côtier. Quelle est l'origine
de cette énergie? De nombreux travaux, depuis plusieurs décennies, se sont attachés à
élucider cette question sur quelques exemples observés dans le monde.
Le cas le plus fréquent est probablement celui de sources côtières dont l'écoulement
naturel s'inverse lors de marées hautes importantes, le niveau de la mer étant alors
temporairement plus élevé que celui des eaux souterraines près de la côte. A marées
moyennes ou basses, le courant s'inverse, l'eau de mer est rejetée, et la source reprend
un écoulement normal. Ceci a été décrit notamment aux Bahamas sur Blue Hole d'Andros
Island (Stringfield & Legrand, 1971; Kohout, 1960a,b). L'inversion de l'écoulement sous
l'effet de la marée a été décrit également sur la côte Yougoslave de l'Adriatique (Kucer,
1950).
Un autre processus est illustré par Spring Bayou en Floride (Stringfield & Legrand,
1969). L'eau saumâtre de l'embouchure d'Anclote River s'écoule souterrainement sur
3 km, vers le lac Tarpon, lorsque le niveau du lac est plus bas que celui d'Anclote River.
L'écoulement s'inverse quand le niveau du lac est supérieur à celui du fleuve. Il s'agit,
dans ces deux exemples, d'absorptions temporaires alternées avec des périodes
d'écoulement sortant.

Plus rares sont: d'une part, les descriptions d'absorptions d'eau de mer permanentes et
d'autre part, les absorptions se produisant simultanément avec un écoulement sortant
d'eau douce par la même cavité karstique. Cette note traite de ces deux types de
phénomènes à partir de deux exemples particulièrement spectaculaires (Fig. 1).
Il s'agit d'abord de la discussion d'une nouvelle hypothèse concernant la cause de
l'absorption permanente d'eau de mer sur le site d'Argostolia, dans l'île de Céphalonia
(Grèce). Ensuite, il sera explicité le mécanisme d'absorption d'eau de mer simultanément
à un rejet d'eau souterraine, exemple d'une source du sud de la France.

Fig. 1 - Situation sur le littoral nord-Méditerranéen, des deux phénomènes

étudiés :
1: Absorption quasi-continue d'eau de mer sur le site d'Argostolia (île
de Céphalonia, Grèce).
2: Absorption discontinue sur le site de la Roubine (sud de la France).

NOUVELLE HYPOTHESE SUR L'ORIGINE DE L'ENERGIE RESPONSABLE DE

LA PENETRATION D'EAU DE MER DANS LE SOUS-SOL DE L'ILE DE
CEPHALONIA (GRECE)

L'île de Céphalonia en Méditerranée, près de la côte occidentale de la Grèce, présente

 Absorption massive d'eau de mer par des aquifères karstiques côtiers 121

l'un des phénomènes hydrologiques les plus étonnants que l'on puisse observer dans le
monde. En effet, sur le littoral sud-ouest de cette île, près de la ville d'Argostolia, un
courant d'eau de mer s'engouffre en quasi permanence dans le sous-sol karstique, par des
cavités (cathavotres) creusées sur des fractures de la roche. Cette roche est constituée de
calcaires et dolomies du Jurassique, Crétacé et Cénozoïque, l'ensemble étant épais de
900 m.

Ce courant, qui peut atteindre un débit de 0.3 m 3 s"1, est suffisamment puissant pour
actionner des roues de moulins (sea mills). Maurin & Zoetl (1967) ont montré à l'aide
d'un traceur, que l'eau de mer absorbée, réapparaissait sur la côte opposée de l'île, par
des sources d'eau saumâtre (près de la ville de Sami). Le trajet souterrain est de 15 km
(Kg. 2).

. 10k ™. Hill 1 ^ 2 * 3 •*
Fig. 2 - Ile de Cephalonia (Grèce). Ecoulement souterrain de l'eau de mer
absorbéea Argostolia vers les sources Sami (d'après Maurin & Zoetl,
1967). Dans le cartouche, trajectoires du courant marin autour de la Grèce
et près de l'île de Cephalonia. 1: Karst, 2: cathavotre, 3: source, 4:
traçages.

Depuis près d'un siècle, diverses hypothèses ont été proposées pour expliquer ce
phénomène (Crosby & Crosby, 1896) sans qu'aucune n'ait pu être étayée par des données
expérimentales. La plus récente suppose l'existence dans le sous-sol de l'île, d'un
"natural ejector working" qui serait actionné par les eaux d'infiltration (Glanz, 1966).
Cette hypothèse souvent citée (Zotl, 1974); Bogli, 1978) est cependant peu vraisemblable
du point de vue hydrodynamique (Cooper et al., 1964; Stringfield & Legrand, 1969). Ces
derniers auteurs pensent, que la différence de densité est à l'origine du courant d'eau
absorbé: l'eau de mer plus lourde, envahirait l'aquifère karstique renfermant de l'eau
douce.

Une nouvelle étude de ce phénomène à eu pour objectifs de reconnaître le rôle éventuel

des écoulements densitaires, mais aussi la raison pour laquelle l'absorption d'eau salée
se fait toujours du même côté de l'île, sans inversion du courant souterrain (Drogue &
Soulios, 1988; Drogue, 1989).
122 C. Drogue

Gradient hydraulique et courants densitaires

Une des premières observations que l'on a pu faire, est l'existence non relevée jusqu'à
présent, d'un courant marin permanent qui traverse l'archipel ionien du SSE vers le
NNW. De ce fait, le niveau de la mer sur la côte située face au courant (où se produit
d'ailleurs l'absorption) doit être naturellement plus élevé que sur la côte opposée (ou
apparaissent les sources). Ce gradient hydraulique entre les deux côtés de l'île,
expliquerait le courant souterrain comme cela avait été envisagé dans un travail très
ancien (Fuller, 1906). Pour vérifier cette hypothèse, un nivellement a été réalisé au
tachéomètre électronique à partir de deux points du réseau géodésique de l'île. Les
mesures montrent qu'effectivement le niveau de la mer sur la côte où se trouve les
cathavotres, est généralement plus élevé que sur la côte opposée. Mais ce n'est pas
toujours le cas, et cet écart, qui n'est pas constant dans le temps, peut s'annuler sur des
périodes de quelques heures. Au cours de cette étude, la différence d'altitude a ainsi
varié entre 0.30 ± 0.02 m et 0.02 ± 0.02 m.

Ce gradient hydraulique ne peut pas être étranger au phénomène d'absorption. Cependant

il n'en constitue pas la cause unique, puisque l'absorption fonctionne lors des périodes
où le gradient est nul. De plus, des relevés dans l'aquifère karstique de part et d'autre
de l'île, mettent en évidence une pente piézométrique apparente qui serait à l'inverse de
l'écoulement des eaux souterraines. En effet, près des points d'absorption, le niveau
d'eau dans une cavité du karst était au cours de l'étude, à - 1 . 7 0 ± 0.02 m par rapport
au zéro géodésique de l'île (soit à —0.73 m sous le niveau moyen de la mer alors que sur
la côte opposée, près des sources, un autre niveau d'eau était à -0.7 ± 0.02 m, donc plus
haut que le niveau amont. C'est apparemment tout à fait paradoxal, sauf si on considère
que les eaux souterraines ont des densités variables, décroissantes vers l'aval.

L'eau de mer a une masse volumique de 1.258 kg dm3 à 25°C, alors que celle de l'eau
saumâtre des sources de Sami n'est que de 1.0002 kg dm3 à 15 °C (températures
respectives de l'eau de mer et des eaux souterraines en juin 1987). Avec ces valeurs,
pour le calcul des charges hydrauliques, il faut connaître l'épaisseur de l'aquifère.

Karstification
On ne possède pas pour l'île de Céphalonia, de données sur les profondeurs maximales
sous le niveau de la mer, des circulations d'eau souterraine. Cependant la
paléogéographie fourni sur ce problème des informations d'un grand intérêt.

On sait que les magasins karstiques peuvent se développer très profondément sous le
niveau des sources, aidés en cela, en zones côtières, par les régressions marines. Ainsi,
la Méditerranée, au Cénozoïque, a évolué en bassin fermé sous climat chaud. Il se serait
alors produit au Messinien (7 à 9 Ma BP) un fort abaissement du niveau marin de
plusieurs centaines de mètres sous le niveau actuel (Hsu et al., 1978; Steckler.& Watts,
1980). Après une remontée générale, une nouvelle régression à —110 m s'est produite
lors de la glaciation Wûrm, soit à 35000 ans B.P. (Monaco et al., 1972). Le magasin
aquifère s'est formé dans la profondeur de la masse carbonatée à la faveur de ces
périodes marines régressives et est aujourd'hui pour l'essentiel sous le niveau de la mer.
Ceci explique que l'on connaisse des sources sous-marines, à —80 m, près des côtes de
la Grèce continentale et à -150 m au Liban (Mistardis, 1965; Moullard et al., 1965;
Burdon, 1965; Kranjc, 1982). Admettons donc une profondeur minimale de karstification
c'est à dire des possibilités de circulation d'eau, à 100 m sous le niveau actuel de la
Méditerranée. La piézométrie près des sources de Sami avec des eaux saumâtres, serait
alors supérieure de 2.56 m à la piézométrie aux points d'absorption d'eau de mer.
Comme l'écart observé n'est que de lm, la différence (1.56 m) permet de justifier
l'écoulement souterrain. Ainsi le phénomène amorcé par la surélévation du niveau de la
 Absorption massive d'eau de mer par des aquiferes karstiques côtiers 123

mer sur la côte d'Argostoli, serait entretenu par les effets densitaires, ce qui explique que
l'absorption se continue alors même que l'écart d'altitudes entres les deux niveaux marins
est nul.
Cependant, en Méditerranée comme ailleurs dans le monde, les autres îles karstiques
placées dans un courant marin, ne présentent pas ce phénomène d'absorption. C'est donc
qu'il y a à Céphalonia, une caractéristique particulière. Il faut considérer que l'eau
saumâtre des sources de Sami est un mélange, dans le sous-sol, d'eau douce due à
l'infiltration des pluies sur les reliefs de l'île (800 à 900 mm par an) et d'eau salée
absorbée à Argostolia. Pour que l'écoulement souterrain ne soit pas inversé à Argostolia,
il faut au coeur de l'île, que la charge hydraulique alimentée par l'infiltration soit plus
faible que celle existant à Argostolia. Pour cela, comme l'eau d'infiltration a une masse
volumique de 0.9990 kg cm"3 à 15°C (mesurée sur un suintement dans une grotte, à 50 m
de profondeur en juin 1987), le niveau des eaux douces souterraines ne doit pas dépasser
3 m au dessus du niveau de la mer à Argostolia(en raisonnant encore avec une épaisseur
minimale aquifère de 100 m) (Fig. 3).

Fig. 3 - Schématisation du fonctionnement de l'écoulement d'eau salée

dans le karst de l'île de Céphalonia (à la température de 15°C).
e: Piézométrie calculée avec la densité de l'eau de mer à Argostolia.
f: Piézométrie calculée avec la densité de l'eau saumâtre des sources de
Sami.
g: Piézométrie calculée avec la densité de l'eau du karst, non contaminée
par l'eau de mer.

Cela suppose de faibles pertes de charges dans les écoulements souterrains et l'existence
de conduits karstiques très perméables reliants les deux cotés de l'île. Ces conduits se
sont formés probablement de la façon suivante (Glanz, 1966) lors des régressions
marines: le fond marneux de la baie de Livadi, où se trouve Argostolia, et dont la
profondeur maximale actuelle est de 28 m, a été émergé. Sur ce fond imperméable, les
eaux de pluie ont formé des ruisseaux qui se sont enfouis dans les calcaires à la limite
124 C. Drogue

des marnes, par les cathavotres. Ces écoulements sont réapparus sur la côte opposée et
ont développés ainsi des conduits karstiques importants, aujourd'hui utilisés par l'eau de
mer.

Conclusion sur l'origine du courant marin

C'est donc l'énergie d'un courant marin qui est à l'origine de l'absorption d'eau de mer
à Céphalonia. Le phénomène est permanent car il est entretenu par des écoulements
densitaires. Il a été rendu possible enfin, par l'existence d'un conduit karstique, mis en
place probablement au Cénozoïque, et orienté comme le courant marin.

ABSORPTION D'EAU DE MER ET ECOULEMENT SORTANT D'EAU

SOUTERRAINE SIMULTANES
Situation de la source étudiée
Un exemple de ce phénomène a été étudié dans le Sud de la France, sur une source de
la côte méditerranéenne: source de la Roubine (Drogue & Bidaux, 1986). C'est une
émergence qui apparaît à la limite d'un massif carbonate mésozoïque, intensément
karstifié, au contact de dépôts argilo-marneux Cénozoïques et Quaternaires de la plaine
littorale. Ces roches carbonatées d'une puissance de 800 m environ jusqu'au Lias
marneux, se poursuivent sous cette plaine et sous la mer. Elle sont très fracturées et
karstifiées (Fig. 4).

3*48 '

S karstic. 2 3 nokmtie . S kg»

Fig. 4 - Situation de la source de la Roubine (France), près d'un étang,
à la limite d'un massif karstique (sud de la France).

Cette source est constituée de plusieurs griffons, répartis sur une superficie de 2000 m 2 ,
dont l'écoulement s'effectue, par un canal de 2.5 km dans un étang séparé de la mer par
un étroit cordon dunaire. Le débit, en fortes crues, peut atteindre 6 m 3 s"1. Il est de 0.1
à 0.2 m3 s"1 à l'étiage en fonctionnement normal de la source, c'est à dire avec un
écoulement sortant.
 Absorption massive d'eau de mer par des aquiferes karstiques côtiers 125

Phénomène d'absorption
L'absorption d'eau salée se manifeste seulement lors des faibles débits. Les premières
observations eurent lieu en 1981 (Derozier, 1984). On assiste alors, en l'absence de toute
pluie, à des soulèvements du niveau de l'eau de plusieurs décimètres, d'une durée allant
de quelques heures à 4-5 jours selon les cas.

Au cours de ces mouvements, un courant d'eau salée venant de l'étang se déplace dans
le fond du canal, sous l'eau douce et en sens contraire, pénètre dans un ou deux griffons
et disparaît dans l'aquifère. Dans le même temps, l'évacuation de l'eau douce se poursuit
par les autres griffons. Cette eau douce s'écoule vers l'étang dans le canal, au-dessus de
l'eau salée (Fig. 5).

tracers
t

/ , -z Ov-. mudoy *and fZ

ând grâvêl"LZ
n/ > salt water ^, » -^ » _ i. _

J\.i
karat limestone /

Fig. 5 - Source de la Roubine (France). Coupe géologique et courants lors

d'une absorption d'eau de mer.

Pour ce type de situation, les débits maximaux simultanés observés (période 1982-1984)
furent de 0.18 m3 s"1 pour l'eau douce et 0.20 m3 s"1 pour l'eau salée. L'interface entre
les deux veines liquides est nette si l'on considère les valeurs de la conductivité
électrique et de la température des eaux. Ces deux paramètres sont, en effet, très
différents pour la source et pour l'étang (Fig. 6).

up stream - * - velocity m s-' *• down stream conductivity m S cm - 1 tenperoture °C

0.4 0.2 0.0 02. 0.4 0.0 40 17 18 n n
spring flow T freed
rre« water

->5<>/troneitiofl un/*//////
W//;A

.0.3
incoming flow

0.4

Fig. 6 - Source de la Roubine (France). A quelques mètres à l'aval de la

source profils des: vitesses, conductivités et températures mettant en
évidence la superposition dans le même canal de deux écoulements de
directions opposées.
126 C. Drogue

La conductivité de l'eau de source est de 4.2 à 4.3 ms cm"1 (à 25°C), alors que celle de
l'étang varie, selon les saisons, de 48 à 56 ms cm"1. Ceci amène à formuler une
remarque: par commodité de langage, nous désignons par eau douce, l'eau émise par la
source, mais en fait, cette eau est contaminée par l'eau de mer. En effet, la conductivité
électrique moyenne d'une eau de source de terrains carbonates loin du littoral est de
0.6 ms cm"1 avec une teneur de 15 à 35 mg l"1 en Cl" (l'étang a une teneur en Cl" de
18 000 mg l 1 ) .

On discutera plus loin de la cause de cette contamination. En ce qui concerne la

température, elle est sur la source, relativement stable, oscillant dans l'intervalle 17.3
± 0.5°C. La température de l'étang, par contre, est soumise à de fortes variations: de
5 à 10°C en hiver à 25-28°C en été. De plus, en un même instant, la température peut
varier, d'une part, avec la profondeur et, d'autre part, d'un secteur à un autre du plan
d'eau.
Compte tenu des remarques ci-dessus, le suivi dans le temps de l'absorption fait
apparaître des particularités remarquables (Fig. 7). On relève, notamment, que le début
du phénomène se manifeste de façon abrupte sur les conductivités. C'est la preuve de la
difficile miscibilité des eaux douces et salées. La température, par contre, varie
progressivement avant même l'apparition de l'eau salée. C'est dû au fait que la
progression de l'eau salée dans le fond du canal repousse, vers la source, des eaux
douces déjà influencées par la température extérieure. La température de l'eau salée
absorbée n'est jamais stable alors que la conductivité (ou salinité) peut atteindre un état
stationnaire.

L'arrêt de l'absorption apparaît nettement à la fois pour les conductivités et les

températures. Cela correspond à la reprise de l'écoulement sortant.

En ce qui concerne le mécanisme hydraulique, il apparaît que les oscillations du niveau

d'eau à la source sont provoquées par des variations du niveau de l'étang sur la rive
Ouest. Les mesures démontrent qu'il ne s'agit pas d'effets de marée. Par contre, on a pu
constater que ces oscillations sont associées à certaines directions de vents.

Les basculements de plan d'eau libre sous l'effet des vents est un phénomène bien connu,
avec abaissement près de la rive au vent et soulèvement sur la rive opposée. C'est ce qui
se produit ici. La source est située sur la rive ouest, les vents de secteurs Est provoquent
un soulèvement sur cette rive et abaissement sur la rive est.

Le soulèvement rapide sur la rive Ouest crée donc un gradient hydraulique dans l'eau
salée, dirigé vers la source. Ce gradient est à l'inverse de celui de l'eau douce qui coule,
elle, vers l'étang. La progression de l'eau salée vers la source est favorisée par le faible
gradient hydraulique de l'eau douce dans le canal, de l'ordre de 0.01%, et par la pente
quasi-nulle du fond du canal.

La superposition des deux écoulements, de sens inverses, et la faible miscibilité

apparente des deux veines liquides, sont dues aux différences des masses volumiques
(pour l'eau de source: 0.9994 kg dm"3; pour l'eau salée: 1.0258 kg dm"3 à 25°C). L'effet
de ce contraste augmente, bien sur avec l'abaissement de la température de l'eau salée
en saisons froides.

Enfin, on est en droit de penser que l'eau salée qui pénètre dans l'aquifère réapparaît
après dilution dans l'écoulement de la source, et serait à l'origine de sa salinité.
Pour vérifier cela, des injections de traceurs furent effectuées dans le flux salé absorbé
(jusqu'à 10 kg de fluorescéine sodique pour une expérience). Sur plusieurs mois
d'observations, aucune trace de colorant n'a été décelée. Il y aurait donc, apparemment,
séparation des deux écoulements au sein même de l'aquifère (Fig. 7).
 Absorption massive d'eau de mer par des aquifères karstiques côtiers 127

lagoon Méditerranée sea

cenozoïc and
water table quarternary

karst
limestone

• 2 km. , fait
vertical exaggeration 100: I

Fresh w a f e r Salt water

Fig. 7- Source de la Roubine (France). Représentation des écoulements

souterrains probables dans le karst soumis à une invasion d'eau de mer.

Il est certain toutefois, qu'une partie d'eau salée se mélange avec l'eau de l'aquifère.
Mais, l'essentiel de la salure de la source a, sans aucun doute, pour origine la
contamination des eaux profondes par la pénétration dans le karst de l'eau de mer sous
la forme d'un biseau salé.

Il est vraisemblable que l'eau de l'étang, absorbée, s'écoule par gravité vers l'interface
du biseau salé. Elle doit être, ensuite, entraînée par l'écoulement de la nappe vers des
sources sous-lagunaires ou sous-marines, ou vers des écoulements ascendants diffus, au
travers des autres aquifères constitués par les terrains de couvertures.

CONCLUSION
Dans les deux cas étudiés, on notera, outre le rôle de certains facteurs (courants marins,
vents, marées océaniques) la permanence des effets dûs aux courants densitaires et le rôle
de la karstification profonde.

Celle-ci, sur tous les karsts du pourtour méditerranéen, a été favorisée par les régressions
qui ont permis la mise en place de magasins karstiques à plusieurs centaines de mètres
sous le niveau actuel de la mer ceci explique que ces échanges eaux douces-eaux salées
soient connues en un grand nombre d'autres régions que celles décrites ici.

REFERENCES
Bogli, A., 1978, Karsthydrographie and Physiche Speldologie. Springer Verlag.
Burdon, D. J., 1965, Hydrogeology of source karstic areas of Greece. In: Hydrology of
Fractured Rocks (Proc. DubrovnikSymp. 1967), vol. I, 237-250. IAHS Publ. no. 74.
Cooper, H. H., Kohent, J. F., Henry, H. R. & Glover, R. E., 1964, Sea water in coastal
aquifers, relation of salt water to fresh ground water. US Geol. Survey Wat. Supply
Pap. 1613-c.
128 C. Drogue

Crosby, F. W. & Crosby, W. O., 1896, The sea mills of Cephalonia. Tech. Quarterly
and Proc. Society, Mass. Inst. Technol., Boston, 9,1, pp. 6-23.
Derozier, 1984, Thesis, Univ. Montpellier, France.
Drogue, C. & Bidaux, P., 1986, Simultaneous outflow of fresh water and inflow of sea
water in a coastal spring. Nature, 322, no. 6077, 361-363.
Drogue, C , 1989, Continuous inflow of sea water and outflow of brackish water in the
substratum of the karstic island of Cephalonia, Greece. /. Hydrol., 106, 147-153.
Drogue, C. & Soulios, G., 1988, Absorption massive d'eau de mer et rejet d'eau
saumâtre dans l'île karstique de Cephalonia (Grèce): nouvelle interprétation du
phénomène. C. R. Acad. Sci. Paris, 307, II, 1833-1836.
Fuller, M. L., 1906, Conditions of circulation at the sea mills of Cephalonia. Geol. Soc.
Am. Bull., 18, 221-232.
Glanz, T. H., 1966, Das phanomen der meer miihle von Argostolion. Steir. Beitr.
Hydrogeologië, no. 17, 113-127.
Hsu, K. J., Montadert, L., Bennovilli, D., Cita, M. B., Erikson, A., Garrison, R. E.,
Kidd, R. B., Melieres, F., Muller, C. & Wright, R., 1978, Initial Report of the
Deep Sea Drilling Project, 42(1), 1052-1078. US Government Printing Office,
Washington.
Kohout, F. A., 1960a, Ocean groundwater interface: sink holes and submarine springs:
Geology, hydrogeology windows in Florida Bahamas platform. Annual Meeting
Minneapolis, Minnesota.
Kohout, F. A., 1960b, Cyclic flow of salt water in the Biscayne aquifer of southeastern
Florida. /. Geophys. Res., 65(67), 2133-2141.
Kranj c,A.A.,1982, Proposition d'une bibliographie yougoslave sur le karst littoral. Comité
Nat. Fr. de Géographie. Coll. Perpignan, sur Karsts Littoraux (mai), pp. 41-50.
Kucer, L, 1950, Karst sources at the sea coast. Razprave. Diss. Sazu, cl. Ill, Ser. A., I,
Ljubljana, Yougoslavie, pp. 138-147.
Maurin, V. & Zoetl, J., 1967, Salt water encroachment in the low altitude karst water
horizons of the island Kephallonia (Ionian islands). In: Hydrology of Fractured Rocks
(Proc. Dubrovnik Symp., 1967), vol. II, 423-430. IAHS Publ. no. 74.
Mistardis, G. G., 1965, Recherches hydrogéologiques dans la région des lacs karstiques
béatiens. In: Hydrology of Fractured Rocks (Proc. Dubrovnik Symp., 1967), vol. I,
162-170. IAHS Publ. no. 74.
Monaco, A., Thommeret, J. & Thommeret, Y., 1972, L'âge des dépôts quaternaires sur
le plateau continental du Roussillon (Golf du Lion). C. R. Acad. Sci., Paris, 21 A,
2280-2283.
Moullard, C , Mijatovic, B., Kareh, R. & Massahad, B., 1965, Exploitation d'une nappe
karstique captive à exutoires sous-marins: Problèmes posés et solution adoptée. Côte
libanaise. In: Hydrology of Fractured Rocks (Proc. Dubrovnik Symp., 1967), vol. I,
237-250. IAHS Publ. no. 74.
Steckler, M. S. & Watts, A. B., 1980, The Gulf of Lion, subsidence of a young
continental margin. Nature, 287, 425-429.
Stringfield, V. T. & Legrand, H. E., 1969, Relation of sea water to fresh water in
carbonate rocks in coastal areas with special reference to Florida: USA, and
Cephalonia, Greece. /. Hydrol, IX(4), 387-404.
Stringfield, V. T. & Legrand, A. E., 1971, Effects of karst features on circulation of
water in carbonate rocks in coastal areas. J. Hydrol. 14, 139-157.
Zotl, J. G., 1974, Karsthydrogeologie. Springer-Verlag.
Hydrogeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990}.
IAHS Publ. no. 207, 1993. 129

ANALYSIS OF SEA WATER INTRUSION ASSOCIATED WITH KARSTIC

CHANNELS BENEATH OVACIK PLAIN, SOUTHERN TURKEY

HATIM ELKHATIB & GÛLTEKÎN GÛNAY

International Research and Application Center for Karst Water Resources Research, Hacettepe University,
Beytepe 06532, Ankara, Turkey

ABSTRACT

A commonly encountered problem in coastal plains where a shortage of water exists is

sea-water intrusion. Fresh water contaminated with saline water cannot be used for water
supply and irrigation purposes. There are many methods described in the literature to
analyse and prevent sea-water intrusion into a granular aquifer. In karstic areas, where
groundwater flow generally is concentrated, it is difficult to estimate the morphology of
the saline water intrusion. Keeping in mind that isotropic aquifers are rare in nature, to
estimate the interface between fresh water and salt water requires multidisciplinary
methods. Moreover, in karstic areas sea-water intrusion is usually associated with
underground karstic channels open to the sea which is the case of the Ovacik plain
located on the Mediterranean coast of Turkey. In this study, by using electrical
conductivity data and water level measurements from 16 boreholes aligned perpendicular
to the shoreline, the structure of the salt-water/fresh-water interface was established. It
is concluded that the structure of salt-water intrusion in karstic areas depends mainly on
the structure of the karstic features, the elevation of the recharge area, tidal effects, and
the geology of the seashore.

INTRODUCTION
The investigation of groundwater in the saturated zone is one of the most important
aspects of hydrogeological studies on shallow water table aquifers in coastal karst
regions. The area and thickness of the fresh water and the salt-water/fresh-water interface
are important when a coastal karstic aquifer is utilized as a fresh-water source. The karst
dynamic features and the tendency of carbonate rock to develop local zones of high
permeability (underground karst channels) sometimes cause many problems.

One of the major problems in coastal karst areas is sea-water intrusion. The present paper
analyses sea-water intrusion associated with karstic channels beneath the eastern
boundaries of the Ovacik plain on the Mediterranean coast of Turkey. Water level
measurements from 16 boreholes aligned perpendicular to the shoreline were documented
and the electrical conductivity of the groundwater was plotted against depth. This study
gave a good idea about the position of fresh-water, the distribution of fresh water
thickness, and variation of the fresh-water/salt-water transition zone.

PROBLEMS AND ANALYTICAL TREATMENTS

Sea-water intrusion is a complex problem because of the large number of internal and
external factors which have a direct influence on coastal karst aquifers and the behaviour
of the brackish water zone. At the eastern boundaries of the Ovacik plain, the external
factors may include the natural loss of fresh groundwater discharged into the sea, tidal
effects, natural recharge from the coastal aquifer (Ada Mountain), and the artificial
discharge of fresh water from the Ovacik plain by pumping. The internal factors are
mainly related to the geometrical heterogeneity of the porous collectors and the
anisotropy of the coastal karst aquifer.
130 Hatim Elkhatib & GUltekin Giinay

According to the mechanism of sea-water intrusion into the mainland, the eastern part of
the Ovacik plain (Ovaciki§ikh formation of dolomitic limestone) is considered to be an
open aquifer towards the sea, which provides a hydraulic link between the groundwater
and sea water. In the Ovacik plain (eastern boundaries) the water-bearing formation is
underlain by an incomplete barrier (Haciishakh formation), thus sea-water intrusion may
occur above this impermeable unit. This could be the main reason for the enlargement
of the underground karstic fracture systems and the development of karstic channels
under the eastern boundaries of the Ovacik plain.

FRESH-WATER/SALT-WATER INTERFACE
The position of the fresh-water/salt-water interface is determined by using electrical
conductivity profiling along the boreholes aligned perpendicular to the Ovacik coastline.
The distribution of groundwater flow is determined from measurements of the water table
at selected boreholes (Fig. 1). The electrical conductivity (EC) of the sea was
55 000 /iS/cm, while that of the fresh water ranges between 350 and 1000 /xS/cm.

The EC of water samples within the transition zone ranged between 2000 and 30 000
fiS/cm. Along the first electrical conductivity profile (Fig. 2), a sharp interface does not
exist and the change in the EC values against depth is locally abrupt. The transition zone
in this profile is fairly thick, and this may be related to the underground high
permeability zones (karst channels) providing a hydraulic connection with the sea.

As seen from the profile, the coastal contamination at SK-3 is noted at depths between
15 and 22 m. Down to this depth the water is almost stagnant. At SK-4, the transition
zone is at depth between 10 and 20 m below sea level. The wells in the first profile
penetrate the Ovaciki§ikh formation which is mainly composed of dolomitic limestone
and wackestone throughout the saturated zone.
Along the second EC profile (Fig. 3), the transition zone may exist at great depth, and
the gradual change in the EC values ranges between 350 and 1000 ftS/cm. This is related
to the clayey and shaly siltstone layers of the Haciishakh impermeable unit which may
retard the coastal contamination to great depth. The depth and thickness of the
impermeable layers throughout the saturated zone, are controlled by the thrust faults
which affect all the lithological units in the area. Therefore, the fresh-water zone through
this profile is quite thick.

Fig. 1 - Groundwater contour map near the Ovacik coastline.

 Analysis of sea water intrusion associated with karstic channels 131

120

Fig. 2 - Cross section through the transition zone along the first selected
profile (numbers on the lines are electrical conductivities).

The distribution of groundwater flow from the mainland may continuously have a steep
gradient throughout the coastal aquifer, via subsurface conduits (karst channels) which
are hydraulically connected to the sea at the eastern boundaries of the Ovacik plain.

The coastal and submarine springs in the study area were effected by the thick transition
zone. A cross section through the Ovacik area shows the main depth and distance from
the shoreline of the different karstic features (caves, channels, and springs) which extend
under the eastern boundaries of the Ovacik plain.

direction
Thrust fault
Impervious

-10
presh ware . . ,L
-20
-3 0
*(C
-40 S> l

-50h * p < ^ '

/ JL j-
490 325 185 86
Fig. 3 - Electrical conductivity profile on the mainland.
132 Hatim Elkhatib & Gilltekin Gtlnay

CONCLUSION
An investigation of sea-water intrusion by using electrical conductivity profiles provided
valuable information about the position of the fresh-water/salt-water interface, the
distribution of the fresh-water zone, and the fluctuation of sea-water intrusion throughout
the karst channels under the eastern boundaries of the Ovacik plain. Thus, it is concluded
that in such coastal karstic aquifers, the structure of the fresh-water/salt-water interface
and sea-water intrusion phenomenon depend mainly on the structure of the karstic
features, the elevation of the fresh water, and the physio-chemical and dynamic
characteristics of this heterogeneous aquifer.
Hydrogeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHS Publ. no. 207, 1993. 133

THE HYDROGEOLOGY OF SOME COASTAL PALEODUNES IN AN

EQUATORIAL AREA AND THEIR KARST RELATED MORPHOLOGIES: THE
CASE OF GESIRA (SOMALIA)

P. FORTI
Department of Geological Sciences, Italian Institute of Speleology, Via Zamboni 67, 40127 Bologna, Italy

F. FRANCAVILLA
Department of Geography, Piazza Scaravilli 2, 40126 Bologna, Italy

ABSTRACT
Close to Gesira village, a few kilometres south of Mogadishu, there are some paleodunes.
The paleodunes are nearly parallel to the seashore and consist of cemented calcareous sand.
Their calcitic and aragonitic grains came from the breakdown of madreporic structures and
in several cases underwent strong cementation; the sea partially eroded the first lines of
paleodunes thus creating an abrasion platform and a cliff, in which several marine karst
microforms are still now actively forming. Several caves, up to 10-20 m in length, have
developed along the cliff, where major cementation took place. At low tide inside the
largest of these cavities, water constantly flows outward and in the first 5-10 m of the
platform, peculiar small canyon-shaped structures up to 1 m deep and with a smooth
rounded bottom have developed. Their genesis cannot be simply related to the marine
abrasion or the biological activity. The genesis of the largest caves, as well as that of the
canyon-shaped structures, is surely related to the presence inside the paleodunes of two
different waters, which mingle together: one, prevailing during high tides, is characterized
by sea water, and the second, prevalent at low tides, is a fresh meteoric or partially mixed
water. In the present paper the first hydrochemical data on both types of water are
discussed and on the basis of the morphological and petrographical analyses a model is
presented to explain the genesis of both the largest caves and the canyon-shaped structures.

INTRODUCTION
The area of coasted paleodunes south of Mogadishu (Fig. 1 A) has previously been studied
from the geomorphological point of view (Sauro, 1980). The relationships between its
fresh water and salt water have also been investigated (Dal Pra' et al., 1984; Dal Pra'
& Hussen, 1989), as well as its geo-electrical behaviour (Benvenuti et al., 1983).

Our study focused on the coastal caves developing along the cliff, partially created by
the mechanical erosion of the sea waves. They are complex hydrogeological systems,
whose evolution is controlled by mixing waters, different in origin and chemistry.
During May and June 1989 the conductibility of the seeping and dripping waters of the
caves was monitored under different meteorological conditions (sunny, cloudy, raining
periods). Thus the mixing conditions were monitored which are responsible for the
evolution not only of the majority of the caves but also the unusual karst microforms.
In the present paper hydrogeological sketch is firstly presented, then the hydrochemical
measurements are discussed and finally, on the basis of the petrographical hydrological
and geochemical data, a model for the genesis of the largest caves as well as for the
micro-forms (among which the canyon-shaped ones are the most unusual) is suggested.

GEOMORPHOLOGY AND HYDROGEOLOGY OF THE AREA

The studied area is a submerged coast and up to several kilometres inland it is
characterized by dune bars parallel to the shoreline (Fig. IB and C). Marine ingression
134 P. Forti & F. Francavilla

|Cemanted dunca| lagoon j Puna bar JAbraaion plaUormj Sea

kHermedate plan
Weathered . witti per uatry , ,
paleodunea j cemenied aaooa | Calcareous cemented dunaa s«a

Fig. 1 - A: studied area. B: model of zones of enhanced solubility inside

the Gesira coastal dunes (cross section) and for the development of
marginal caves and micro-canyons (plan). C: cross section of the studied
area. D: model of the behaviour of fresh, brackish and salt waters along
the coast, south of Mogadishu.

partially eroded the external bars, consisting of carbonate eolian deposits (eolianite):
their calcitic and aragonitic grains came from the breakdown of madreporic structures
and in several cases underwent strong cementation.

In the inner part, red to red-brown dunes, consisting of strongly weathered quartz grains,
gradually take the place of the white eolianitic dunes; few and small active dunes are still
present along the coast. The shoreline morphology is characterized by: (a) a cliff several
metres high; (b) a marine abrasion platform, several tens to hundreds metres wide (in the
low tide periods) and gently dipping toward the sea; (c) some madreporic structures a few
hundred metres from the seashore.

The presence of small depressions and ridges in the platform, as well as reefs and small
islands parallel to the shoreline, testifies to an older, quite completely eroded, eolic
morphology. All the intertidal zone undergoes marine abrasion while the eolianitic cliff
is shaped by both meteoric agents (rain, wind) and karst.

Part of the area shaped by these processes is higher than the high tide level and it is
reached only by breakers and spray; in these zones karst macroforms like cavities of
different sizes are common, the larger of which may reach 10-20 m in length; sometimes
tunnels, crossing the eolianite bar, may be formed. Microforms are widespread and
exceptionally developed, among the most common are pinnacles, kamenitzas, small
pockets and crests etc. (Sauro, 1980).
 Hydrogeology of some coastal paleodunes and their hirst related morphologies 135

The coastal dunes have groundwater which is recharged from the Sheebeli River valley
in the north: the waters firstly cross the red sands of the quartz paleodunes and then the
white grains of the eolianites (Fig. ID).

Close to the shoreline the salt water wedge goes back and forth inside the coastal aquifer
according to the tide, thus causing the potentiometric surface to oscillate. Close to Gesira
this surface rises up to 35 cm during the highest tides (1.6 m high), but the variation in
the level of the brackish-salt water interface is far larger, reaching values of 9 m.

The tides cause not only vertical variations but also horizontal displacements. In fact,
during the maximum tide, sea water may flood places over 100 m from the shoreline. As
a consequence we observed noticeable variations in the thickness of the fresh water lenses
(close to Gesira, it may change from 10 to 19 m from the high to the low tide). The
piezometric gradient is about 13/1000, while that of the interface is over 3 times larger.

Waters inside the coastal dune bars, as those of the islands, continuously undergo
complex mixing processes depending on the fluctuation of the piezometric surface and
tidal movement of the salt wedge.

The above-mentioned factors are the main causes for the variation of the temperature and
the salinity in the coastal aquifers, but other important factors to be considered are the
meteoric waters seeping inside the eolianitic dunes during monsoon rainstorms and the
salt water coming from the breakers or from the spray.
Finally, condensation water represents an important factor in the salinity balance of the
seeping water in the highly porous aquifers of the eolianitic dunes as for that flowing
inside the coastal caves.

THE MIXING MICRO-CANYONS

An usual karst form of the Gesira coastal paleodunes is represented by the cuts or
"micro-canyons" developing from the entrance of the largest caves and perpendicular to
them (Fig. 2). They are located in the abrasion platform toward the open sea and are
quite uncontrolled by fractures eventually existing in the eolianitic formation. The largest
micro-canyons are 10-15 m in length and have a maximum depth of 1-1.5 m (see Fig. 3)
with the central part always the deepest. They are completely covered by marine
organisms except in the deeper zone where the rock is polished.
Petrographical analyses provide evidence of a clear difference between the areas with
biological cover and the polished areas: the first areas are distinguished by a lot of
micrometric tubes as well as few larger ones, clearly related to past biotic corrosion
(Ciabatti et al., 1980), while in the second areas these tubes are completely absent.

The micro-canyons only develop along the main discharge directions for the cave waters,
as verified at low tide, when the abrasion platform partially emerges.

In fact the genetic mechanism by which these peculiar karst forms are developed, is
controlled by two different kinds of water mixing together in the abrasion platform: the
salt sea water and the brackish water which comes from four different sources, whose
contribution may change in time and comprises of:

(a) The seepage of fresh meteoric waters (see Fig. ID) in the eolianitic seashore dunes
which is prevalent during or immediately after rainy periods.
(b) The discharge coming from the coastal dunes and their backlands, which host a
groundwater fed by waters coming from very far off.
136 P. Forti & F. Francavilla

f-~
- ^fSf" "

^ . *r -J."--'
'*£&Z *.~, • ;« -''.•-•rJfV-'.'ï
- ^Z'^'^^W' ' • :"•*; 'f.i.L-
^ifei^"^* * ' *• -~i
'rf, 3&%! &~*Kf* \ - • , • . .
.;4*':-%?
- » * - • _

a£r,_"-•*'. • •" •" -•

""; / r .: **
' .'z'- ' - ' • i/^v" t>y.
%. '.^r^^^L^ ^
" ~*

. ** «
" . - • ' ' "
^/~
.* * " v
i
'> ï^^»^ &
'"^•^WI^VBB^'

Fig. 2 - A mixing micro-canyon starting at the entrance of one of the

largest sea-caves of Gesira.

POLISHED
SURFACES^

Fig. 3 - Longitudinal and cross sections of a mixing micro-canyon: the

shape of the central part is a V with flat bottom which a corrosional
rounded tube is developed.

(c) The sea water which feeds directly the lagoon just behind the Gesira shoreline and
seeps from the top of the coastal dunes due to breakers and sprays: this component
prevails during high tides and/or stormy periods.

(d) The water from condensation processes inside the caves as a consequence of the high
difference of temperature between the open air and that of the caves: this component
 Hydrogeology of some coastal paleodunes and their karst related morphologies 137

is important only in the diurnal low tides with sunshine.

The brackish water resulting from the mixing of these four sources has different salinity
in time, as shown by conductivity measurements (Forti & Francavilla, 1990), but it is
always undersaturated with respect to calcium carbonate and therefore aggressive (Forti
& Perna, 1986; Lohmann, 1988). The brackish water is responsible for the clear
corrosion morphologies which develop inside the caves more frequently than erosional
ones. This means that all the largest caves of the eolianitic dunes have to be considered
real karst cavities and not meteoric ones. Moreover, at the entrance of the caves, the
mixing of the brackish water with the salt water causes the genesis of the micro-canyons.

As said before, the canyon-shaped structures are always independent of the fractures in
the eolianite formations and cannot be produced by erosional processes as testified by
their shapes closed toward the sea and the cave, the algal cover over the parts more
exposed to the waves, and the lack of sand or rounded gravels inside. Therefore all the
morphological evidence is in favour of a chemical genesis.

Moreover, the existence in the deepest central part of rounded smoothed surfaces
confirms the hypothesis that corrosion is active only there. The probable mechanism for
the evolution of the micro-canyons is based on the slight difference of salinity between
the karst brackish water and the sea water: in fact, due to this slight difference their
relative density practically depends only on their temperatures. This allows the brackish
water to become lighter or heavier than the salt water depending on the boundary
conditions (low tide, starting of high tide) which control their relative temperatures.

At low tide (Fig. 4), the salt water trapped inside the micro-canyons is warmed and its
density becomes lower than that of the brackish water flowing from the caves. The karst
water therefore concentrates in the bottom of the micro-canyons and, due to convective
movements, it will mingle with the sea water. This mixed water, being aggressive, will
react almost only in the deeper parts of the micro-canyons. This mechanism, will be
active only during diurnal low tides. On the contrary, at the beginning of the rising tide
(Fig. 4) the heavier sea water will flow below the brackish water thus causing the mixing
and therefore the corrosion effects in the same places as before: in this case the
mechanism will be effective every time. The simple mixing corrosion activated by the
succession of low and high tides explains not only the rounded cross section of the micro-

OWE 1 SEA

COLD BRACKISH WATER

Fig. 4 - Mechanism by which mixing corrosion is activated in the central

part of the micro-canyons, the uncovered part always has a rounded shape,
resembling that of phreatic tubes; the shape of the other parts of the
micro-canyons is a V with a flat bottom.
138 P. Forti & F. Francavilla

canyons but also the lack of algal cover and the presence of the polished smooth surfaces
in the deeper parts where chemical corrosional processes are often active.

FINAL REMARKS
A preliminary analysis of the karst hydrogeological systems of the eolianitic coastal
dunes south of Mogadishu brought in evidence the existence of aquifers in which waters,
different in origin and in chemical content, mingle together: as a consequence the karst
and hyperkarst (Forti & Perna, 1986) corrosion is magnified with the evolution of
complex macro- and micro-forms. The most unusual of these, the mixing micro-canyons,
have been here analysed morphologically and genetically.

ACKNOWLEDGMENTS
The authors thanks their colleagues of the National University of Somalia, and in
particular Professor Omar Shirey, for their warm and friendly hospitality. The work was
partially supported by grants from M.P.I.

REFERENCES
Benvenuti, G., Abdulkadir, S. D., De Florentiis, N. & Ripolla, A., 1983, Risultati
preliminari di un'indagine geoelettrica nella zona costiera nei dintorni di Gesira
(Mogadiscio). Quad. Geol. Somalia, 7, 75-84.
Ciabatti, M., Ferrari, A. & Rabbi, E., 1980, Ruolo di microorganismi nell'attacco delle
roccie carbonatiche e nella precipitazione della calcite in ambiente lagunare (Lago
di Terlago, Trentino). Grotte dItalia, 4, 9, 29-48.
Dal Pra', A., De Florentiis, N., Mumin, M. G. & Hussen, S. M., 1984, Oscillazione
della superficie piezometrica della falda costiera provocata dalle escursioni di
marea lungo il litorale di Mogadiscio (Somalia). Mem. Soc. Geol., 36, 371-375.
Dal Pra', A. & Hussen, S. M., 1989, Ricerche sperimentali sui rapporti tra acque dolci
di falda e acque salate di intrusione marina lungo la costa della Somalia Centrale
nella zona di Jasira (Mogadiscio). Geologia Tecnica, 4 (3), 14-28
Forti, P. & Francavilla, F., 1990, Due nuove forme carsiche delle grotte costiere
somale: i solchi di miscela e le concrezioni tubiformi. Atti XVI Congr. Naz.
Speleologia, Udine, in press.
Forti, P. & Perna, G., 1986, L'ipercarsismo con particolare riguardo all' Iglesiente
(Sardegna Sud Occidentale). Nat. Alp., 36 (2-3), 85-100.
Lohmann, K. C , 1988, Geochemical patterns of meteoric diagenetic systems and their
application to studies of paleokarst. In: Paleokarst (ed. by N. P. James & P. W.
Choquette), 58-80. Springer-Verlag.
Sauro, U., 1980, Appunti sulla morfologia costiera della Somalia (Zona di Mogadiscio).
L'Universo, 60(4), 617-646.
Hydrogeologiad Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHSKibl.no. 207, 1993. 139

GEOHYDROLOGICAL BEHAVIOUR OF THE CERRITOS WELL SYSTEM,

NORTHEASTERN MEXICO

JAIME A. TINAJERO GONZALEZ & L. VELAZQUEZ AGUIRRE

National Water Plan Commission, SARH, Bovedas 78-Jardines del Sur, 16050 Mexico City, Mexico

ABSTRACT

The Cerritos well field in the state of Nuevo Leon in northeastern Mexico is in a
semiarid region with a mean annual rainfall of 630 mm. The lithostratigraphic
outcroppings in the area are the Cupido, La Pena, Aurora, Cuesta del Cura and Eagle
Ford formations, the lower parts of which are covered with alluvial sediment. Among
them, the best possibilities for forming aquifers are the Cupido and Aurora limestone
formations of Lower Cretaceous age where rainwater has filtered through fractures,
resulting in underground drainage leading to varying degrees of karst. In order to supply
water for the Cadereyta oil refinery, 12 exploration/exploitation wells were drilled
between 1978 and 1981. However, only six wells were productive. In order to determine
the effective capacity of the field, studies were made on which a number of
recommendations for supplying water to the refinery were based.

INTRODUCTION
The Cerritos well system is located in the state of Nuevo Leon, 45 km northeast of
Monterrey and 53 km north of Cadereyta (Fig. 1). At the beginning of this decade
producing wells in this system were the main source of water for the Petroleos Mexicanos
(PEMEX) refinery in Cadereyta. The system is situated in a semiarid region with a mean
annual rainfall of 630 mm of which more than 60% falls in the June-September period.
From a geomorphological standpoint, the region is hilly, with long narrow low ranges
(sierras), with asymmetrical sides, separated by broad rolling valleys.
The geological formations with aquifer capabilities in the region are the limestone
formations known as Cupido and Aurora, that date from the Lower Cretaceous age. The
first wells were drilled during the second half of 1978, but the water supply to the
refinery was not put into operation until the end of 1979. In June 1980, with completion
of the last productive well, a supply of 250 1/s was achieved; however, over the next 13
months the water level in the wells dropped an average of 20 m, and production
decreased to less than the rated capacity; moreover, two of the main wells started to
show suspended clay.

This situation caused the superintendent of the Cadereyta refinery to consult the National
Water Plan Commission (Comision del Plan Nacional Hidraulico), concerning a strategy
to be followed that would assure the necessary water supply for that facility. Here we
present the results of the study of the geohydrological behaviour of the aquifer in the
Cerritos Well System.

Hydrogeology
In order to define hydrogeological systems in areas where limestone-type sedimentary
rocks are the predominant lithological element, it is necessary to relate the structural and
stratigraphie characteristics of the region at issue to the geochemical properties and
isotopic content of the water held in the rocks. That is, a system will be defined by its
geological characteristics (Fig. 2) and it will be differentiated by its hydrogeochemical
140 Jaime A. Tinajero Gonzalez & L. Velazquez Aguirre

""~*=—-.. x Rivir or stream pg^jEXICQ-?

i_J W@ath@r station
•^5^sc~1

Fig. 1 - Location of the Cerritos well system.

SeolopieaS mtap towfarwaess» J

Symbols u s cad

. Anticlinal mxim , , CroEa-ssofcion lin®

—|— Synclinal mxlm > > ^ I n m - i a faul&

Racsharoo anaa of thss

Wt-OT m
IV
hidrogsao l o g i c a l
tsycstam

Fig. 2 - Hydrogeological system in folded limestones.

 Geohydrological behaviour of the Cerritos well system, northeastern Mexico 141

and isotopic relationships. It is well known that the hydraulic characteristics of limestone
answer to geological processes that occur after the sediment is deposited and that cause,
among other phenomena, fractures and dissolution fissures where the water can pass
through. In this manner the ionic concentration of the water in an aquifer in limestone
formations is directly proportional to the length of the path of flow.
Hydrogeological systems with long flow paths are associated with high ionic
concentrations, due to the fact that the water has sufficient resistance time in the
medium. If the preceding is correlated with the isotopic content, in this case tritium, it
may be established that the amount of this compound in the groundwater is inversely
proportional to the ionic concentration and, therefore, to the length of the path of flow.
We are now in a position to distinguish between the hydrogeological systems (Fig. 3).

RA. + K ( $ 5 S E )

100- SYSTEMS

_ 1 MINA
mm 2 BUENOS AIRES
50 — 3 TOPO cmco
— 4 PICACHOS
— 5 JARDIN
« • 6SABINAS
— 7 BUSTAMANTE
• - 8 HIGUERAS-MAMULIQUE (CERRITOS)

10

. TRITIUM (U.T.)
1 5 10 1000

Fig. 3 - Sodium (Na) plus potassium (K) content vs. tritium content.

Maxey & Mifflin (1966) applied this principle, classifying hydrogeological systems as
regional, local, and very local. In Fig. 4 we show the hydrogeochemical criterion for
differentiating hydrogeological systems in limestone rocks, as well as the classification
of the systems in the state of Nuevo Leon; as can be seen, the aquifer that the wells in
the Cerritos area exploit is considered a local hydrogeological system (Comision del Plan
Nacional Hidraulico, 1980). A system of this type is characterized by having short flow
paths, recharge areas near the exploitation area, low dissolved salt content, a water
temperature similar to that of the environment and significant amounts of tritium.
Planning long-term exploitation of this type of aquifer is quite risky, for even though its
production is attractive in principle, it is quite probable that in a short time it will fall
off abruptly because these systems have a small storage capacity in comparison with
regional aquifers. Moreover in northeast Mexico the productive Cretaceous limestone is
confined; hence the storage coefficient of the aquifers is small, somewhere between 10~3
and lO"5.

A hydrogeological setting: Classification of the Cerritos well system as a local limestone

hydrogeological system necessarily implies that its flow path is short; for this reason its
142 Jaime A. Tinajero Gonzalez & L. Velazquez Aguirre

REGIONAL
KsnuBirnoN OF SAHPLJS
BY S Y S T Ï 2 S
• MIMA
o HONTEBREY
• BUENOG AIRES
• TOPOCHICO
VEKY LOCAL
• COMAS
m LAMPAROS
^SABWAS
-CERRITOS
» PICACHOS ORIENTS
O LOW TRITIUM SAMPLES
O HIGH TRITIUM SAMPLES

a. +s»4 («pn>)

Fig. 4 - Hydrogeochemical classification of hydrogeological systems in

limestone rock.

recharge area is located in nearby areas specifically, in the outcroppings of the Cuesta
del Cura formation on the east side of the Sierra de Mamulique. Moreover, the flow of
hydrogeological folded limestone systems, like that of Cerritos, follows a trajectory that
is transversal with respect to the structure in the flanks of the anticlinal axis, according
to the depth of the beds, and it is longitudinal in the lower parts of the synclinal axis,
following the regional plunge of the structure.

The boundaries of the Cerritos system (Fig. 5) are geological limits (structural and
stratigraphie). Structurally, the system is limited on the surface by the conjunction of the
high part of the ranges with the position of the anticlinal axis, and underground by the
depth at the anticlinal axis corresponding to the structural crest of the impermeable unit
that makes up the foundation of the aquifer (La Pena formation), which constitutes a
stratigraphie limit.

Another limit to the transversal continuity of the structure is represented by an inferred

fault, whose direction is practically north-south, that cuts the synclinal axis obliquely and
separates the west side of the Sierra de Higueras from the proposed system. Nevertheless,
by virtue of the fact that the Sierra de Higueras does not correspond to the symmetrical
structure of the Sierra de Mamulique — because the former is structurally displaced to
the southeast — the contribution from the western flank of this range would be minimal
if the inferred fault did not exist. This is because at the conjunction of the adjacent
structural axes (anticlinal and synclinal) a high structure is formed that prevents the
passage of groundwater recharged on this side of the Sierra de Higueras.

In addition to all this is the fact that in the northern and northeastern portions of the area
being studied, on the western flank of the Sierra de Picachos, there are intrusive igneous
rocks with a large number of dikes associated with them. One of these dikes is located
in the northwestern part of the Sierra de Higueras, which possibly influences the cut of
the structural continuity of the system.
 Geohydroîogical behaviour of the Cerritos well system, northeastern Mexico 143

Açuifor limes terta

Fig. 5 - Geological boundaries of the Cerritos system.

Hydrogeologieal analysis: Six of the 12 wells sunk in this area were dry or unusable
(wells 2, 3, 6, 7, 8 and 10); one had poor quality water (well 3), four were situated
outside the Cerritos hydrogeologieal system — that is, they were not part of the same
structure — and one other (well 6) because it is located on the western side of the Sierra
de Higueras, there apparently being no hydraulic connection between it and the eastern
side of the Sierra de Mamulique where the producing wells of the Cerritos
hydrogeologieal system are situated (Fig. 6).

SI Chilarlto

@ Operating M i l l
•^ Owt-e?-e»i*vies» walla
O Cry c»alla
TTtl Poult
O Villas»

Fig. 6 - Location of the Cerritos wells.

144 Jaime A. Tinajero Gonzalez & L. Velazquez Aguirre

In Fig. 7 we have a geological section of that system, showing the location of the
producing wells, their depth, and location of pump chambers. A study of this section
reveals that wells 9 and 5 have shallow pump chambers; moreover they are located in the
Cuesta del Cura formation, generally speaking highly fractured and with drag folds, and
therefore very structurally unstable. For this reason both wells were reported as silted.
Well 1 has a pump chamber in the same formation, although much deeper; nevertheless,
by the end of 1981 small amounts of suspended clay were appearing, and the pump was
replaced by one of lesser capacity.

Fig. 7 - Transversal geological cross section of the Cerritos

hydrogeological system and location of wells.

It should be pointed out that this is the only producing well that totally penetrates the
aquifer formations; most of the others are classified as partially penetrating, except for
wells 11 and 12 that have deeper pump chambers and penetrate the aquifer to about 85 %.

Geohydrological aspects

Estimation of the recharge volume: Using the precipitation data from the Higueras and
Cienega de Flores weather stations, an analysis of the period from 1938 to 1981 was
made. The monthly precipitation for the period was computed for both stations, and two
types of statistical years were estimated: one rainy and one dry. In the first case the mean
monthly precipitation was calculated from all rainfalls greater than the historical monthly
mean, while the calculations for the dry year were based on the values that were less than
the historical mean. The estimated recharge analysis was made using Knisel's (1972)
equations and methodology, where the recharge is expressed as a function of the
precipitation; this function is of the type shown in Fig. 8.

The equation of the curve is:

y = abx/(a 2 + x 2 ) (1)

Substituting the recharge r for y and the precipitation p for x, the equation becomes:

r = abp/(a 2 + p 2 ) (2)

where a and b are the shape and scale parameters, respectively; in the case at hand, the
results were 76.44 for a and 1.604 for b, and represent the amount of rainfall at which
surface runoff starts. Taking the above into account, the equation for r would be:
 Geohydrological behaviour of the Cerritos well system, northeastern Mexico 145

1 2 2 . 8 1 p
SB43>P2

Prseipit&tien in m (p)

Fig. 8 - Recharge as function of precipitation.

r = 122.61 p/(5843 + p2) (3)

where r is a percentage of the precipitation p. In order to obtain the total recharge, both
sides of the equation must be multiplied by the precipitation p, giving:
R = 122.61 p2/(5843 + p2) (4)

Using this equation and taking as p the maximum, mean, and minimum monthly rainfall
correlated from the two weather stations, Table 1 was prepared, which shows the values
corresponding to the recharge estimated for a rainy year, an average year and a dry year.

The maximum, mean and minimum monthly recharge volumes (Table 2) were obtained
by multiplying the values of the recharge depth by the recharge area of the region, which
was taken as 2.15 km. Thus the annual volumes for the three types of year considered
were as follows: rainy year, with a maximum recharge of 1.7 X 106 m , average year,
with a recharge of 1.0 X 106 m3; and a dry year with a minimum recharge of 0.4 X 106

Table 1. Recharge depth in mm.

Type of year JAN FEB MAR APR MAY J UN JUL AUG SEP OCT NOV DEC

Rainy 36.7 30.9 25.4 70.9 75.3 95.5 81.6 109.4 110.8 78.3 36.4 30.1

Normal 11.17 8.83 9.4 48.1 48.1 64.0 43.6 80.1 95.0 48.2 13.6 10.2

Dry 2.186 1.903 1.676 15.425 15.425 24.910 16.174 27.494 62.460 18.769 2.901 1.942

Table 2. Recharge volume in m3 in the Cerritos system, estimated.

Type of year JAN FEB MAR APR MAY JUN JUL AUG SEP OCT NOV DEC TOTAL

Maximum 78895 66435 54610 152435 161895 205325 175440 235210 236220 168345 78260 64715 1,689788

Mean 24015 18984 19995 103415 56975 137600 93740 172215 204250 103630 29240 21930 985989

Minimum 4700 4091 3603 10025 33164 53556 34774 59112 134289 40353 6237 4175 381842
146 Jaime A. Tinajero Gonzalez & L. Velazquez Aguirre

Forecast hydrodynamic behaviour: The median drawdown curve for the system shows
various different slopes produced by the variations in extraction and the influence of the
recharge as a function of the precipitation (Fig. 9). Increases in the slope are related to
the increase in extraction and scant rainfall. Decreases are linked to the presence of
rainfall in the region, when extraction is kept constant. Nevertheless, the greatest
increases in drawdown correspond directly to the increase in extraction (for example, in
the months of July 1980 and August 1981).

loss I iaai
J*"™ ID E I F l u I A I U I J I J |A|tt|cje<{>s|z|F »|«|»|JW»M. IN E BOHTHE
^Ml^/i"Mi-VSv$fi6\ •••••* ••••••
•••••• ••••••

1 aee

e
i
! 10 !
as
30
48

"I 50
OS ï
70

1
!
-
11111111 IT
• »••••::•-"*•••••
rev-' r, •
••••••
••••••

MONTHS l l l l l l l l l l nKnEtEinnSinSSniiilEiliiilic KBMTHS

— isoi

Fig. 9 - Volume of extraction, mean drawdown and average precipitation.

With the purpose of determining the mean drawdown during the period from March 1980
to September 1981, the mean percentage of increase in drawdown was plotted against the
monthly percentage of increase in the volume of extraction. The points obtained fit an
exponential curve with a correlation coefficient of 0.83, which simulates the behaviour
of constant exploitation (Fig. 10).
The recharge was implicitly taken by grouping the variations in the levels for the months
with little drawdown in a longer interval (2 to 3 months) that showed a net drawdown.
From the respective graph it can be seen that the monthly increase of drawdown levels
is 11.52%, with sustained extraction.

200
c
I
•o 13©
1
L
ft.
1©© il.52 •
m .©3
8

5®

•f**^
y
4© @© BB %©m %
% o f velt&R@ oitferec&sd

Fig. 10 - Drawdown vs. extracted volume.

 Geohydrological behaviour of the Cerritos well system, northeastern Mexico 147

The previous analysis lacked a parameter that would explicitly consider the possible
recharge of the aquifer. This point remains as an alternative to predict the hydrodynamic
functioning of the well system, since a certain similarity was observed between the
forecasts obtained by using the first technique and those that were made by using the
method described below.

In order to arrive at a simulation that is closer to the behaviour of the aquifer system,
the monthly estimated recharge was calculated using the Knisel method described above.
Thus an equation of behaviour was suggested. Taking as a basis the cause and effect
relationship, the basic equation is:

AV = SA(y) AY (5)
In this equation AV is the variation in volume of the system, and is equal to the recharge
less the extraction (R — E); S is the storage coefficient; A(y) is the drained area (in
overview), which is a function of the depth (y), AY being the variation in depth of the
water levels, that is, the increase of the lowering. Solving for AY:

AY = AV/SA(y) (6)
AY = (R - E)/SA(y) (7)
Because the area factor SA(y) was not known, an attempt was made to determine it,
relating the increases in volume and lowering levels (as a quotient) with the variation in
depth.

In Fig. 11 it can be observed that the behaviour of the area factor is exponential and has
two values: one "optimistic" value of 0.05 x 106 m, and one "pessimistic" value of 0.02
X 106 m. Both values are towards the asymptotic portion of the curve between the last
two known values. Both values for the area factor were used in the above equation in
order to ascertain the increase of lowering as a function of the changes in the volume
DV, using a constant monthly extraction volume. The following step was to predict the
water levels for the year 1982. In order to do so, the monthly storage change to area
factor ratio was calculated. The changes in storage are the difference between the
estimated monthly recharge and the monthly extraction.
v
37B

242 .49 - 18 .ESS lnx

1456

A V 3 2
X 10 m
AY
•476 3 0 0
100 200 X BJtsr* cbova n lml
Fig. 11 - Criterion for determining the area factor.
148 Jaime A. Tinajero Gonzalez & L. Velazquez Aguirre

On forecasting the water levels for the "pessimistic" value (SA(y) = 0.02 x 10 m), the
result agreed with the behaviour of the system at the end. In order to make this forecast
two basic assumptions were used:
- constant extraction of 830 000 m3 per month,

— equal distribution of this volume among the three wells that were productive on the
date of the forecast (wells 1,11 and 12).

It should be pointed out that any change in one or both of these assumptions would
modify the forecast, for the system would be subjected to conditions of
exploitation/evolution of levels that were not considered in this study (Fig. 11). The
results for the producing wells up until 1982 were the following:

Well number 12: The last record for this well is dated November 1981; therefore the
drawdown for December of that year was taken as equal to the drawdown predicted for
that month. According to Fig. 12, concerning the drawdown forecast, the well should
cease to be productive at the end of January 1982 in the case of a dry year, or at the
beginning of February of the same year, in the case of a rainy year. This forecast was
accurate, for the well ceased to produce during the final days of January 1982.

s*Cyî«o.oa» to B

nantho j (tenths | nsntha |

lass isei loaa

— . — . ^ . Avoraso r a i n f a l l yaar ^ Dry yaar

Fig. 12 - Hydrograph and estimates for well 12.

Well number 11: In the case of this well, the available data regarding levels went up to
December 1981, and in that month work started on the prediction for 1982 (Fig. 13). The
forecast based on the "pessimistic" value indicated the well would cease operation in
April 1982, no matter what type of year was the case.

Well number 1: The last data available regarding this well are from November 1981, the
month in which the forecast was made. The first lowering obtained corresponded to
December 1981; since there was no better approximation, the same technique was used
as in the case of well number 12. Taking the "pessimistic" value as a basis, it was
forecast that the well would cease to produce in February 1982, because the lowered
level would reach the bottom of the pump chamber (Fig. 14).
 Geohydrological behaviour of the Cerritos well system, northeastern Mexico 149

Fig. 13 - Hydrograph and estimates for well 11.

- Racartlad d a t a
- Awaiaua r a l n T a l l ymmr

Fig. 14 - Hydrograph and estimates for well 1.

CONCLUSIONS
(a) The aquifer exploited by a group of wells in the Cerritos region is a limestone
hydrogeological system that was local in character, with a small storage capacity.

(b) Of the total number of wells sunk, wells 2, 3, 6, 7, 8 and 10 failed to produce water
because they were located outside the Cerritos hydrogeological system and were in
an area where no hydraulic continuity exists.
150 Jaime A. Tinajero Gonzalez & L. Velazquez Aguirre

(c) The lack of an adequate exploitation policy, the local character of the aquifer and its
low recharge capacity resulted in continuous drawdown of the water levels in the
wells.

(d) Of the five productive wells (1, 4, 5, 9 and 11), wells 4, 5 and 9 rather quickly
ceased to operate, because their pump chambers were shallow and they were located
in an unstable geological formation.

(e) The fact that the water supply to the refinery was kept up by increasing the flow
from the operating wells — with an average volume of 830 000 m3/month, caused the
wells whose pump chamber depth was exceeded to go out of service.
(f) On comparing the percentages of increase in the volume of extraction with those of
the mean drawdown, it was found that when the extraction flow was kept constant,
the monthly increase of mean drawdown was 11.52%.

(g) The results of the forecast for the producing wells up to 1982 was as follows:
Considering the "pessimistic" value of the area factor, well 12 would cease to
produce at the end of January 1982, if it were a dry year, or at the beginning of the
following month, if it were a rainy year. With respect to well 11, it was forecast that
with the same extraction (830 000 m3/month), it would close down operations in
April 1982, regardless of the prevailing rainfall conditions. Well 1, under the same
conditions, would cease to produce in February 1982.

(h) Unfortunately, the forecast for well 12 came true, and with the extraction system
thus affected, a new study had to be made.

RECOMMENDATIONS
(a) It is necessary to rehabilitate wells 4, 5 and 9, which have been reported as silted,
and this work should include deepening the wells, as they only cut through the
Cuesta del Cura formation, which is structurally unstable and has reduced aquifer
capacities.

(b) The recommended depth for these wells would be determined by contact between the
Aurora and the La Pena formations, for the latter is the foundation of the aquifer.
(c) In the future, a complete detailed geohydrological study should be carried out in
order to identify any aquifer exploitation area.

(d) In accordance with the results of the geohydrological study at issue, and considering
the behaviour of the Cerritos well system aquifer, the Cadereyta refinery must make
provision for alternate sources that will assure a sufficient supply of water for its
processes.

(e) It must be borne in mind that it is not recommendable to plan long-term exploitation
of folded limestone hydrogeological systems that are local in character, such as those
located to the north of Monterrey.

(f) Before locating a large production facility (such as a refinery) studies should be
carried out that consider both the volume of water required and the volume of water
available in the region.
 Geohydrological behaviour of the Cerritos well system, northeastern Mexico 151

REFERENCES
Comision del Plan Nacional Hidraulico, 1980, Determination del funcionamiento
geohidrologico de los acuiferos de la zona de Monterrey: N.L., Direction de
Inventarios de Agua y Suelo, Mexico.
Knisel, W. G., 1972, Response of karst aquifers to recharge. Hydrology Papers,
Colorado State University, Fort Collins, Colorado, no. 60.
Maxey, D. G. & Mifflin, D. N., 1966, Occurrence and movement of groundwater in
carbonated rocks of Nevada. Nat. Speleol. Soc. Bull., 28(3).
Hydrogeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHS Kbl. no. 207, 1993. 153

ALIMENTATION ARTIFICIELLE DES SYSTEMES AQUIFERES DANS LES

REGIONS KARSTIQUES

BORIVOJE F. MIJATOVIC
Université de Novi Sad, Geozavod, Belgrade, Yougoslavie

RESUME
L'utilisation la plus économiques des ressources en eau dans les régions karstiques est
un problème en liaison avec la possibilité d'une recharge artificielle des aquifères du
karst. Cependant cette opération nécessite une connaissance précise de l'hydrogéologie,
de l'hydrologie et de l'hydrodynamique des réservoirs dans le karst. Même si les
problèmes théoriques dans ce domaine peuvent être considérés comme résolus, il reste
à examiner de nombreux aspects pratiques concernant les sites de recharge artificielle et
en particulier l'aptitude à la création de réservoir pour certaines zones karstiques. Le but
à atteindre est d'aboutir à la capacité maximum de mise en réserve de l'eau qu'un
réservoir dans le karst pourrait avoir pour obtenir une augmentation du débit d'eau
souterraine par rapport à son état naturel.

ABSTRACT
An economic use of water resources in karst regions is a problem in conjuction with the
possibility for the artificial recharge of karst water-bearing aquifers. However, a precise
knowledge of the hydrogeology, hydrology and hydrodynamics of the karst water
resources is required. Even though the theoretical problems in this field could be
assumed solved, there are numerous practical aspects regarding artificial recharge sites
and particularly the capacity of certain karst reservoirs. The aim is to achieve the
maximum water storage capacity that a karst reservoir could have to obtain an additional
quantity of groundwater flux in relation to its natural state.

INTRODUCTION
L'objectif de l'alimentation artificielle dans les régions karstiques est d'utiliser la
capacité de stockage que peut offrir une roche calcaire en lui fournissant une quantité
d'eau supplémentaire par rapport à celle qu'elle reçoit dans les conditions hydrologiques
habituelles auxquelles elle est soumise, afin d'en retirer un bénéfice a plus ou moins long
terme. C'est donc sur le plan de la régularisation des ressources en eau que l'alimentation
artificielle parait à première vue séduisante, mais c'est l'état des connaissances sur la
géométrie du réservoir sur de la dynamique des eaux souterraines sur lesquelles se
fondent les conceptions théoriques, qui permet de répondre aux questions pratiques sur
les problèmes d'aménagement et de gestion dans le domaine des eaux karstiques.

D'une manière générale, on peut dire que l'alimentation artificielle des systèmes
aquifères karstiques peut avoir pour objet de reconstituer des nappes déprimées, soit par
l'injection dans le sous-sol de l'eau excédentaire des fleuves ou rivières, soit par certains
travaux d'aménagement des sources karstiques en vue de modifier artificiellement leur
régime.

Quelles que soient les modalités de l'alimentation artificielle dans le karst, que l'on
puisse concevoir, leur réalisation implique que l'on se soit assuré au préalable:

— de la capacité du réservoir aquifère a permettre le prélèvement ou le stockage

154 Borivoje F. Mijatovic

envisagé compte tenu de sa structure, de ses paramètres hydrogéologiques, du régime

des écoulements qui l'affectent et des caractéristiques physico-chimiques de ses eaux
(Fig. 1), et

— de la capacité du site de captage à assurer pleinement son rôle de liaison avec le

réservoir des systèmes aquifères pour que les échanges entre celui-ci et l'extérieur
puissent être assurés avec les débits, la durée et la qualité souhaités par l'exploitant.

C'est donc par deux approches, l'une globale, l'autre locale, effectuées conjointement,
que l'hydrogéologue devra chercher à déterminer la faisabilité et l'intérêt pratique de
chaque projet concernant la régularisation du régime d'écoulement d'une nappe karstique.

Fig. 1 - Deux différentes structures calcaires: la susceptibilité au stockage

souterrain se manifeste différemment.

CONDITIONS HYDROLOGIQUES ET HYDROGEOLOGIQUES DE

L'ALIMENTATION ARTIFICIELLE

Les caractéristiques dont il faut tenir compte dans le cas d'un réservoir souterrain à
alimenter sont un peu plus complexes que les conditions relatives aux eaux de surface a
injecter. C'est ce qui explique qu'au sujet des possibilités d'alimentation artificielle des
nappes, certains spécialistes aient parfois des idées un peu trop simplistes, ou
exagérément optimistes.

Notons d'abord que dans le zone climatique tempérée les formations karstiques
comportent en général peu de réservoirs non saturés disponibles; la répartition assez
régulière des pluies au cours de l'année, la faible valeur moyenne de l'évapotranspiration
ont pour résultat que la saturation des formations du sous-sol susceptibles d'emmagasiner
de l'eau s'effectue aisément, autrement dit que dans beaucoup de régions la profondeur
des surfaces des nappes karstiques est relativement faible (elle excède rarement plusieurs
dizaines de mètres).
 Alimentation artificielle des systèmes aquifères dans les régions karstiques 155

En zone aride par contre, les formations karstiques qui possèdent des réservoir non
saturés sont nombreux, souvent très épais, et constituent de ce fait de grand capacité de
stockage. On pourraient en citer beaucoup, dans les pays méditerranéens et notamment
à la bordure nord du Sahara.
D'autre réservoirs, en pays aride ou sub-aride, proviennent de la surexploitation des
nappes dans les régions d'irrigation intensive, comme au sud de l'Espagne et en Grèce;
cette surexploitation s'est traduite par la mise hors d'eau de formations calcaires étendues
et puissantes.
Malheureusement, on dispose rarement dans ces régions d'eaux de surface utilisables
pour l'alimentation artificielle. Quand elles existent, il s'agit bien souvent de crues
rarissimes, soudaines et très violentes, et par la même difficiles à utiliser; en outre elles
charrient régulièrement un débit solide important, d'où des risques de colmatage
généralement insurmontables.

Il faudrait aussi parler des nappes des secteurs fortement urbanisés ou industrialisés. Leur
exploitation intensive se traduit souvent par une dépression accentuée et étendue; De
telles nappes karstiques surexploitées ouvrent des perspectives assez larges pour
l'alimentation artificielle, car le nombre comme l'importance de ces surexploitations
s'accentuent sans cesse. Dans ces cas-là, l'alimentation artificielle permet une épuration
naturelle des eaux, c'est a-dire une amélioration significative de la qualité des eaux; cette
épuration est beaucoup plus efficace dans l'injection par bassins que dans l'injection par
puits ou forages.

De toute évidence, on ne peut injecter de l'eau dans le sous-sol karstique que si on

dispose de ressources, et donc d'abord des débits nécessaires. Reste le problème de
qualité acceptable de l'eau utilisée pour l'alimentation artificielle, car les circulation en
fissures et conduits karstiques diminue considérablement les possibilités d'auto-épuration
des aquifères karstiques. De ce fait on a conçu que l'eau pour l'injection doit être de
qualité acceptable; il est souhaitable qu'elle soit, sinon pure, du moins peu polluée,
surtout si on doit l'utiliser ultérieurement pour l'alimentation humaine. Mais il faut, aussi
et d'abord, qu'elle ne contienne que peu de matières en suspension lorsqu'on se propose
de l'injecter brute, sauf, dans une certaine mesure, en cas d'injection dans des karsts a
très forte perméabilité.
D'une manière générale, les eaux de fonte des neiges, celles des massifs calcaires et des
terrains cristallins, et en fait la plupart des eaux de surface en période d'étiage présentent
souvent les qualités requises. Mais, surtout si le bassin versant comporte des terrains
argileux, la moindre crue peut rendre l'eau inutilisable a l'état brut. Or, c'est
précisément l'eau des fortes ou moyennes crues qu'il serait intéressant d'utiliser, en
raison des volumes importants qu'elle représente et parce que cette eau est la plupart du
temps inutilisée.

La possibilité de traiter l'eau avant son injection soit totalement, soit en se bornant à un
traitement sommaire (floculation suivie d'une filtration rapide) risque d'abaisser
beaucoup l'intérêt de cette opération si les sujétions économiques ne sont pas bien
respectées. Tout au plus pourra-t-on, pour des raisons d'hygiène, et parce que son coût
est plus 'élevé, épurer l'eau destinée à la consommation humaine, mais non celle d'une
nappe karstique destinée à l'irrigation ou à l'industrie.

En ce qui concerne les conditions hydrogéologiques il faut bien souligner, que la nature
du karst se montre très favorable pour une opération telle que l'alimentation artificielle:
la transmissivité élevée et l'épaisseur importante, favorisent une percolation, voire un
vrai ruissellement souterrain des eaux au sein des formations aquifères. Si la couche
aquifère par contre est mince, un tel réservoir, même étendu, se révèle presque
156 Borivoje F. Mijatovic

inutilisable; en outre, si la perméabilité de l'aquifère karstique est faible (craie peu

fissurée par exemple), l'opération nécessitera un dispositif trop important pour s'avérer
rentable. Pour être efficace, l'alimentation artificielle dans le karst nécessite que
l'augmentation de la réserve de la nappe qu'elle détermine soit conservée pendant un laps
de temps suffisant, compatible avec le délai désiré de reprise de l'eau.

Il est essentiel de bien comprendre, à ce sujet, que le maintien de l'élévation de niveau

ou de pression de la nappe importe plus que la conservation de la masse d'eau injectée
elle-même: cette masse d'eau pourrait séjourner dans le terrain assez longtemps, et son
introduction aurait cependant provoqué une onde se propageant rapidement jusqu'aux
limites d'émergence de la nappe, déterminant un accroissement du débit sortant à ces
limites. En ce cas, l'effet de l'alimentation artificielle ne serait pas très important. Ceci
à cause de la vitesse moyenne des écoulements préférentiels, fonction de la transmissivité
et de la pente hydraulique. Ces vitesses sont beaucoup plus grandes dans les conduits que
dans les fissures et ce fait explique une vidange rapide de la réserve injectée. Si ceci
prouve que le rapport de la transmissivité sur l'emmagasinement (T/S, c'est-à-dire la
diffusivité) est grand, l'effet d'une recharge sur le débit des écoulements aux émergences
devient trop rapide (Fig. 2).

Fig. 2 - Coupe géologique présentant deux situations typiques d'un massif

calcaire. A gauche, il peut y avoir de réserves d'eau souterraines, et à
droite, par contre non, même si la roche est fissurée.

Pour qu'un réservoir karstique ne se vide pas rapidement de son remplissage, il faut des
conditions favorables qui dépendent non seulement des caractéristiques de la formation
aquifère, mais aussi de l'étendue de la nappe, qui permet de choisir des points d'injection
assez loin des limites lorsque l'on veut provoquer un effet durable, et enfin du caractère
de ces limites, de leurs conditions hydrodynamiques. Cette notion de "conditions aux
limites" est fondamentale dans tous les cas d'alimentation artificielle, autant que lorsqu'il
s'agit d'aménagement d'eau souterraine (Archambault et al., 1968).

En terrain calcaire karstifié, les conditions sont évidemment particulières. Dans les
massifs très accidentés, c'est la forte dénivelée jointe à la forte diffusivité qui provoque
l'évacuation, rapide ou même quasi immédiate, des eaux d'infiltration. Mais en plaine,
ou sous les grands plateaux, la pente hydraulique des nappes en réseau peut être très
faible ou presque nulle. C'est le cas des karsts sous-jacents, recouverts par des terrains
néogènes ou lorsque le karst est très développé, avec des chenaux largement ouverts; la
vidange étant très rapide prouve que dans les massifs calcaires de très grande dimension
un stockage d'eau d'une durée appréciable pourra être envisagé (Fig. 3).

Il semble donc que les massifs calcaires, en dépit de certaines considérations restrictives
et contre-indications évidentes, puissent offrir des perspectives intéressantes pour
l'alimentation artificielle. Mais les données du problème sont souvent fort complexes,
exigeant des recherches préalables approfondies, pour déterminer les conditions pour qu'
une opération d'alimentation artificielle puisse être envisagée dans de tels terrains ou
non.
 Alimentation artificielle des systèmes aquifères dans les régions karstiques 157

Fig. 3 - Une structure géologique favorable a priori à l'existence du

réservoir, notamment si le calcaire en profondeur est barré. Ce n'est pas
toujours le cas, et le forage peut être sec.

Nous nous bornerons ici a une présentation rapide des phénomènes mis en jeu et des
difficultés que l'on peut rencontrer lors de l'étude de différents projets. L'expérience
acquise jusqu'à présent montre qu'il et possible d'établir des prévisions de débits de
pompage pour un système karstique a l'échelle d'une exploitation "minime" des eaux
injectées, pouvant descendre à grande profondeur, en s'appuyant sur les conditions
hydro géologiques.

De même, une action sur l'écoulement des nappes karstiques, qui modifie artificiellement
leur régime, en modulant leur débit par un aménagement approprié, peut constituer un
moyen de gestion du réservoir karstique correspondant, analogue à son exploitation par
forages et pompage. Cet aménagement peut mettre en oeuvre des techniques diverses
(Margat, 1981):

— action sur le niveau des sources (abaissement ou relèvement temporaire et parfois

alterne), et

- action directe sur le débit d'écoulement de nappe (pompage périodique de débit

supérieur au débit naturel).

ETUDES ET ESSAIS
Les éléments d'appréciation de faisabilité résultent d'études dont les méthodes sont
largement similaires à celles requises pour tout projet d'exploitation d'eau souterraine.
Aussi ne s'attachera-t-on ici qu'aux aspects spécifiques à ces aménagements.

Un projet d'alimentation artificielle d'une nappe karstique captive à partir des

forages d'injection d'eau d'une rivière proche (région Lajkovac, près de Belgrade)

Dans une région karstique, près de Belgrade, sous l'effet de pompages intensifs, la nappe
est fortement déprimée (le rabattement moyen au centre de station de pompage est de 50
à 60 m) et sa remise en pression, en vue de restaurer les niveaux piézométriques
déprimés par l'exploitation s'avère nécessaire.

La Fig. 4 fournit des données sur l'étendue du contexte géologique (au moins celle de
secteur étudié), les pendages, l'épaisseur, etc. Les limites est sont imperméables et bien
définies. Il s'agit alors d' un réservoir karstique du Trias supérieur sous-jacent au dépôt
néogène dont la puissance peut atteindre 300 m environ. La nappe est captive ou libre
suivant les secteurs: tous les forages d'exploitation (au nombre de 5) se trouvent dans la
zone ou la nappe karstique est en charge (captive).
158 Borivoje F. Mijatovic

La perméabilité de ces roches carbonatées provient de l'association de la fissuration et

de karstification, pouvant donner lieu à de très grands débits par forages d'exploitation
(le débit moyen est de 40 à 50 1/s). Avec un débit de prélèvement total de 150 1/s, depuis
5 ans les niveaux piézométriques décroissent sans cesse, ce qui prouve que la nappe se
trouve dans un état de surexploitation, les rabattements progressifs dans certains ouvrages
ont dépasse jusqu'à présent 50 m de son niveau primitif. C'était donc une raison pour
actualiser un projet d'alimentation artificielle à partir de la rivière Kolubara (qui coule
sur le Néogène à travers la région), pouvant offrir une quantité d'eau supplémentaire par
rapport à celle que l'aquifère reçoit dans les conditions hydrologiques habituelles.

La transmissivité est comprise entre 1.10 et 5.10 m/s, le coefficient d'emmagasinement

de 0.01 à 0.03. La simulation des écoulements sur un modèle maillé (Fig. 5) nous a
permis en outre d'effectuer la synthèse des données hydrologiques sur les débits
d'alimentation naturelle, et d'étudier les liaisons du réservoir karstique avec les autres
niveaux aquifères. ..__

Fig. 4 - Carte hydrogéologique schématique de la région de Lajkovac.

 Alimentation artificielle des systèmes aquifères dans les régions karstiques 159

Pour résumer ces résultats il faut dire que le débit naturel disponible est de l'ordre de 80
à 100 1/s; l'exploitation intensive de l'eau avec un débit de 150 1/s, et qui prochainement
doit atteindre 200-250 1/s, exige alors une quantité d'eau supplémentaire de l'ordre de
5 000 000 m3 par an. L'expérimentation de type pompage-injection dans un forage
d'exploitation (BLa-8) a permis de vérifier que l'écoulement transitoire pur qui intervient
lors de l'injection d'un débit Q est régi par la même loi d'évolution que pour l'exhaure
de ce même débit. L'équation représentative des deux phénomènes est la même, aux
changements près suivants:

Pompage: Injection:
Débit - Q (constant) Débit - Q (constant)
Rabattement: s Relèvement: h = - s
Coefficient d'emmagasinement: S Coefficient de remplissage: R

"x* A l _ = i00m M-N=400 no«uds

AB=l900m
Fig. 5 - Modèle maillé du réservoir karstique sous-jacent.

Les recherches pour justifier ce projet sont en cours. Pour des raison évidentes
d'empêcher le colmatage du réservoir karstique (problème majeur de l'alimentation
artificielle), on a décidé de faire un prétraitement réduit dans des bassins de décantation.
On compte en effet sur l'épuration par filtration à travers la couche de sable disposée sur
le fond des ouvrages et le terrain. Le colmatage devrait être combattu par un entretien
constant des installations. Un tel prétraitement de l'eau permettra d'éliminer les causes
de colmatage ou du moins d'en atténuer les effets, ce qu'on estime suffisant, les
installations pour deferrisation et demanganisation étant en marche depuis deux ans.

Mais le colmatage dû à l'air entraîné reste un cas particulier qui nécessite des précautions
spéciales. Il s'agit des résultats obtenus sur les essais d'injection dans les dolomies et
calcaires du Cénomanien-Turonien d'Israël, contenant des cavernes karstiques (Sternau,
1967). Dès l'arrêt de l'injection, de l'air a jailli du puits: la pression du jet a augmenté
progressivement pour atteindre environ une atmosphère une minute après l'arrêt,ce qui
160 Borivoje F. Mijatovic

a d'ailleurs entraîné la ruine de l'ouvrage (par explosion).

Pour éviter ce phénomène l'arrivée en chute libre de l'eau injectée dans les forages, les
turbulences doivent être empêchées. Dans ce but l'eau doit être introduite dans l'ouvrage
par l'intermédiaire d'un tube plongeant sous le niveau dynamique. Des pertes de charges
suffisantes sont créées soit par le choix d'une dimension appropriée du tube, soit en le
munissant, à la base, d'une valve à contrepression; soit encore en fermant son extrémité,
la partie inférieure étant crépinée, etc.

La Fig. 6 représente schématiquement la disposition des forages d'injection avec le

captage d'eau de la rivière Kolubara, le long d'un front situé bien en amont de la zone
déprimée de la nappe karstique. Pour ce projet on a déjà élaboré et défini la coupe
hydrogéologique sur une distance de 1 km. On a prévu une prise d'eau de rivière
comportant des traitements préventifs d'une capacité de 1 m3/s et des forages d'injection.
Ces travaux sont actuellement en cours de réalisation.

Fig. 6 - Disposition schématique des forages d'injection avec le captage

d'eau de la rivière de Kolubara.

Un projet de la régularisation artificielle du régime d'écoulement de la source

Uganjska Vrela (prés de la ville de Cetinje)
Un projet d'aménagement et de la maîtrise des eaux souterraines karstique ne peut
valablement se concevoir en respectant seulement l'état naturel de leurs écoulements,
c'est-à-dire, des débits consécutifs d'une source ou d'un flux souterrain. En effet, il faut
tenir compte également des influences importantes sur les nappes karstiques, des
variations saisonnières du débit. Ce sont donc ces influences qui créent une sensibilité
aux limites du réservoir karstique.

C'est donc d'après cette conception que dans une région au sud-ouest des versants de
montagne une zone de stockage des eaux d'infiltration a été identifiée. En effet, c'est une
brahisynclinale constituée de dolomies et de calcaires du Trias supérieur, dont la position
par rapport aux roches imperméables sous-jacentes du Trias moyen, permet un
écoulement concentré dans la zone privilégiée des sources à Uvala Uganjska Vrela. En
raison du régime hydrologique irrégulièrement réparti dans le cycle annuel sur l'ensemble
du bassin versant, ainsi que du régime des écoulements souterrains avec un rapide
tarissement dans l'aquifère lui-même, toute tentative d'un captage à grand débit a été
abandonnée a cause des résultats médiocres.
 Alimentation artificielle des systèmes aquifères dans les régions karstiques 161

Fig. 7 - Coupe schématique du site d'aménagement du réservoir karstique

de la source Uganjska Vrela.

C'est pourquoi, l'idée de régulariser le régime naturel des écoulements artificiellement,

par l'aménagement d'un réservoir karstique, s'avérait la seule solution optimale pour le
captage, pendant la période d'étiage. L'essentiel est la construction d'un écran
imperméable-contrefort au versant du réservoir karstique, fondé dans les roches
imperméables sous-jacentes (Fig. 7).
Q(mVs)

19S3 19»

Fig. 8 - Source karstique Ain Bou Merzoug, Algérie. Pompages 1953-

1954.

Les résultats de recherche jusqu'à présent ont justifié la suite de ces travaux
d'aménagement, le volume utile de ce réservoir étant de l'ordre de 2 à 3 millions de m3
d'eau (Mijatovic et al., 1987). Un grand nombre de sources karstique de type à
déversement, notamment dans les régions du "karst barré "(formé par les dépôts
quaternaires et néogènes dans les poljés, appuyées contre des calcaires karstifiés),
pourront être surexploitées temporairement, permettant ainsi de tirer parti de la capacité
de stockage qu'offrent certaines catégories de réserves d'eau que l'on trouve dans la
littérature concernant la régularisation par pompage de la source Bou Merzoug, en
Algérie.
Les premières expériences ont été réalisées en 1953-1954, par pompages dans des forages
profonds de 40 m proches de la source, d'un débit total de 0.5 puis 0.8 m 3 /s pendant près
de 100 jours, le débit naturel étant alors d'environ 0.45 m 3 /s: soit un gain global de
162 Borivoje F. Mijatovic

2 200 000 m3 (0.25 m3/s en moyenne). L'abaissement de niveau qui en a résulté a duré
au maximum de trois mois, après l'arrêt des pompages (Fig. 8). Le rendement instantané
a varié de 20 à 43%. La capacité apparente du réservoir karstique s'évaluait ainsi à
700 000 m3 par m de hauteur (Margat, 1981).

REFERENCES
Archambault, J., Bize, J. & Margat, J., 1968, Alimentation artificielle des nappes
souterraines. Bull, de BRGM, no. 1, Paris.
Margat, J., 1981, Gestion des ressources en eau souterraine par aménagement de source.
II Symp. ACSAD, Rabat.
Mijatovic, B. F., Pavlovic, P. & Martinovic, P., 1987, Avant-projet hydrogéologique
d'un réservoir karstique dans Uvula Uganjska Vrela en vue d'aménager le régime
d'une source. IX Symp. Yougoslave sur l'Hydrogéologie, tome I, Pristina.
Sternau, R., 1967, Artificial recharge of water through wells. Experiences and
techniques. In: Symposium of Haifa: Artificial Recharge ofAquifers and Management
of Aquifers, 91-100. IAHS Publ. no. 72.
Hydrogeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHS Kibl. no. 207, 1993. 163

HYDROGEOLOGICAL CHARACTERISTICS AND GROUNDWATER

MISMANAGEMENT OF KAZERUN KARSTIC AQUIFER, ZAGROS, IRAN

PETAR MILANOVIC
Energoprojekt, Lenjinov Bulevar 12, 11070 Belgrade, Yugoslavia

BIJAN AGHILI
Mahab Ghodes, 17 Takharestan Alley, Shadid Dastgardy Avenue, Tehran, Iran

ABSTRACT
According to the structural and hydrogeological characteristics, the Kazerun area in
southern Iran (Zagros mountains) is divided into seven karstic and five alluvial
subaquifers. All of them are, more or less, interconnected making one complex
hydrogeological entity. But very strong recent tectonic activity has caused structural
block movement, i.e. the partial separation of the aquifers into different structural units.
It is concluded that water management should be under strict control. With ever
increasing demands for groundwater in the future, artificial recharge and regulation of
aquifer use is very important for regional groundwater management.

REGIONAL GEOLOGICAL AND HYDROLOGICAL FEATURES

The geology of the Zagros geotectonic unit in the Kazerun area of southern Iran is
characterized by long and regular anticline and syncline folds (Fig. 1). The anticlines are
normally mountain ridges and the synclines are valleys and plains (Fig. 2). The oldest
formations in the Kazerun area are Cretaceous shales and marls with argillaceous and marly
limestones. Eocene-Paleocene (Pabdeh formations) are represented by grey shales and thin
argillaceous limestones. From the hydrogeological point of view the most important
sediments are the Oligocene-Miocene Asmari thick massive limestones with medium-bedded
dolomite, and the Gachsaran (anhydrite alternating) formations. The Upper Miocene (shaly
limestones) and the Pliocene sandstone, red marl and gypsum beds are overlain by the Upper
Pliocene Bakhtyar formation (massive conglomerates and conglomeratic sandstones).

All karstic and alluvial subaquifers of the Kazerun area are hydrogeologically and
hydrologically connected to form one complex hydrogeological unit. The broad Kazerun
catchment cannot be defined accurately. The northern catchment with the highest
subaquifers is especially questionable, as well as the Arzhan Lake area at an altitude
1990 m a.s.l. The Arzhan Lake is a temporarily flooded karst polje. The main inflow into
the lake is through the Arzhan spring (Q max > 750 1/s; Qmin < 100 1/s). The lake
fluctuates 3-4 m. At the lowest part of the polje there exists a ponor (swallow-hole) in
the form of a circular depression with diameter 10-15 m and depth of about 10 m. The
direction of groundwater flow sinking into the ponor is assumed but not proved. Also,
the swallowing capacity of the ponor is not known. According to the general geology,
the possibility of a hydrogeological connection between the Arzhan Lake and the Kazerun
area cannot be excluded. Probably, only part of the sinking water flows from the Arzhan
Lake depression in the direction of the Kazerun subaquifers.

The high altitude natural karstic reservoir consists of the Renjan and the Sasan
subaquifers. Both subaquifers discharge at the surface of the Tange Chugan area and form
the Shapour River. These two subaquifers and two adjacent alluvial subaquifers represent
one hydrogeological gauged river flow at Bushigan. The influence of the inflow from the
Dashtak subaquifer on the flow at Bushigan is insufficient. The average annual capacity
of this hydrogeological unit is Q = 9.12 m 3 /s.
164 Petar Milanovic & Bijan Aghili

KAZERUN KARSTIC
CATCHMENT AREA

46° 48° 50° 52» 54° 56° 58° 60° 62°

APR SCALE .' I/25COO00O

SCHEMATIC OF GENERAL DIRECTION OF SROUND WATER

IN KAZERUN ALLUVIALLS AND KARTIC SUBAQUIFERS

DAD I N OUTLET
Q = 4 . 4 mVs
BOUSHIGAN OUTLET
Q = 9 . 12 m / » GENERAL DIRECTION OF GROUND WATER COURSE IN LIMESTONE AQUITER
PROBABLE GROUND WATER COURSE FROM THE AREAS WEHICH RECENTLY
ARE NOT INCLUDED tN KAZERUN CATCHMENT AREA
APP. SCALE.' 1/360.000 .. PROBABLE GROUNDWATER COURSE THROUGN KARSTIC AOUIFER BELOW
< '^> IMPERVIOUS FORMATIONS
- — — » GENERAL DIRECTION OF GROUND WATER COURSE IN ALLUVIAL AOUIFER

Fig. 1 - Geological map of the Shiraz-Kazerun area.

Another complex hydrogeological unit consists of the Dashtak, Sarvak, Parishan,

Sarbalesh and Dadin karstic subaquifers. Five alluvial subaquifers are developed in
syncline valleys (Shapur, Kazerun, Famour-Parishan and Chenar Shahijan sub-aquifers).
The retention capacity (capability) of karstic subaquifers is relatively good and resident
 Hydrogeology and groundwater mismanagement ofKazerun karstic aquifer, Iran 165

time is relatively long. This unit discharged through the Dadin spring zone an annual
average discharge Q = 4.44 m 3 /s. Approximately 0.7 m3/s of water is transferred (by
irrigation canals) from the Bushigan unit to the Dadin unit. All over the Kazerun area the
quantity of water used for irrigation and water supply purposes is Q = 7.23 m 3 /s
(without returned flow). According to these data, the total water capacity of the Kazerun
catchment area can be estimated at Q = 20-21 m 3 /s. Considering that 5 m 3 /s of the total
water quantity is used for different purposes, the available capacity of the Kazerun area
can be estimated as Q = 13-14 m 3 /s. It means that the available water potential is
controlled at two concentrated outlets: the Bushigan gauging station and the Dadin spring
zone.
ELIml

Fig. 2 - Simplified cross section through the Zagros mountains.

SASAN KARSTIC SUBAQUIFER

The Susan subaquifer is the most important underground reservoir (with good quality
water) in the whole Kazerun area. The aquifer discharges through two main outlets: the
Sasan spring and the Sarab Sheer spring. The average annual discharge from the Sasan
subaquifer is Q = 6.3 m 3 /s. The Sasan spring main outlet is at 813.56 m a.s.l. and the
Sarab Sheer spring is at 806 m a.s.l.
The hydrogeological characteristics of the Sasan aquifer have been formed during a long
evolutionary process, as a consequence of very dynamic new tectonic activity in the
Zagros geotectonic unit. Recent seismic activity indicates a continuation of this new
tectonic activity. It means the evolutionary processes of Sasan (and all the other) karstic
subaquifer is continuing. The position of the folded structures had a decisive role in the
formation of the pre-karstic drainage network, i.e. directions of the main surface
outflow. As a result of the new tectonic activity, the homogeneous anticline structure has
been fractured transversely and huge separate blocks have been formed. The individual
movement of these separate blocks has played a crucial role in the karst erosion process.

At the beginning of the karst aquifer development, the most important joint system was
that which has been formed by tensile stresses along the anticline axis. Along this system
the intensive karstification process occurred in the Asmari limestone. As a consequence
of that stage of karst evolutionary process the huge Shapour cave system has developed.
By lowering the erosion base level, i.e. by cutting into the Tange Chugan valley, the
karst aquifer discharge zone has adjusted to the level of the new base of erosion (Fig. 3).
A few stages of the Sasan aquifer evolution process, presented in Fig. 3, can be clearly
seen at an almost vertical outcrop on the northwest side of the Tange Chugan valley. The
166 Petar Milanovic & Bijan Aghili

cave openings are situated very near the geological boundary between the impervious
Pabdeh and the very karstified Asmari formation. The cave openings are observed at an
altitude between that of the Shapour cave and the recent aquifer outlet - the Sasan spring
zone.

Recently, the aquifer discharge level is in the process of adjusting to the altitude of Sarab
Sheer, as shown by the slow "dying out" of the Sasan spring. By lowering the
groundwater level within the aquifer, the channels at higher altitude begin to loose their
permanent function in the drainage system. Only when the water table is high (in rainy
periods) is water conveyed through some of these shallow channels. The highest channels
remained permanently above the water table (Shapour cave system and a few levels of
karstification below it).

Fig. 3 - Reconstruction of the evolution of the Sasan subaquifer.

In Tange Chugan Asmari limestone there exists the highly developed Sasan karst aquifer
with huge transmissibility. After a precipitation of more than 30 mm in 24 h in the
vicinity of the discharge zone, the water pressure in the aquifer changed abruptly and
rapidly. Because of a huge infiltration capacity, the lowest part of the aquifer reacts to
such precipitation after only 24 h or less. It means that the part of the karst conduit
system which discharged through the Sasan springs gets under pressure very quickly (Fig.
4). The discharge rapidly increased by 3 to 4 times.

The aquifer background water table in the analysed period is unknown owing to the fact
that this part of the aquifer is not controlled by the piezometric borehole. It can be
assumed that the maximum water table in the Tange Chugan can react to the altitude of
 Hydrogeology and groundwater mismanagement ofKazerun karstic aquifer, Iran 167

9TT30

26DOC19S4 01.JQM985 tOJan. 20 Jan. 30Jan. 9 Fob- 19F«b.

Fig. 4 - Sasan spring discharge.

860-870 m a.s.l. The above aquifer zone (aeration zone) is also very karstified and never
saturated.

RENJAN SUBAQUIFER
In comparison with the Sasan aquifer, the Ranjan subaquifer belongs to the upper part
of the catchment area. The aquifer is developed along the anticline flank and confined
by the anticline core (Pabdeh-Gurpi formations) to the southwest, and by the Gachsaran
formation in the northeast. The discharge zone is at an altitude of 920 m. The water table
fluctuates by as much as 60 m. The aquifer discharge varies from 200 to 1500 1/s. The
aquifer is mostly discharged through the Shapour River bed. The upstream part of the
river flow is intermittent and exists only in rainy periods.

DASHTAK SUBAQUIFER
The Dashtak karstic subaquifer is lower in comparison with the Sasan subaquifer. The
subaquifer on the southwest flank of the anticline structure between the Kazerun and the
Tange Chugan gorge in the Asmari limestone. This aquifer naturally discharged only
through the Sarab Douktaran spring zone at an altitude of 804 m. The average annual
discharge from this spring zone in 1985/1986 was Q = 300 1/s. The other controlled
outflow is groundwater abstraction for municipal water supply (Q = 135 1/s annual
average).

The water table slopes in the structures direction, from the southeast to the north-west.
The long-term groundwater pumping of municipal wells and a few pumping tests from
the wells along the structure indicate that the aquifer has a capacity of more than 0.5
m 3 /s. Karst phenomena were discovered by boreholes at altitudes of 753-693 m, i.e.
about 130 m below the water table. This indicates the depth of karstification is as much
as 250 m.

SARVAK SUBAQUIFER
The Sarvak subaquifer has developed in the Illam-Sarvak formation. The karstified Illam-
Sarvak limestone is overlain by impervious Gurpi and Pabdeh formations which act as
a barrier between the anticline karstified core and the Dashtak subaquifer in the Asmari
168 Petar Milanovic & Bijan Aghili

limestone. The water table is at a depth of 861 m, i.e. about 30 m higher than the water
table of the nearest Asmari aquifer.

The groundwater flow direction and the zone of aquifer depletion can be estimated on the
basis of very few data. The most probable direction is along the structure toward the
spring zone that fed the Parishan Lake. During high water tables, an old qanat system
collects the aquifer overflow and conveys it to the Asmari formation area.

PARISHAN SUBAQUIFER
The Parishan subaquifer is a complex hydrogeological unit which is characterized by the
strong interrelationships of surface and underground waters. The aquifer has developed in
the Asmari limestone. The bottom of the Parishan Lake is at the erosion base level of the
Parishan subaquifer. In spite of the fact that the surrounding Asmari limestone is fairly
karstified, the underground water circulation can be only in the direction from the aquifer
to the lake. The geological structure, made impossible the development of the karst aquifer
below the bottom level and the direct connection with the lower karst aquifers.
The area between Parishan and Arzhan Lake belongs to the Parishan catchment.

SARBALESH SUBAQUIFER

The Sarbalesh subaquifer has developed in part of the Asmari limestone anticline
structure. In the southwest the aquifer is bordered by an impervious anticline core
situated between the aquifer and the lowest erosion base levels. In the other direction,
the boundary with alluvial aquifers in the Shapour and the Kazerun plains is not
definitely determined.

According to an analysis of well logging and geoelectrical sounding data it can be

concluded that the Asmari limestone at the Sarbalesh anticline is fairly karstified. The
deepest karst feature observed by piezometric borehole logging is at a depth of 420-430
m, which represents extremely deep karstification. Generally, the base of the
karstification in the Sarbalesh Asmari limestone is at a depth of less than 250 m.

In recent years, monitored water table fluctuations indicate a slow, but permanent, trend
of decreasing water tables. It may be the consequence of a few dry years, or influence
of overpumping the alluvial aquifer. During four years, the water lowered by
approximately 4 m (Fig. 5).

On the basis of a pumping test, this aquifer is estimated to be a considerably productive

hydrogeological reservoir. The most productive single well has a capacity Q = 115 1/s
for S = 0.46 m. By simultaneous pumping tests in four wells the pumping rate was found
to be Q = 214 1/s.

1983 | 1984 | 1SB5 | 1986 | 1387

Fig. 5 - Water table fluctuations in the Sarbalesh karstic subaquifer.

 Hydrogeology and groundwater mismanagement ofKazerun karstic aquifer, Iran 169

The hydrogeological model of natural recharge and discharge for this aquifer is not clear.
Groundwater can flow only in the direction of the anticline axis. Due to the impervious
anticline core, hydraulic continuity with the southeastern flank of the Sarbalesh anticline
is not feasible. Available data do not indicate that water flows through the anticline core
zone but the possibility cannot be excluded.

The way the Sarbalesh subaquifer is recharged is a key hydrogeological problem. In the
total water balance of this subaquifer the immediate infiltration is less important than the
subaquifer replenishment as a consequence of its direct hydraulic connection with other
karstic and alluvial subaquifers. It is very important to determine the supposed
connection between the Dashtak and the Sarbalesh subaquifers through the synclinal part
of the structure below the Gachsaran formation and the hydraulic explanation of the
> 100 m difference between the piezometric levels of these two subaquifers (about 10 km
apart). The influence of the alluvial subaquifer on the Sarbalesh recharge rate is also a
very important supposition which needs to be proved. The lack of the necessary data
makes it impossible to prove this assumption as a final hydrogeological model of the
Sarbalesh subaquifer recharge.

The groundwater flow associated with the direction of the folding axis should be accepted
as dominant hydrodynamic features causing the subaquifer outflow in both structural
noses. The amount of filtration toward the Dadin subaquifer area is more developed than
in the opposite direction.

DADIN SUBAQUIFER
The Dadin subaquifer discharges through the long spring zone along the deeply eroded
part of the Sarbalesh anticline. The length of that zone is approximately 12 km. It is the
lowest discharge zone of the Kazerun aquifer area, 727-660 m above sea level. The
average annual discharge is Q = 3.35 m^/s. The fluctuation of discharge is small, i.e.
the Dadin spring zone yield is almost constant. The "regulated water" from a large part
of the regional Kazerun aquifer discharges through this zone. The subaquifer has
developed in the karstified Asmari limestone at the extreme southeast part of the
Sarbalesh anticline.

A very important area of the Dadin aquifer recharge is the upstream catchment area of
the Dalaki River. The alluvial aquifer of the Dalaki River is hydraulically connected with
the karstified Asmari limestone which transmits the groundwater flow to the Dadin spring
zone. The Dadin subaquifer shows a partial deterioration in water quality. This might be
due to its probable indirect hydrogeological connection with the Parishan karstic
subaquifer and the Famour-Parishan alluvial subaquifer.

PROBLEMS OF GROUNDWATER MISMANAGEMENT

A considerable amount of Kazerun waters run through more or less hydrogeologically
interconnected subaquifers, which partially retard the filtration and prolong the water
retention time.

The water table altitude of 861 m (observed in the Sarvak subaquifer) is the highest
aquifer water table in the hydrogeological unit that discharges in the Dadin spring zone
(altitude 657-601 m). The groundwater flows from upper to lower hydrogeological
structures, through karstic and alluvial aquifers. The complexity of the folded structures
and the very complicated hydrogeological structure, has resulted in extensive of
groundwater flows from the recharge to the discharge areas. This makes the final
discharge very equalized, but it also results in a partial deterioration of water quality.
170 Petar Milanovic & Bijan Aghili

Two ways of groundwater development are in operation:

— by partial exploitation of the karst subaquifer dynamic reserves which
discharge through the springs.

— by pumping from shallow wells (mostly from alluvial dug wells).

In the second case, there is a possibility of uncontrolled groundwater abstraction, i.e. for
partial pumping of static water reserves. Since 1978/1979 the number of wells in the
Kazerun area has increased tremendously. The quantity of water pumped was larger than
the annual recharge from precipitation. Consequently there was a continuous lowering
of the water table over the next seven years. Figure 6 presents the water table
fluctuations compared to the number of wells and precipitation data. The number of wells
is used for this analysis instead of the total quantity of pumped water, because this is not
known. A graphical approximation of the trend of decreasing water tables indicates a
permanent annual average water table drawdown of approximately 0.75 m.

The annual lowering of the water table in the Sarbalesh anticline is approximately 1 m.
This is more than in the alluvial aquifer, which is under the immediate influence of a
tremendous pumping. This can be explained by the impact of the mismanagement of the
Kazerun alluvial subaquifer.

Fig. 6 - Water table decrease as a consequence of the increasing number

of wells.

WATER MANAGEMENT STRATEGY

The Kazerun catchment area is relatively rich in water resources but the natural water
balance of the alluvial and karstic aquifers is affected by irrigation and water supply.
Also the influence of an adjacent catchment with a very karstified salt plug results in
water pollution. Two flows from that area transport to the downstream part of the
Shapour River 422 000 tons of dissolved material per year. In that part of the Kazerun
catchment the natural deterioration of groundwater and surface water is a serious
problem. The human influence of groundwater pollution by fertilizers is also very
intensive.
 Hydrogeology and groundwater mismanagement ofKazerun karstic aquifer, Iran 111

The water resources should be strictly managed. In the future, with ever increasing
demands for groundwater, the artificial recharge and regulation of aquifers, for the
purpose of maximum benefit, should be very important for regional groundwater
management strategy. The winter surplus of the Sasan subaquifer discharge should be
used to replenish and rehabilitate subaquifers that are overpumped during the dry and
vegetation periods.

A variety of methods and manmade structures could be applied to recharge the aquifers
(canals, qanats and wells). The direct connection between the spring and the recharging
point through underground closed systems may be the best solution. This solution should
simulate the direct underground overflow from one subaquifer (Sasan plus Sarab Sheer)
to another (Kazerun subaquifer). The induced infiltration could be combined with other
hydrogeological and geotechnical structures: exploratory galleries, exploratory wells,
grout curtains and cutoff walls. iv

Two main water management problems in the Kazerun area are: local disequilibrium
between groundwater extraction and natural aquifer recharge and water pollution in some
parts of the aquifer. The regional water management strategy can be summarized as an
optimal water development for the needs of the Kazerun area, and protection against
pollution of all remaining water transported and used in downstream areas, particularly
near the seashore where there is a permanent water shortage. In comparison with the high
developed karst aquifers, part of the Zagros tectonic unit has convenient hydrogeological
features for underground water extraction, as well as for artificial improvement and
regulation of aquifer discharge in some hydrogeological subunits.

The Sasan-Sarab Sheer spring zone can be controlled and tapped in a way which is
technically and economically most acceptable. The Dashtak and the Sarvak subaquifers
are important for the Kazerun municipal water supply. The Parishan and Famour Parishan
subaquifers should be kept for local needs. The Renjan subaquifer is also important for
local needs, and for partial water transfers during periods of local water surplus. The
Kazerun alluvial subaquifer can be jeopardized by groundwater mismanagement and
should be managed with special care. The existing exploitation of groundwater in the
Kazerun subaquifer has a deteriorative influence on the Sarbalesh karstic subaquifer.
According to the regional hydrogeology it can be assumed that the influence of the
mismanagement due to overpumping will be expanded to the Dadin and (possibly) to the
Dashtak subaquifers if the present water management regime continues. Artificial
recharge is necessary for the replenishment of the Kazerun and the Sarbalesh subaquifers.
The limitation of pumping for a few years is another way of replenishing the subaquifers.

In very humid years the rehabilitation treatment can be omitted. The Dadin subaquifer
can be estimated only as a good potential subaquifer, but, because of the very complex
hydrogeology of the area, new data and additional analyses are necessary. The Shapour
alluvial subaquifer is mostly saturated with water from the Sasan karstic subaquifer
through the surface river channels and through direct underground connections.
Hydrogeological Processes in Karst Terrtmet {Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHSPubl.no. 207, 1993. 173

KARSTIC HOT WATER AQUIFERS IN TURKEY

SAKÎR SÎMSEK
International Research and Application Center for Karst Water Resources, Hacettepe University, 06532 Beytepe,
Ankara, Turkey

ABSTRACT
The purpose of this paper is to determine the common properties of karstic hot water
aquifers based on hydrogeology and drilling results of selected geothermal fields of
Turkey. During the last 20 years, many geothermal fields were discovered in Turkey.
Different types of hot water aquifers (geothermal reservoirs) were identified in these
fields. The hydrogeological properties of the hot water aquifers such as lithology,
thickness and horizontal extend have an important role in geothermal energy production.
Most of the hot water aquifers in Turkey are karstic limestone (Mesozoic, Cenozoic) and
marble (Paleozoic). Karstic hot water aquifers are in the Sakarya-Kuzuluk, Yozgat-
Bogazhyan, Aydin-Salavath, Konya-Ilgin, Îzmir-Çe§me, Ankara-Meliksah, Denizli-
Pamukkale and Afyon geothermal fields. These fields are characterized by high
productivity, medium to low enthalpy fields, and geothermal fluid with a high C 0 2
content that creates a scaling problem.

INTRODUCTION
As a result of exploration studies, many hot water aquifers (geothermal reservoirs) have
been discovered by the General Directorate of Mineral Research and Exploration of
Turkey (MTA). Exploration studies include prospecting, hydrogeology, geochemistry,
geophysical surveys (gravity, resistivity, seismic), drilling, testing and prefeasibility
studies. Many technical experiments were carried out on geothermal energy based on the
results of these activities.

After the completion of the preliminary studies, drilling had been done in approximately
40 geothermal fields. The aquifers of 75% of these fields are composed of carbonate
rocks, mainly limestone. The important calcareous hot water reservoirs (Fig. 1) and
discovery dates are as follows: Denizli-Kizildere (1968), Afyon-Ômer, Gecek (1970),
izmir-Çesme (1974), Ankara-Meliksah (1975), Sivas-Sicakçermik (1976), Konya-Ilgm
(1977), Kir§ehir-Terme (1987), Ankara-Haymana (1987), Sakarya-Kuzuluk (1987),
Çankm-Çavundur (1988), Aydin-Salavath (1988), Kiitahya-Yoncah (1989) and Yozgat-
Bogazhyan (1989).

GENERAL PROPERTIES OF THE KARSTIC HOT WATER AQUIFERS

Soluble rocks such as carbonates (limestone, dolomite, dolomitic limestone) sulphates
(gypsum, anhydrite), saline rocks (NaCl and KC1) can be karstified. But these features
are more common in limestone. Widespread carbonate outcrops are in the Alpine-
Himalayan orogenic belt, and one third of Turkey is covered by carbonate rocks (Eroskay
& Giinay, 1980). Other countries too (China, Algeria, Czechoslovakia, Switzerland,
USSR, Belgium, Yugoslavia, Bulgaria, Greece, Italy, Germany, France, Hungary and
Israel) have karstic hot water aquifers (Boldizsar, 1976; Garagunis, 1976; Shaterev,
1976; Antonenko & Mavritsky, 1978; Haenel, 1985; Quilong et al., 1985; Arad, 1988;
Luoetal., 1988; Zhang, 1988; Fekraoui, 1990; Frankoetal., 1990; Konokov & Drokov,
1990; Rybach & Hauber, 1990; Vandelberghe, 1990).

General properties of karstic hot water aquifers are similar to those of cold water
174 §akir §im§ek
 Karstic hot water aquifers in Turkey 175

aquifers, but hot water aquifers have a higher dissolution capacity due to the high
temperature, pressure and chemical composition. There is an important relationship
between temperature and hydrothermal cave development (Ford, 1988). As an
interpretive result of the hydrogeological studies it can be that main common properties
of the karstic hot water aquifer in geothermal fields are as follows: high productivity (10-
300 1/s for each well) characterizes the medium (0-150°C) and low enthalpy fields (30-70

NE

Fig. 2 - Active fault zone and travertine deposits in the Sivas-Sicak

Çermik field due to the karstic limestone aquifer.

•:,,.j «t-A 3

Fig. 3 - Travertine terraces and deposits in the Pamukkale Geothermal

field.
176 §akir §im§ek

I
9 5
I d
i 1 D D
I j 1 o" 6*

ANIONS
S d" ! | d
\
w
A
! 8" ! V. o
8" 8"
A
R FI H H 7 H £ î; H H H H H H K
1 X B X X 0 X X X X X X X X 0
•8 u o u
* u:
M a «
|
t/3
Z
a M i4 M ¥ i !

z
1 1
Z

î ? S
1 1 11 ! 1 6
Z.

f. f Iè i S E

i
â 8 â & Z
aR1 â â 0 1 à Z z
s •c
s R s s s R S S R P % S
o O O o H

Si 5 5 R 8 8 R R 8 R 8 S R R 8 8
b ^ « g Si s 5 g R a S § ^

s g g
a S
I8 S S 5 Ê 5
S

S S
s s S S R "R^ 5 R R R S ? R
o o O o o O O
§ °
M o o «n t- o O
(
s " 8 S R
§ 1 1 5 à
R g « g
8' Kl

o" r-
u 1 3 H a 5 1 S S 1 i ! S 1 S
A
rf s s s R O
R 8 8 8 8 8 8 R R
o c o
o
8 8 8 S 8 8 R 8 S 8 R
•p-
if O
O
O O O o O

I z
s S. 8 8 8 S R R 8 8 8 S 2 o
<S O O O O
« "
.3
in
R 8 s 2 8 S 8
O
R
O
S R
O

1
I
•?

u
R
o

s
o
O

5
O
R

R
o

3
R
°
S
•o
8

S
8

S
8

S
S

2
R
O

S
O
R
O

S
o
R

R
=
H
o

S
R
o

R
O
O
o

s
o
I
»
s
o
8
R
S
S
8
a
8
3
8
S
2
a
8
5
?
R
8
5
8
Ci
8
rR
8
R
8
a
S
S
R
«
H
R I
1
R
O
B S 8 8
O
8 8 8 8 2 R
O
8 s
o
S
O
o
o
"*" =
R 8 R 8 8 R 2 8 R g s R R o

if O O
~"
O O o o O o

i S 8 R R 8 8 R S S 8 8 R R R R R R 1
s «3 S S S R S 5 5 R a 5 S S S
R S f-1
1
à § 3 S s S § S
§ S 5
O o
Ç â 1 S I I
X
t^ r- **} o 5 " S S s ~ O
5 3 s ? ; s

,
2
s s a R S 8
i-,
S a R o «, ? R S s o I
s •^ " 1
.6
& M S 1 gS 1 ^ S I
g S P 1 S S I i 1
S
1 S g S 1 1 I 11 ! g p 1 I S •S

s o
8. § 1 g § § g s s
u o
s s
r»
g
vD
s R 8
r
g S R s 2 R
«S
8 s 8
>e
8 ? c
«e
" * ^
s . i
I
o
"* 8 a S * S a $ S Si s ï »
s § a
n •» S s R R « s s s « £ 3 2 « 5 s « 1
u S * £ i X * z X *

E
1 5
3 s
if S
g ,1
.S-

II II îl
<
îi 1 Î
< 1"
O
c

1
I
_ _ ** « n *3" V,
» *" - O s
 Karstic hot water aquifers in Turkey 177

°C), geothermal fluids have a high C 0 2 content that creates a scaling problem and
travertine deposits around hot spring sites (Figs 2 and 3).

The analyses of karstic hot springs and water from wells of geothermal fields (Table 1)
show high Ca + + and HC0 3 2 " values are the main characteristics as follows:

Ca > Na or Mg > K and HC0 3 > S0 4 or Cl

Because of sea-water intrusion at the izmir-Çesme field and effect of evaporitic rocks at
the Yozgat-Bogazhyan field, these fields differ from the others.

Lithology and tectonic movements are the main controls in Turkey on karstification in
the crystalline limestones and marbles from Paleozoic to Cenozoic age. There is a large
stratigraphie gap between aquifers and impervious cap rock units. Heating systems have
developed as a result of neotectonic activity and geothermal fields have occurred along
the active tectonic zones especially on graben structures. One example is the well-known
tourist area of the Denizli-Pamukkale Geothermal field. According to the hydrogeological
model, the main aquifers (geothermal reservoirs) in this area are limestone and marbles
with high permeability characterized by a network of joints, fractures, faults and karstic
features. The alternation of Pliocene claystone, marl and sandstone and Paleozoic schists
acts as an impermeable cap rock in the region. The recharge rate in the area is quite
large related to the permeability of the rocks forming horsts and grabens. The recharge
is mainly from surface, subsurface and underground waters, which originated at rainfall,
infiltrating the basin (Fig. 4).

YE MCE HORST
CÛRÙKSU GRABEN

Mil PAMUKKALE
59'C
GEOTHERMAL FIELD
-A?5"C
B
SW 28'C

AlluviLm, travertine
QUATERNARY

CENOZOIC <^

Limestone MESOZ01C ?
Marble
. . r ,_. PALEOZOIC
E P S ) Sch.st
<gfc Thermal spring

ANALYSES OF PAMUKKALE AND MlS TUP. A. SPRINGS ( ir ppm ) nmm^ Geothermal water
= ^ > Water supply (Recharge)
Hot
s p r i n g PH K Na NH
4
Ca Mg Fe
K B SiO, HCQ S0
4
CI F S t C' gâ Horizontal ground water
<tolo1> losal ',
P.Kole
38 0.0 4 5 0 90 0.0 0 0 085 41 1060 630 15 13 0 0 9 35 ilow into plane of paqe.
TermdK 6.15 18
P.Kate
7.7 2 2 130 0.5 530 110 1.4 0.0 2.7 43 1415 830 34 25 0.06 59
3oz Pn 6.4 8.4 54 0.0 390 104 2.0 M. 23 63 915 728 20 2.0 0.12 28

Fig. 4 - Geological cross section of the Pamukkale Geothermal field.

178 Çakir §im$ek

IMPORTANT KARSTIC HOT WATER AQUIFERS AND THEIR CLASSIFICATION

ACCORDING TO AGE
The simple classification based on the age of the karstic aquifers is given in Table 2.

Table 2. The age of selected karstic aquifer formations in Turkey.

CENOZOIC:
Quaternary Karahayit*, Bursa
Neogene Denizli-Kizildere* - one aquifer, Konya-Ilgin* - partly
Tertiary Ankara-Melik§ah*
Eocene Yozgat-Bogazhyan*
MESOZOIC izmir-Ce§me*, Ankara-Haymana*, Sakarya-Kuzuluk*, Diyarbakir-Çermik*,
Çankin-Çavundur, Amasya-Hamamôzû, Eskçehir-Inônii

PALEOZOIC Sivas-Sicak Çermik*, Denizli-Kizildere*, Denizli-Pamukkale, Afyon-Ômer*,

Gecek*, Heybeli*, Manisa-Urganh, Kûtahya-Yoncah*, Kiiçehir-Terme*,
Kiitahya-Simav*, Aydin-Salavath*, Bursa

*determined by drilling.

Cenozoic limestone formations

The main Cenozoic aquifers are Quaternary travertines and Tertiary limestone formations

d= 3 0 0 . 5 0 m
Q
= HO It / s o c t = •J2-C
O.SYSTEM

SYSTEM

DEPTH
SERIE

LITHOLOGY EXPLANATION

2 Alluvium
1 QUATERNARY

I
I 1
Red colour c l o y ,
oi It el one limestone
alternations
R

1 |
80 •
E
F

o 1
I

1 1
>
U

1 1
Q

UJ 1
a.
E

1
z 0
A

o 1
< J' ** 1 1
N

Ul 1 Limestone portly
1 1 karstic
Kl 1
G E

o
S E C O N D A R Y

1 1
I
O o 1 1
1
0

z a. -1 150 _ |J|_
~—p"*™
E

UJ UJ i i
a. i
N

i i
w

I I
i
i i
200 - i
i i
i
i i
PALEOZOIC

r
A QUI FER

T
MAIN

IS^ÉnÉHUl Morblfi
{kantic}
j ~ h-
Fig. 5 - Lithological log of Konya-Ilgin well no. 1 (after Canik, 1981).
 Karstic hot water aquifers in Turkey 179

(Neogene, Eocene). The Quaternary travertine beds are karstic but generally they are
secondary aquifers. The Neogene limestone formations which have widespread outcrops
in Turkey are generally secondary hot water aquifers. Karstic limestones (Pliocene) form
the hot water aquifer in the Konya Ilgin geofhermal field (Fig. 5).

Karstic features were determined by hydrogeological survey in the region (Canik, 1981).
The flow rate of the well is 110 1/s at 42°C from compressor testing. In the Denizli-
Kizildere geothermal field, Sazak limestone (Pliocene) forms a reservoir which is partly
karstic. From the first well, steam and hot water mixture were obtained from a depth of
540 m at temperatures of 198°C. Also, the marble formation (Paleozoic) in this region
has karstic properties. Karstic structures are determined through the bottom level of the
Miocene limestone in the MH-l/A well drilled to a depth of 594 m in the Ankara-
Meliksah geothermal field (Fig. 6). According to the first results from the well a flow
rate of 300 1/s was obtained at 39°C output temperature.

The karstic Eocene limestone formations which are widespread in central and eastern
Anatolia, are reservoirs of hot water as well as of cold water in the geothermal areas
(Fig. 7). In the Yozgat-Bogazhyan geothermal field, two wells were drilled (BB-1 and
BB-2) with flow rates of 125 and 100 1/s respectively at temperatures of 32-46°C
(Ôzmutaf & Yiice, 1989).

Mesozoic crystalline limestone formations

Some of the low temperature geothermal aquifers consist of Mesozoic crystallized

limestone formations. One hot water aquifer is of Lower Triassic limestone at a depth
of 282 m (Ç-1) in the Izmir-Çesme geothermal field (Sahinci, 1988). In this well 42 1/s
of hot water was produced by pumping, with a temperature of 56°C (Fig. 8).

Q = 300lt/sn 1:39C
U.SYSTEM

SYSTEM

DEPTH
SERIE

(m)

IfTHOLOGY EXPLANATION

F SandMone
PLIOCENE

438«
•::*-.•.•:
Clay O
~_ o
too- a.
>-
u <
— —
or .J . . .
Sondslone
.. -. . : ..
1
E

o < 1 limestone
1
300-
. 1. . - . 1. .
Conglomerate
N

N
1 1 Limestone
J 1
E

UJ
o -*3S3
400- . . . . Conglomerate
C

U-
1
z
JL^^-aa!—
Karstic
0

500- ^ r ^ T " " ^ — Limestone

UJ UJ
?
l

<
M

in i-
594-

Fig. 6 - Lithological log of Ankara-Melik§ah well no. MH-1A.

180 Çakir §imsek

Q = 125 I f / «n
t:32°C

U-SYSTEM
z

SYSTEM

SERIE
Ei LITHOLOGY E X P L A N A T I O N

OUA. Alluvium
•o. •: <û. °-
zo
———

PLIOCENE
—. — . — C l a y , sandstone
40
—J — —_

R O C K
- 0 a conglomerate
— — „ — at t a r n a t i o n s
U
_— — _
80 - . .—
o
>- _
l —| _1 —| _1
_1 —| _1 —| 1_ c l o y
E O C E N E

NJ and m a r l

C A P
100'
o < - 1 —1 - — alternations

z \- IZO - i — i
UJ
OC 13? — — korsffc limestone
E
UJ
to
I
0 = 00 It/sn t=A6uC
U-SYSTEM

SYSTEM

0 EPT N
SERIE

(m)

U.TMOLOGY E X P L A N A T I O N

QUA. Alluvium
)B •'?•'•'%<?•

UJ
Cloy, sandstone
z 40- 6 V •* O
conglomerate
UJ
CJ
— —— alternations
R O C K

o 60
C

-J
_ —
Y

a.
1

IS
R
0

- l - l - -
A

_!_!._
Z

100 1 — 1 —
I

— i— | — clay and marl

T
0

1— 1 - 1
UJ
alt e m o t i o n s
R

C A P

IÎ0- — 1— 1— H-
S E N

z
E

t —
T

UJ 140
o
o 160 1 1 1
fcaretic limestone
»
UJ
j ^ p ^ . .
177 !
Fig. 7 - (a) Lithological log of Yozgat-Bogazliyan well no. BB.l; (b)
lithological log of Yozgat-Bogazliyan well no. BB.2 (after Ôzmutaf &
Yiice, 1989).

Another important hot water aquifer (Upper Cretaceous limestone) was discovered in the
Ankara-Haymana field (Fig. 9). The H-4 well was drilled to a depth of 221 m and
produced 52 1/s at 45°C temperature.

Paleozoic marble formations

In many of the geothermal fields, karst is developed in marble of Paleozoic age. In both
the Konya-Ilgin and Denizli-Kizildere geothermal fields, Neogene limestone and
 Karstic hot water aquifers in Turkey 181

Q t 42 I t / t « e t»36*C d« 212.40m

U.SYSTEM
X
w w

« c Et LfmOLO«Y CÏPLANAT10H
Q
>
u M
QUA AfluvttMI.
:• ;-:.•;
« a
TER NEO. Tuff
, . < j , .*.
•1—«r:~
i i
«0 • i i kl
i i k.
• i i i
3
i i O
J U R A S S I C
i i
i i
• • Lime»! on»
l i >
iao - i_J. Ci«y fnttrmadtott
«
i i
u i i O
Z
• i
i i o
• • o
o w
ISO M
N

o i
•J 1"k- Y
ca U * * I I I O A « . T94 ctay
i i

ROCK
otferiMtpono
«i
200 - l - l -
W
T R I A S

O £*
X
«. £» »
Brtchio
£> <»

CAP
a.
3 <£J o
O O
240
• i \
i l AQUIFER
Π1
kl _^-
MAIN

^rS
rr^
O
-J
2«î i i

Fig. 8 - Lithological log of Îzmir-Çe§me well no. 1.

Paleozoic marble form geothermal reservoirs (Canik, 1981; §im§ek, 1985). In Aydin-
Salavath geothermal field, a marble formation has flow rates of 300 t/h at each well
(AS-1 and AS-2). The wells are 1500 and 962 m deep and the temperatures are 162°C
and 171°C respectively. In the Kir§ehir-Terme field, a marble formation is well
developed as a karstic aquifer and hot water is produced from a depth of 333 m with a
flow rate of 45 t/h (Ôzgûr et al., 1987). In Sivas-Sicakçermik field, a total flow rate of
400 t/h hot water was produced from three shallow wells from depths of 200-350 m.
Other important fields are: Aydin-Germencik (232°C), Kutahya-Simav (162°C), Afyon-
Omer-Gecek (98°C) and Heybeli (56°C).

EXPLOITATION PROBLEMS IN KARSTIC HOT WATER AQUIFERS

Among the factors affecting the production of hot water and steam, the most important
problem is scaling. The most common deposition, CaC0 3 , causes a reduction in the
diameter of production and transmission pipe lines and drawdown in the reservoir. To
solve this problem, downhole heat exchangers and chemical inhibitors are applied to the
wells. At present, downhole heat exchangers are used in the Afyon-Ômer, Sakarya-
Kuzuluk and Kutahya-Simav geothermal fields. In the Denizli-Kizildere, Afyon-Ômer-
Gecek, Gazhgol and Manisa-Salihli fields, chemical inhibitors have been successfully
used. Thus, this important problem which is seen in karstic type geothermal reservoirs
during commercial usage can be eliminated.
182 §akir §im§ek

U.SYSTEM

SYSTEM

DEPTH
SERIE
LITHOLOGY EXPLANATION

Tiiïïintïiiiï.fr

QUATER_
cloy, troverllne
SENOZOIC

NARY
10
JtUUiUU uj-*U±
20
cloy
_ — —
/ 40 — •"—'•" cloy, sandstone

SC

-L-Ss- cloy
ao

- 1 1

K
— 1 _
100

C
O U S

clay ~ mart
-— —
IIS

O
— 1- - alternation

R
140
o 1 — 1 —
E

160 1 -
C

« _
_ .. __

CAP
o elay , marl
A

•
1
•
-
1
1
eandetone,
N 100 marl
T

O
E

O.
M too 1 -1
1 — • 1 clay - marl
R

3
U .— ... eatujefone
2 J17 ~^~^^~*~
C

J2I ÇSS3SZC& LIMESTONE

Fig. 9 - Lithological log of Ankara -Hay mana well no. 4.

J_
Another major problem that might be faced during drilling is the mud leakage into the
formation, where mud cement-control is impossible. So if there are two karstic zones,
only the shallower, low temperature aquifer can be used, instead of the high temperature'
high flow rate deeper aquifer.

RESULTS AND SUGGESTIONS

The important characteristics of karstic geothermal reservoirs in Turkey are as follows:

— they have a high production rate,

— they generally form low and medium enthalpy geothermal fields,

— in a high enthalpy field, they produce a high rate of C 0 2 for commercial usage such
as the Denizli-Kizildere field,

— scaling in the well-bore and transmission pipe lines may be a problem, but this can
be resolved by using downhole heat exchangers and chemical inhibitors.
Due to their high production rates, the karstic reservoirs could be used for heating,
balneological and tourist purposes.

ACKNOWLEDGEMENTS
I express my gratitude to the General Directorate of Mineral Research and Exploration
for permitting and supplying some of the data for this paper.
 Karstic hot water aquifers in Turkey 183

REFERENCES
Antonenko, G. K. & Mavritsky, B. F., 1978, The geochemical features of thermal and
mineral groundwaters of the Rionion artesian basin and perspectives of their
utilization. Proceeding of Hydrogeochemistry of Mineralized Waters Conference
(Cieplice Spa, Poland), 263-270, Warsaw, Poland.
Arad, A., 1988, B, F and Sr as tracers in carbonate aquifers and in karstic geothermal
systems in Israel. In: Karst Hydrogeology and Karst Environment Protection (Proc.
21st Congress of IAHS, Guilin, China, October 1988), vol. 2, 922-933. IAHS Publ.
no. 176.
Boldizsar, R., 1976, Production of energy from the Pannonian geothermal anomaly. In:
Proceedings of the Int. Congress on Thermal Waters, Geothermal Energy and
Volcanism of the Mediterranean, vol. 1, 115-129. Athens.
Canik, B., 1981, Konya-Ilgin Sicaksu Kaynaklanmn Hidrojeoloji încelemesi. S.U. Fen
Fakultesi Dergisi Sayi 1. seri A.s. 1-18, Konya.
Eroskay, S. O. & Gûnay, G., 1980, Tecto-genetic classification and hydrogeological
properties of the karst regions of Turkey. In: Proceedings of the Int. Seminar on
Karst Hydrogeology (Oymapinar), 1-41. DSI-UNDP Project Publ. DSL
Fekraoui, A., 1990, Outline of geothermal features in Algeria. Geothermal Resources
Council Transactions, vol. 14, Part 1, 3-8, Hawaii, USA.
Ford, D., 1988, Characteristics of dissolutional cave system in carbonate rocks, In:
Paleokarst, 25-36. Springer-Verlag, New York.
Franko, O., Bodis, D., Fendek, M. & Remsik, A., 1990, Outline of geothermal activity
in Czechoslovakia. Geothermal Resources Council Transactions, vol. 14, Part 1, 31-
40, Hawaii, USA.
Garagunis, C. N., 1976, Geothermal areas in the Magic Gulf region in Greece. In:
Proceedings of the Int. Congress on Thermal Waters, Geothermal Energy and
Volcanism of the Mediterranean Area, vol. 1, 241-274, Athens.
Haenel, R., 1985, Present status on utilizing geothermal energy in the Federal Republic
of Germany. In: Int. Symp. on Geothermal Energy (Hawaii, USA), 69-77.
Konokov, V. I. & Drokov, I. M., 1990, Utilization of geothermal energy USSR.
Geothermal Resources Council Transactions, vol. 14, Part 1, Hawaii, USA.
Luo Xiangkang, Li Hongjue, Jiang Bing, Lu Duanfu, Wu Wenxun, Lin Shigui & Wang
Yujie, 1988, The characteristics and resource assessment of the thermal groundwater
from the Lower Triassic series karst in Chongqing Xiaoquan Hotel area. In: Karst
Hydrogeology and Karst Environment Protection (Proc. 21st Congress of IAH,
Guilin, China, October 1988), vol. 2, 983-991. IAHS Publ. no. 176.
Ôzgûr, C , Uzel, Ô. F. & Tamgaç, Ô. F., 1987, Kir§ehir-Terme Kaphcasi Terme 2-3
Sicaksu Sondajlan Kuyu Bitirme Raporu. MTA Report no. 8222, Ankara.
Ôzmutaf, M. & Yûce, G., 1989, Yozgat-Bogazhyan-Bahariye (Cavlak) Kaphcasi Sicaksu
Sondajlan (BB-1, BB-2) Kuyu Bitirme ve Korunma Alanlari Etûdû Raporu. MTA
Report no. 8216, Ankara.
Quilong, Y., Kuide, X. & Zhang Zhenguo, 1985, Preliminary assessment of the
geothermal resources of China. In: Int. Symp. on Geothermal Energy (Hawaii, USA),
43-52.
Rybach, L. & Hauber, L., 1990, Swiss geothermal energy update 1985-1990. Geothermal
Resources Council Transactions, vol. 14, Part 1, 239-246, Hawaii, USA.
Shaterev, K. D., 1976, Les acrotothermis de la péninsule Balkanique. In: Proceedings
of the Int. Congress on Thermal Waters, Geothermal Energy and Volcanism of the
Mediterranean, vol. 2, 519-533, Athens.
Sahinci, A., 1988, Çesme yôresindeki kaplicalann hidrojeolojik - jeokimyasal ôzellikleri
ve soguk sularla ili§kileri. Ulusal I. Hidrojeoloji Simpozyomu. Bildiriler kitabi, s.
103-109, Ankara.
§im§ek, §., 1985, Present status and future development of the Denizli-Kizildere
geothermal field of Turkey. In: International Symposium on Geothermal Energy
(Hawaii, USA), 203-210.
184 §akir §imsek

Vandelberghe, N., 1990, Geothermal energy in Belgium: an overview. Geothermal

Resources Council Transactions, vol. 14, Part 1, 9-14, Hawaii, USA.
Zhang Zhenguo, 1988, An assessment of karst geothermal resources of the North China
basin. In: Karst Hydrogeology and Karst Environment Protection (Proc. 21st
Congress of IAH, Guilin, China, October 1988), vol. 2, 968-974. IAHS Publ. no.
176.
4 Hydrology
Hydrogeological Processes in Karst Terrtmes (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHS Publ. no. 207, 1993. 187

FLOOD FREQUENCY ANALYSIS IN KARST RIVER BASINS

ERTUÔRUL BENZEDEN, M. SIBEL OLCAY & K. BAÔALI

Department of Civil Engineering, Dokuz Eyliil University, Faculty of Engineering and Architecture, 35100
Bornova-fcmir, Turkey

ABSTRACT

The frequency distribution models for the annual peak flows of 21 streamgauging stations
in seven karst river basins of Turkey are investigated. Furthermore, annual peak flows
observed at seven representative stations are separated into their surface flow and baseflow
components by introducing a graphical procedure based on annual hydrographs of daily
flows. Then, frequency analyses with eight theoretical distribution models were carried out
on the surface flow and baseflow components in order to infer on how the baseflow affects
sample statistics and frequency distributions of the total peak flows. Those physically
meaningful moment and maximum likelihood estimates of the lower bounds are compared
with the average karst spring effluents estimated previously by various authors. The
general conclusions of the study are that almost all the available theoretical distribution
models, including the normal distribution, may be used to represent the observed
frequencies of annual peak flows in karst river basins, and further that the models with
lower bounds have a physical justification. Regression and correlation analyses performed
between surface flow and baseflow components revealed that these components are almost
uncorrelated. Apart from this fact it has been shown that variation coefficients of total peak
flows are considerably reduced with respect to variation coefficients of surface flow
components. Skew and variation coefficients of peak floods, and of surface flow
components as well, are significantly correlated while these two descriptors of the baseflow
components are almost independent. Relationships between these two descriptors appear
to have a parallel downwards shifted form of the theoretical relationship of the lognormal
distribution.

INTRODUCTION
Almost one third of Turkey's surface area is covered by karstified limestones (Fig. 1).
According to a recent study (Baran, 1987) karst spring contributions amount to one third
of the water potential in karst river basins and one fifth of the overall surface water
potential of Turkey. Taking into account this significant share (1100 m 3 /s), great efforts
have been spent on karst research studies in Turkey during the last decade (Ôzis &
Harmancioglu, 1980; Giinay, 1980); and governmental agencies as well as politicians
have emphasized that the karst water resources of Turkey should be developed during the
next few decades. Therefore, immediate solutions to the hydrogeological problems of
karst must be derived, so that the limited financial resources of the country can be used
efficiently.
The effects of karst on floods is not as easily perceived as its effects on mean and low
flows. However, it is quite logical to expect that, to some extent, the karst media, as
well as the karst baseflows, may affect the distribution of flood peaks. Another reason
for this study is that the current literature does not provide a comprehensive study related
to the frequency analysis of peak floods in karst river basins.

Furthermore, it is a controversial matter whether the information extracted through

regional frequency analysis of non-karst, or even of a specific karst region, may be
directly applicable at a given karst project site, unless the physical factors governing the
flood mechanism and the baseflow characteristics of that river basin are considered
thoroughly.
188 Ertugrul Benzeden et al.

Fig. 1 - Karst river basins and representative streamgauging stations in the

southern part of Turkey.

This study investigates the statistical and distributional characteristics of flood peaks (as
annual maxima) over the entire karst zone of Turkey. This paper is a revised and
concentrated form of two theses (Olcay, 1987; Bagah, 1988).

REPRESENTATIVE RIVER BASINS AND AVAILABLE DATA

Beginning with the Dalaman River in the west and including the Tigris River in the east,
the entire southern Anatolia is highly karstified. Some of the river basins in this zone are
the E§ençay, Aksu, Kôprùçay, Manavgat, Gôksu, Tarsus, Seyhan, Ceyhan, Euphrates,
and the Tigris (Fig. 1).
Twenty-one streamgauging stations in seven representative karst river basins were
selected and the peak flow data (as annual maxima) available were collected (EIE, 1956;
1956-1985).
Despite the fact that a considerable amount of snowmelt is measured at some stations
(such as, 2103, 1708, and 2015) and that the sample sizes at various stations (such as,
811, 1712, 1714, 2010, and 2123) are relatively short, stations of both types are included
in the study in order to achieve a more general discussion of the results. Certain
hydrologie properties of the selected representative stations are presented in Table 1.

BASEFLOW SEPARATION
In principle, the annual peak flow observed at a karst river site is composed of three
basic components: surface flow, the subsurface flow, and groundwater flow. Taking into
account the fact that the surface and baseflow components in karst river basins are
relatively greater than the subsurface component, this study considers the peak flows
simply as the sum of surface flow and baseflow components. The recorded annual
maxima (the total peak flow) at any station is simply separated into its components by
introducing a graphical procedure based on the analysis of annual hydrographs of daily
flows on semi-logarithmic paper.
 Flood frequency analysis in karst river basins 189

Table 1. Representative streamgauging stations and their various hydrologie properties

(EÎE, 1956, and 1956-1985).
In the Period of Obs.
Station Drainage Eleva- RprfyrH
no. STATION TITLE RIVER Area tion Length Mean Flow
Plow Modulus
(km2) (m) (years) (m3/s) (1/s/km2)

811 Suçati Dalaman 3856.4 594 21 18.87 4.89

901 Homa Manavgat 928.4 35 43 151.94 163.65
902 Beskonak Kôpriiçay 1942.4 117 43 87.10 44.84
1708 Muhat Tarsus 1416.0 50 27 42.81 30.23
1712 Bucakkisla Goksu 2689.2 397 21 31.88 11.85
1714 Karahacih Goksu 10043.6 24 22 127.08 12.65
1801 Himmetli Goksu 2596.8 665 47 31.19 12.01
1805 Gôkdere Goksu 4242.8 350 44 61.54 14.50
1806 Ergenusagi Zamanti 8698.4 347 37 67.70 7.78
2001 Kilavuzlu Ceyhan 8484.0 450 43 85.51 9.66
2005 Akçil Ceyhan 4202.0 1114 28 26.93 6.41
2006 Karaahmet Goksun 739.2 1324 29 9.52 12.90
2007 Çukurkôprû Sanbos 620.0 35 25 11.78 19.00
2009 Poskoflu Goksun 1387.2 1040 28 13.45 9.70
2010 Kurtleravsan Aksu 3498.8 414 22 33.75 9.64
2015 Tarar Hurman 915.2 1180 26 8.68 9.48
2103 Keban Euphrathes 63873.6 688 42 656.66 10.28
2123 Çmarkôy ÇagÇag 863.6 560 20 10.67 12.35
2603 Besiri Garzan 2450.4 545 36 50.55 20.63
2605 Diyarbakir Tigris 5655.2 570 35 73.93 13.07
2610 Baykan Bitlis 640.4 919 28 20.17 31.50

t (ctoysî

Fig. 2 - Graphical procedure of baseflow separation.

190 Ertugrul Benzeden et al.

As shown at Fig. 2, this approach assumes that both limbs of the annual baseflow
hydrograph are linear functions of time. This approach is much more realistic for the
recession limb and has been used in many studies (Keloglu, 1984; Ati§, 1988; Benzeden
& Tathoglu, 1986; Birsoy, 1989; Baran, 1989). The error introduced into the baseflow
estimates under linearity assumption is comparably as small as the error introduced by
any other assumption.

FREQUENCY ANALYSIS

A truly efficient design of any hydraulic structure can be achieved only as a result of
studies which relate cost to risk and frequency analyses. Parameters of any hypothetical
distribution function must be estimated from data samples. Therefore, all frequency
techniques are fully data dependent.

Within the scope of this study, five theoretical distribution models with two parameters
(normal, lognormal-2, Gumbel or type I extremal, log-Gumbel, and gamma-2), and,
taking into account the fact that the physically meaningful lower bounds may well be
justified in karst river basins, the models with lower bounds as a third parameter
(lognormal-3, gamma-3, and log-Pearson type 3) were applied. Parameters of the
postulated frequency distribution functions are estimated by the method of moments and
the method of maximum likelihood (Yevjevich, 1972; Kite, 1977).

In order to avoid the loss of information due to grouping suffered by the chi-square test,
the Kolmogorov-Smirnov test which is based on the absolute maximum difference
between the empirical and specified hypothetical frequencies (Yevjevich, 1972; Bayazit,
1981), is applied. For the sake of brevity, mathematical expressions related to the well
known density functions and the formulae available for parameter estimation are not
included in the paper. But, the various symbols used are defined as follows:

N sample size
x arithmetic mean of recorded events
Sx standard deviation of recorded events
C v = S x /x variation coefficient of recorded events
Cs skewness of recorded events
y logarithmic mean of recorded events
Sy standard deviation of log transformed data
C vy = S y / y variation coefficient of log transformed data
C sy skewness of log transformed data
D computed absolute maximum probability departure of the hypothetical and
empirical frequency distributions
Dc critical value of the Kolmogorov statistic at 5% level of significance

APPLICATIONS ON THE KARST RIVER BASINS IN TURKEY

Frequency analysis of annual (total) peak flows

As a first step of the study, the basic statistical descriptors of annual peak flows, such
as the mean (x), the standard deviation (S x ), the variation (C v ) and skew (C s ) coefficients
are computed. Similar sample statistics are computed for the log transformed samples.
All these sample statistics are given in Table 2. When the data samples recorded at
stations 1708, 1801, 1806, 2009, and 2015 are plotted against their empirical
nonexceedance probabilities, computed according to Weibull's plotting positions, P m =
m/(N + 1), where m is the order number and N is the sample size, it is seen that the
observed maximum event of each data set is too far from the expected trend. Assuming
 Flood frequency analysis in karst river basins 191

that these observations might have large measurement errors, and hence that they may
be discarded from the data set, computations are repeated with N — 1 number of
samples. The results relating to these computations have an asterisk symbol on the right
of the station number.

Using another computer program, both moment and maximum likelihood estimates of the
parameters for eight hypothetical distribution functions are computed. The direct, as well
as the indirect method of moments is applied for the log-Pearson type 3 distribution
(Bobée, 1975). Parameter estimates by the direct method of moments cannot be obtained,
except for those stations 1801, 1806, and 2009, since the characteristic ratio B (which
is a function of the first three sample moments about the origin) is smaller than 3 (Kite,
1972). Apparently, parameter estimation by the method of moments, neither for the
lognormal-3 nor for the gamma-3 distributions, is possible for the stations with
negatively skewed data samples. The same rule holds for the log-Pearson type 3 if the
log transformed data is negatively skewed.

Although the method of maximum likelihood almost always produces efficient estimates
for the parameters, the numerical solution of the set of nonlinear likelihood equations for
the parameters may or may not converge in iterations. As an example, for very small
values of sample skewness, a likelihood solution for gamma-3 may not be possible

(Matalas & Wallis, 1973). And further, a likelihood solution is not possible for the
gamma-3 distribution when the sample skewness exceeds 2.

Table 2. Sample statistics of the observed annual peak flows and of their logarithms.
Station N X Sx cv Cs y Sy c c
»y
no.

811 21 196.78 106.01 0.5387 0.8326 5.1141 0.6493 0.1270 -1.0195

901 43 699.98 189.65 0.2709 0.3869 6.5129 0.2868 0.0440 -0.6340
902 43 895.09 493.95 0.5518 1.4350 6.6602 0.5307 0.0797 0.0265
1708 27 278.66 281.57 1.0104 2.3394 5.2896 0.7885 0.1491 0.7489
1708* 26 242.38 213.29 0.8800 2.2436 5.2197 0.7136 0.1367 0.6346
1712 21 250.78 99.49 0.3967 0.9747 5.4503 0.4011 0.0736 -0.2719
1714 22 926.59 321.48 0.3470 -0.2597 6.7557 0.4372 0.0647 -1.6230
1801 47 207.76 126.78 0.6102 2.2200 5.1912 0.5387 0.1038 0.0387
1801* 46 196.88 103.67 0.5266 1.8007 5.1614 0.5039 0.0976 -0.2434
1805 44 665.02 361.26 0.5432 1.6758 6.3705 0.5200 0.0816 -0.2012
1806 37 293.35 166.23 0.5667 2.7811 5.5739 0.4424 0.0794 0.8048
1806* 36 274.56 122.39 0.4458 2.3677 5.5377 0.3892 0.0703 0.3433
2001 43 577.84 283.32 0.4903 1.4699 6.2525 0.4672 0.0747 0.0455
2005 28 112.75 48.92 0.4338 0.9239 4.6366 0.4319 0.0932 -0.0660
2006 29 60.60 30.02 0.4954 0.7935 3.9784 0.5284 0.1328 -0.3587
2007 25 66.37 26.72 0.4027 1.0913 4.1154 0.4223 0.1026 -0.6744
2009 28 103.13 89.39 0.8668 3.6021 4.4264 0.6094 0.1377 0.7993
2009* 27 88.58 46.28 0.5224 1.3123 4.3604 0.5091 0.1168 -0.0024
2010 22 295.18 159.60 0.5407 1.1453 5.5153 0.6816 0.1236 -1.6062
2015 26 54.44 57.29 1.0523 3.3291 3.6814 0.7587 0.2061 0.5795
2015* 25 45.02 31.86 0.7076 1.8793 3.6018 0.6544 0.1817 0.0188
2103 42 3508.80 1279.53 0.3647 0.2779 8.0919 0.3942 0.0487 -0.4677
2123 20 25.13 21.17 0.8422 1.9529 2.9865 0.6481 0.2170 1.2366
2603 36 525.38 207.52 0.3950 1.1674 6.1932 0.3805 0.0614 0.1208
2605 35 1273.69 710.15 0.5576 0.9847 6.9974 0.5724 0.0818 -0.1617
2610 28 215.63 101.05 0.4686 0.4803 5.2631 0.4848 0.0921 0.0017
192 Ertugrul Benzeden et al.

The computational results showed that the likelihood solution does not converge for
gamma-2 at one station, for lognormal-3 at thee stations, for gamma-3 at eight stations,
and for log-Pearson type 3 at 15 stations. The computed (D) and critical (Dc) values of
the Kolmogorov statistics are given in Tables 3 and 4.

The physically meaningful lower bounds, which are specially emphasized in this study,
together with the average karst baseflows estimated previously by various authors are
given in Table 5.
The abbreviations NOR, LN2, G2, GUM, LGUM, LN3, G3, and LP3 in tables are used
for normal, lognormal-2, gamma-2, Gumbel, log-Gumbel, lognormal-3, gamma-3, and
log-Pearson type 3, respectively.

Frequency analysis of surface and baseflow components

Annual peak flows observed at seven highly karst representative stations are separated
into their surface flow and baseflow components using a graphical procedure based on
annual hydrographs of daily flows (Bagah, 1988). Then, similar computations, carried
out earlier for the total peak flows, were performed on these components.

Table 3. Kolmogorov statistics for the hypothetical distribution functions with moment
estimators.

Station
no. N Dc NOR LN2 G2 GUM LGUM LN3 G3 LP3

811 21 0.285 0.072 0.135 0.095 0.073 0.165 0.081 0.096

901 43 0.204 0.108 0.161 0.144 0.167 0.212 0.133 0.138 ...
902 43 0.204 0.140 0.059 0.066 0.078 0.124 0.056 — 0.056
1708 27 0.253 0.258 0.094 0.162 0.216 0.068 0.146 ... 0.061
1708* 26 0.258 0.212 0.072 0.128 0.186 0.061 0.103 — 0.068
1712 21 0.285 0.122 0.076 0.072 0.092 0.144 0.070 ... ...
1714 22 0.279 0.120 0.186 0.166 0.178 0.244 — — ...
1801 47 0.195 0.152 0.087 0.107 0.118 0.138 0.091 ... 0.071
1801* 46 0.198 0.103 0.092 0.076 0.089 0.150 0.095 ... —
1805 44 0.202 0.175 0.089 0.103 0.116 0.139 0.090 ... ...
1806 37 0.220 0.205 0.117 0.145 0.167 0.111 0.101 ... 0.099
1806* 36 0.223 0.134 0.089 0.097 0.113 0.133 0.125 ... 0.112
2001 43 0.204 0.140 0.054 0.075 0.079 0.087 0.055 ... 0.056
2005 28 0.249 0.131 0.070 0.076 0.082 0.099 0.074 0.065 ...
2006 29 0.245 0.085 0.084 0.052 0.046 0.112 0.054 0.051 —
2007 25 0.263 0.192 0.182 0.173 0.191 0.257 0.178 ... ...
2009 28 0.249 0.215 0.139 0.164 0.204 0.123 0.136 — 0.168
2009* 27 0.253 0.122 0.076 0.076 0.082 0.129 0.081 ... ...
2010 22 0.279 0.106 0.081 0.087 0.108 0.207 0.077 ... ...
2015 26 0.258 0.242 0.141 0.219 0.214 0.131 0.166 — 0.107
2015* 25 0.263 0.181 0.081 0.122 0.135 0.129 0.099 — 0.099
2103 42 0.207 0.099 0.109 0.095 0.096 0.155 0.094 0.093 ...
2123 20 0.292 0.235 0.163 0.225 0.243 0.139 0.200 0.165 0.118
2603 36 0.223 0.108 0.092 0.071 0.056 0.106 0.088 0.102 0.095
2605 35 0.227 0.161 0.092 0.092 0.101 0.092 0.108 0.089 ...
2610 28 0.249 0.146 0.133 0.120 0.102 0.120 0.132 0.127 0.128
 Flood frequency analysis in karstriverbasins 193

Table 4. Kolmogorov statistics for the hypothetical distribution functions with maximum
likelihood estimation.

Station
no. N Dc NOR LN2 G2 GUM LGUM LN3 G3 LP3

811 21 0.285 0.070 0.125 0.093 0.087 0.147 0.086 0.085

901 43 0.204 0.108 0.160 0.144 0.159 0.175 ... 0.128 —
902 43 0.204 0.140 0.062 0.073 0.062 0.121 0.076 0.088 —
1708 27 0.253 0.258 0.092 0.149 0.165 0.071 0.058 ... 0.060
1708* 26 0.258 0.210 0.082 0.128 0.130 0.073 0.057 ... 0.067
1712 21 0.285 0.123 0.081 0.073 0.070 0.143 0.067 0.071 ...
1714 22 0.279 0.120 0.197 ... 0.172 0.204 _ ... ...
1801 47 0.195 0.151 0.076 0.079 0.079 0.120 0.072 0.073 ...
1801* 46 0.198 0.101 0.091 0.066 0.076 0.126 0.077 0.067 ...
1805 44 0.202 0.174 0.091 0.107 0.093 0.154 0.084 0.095 ...
1806 37 0.220 0.204 0.095 0.130 0.088 0.107 0.088 0.111 0.086
1806* 36 0.223 0.132 0.083 0.080 0.083 0.120 0.093 0.080 0.098
2001 43 0.204 0.140 0.054 0.082 0.071 0.094 0.056 0.054 ...
2005 28 0.249 0.130 0.066 0.076 0.066 0.104 0.067 0.169 ...
2006 29 0.245 0.088 0.056 0.054 0.059 0.120 0.057 0.064 ...
2007 25 0.263 0.190 0.194 0.169 0.176 0.254 0.160 0.161 ...
2009 28 0.249 0.214 0.075 0.124 0.115 0.114 0.107 ... 0.097
2009* 27 0.253 0.125 0.071 0.095 0.098 0.131 0.079 0.128 ...
2010 22 0.279 0.107 0.146 0.106 0.075 0.220 0.068 ... ...
2015 26 0.258 0.242 0.092 0.155 0.157 0.122 0.104 ... 0.096
2015* 25 0.263 0.180 0.084 0.100 0.094 0.135 0.086 0.156 ...
2103 42 0.207 0.101 0.108 0.088 0.098 0.131 ... 0.091 ...
2123 20 0.292 0.237 0.202 0.215 0.231 0.170 0.065 ... ...
2603 36 0.223 0.109 0.096 0.089 0.099 0.107 0.098 0.087 0.097
2605 35 0.227 0.162 0.081 0.099 0.099 0.088 0.081 0.071 ...
2610 28 0.249 0.150 0.124 0.125 0.129 0.130 0.133 — 0.150

Various sample statistics for the surface flow and baseflow components and of their
logarithms are presented in Table 6, while Kolmogorov statistics for the eight
hypothetical distribution are given in Tables 7 and 8.

Correlation analyses between surface flow and baseflow components for each of the
representative stations revealed that these two components are almost uncorrelated.
Considering this fact, the search for the physically meaningful lower bounds for both
components is found to be worthwhile. Those lower bounds which are positive and
smaller than the observed minimum flows are given in Tables 9 and 10.

DISCUSSION OF THE RESULTS

Models for the frequency distribution of total peak flows
When the Kolmogorov statistics given in Table 3 and 4 are considered, it can be seen that
observed frequencies of total peak flows may be defined satisfactorily by many
hypothetical distribution functions. Even the normal distribution model is not rejected at
a 5% level of significance, except for station 1708.
194 Ertugrul Benzeden et al.

Table 5. Physically meaningful lower bound estimates for the annual peak flows and
average karst spring contributions estimated by various authors.
Moment Estimators Maximum Likelihood Estimators Estimated Average Karst
Station Spring Contributions
no. LN3 G3 LP3 LN3 G3 LP3 (m 3 /s)

811 ... 1.72 ... — ... ... 5.0 (Ansoy, 1986)

901 ... ... ... — ... ... 113.0 ( Keloglu, 1984 )
902 ... ... ... 80.40 267.67 ... 64.0 (Alkan, 1986)
1708 ... ... 36.40 51.22 ... 46.89 11.9 (Atas, 1980)
1708* _ _. 30.63 48.58 ... 41.99 11.9 (Atas, 1980)
1712 ... _. ... ... 35.91 ... 5.1 (Ata?, 1980)
1714 ... ... ... — ... ... 35.8 (Ata ? , 1980)
1801 12.39 ... ... ... 27.24 ... 14.2 (Alkan, 1986)
1801* 7.00 ... ... ... 17.93 ... 14.2 (Alkan, 1986)
1805 ... ._ ... ... 89.99 ... 18.7 (Alkan, 1986)
1806 78.32 ... 13.87 59.67 93.05 41.70 44.6 (Alkan, 1986)
1806* 95.38 ... 44.08 31.62 80.63 7.10 44.6 (Alkan, 1986)
2001 ... ... ... 30.28 148.87 ... 20.0 (Bagah, 1986)
2005 ... 26.97 ... 0.63 45.62 ... 13.0 ( B a | a h , 1986)
2006 ... — ... ... 13.35 ... 3.0 (Bagah, 1986)
2007 ... ... ... — ... — 2.0 (Baran, 1987)
2009 7.16 ... ... 21.49 ... 14.52 3.3 (Baran, 1987)
2009* ... ... ... 6.71 29.86 ... 3.3 (Baran, 1987)
2010 ... ... ... — ... ... 6.8 (Baran, 1987)
2015 ... ... 4.90 5.00 ... 1.26 4.2 (Baran, 1987)
2015* ... ... ... 0.74 9.59 ... 4.2 (Baran, 1987)
2103 ... ... ... — ... ... 24.5 (Baran, 1987)
2123 ... 8.70 8.95 10.40 — ... 8.5 (Çetiner, 1986)
2603 ... ... 2.40 97.62 206.12 11.39 12.0 (Ôzdaglar,1980 )
2605 ... 61.54 ... ... 281.48 ... 22.0 (Ôzdaglar,1980 )
2610 — ... — 75.37 ... 84.57 10.0 (Baran, 1987)

Table 6. Sample statistics of the surface flow and baseflow components and of their
logarithms at seven representative stations.
Station
no.
N X s, cv Cs y s
y Cv, Cy

SURFACE FLOW COMPONENTS

901 43 619.16 188.47 0.3044 0.4560 6.3798 0.3253 0.0510 -0.6553

902 43 844.67 491.65 0.5821 1.4592 6.5843 0.5693 0.0865 -0.0758
1801 47 180.09 125.47 0.6967 2.2379 4.9917 0.6664 0.1335 -0.6306
1806 37 233.95 167.87 0.7176 2.8443 5.2821 0.5734 0.1086 0.3517
2001 43 497.84 282.30 0.5671 1.4849 6.0636 0.5550 0.0915 -0.1181
2015 26 44.15 54.12 1.2260 3.2450 3.3399 0.9236 0.2765 0.3801
2605 35 1232.60 712.95 0.5784 1.0146 6.9536 0.5925 0.0852 -0.1398

BASE F I,owco MPONEN rs

901 43 80.81 29.73 0.3678 1.3658 4.3329 0.3432 0.0792 0.2938

902 43 50.41 14.21 0.2818 0.7971 3.8833 0.2734 0.0704 0.2127
1801 47 27.67 10.19 0.3684 0.5228 3.2526 0.3772 0.1160 -0.1282
1806 37 59.41 21.14 0.3559 0.6165 4.0236 0.3536 0.0879 0.1018
2001 43 80.00 40.43 0.5053 0.7471 4.2487 0.5412 0.1274 -0.3427
2015 26 10.30 5.58 0.5414 1.1255 2.1919 0.5499 0.2509 -0.1202
2605 35 41.09 22.62 0.5504 1.4878 3.5866 0.5112 0.1425 0.1446
 Floodfrequencyanalysis in hirst river basins 195

Table 7. Kolmogorov statistics of the distribution functions hypothesized for the surface
flow components.

Station
no. N Dc NOR LN2 G2 GUM LGUM LN3 G3 LP3

a) Through use of moment estimators as function parameters

901 43 0.204 0.079 0.138 0.120 0.141 0.199 0.109 0.114 ...
902 43 0.204 0.130 0.072 0.059 0.070 0.142 0.058 ... ...
1801 47 0.195 0.137 0.065 0.093 0.114 0.117 0.068 ... ...
1806 37 0.220 0.226 0.123 0.158 0.194 0.126 0.115 ... 0.119
2001 43 0.204 0.140 0.067 0.065 0.078 0.096 0.056 ._ ...
2015 26 0.258 0.237 0.074 0.181 0.232 0.082 0.141 ... 0.051
2605 35 0.227 0.154 0.095 0.082 0.094 0.089 0.099 0.079 —

b) Through use; of likelihood estimators as function parameters

901 43 0.204 0.079 0.141 0.121 0.137 .__ 0.102 —

902 43 0.204 0.131 0.080 0.059 0.056 0.136 0.083 0.098 ...
1801 47 0.195 0.136 0.075 0.074 0.084 0.168 0.063 0.074 ...
1806 37 0.220 0.224 0.111 0.141 0.106 ... 0.110 0.135 0.106
2001 43 0.204 0.141 0.062 0.074 0.067 0.097 0.062 0.050 .._
2015 26 0.258 0.237 0.074 0.146 0.170 0.080 0.061 ... 0.053
2605 35 0.227 0.156 0.084 0.089 0.091 0.086 0.084 0.073 ...

Table 8. Kolmogorov statistics of the distribution functions hypothesized for the baseflow
components.

Station
no. N Dc NOR LN2 G2 GUM LGUM LN3 G3 LP3

a) Through use of moment estimators as function parameters

901 43 0.204 0.116 0.046 0.068 0.070 0.054 0.042 0.046

902 43 0.204 0.108 0.054 0.069 0.046 0.082 0.058 0.047 0.049
1801 47 0.195 0.095 0.102 0.085 0.082 0.151 0.080 0.081 ...
1806 37 0.220 0.141 0.134 0.129 0.111 0.101 0.132 0.128 0.126
2001 43 0.204 0.081 0.101 0.073 0.064 0.146 0.055 0.063 —
2015 26 0.258 0.120 0.097 0.064 0.061 0.109 0.072 0.074 ...
2605 35 0.227 0.189 0.087 0.117 0.133 0.111 0.102 ... 0.080

b) Through iise of likelihood estimators a s functio i parameters

901 43 0.204 0.116 0.055 0.069 0.059 0.053 0.045 0.040 0.048
902 43 0.204 0.108 0.056 0.075 0.053 0.082 0.053 0.052 0.053
1801 47 0.195 0.094 0.089 0.088 0.089 0.149 0.089 0.125 ...
1806 37 0.220 0.144 0.130 0.138 0.132 0.106 0.100 0.098 0.103
2001 43 0.204 0.083 0.102 0.073 0.079 0.126 0.086 0.098 ...
2015 26 0.258 0.121 0.077 0.080 0.088 0.110 0.079 ... ...
2605 35 0.227 0.188 0.087 0.121 0.098 0.113 0.080 0.095 0.075
196 Ertugrul Benzeden et al.

Table 9. Physically meaningful lower bounds for the surface flow components (in m 3 /s).

Station Moment Estimates Maximum Likelihood Estimates Observed Smallest

no. Surface Plow Comp.
LN3 G3 LP3 LN3 G3 LP3 (m 3 /s)

901 230.0
902 ... ... ... 18.79 215.77 — 238.0
1801 ... ... ... ... 1.64 ... 14.0
1806 20.40 — 0.01 11.72 41.42 0.88 49.0
2001 ... ... ... ... 99.76 ... 129.0
2015 ... ... 0.58 2.11 ... 0.14 4.4
2605 — 51.63 — 3.70 275.06 — 309.0

Table 10. Physically meaningful lower bounds for the baseflow components (in m 3 /s).

Station Moment Estimates Maximum Likelihood Estimates Observed Smallest

no. Base Flow Component
LN3 G3 LP3 LN3 G3 LP3 (m 3 /s)

901 11.51 0.06 17.06 31.12 8.74 37.0

902 ... 19.58 5.26 16.02 25.97 11.94 29.0
1801 ... ._ ... ... 11.75 ... 12.0
1806 ... 1.29 0.16 21.98 30.40 21.16 31.0
2001 ... ... ... ... 18.11 ... 23.0
2015 ... 2.38 ... 0.47 ... ... 3.3
2605 — — 0.10 2.98 10.30 0.03 12.0

Since peak flows are positive valued random variables, frequency models with physically
meaningful lower bounds have a hydrologie justification and may be preferred in
practical applications. The estimated lower bound magnitudes in Table 5 are generally
greater than the average karst spring contributions. This result may be attributed to the
fact that the baseflow in flood seasons is much greater than the annual average base flow.
As a consequence of the method of maximum likelihood, in most cases one obtains
physically meaningful lower bound estimates.

The models with the smallest Kolmogorov statistics, as well as the appropriate models
with physically meaningful lower bounds, are selected and presented in Table 11.
Abbreviated forms of the distribution functions and the estimation methods, together with
the computed Kolmogorov statistics (in brackets), are included in this table.

Models for the distribution of baseflow and surface flow components

The Kolmogorov statistics given in Tables 7 and 8 show that none of the models
postulated for the frequency distributions of the surface flow and baseflow components
 Flood frequency analysis in karst river basins 197

Table 11. Frequency distribution models for the annual peak flows with smallest
Kolmogorov statistics, and selected models with physically meaningful lower bounds.
Station Model with smallest Selected model with a physically
no. Kolmogorov statistic meaningful lower bound

811 NOR (0.070) / MOM G3 (0.096) / MOM

901 NOR (0.108) / MOM / MOY ....
902 LN2 (0.059) / MOM LN3 (0.076) / MOY
1708 LN3 (0.058) / MOY LN3 (0.058) / MOY
1708* LN3 (0.057) / MOY LN3 (0.057) / MOY
1712 GUM (0.070) / MOY G3 (0.071) / MOY
1714 NOR (0.120) / MOM / MOY ...
1801 G3 (0.073) / MOY G3 (0.073) / MOY
1801* G2 (0.066) / MOY G3 (0.067) / MOY
1805 LN2 (0.089) / MOM G3 (0.095) / MOY
1806 LP3 (0.086) / MOY LP3 (0.086) / MOY
1806* G2 (0.080) / MOY LN3 (0.093) / MOY
2001 LN2 (0.054) / MOY G3 (0.054) / MOY
2005 G3 (0.065) / MOM G3 (0.065) / MOM
2006 GUM (0.046) / MOM G3 (0.064) / MOY
2007 G2 (0.169) / MOY ....
2009 LN2 (0.075) / MOY LP3 (0.097) / MOY
2009* LN2 (0.071) / MOY LP3 (0.079) / MOY
2010 GUM (0.075) / MOY ....
2015 LN2 (0.092) / MOY LP3 (0.096) / MOY
2015* LN2 (0.081) / MOM LN3 (0.086) / MOY
2103 G2 (0.088) / MOY ....
2123 LN3 (0.065) / MOY LN3 (0.065) / MOY
2603 GUM (0.056) / MOM G3 (0.087) / MOY
2605 G3 (0.071) / MOY G3 (0.071) / MOY
2610 GUM (0.102) / MOM LN3 (0.133) / MOM

Table 12. Frequency distribution models for the surface flow and baseflow components
with smallest Kolmogorov statistics, and selected models with physically meaningful
lower bounds.
Station Model with smallest Selected model with a physically
no. Kolmogorov statistic meaningful lower bound

a) Surface flow components

901 NOR (0.079) / MOM/MOY

902 GUM (0.056) / MOY LN3 (0.083) / MOY
1801 LN2 (0.065) / MOM G3 (0.074) / MOY
1806 GUM (0.106) / MOY LN3 (0.110) / MOY
2001 G3 (0.050) / MOY G3 (0.050) / MOY
2015 LP3 (0.051) / MOM LP3 (0.051) / MOM
2605 G3 (0.079) / MOM G3 (0.079) / MOM

b) Base flow components

901 G3 (0.040) / MOY G3 (0.040) / MOY

902 GUM (0.046) / MOM G3 (0.047) / MOM
1801 GUM (0.082) / MOY G3 (0.125) / MOY
1806 LN3 (0.100) / MOY LN3 (0.100) / MOY
2001 GUM (0.064) / MOM G3 (0.098) / MOY
2015 GUM (0.061) / MOM G3 (0.074) / MOM
2605 LN3 (0.080) / MOY LN3 (0.080) / MOY
198 Ertugrul Benzeden et al.

Table 13. Effect of karst baseflows on the variation coefficient under the assumption that
surface flow and baseflow components are uncorrected.

Station Standard Deviation Reduction Coeff.of Variation Relative Diff.

no. Facto r

s, $x+y ny
d "z n
x+y = d n
y (nz - n x + y )/n z

901 189.65 190.80 0.3044 0.8955 0.2709 0.2726 -0.006

902 493.95 491.86 0.5821 0.9441 0.5518 0.5495 0.004
1801 126.78 125.88 0.6967 0.8697 0.6102 0.6059 0.007
1806 166.23 169.20 0.7176 0.8038 0.5667 0.5768 -0.018
2001 283.32 285.18 0.5671 0.8703 0.4903 0.4936 -0.007
2015 57.29 54.41 1.2260 0.8151 1.0523 0.9994 0.050
2605 710.15 714.13 0.5784 0.9682 0.5576 0.5600 -0.004

is rejected at a 5% level of significance. From comparisons of the values in Tables 5 and

9, it can be seen that the meaningful lower bound magnitudes of the surface flow
components are considerably reduced or diminished, while the baseflow components have
lower bounds in comparable magnitudes with respect to their means. As we did for total
peak flows, those models with smallest Kolmogorov statistics, as well as the appropriate
models with physically meaningful lower bounds are selected and presented in Table 12.

In contrast to what is expected, the best fit model types for the total flows and for the
surface flow components do not match, except for the stations 901 and 2605. It is also
interesting to note that the meaningful lower bounds of the baseflow components at
certain stations imply that the population of total peak flows must have a lower bound,
though the sample does not. A typical example for such a case is seen at Homa (901).
Baseflow components at Homa are distributed as gamma-3 with a physically meaningful
lower bound of 31.12 m 3 /s, while the total peak flows as well as the surface flow
components are normally distributed with no bounds.

The role of karst on sample statistics

Let x and y denote the base and surface flow components of floods recorded at any site.
The sum of these two components, z = x + y, has the expectation E(z) = E(x+y) =
E(x) + E(y). If x and y are mutually independent, then the variance of x + y is:

var(x + y) = var(x) + var(y)

Denoting the sample estimates of the population means and population variances by x,
y, z, and Sx, Sy, S x + y , respectively, the sample coefficient of variation of the composite
process x + y may be defined as:
n
x + y = S x+y /(x + y)
where S x + y is computed by:
= +
Sx+y (S* Sy )

It is clear that S x + y will be close to Sy if the variance contribution of x on x + y is not

 Flood frequency analysis in karst river basins 199

significant, or if x is a constant process. In such circumstances, the variation coefficient

of the surface flow components yet can not be maintained, that is n x + y < iiy, or d < 1
where d = n x+y /n y is a reduction factor.
On the other hand, the variation coefficients n z = Sz/z and n x + y will differ more or less
from each other due to sampling errors and/or due to the correlation between x and y.
As a dimensionless measure of the discrepancy between nz and n x + y , the relative
difference (nz — n x+y )/n z is used in this study.

The above defined statistics and measures are calculated for the seven representative
stations and are given in Table 13. The d values in Table 13 indicate the proportional
reduction in the variation coefficient of surface flow components, whereas the figures in
the last column indicate the extent up to which the assumption of independence is
satisfied.

Since for certain hypothetical distribution functions, such as the lognormal and
gamma, there exists a theoretical relation between skewness and variation coefficients,
a very simple but useful analysis of skew coefficients is carried out by regressing them
on the variation coefficients. As shown in Fig. 3, a simple linear relationship with a
correlation coefficient 0.821 is found between these two descriptors of the total peak
flows:

g lz = 3.85nz - 0.74
At a first glance, Fig. 3 gives an impression that the relation between sample skewness
and variation coefficients of total peak flows is a parallel downwards shifted form of the
theoretical relation of the lognormal distribution, especially in the range 0.2 < n < 0.8.
It should be noted that the above relation is based on sample statistics estimated through
unequal sample sizes, and hence, is subject to large and/or nonhomogeneous sampling
errors.

A similar relation in the form of g l y = 3.16 ny — 0.29 with a correlation coefficient

0.885, is found for the surface flow components, while there is no significant relation
between these two descriptors of the baseflow components.

Sl

. ^ o„-- , » • • .,-- ,i»

*"
~~ CUMBEL 9 , - i ; ! * ._
'- - . - - * » •""b o
o
- ^8-- o NORMAL g, > B

I I 1 1
. « . - . 1_.
s.2 o.« o.6 e.a i.ta
cosf. of variation

Fig. 3 - Observed and various theoretical relations between skewness and

variation coefficients.
200 Ertugrul Benzeden et al.

CONCLUSION
Much attention has been invested in the choice of model for annual peak flows, within
the context of flood frequency analyses. It is important to stress that the purpose of
this study was not to seek the best distribution function for flood peaks but to focus
on the fact that goodness of fit is a necessary but not a sufficient condition for
acceptance of a certain model, and further that it should be supported by hydrological
justification.

This study has shown that it is not possible to state with certainty that flood peaks in
karst river basins follow a specific distribution. This result is not surprising, and the
essence of the dilemma of parametric modelling of annual peak flows has already been
discussed by Linsley et al. (1982). Nevertheless, the results of this study revealed that
the models with physically meaningful lower bounds may satisfactorily be applied for
representing the observed frequencies of annual peak flows in karst river basins, and
further they have a hydrological justification.

Frequency analyses of the surface flow and baseflow components at seven representative
stations have shown that in most cases these components are from different populations,
and either one or both may have a physically meaningful lower bound. More frequently,
the baseflow components have lower bounds of significant magnitudes.

On the other hand, correlation and regression analyses performed between surface flow
and baseflow components revealed that these two components may be assumed
uncorrelated. Under this assumption it has been shown that various statistical and
distributional features of the surface flow components are affected by the baseflow
components and are reflected in the total peaks. As a result of karst baseflows, means of
the peak flows are significantly increased compared to their variance, and hence, a
smaller coefficient of variation is produced. The relation between the sample skewness
and variation coefficients at 21 streamgauging stations gives the impression that the
theoretical relationship of the lognormal distribution is parallel shifted downwards.

Although, at first glance, analysis of its components instead of direct analysis of the total
peak floods seems to be a wasted effort, using this approach, and under the assumption
that surface flow and baseflow components are mutually uncorrelated, Benzeden (1989)
computed densities and cumulative probabilities of the composite events (namely, of total
peak flows) by numerical integration of a bivariate joint density function of two
independent variables. He showed that this approach gives a better insight into the
problem and can be used as a valuable tool especially when the components of peak flow
are from different populations and when baseflow components reflect significant
magnitudes with respect to total peak flows.

Surface flow components must be estimated more accurately especially when deriving
rainfall-runoff relations and when extracting regional information through data available
in karst river basins. For such purposes individual hydrographs of annual peak floods
must be analysed for baseflow separation.

Methods and approaches used in this study are applicable to mean and low flows. If the
low flows as annual minima are of special interest, they may be regarded as only the
karst spring contributions, and their frequency distributions are most probably bounded
at one or both ends.

The more intriguing challenge of frequency analysis in karst river basins may be the
treatment of flood frequencies by parametric/nonparametric mixture density approaches
(Schuster & Yakowitz, 1985) which allow for the incorporation of experience, regional
data, or other physical factors.
 Flood frequency analysis in karstriverbasins 201

ACKNOWLEDGEMENTS
This paper is based on the MSc theses (Olcay, 1987; Bagah, 1988) which were completed
within aie scope of a research project (Benzeden, 1989) supported by The Research Fund
of Dokuz Eyliil University. This support is gratefully acknowledged.

REFERENCES
Alkan, A., 1986, Determination of karst base flows at the stream gauging stations in
Seyhan Basin based upon reversible regression analyses between monthly flows (in
Turkish). Izmir, DEÛ, Inst, of Techn. Sciences, Term Paper presented in the course
"Karst Water Resources", pp. el-1 to el-7.
Arisoy, Y., 1986, Determination of karst baseflows at the Dalaman and Esencay basins
based upon reversible regression analyses of monthly flows (in Turkish). Izmir,
DEÙ, Inst, of Techn. Sciences, Term Paper presented in the course "Karst Water
Resources", pp. a 1-1 to a 1-29.
Atis, L, 1980, Investigation of karst spring effluents at the rivers of the Eastern
Mediterranean Basin (in Turkish). Izmir, DEÛ, Inst, of Techn. Sciences, Term Paper
presented in the course "Karst Water Resources", pp. la-1 to la-8.
Atis, L, 1988, Determination of surface flow components in river basins with karst
spring effluents, by separation of groundwater and interflow components based on
mathematical models as simple exponential discharge functions of time (in Turkish).
Izmir, DEÛ, Inst, of Techn. Sciences, PhD Thesis no. 9.
Bagah, K., 1986, Investigation of karst spring effluents at the streams in Ceyhan Basin
(in Turkish). Izmir, DEU, Inst, of Techn. Sciences, Term Paper presented in the
course "Karst Water Resources", pp. 2cl-l to cl-13.
Bafah, K., 1988, Analysis of surface flow components of floods in karst river basins (in
Turkish). Izmir, DEU, Inst, of Techn. Sciences, MSc Thesis no. 21.
Baran, T., 1987, Gross hydropower potential of river basins in the southern Turkey (in
Turkish). Izmir, DEU, Inst, of Techn. Sciences, MSc Thesis no. 15.
Baran, S., 1989, Analysis of flows during recession periods on the example of Manavgat
River basin (in Turkish). Izmir, Inst, of Techn. Sciences, MSc Thesis no. 15.
Bayazit, M., 1981, Statistical methods in hydrology (in Turkish). Istanbul, ÎTU, Fac. of
Civil Engng, Publ. no. 1197.
Benzeden, E., 1989, Mathematical modelling approaches for peak floods in karst river
basins (in Turkish). Izmir, DEU, Research Project no. 0922-87-04-02.
Benzeden, E. & Tathoglu, E. (Unal), 1986, Hydraulicity and aquifer discharge
coefficient in river basins with significant karst spring effluents. In: Karst Water
Resources (Proc. Ankara/Antalya Symp., July 1985) (ed. by G. Gûnay & A. I.
Johnson), 77-86. IAHS Publ. no. 161.
Birsoy, Y., 1989, A new theoretical approach to the spring discharge problem (in
Turkish). Izmir, DEU, Dept Geol. Engng.
Bobée, B. 1975, The log-Pearson type 3 distribution and its application in hydrology.
Wat. Resour. Res., 11 (5), 681-689.
Çetiner, A., 1986, Investigation of karst spring effluents in the Lower Euphrathes basin
(in Turkish). Izmir, DEU, Inst, of Techn. Sciences, Term Paper presented in the
course "Karst Water Resources", pp. fl-1 to fl-8.
EÏE, 1956, Discharge measurements of the water years 1935-1953. General Directorate
of Electrical Works Survey Authority, Ankara, ElE Hydrography Publ. no. 2.
EÎE, 1956-1985, Discharges of 1955 thru 1982 Water Years. General Directorate of
Electrical Works Survey, Ankara, 18 separate vols.
Gûnay, G. (éd.), 1980, Karst Hydrogeology. Oymapinar 1979, Proceedings, DSÎ-UNDP
Project, Ankara.
Keloglu, N., 1984, Mathematical simulation models of monthly flows with karst spring
effluents, based upon exponential base flow functions (in Turkish). Izmir, DEÛ, Inst.
202 Ertugrul Benzeden et al.

of Techn. Sciences, PhD Thesis no. 8.

Kite, G. W., 1977, Frequency and Risk Analysis in Hydrology. Water Resources
Publications, Fort Collins.
Linsley, R. A., Jr, Kohler, M. A. & Paulhus, J. L. EL, 1982, Hydrology for Engineers,
3rd edn. McGraw-Hill, New York.
Matalas, N. C. & Wallis, J. R., 1973, Eureka! It fits a Pearson type 3 distribution. Wat.
Resour. Res., 9 (2), 281-289.
Olcay, M. S., 1987, Effects of karst on the distribution of peak floods (in Turkish),
Izmir, DEÙ, Inst, of Techn. Sciences, MSc Thesis no. 17.
Ôzdaglar, D., 1980, Investigation of karst spring effluents in the Tigris Basin (in
Turkish). Izmir, Ege Univ., Fac. of Civil Eng., Term Paper presented in the course
"Karst Water Resources", pp. lb-1 to lb-3.
Ôzi§, N. & Harmancioglu, 1980, An outlook to karst hydrology studies in Turkey.
Ankara. In: Oymapinar 1979, Proceedings (ed. by G. Gûnay), DSÎ-UNDP Project,
294-304.
Schuster, E. & Yakowitz, S., 1985, Parametric/nonparametric mixture density estimation
with application to flood frequency analysis. Wat. Resour. Bull., 21 (5), 797-804.
Yevjevich, V., 1972, Probability and Statistics in Hydrology. Water Resources
Publications, Fort Collins.
Hydrogeological Processes in Karst Terrenes (Proceedings of the Antalya Symposium and Field Seminar, October 1990). ^„
IAHS Publ. no. 207, 1993. 203

SPATIAL DISTRIBUTION OF PRECIPITATION IN THE WESTERN TAURIDS

LEVENT TEZCAN & ALPARSLAN ARIKAN

International Research and Application Center for Karst "Water Resources, 06532 Beytepe, Ankara, Turkey

ABSTRACT

In order to estimate the spatial behaviour of the orographic precipitation over the
important karst areas in southern Turkey, co-kriging interpolation technique is used.
Since the spatial distribution of the precipitation in this area greatly depends on the
altitude, the distribution is estimated in three dimensions by the co-kriging method that
takes the spatial correlation between the variables into account. This method can be used
also by considering other variables affecting the precipitation distribution, such as
distance from the sea, slope directions, etc.

INTRODUCTION
It is important to estimate the spatial behaviour of precipitation in order to determine the
hydrogeological budget of the karst aquifer systems or input distributions of hydrological
models. Most of the karstic aquifers in Mediterranean countries are fed by orographic
precipitation, and the rainfall intensity is highly related to altitude. The distribution of
precipitation is highly variable along the southern coasts of Turkey. In some parts of the
region the precipitation increases with altitude while in other parts it decreases. Using
an average value causes uncertainties in water budgets and model calibration.

A number of spatial interpolation techniques such as the Thiessen polygon method, least
squares approach, polynomial interpolation, Lagrange approach, inverse distance
interpolation, multiquadratic interpolation, optimal interpolation and kriging interpolation
techniques are available. The review of Tabios & Salas (1985) gives a detailed
description and comparison of these techniques. All these techniques are based on a two-
dimensional data field obtained from observation stations without taking into account the
physical properties of each site. In this study a three-dimensional field obtained by
locating observation stations in three-dimensional space and using the data gathered at
these stations is used.

The co-kriging technique is used to estimate the spatial distribution of precipitation,

which is the extension of the kriging technique to several variables whereby several
variables are estimated jointly utilizing a best linear unbiased estimator (BLUE). In the
Taurids karst region, the spatial distribution of precipitation could not be interpolated
using a two-dimensional field since the precipitation intensity observed even at nearby
stations can be very different. Therefore, the co-kriging system is constructed to show
the inaccuracy of the two-dimensional interpolation techniques.

Co-kriging
Co-kriging is a geostatistical technique developed to improve the estimation of a variable
using information on other spatially correlated variables which are generally better
sampled. With this technique no assumptions are made on the nature of the correlation
between variables. The degree of this correlation as well as its spatial structure are taken
into account by the cross variogram.
Let Z(x;) be the first variable and Y(xk) the second variable (i = l...n, k = l...m).
204 Lèvent Tezcan & Alparslan Ankan

Then, the estimated value of Z at any point x0 is given by:

m m
Z'CxJ^XjzCx^^YCx,)
i=l k=l
where
n m

i=l k=l
Minimizing the variance of the estimation error results in the following co-kriging
system:
n m
-X* 1 1 + v ^ X\ 2 12
E
j=l
j7ij . L k 7 i k + M i = 7;o
k=l
i = l...n

n m
12
E % 1 12 v^ \ 2 2 7ko k = l...m
i=i 1=1
The experimental variograms (7) and cross-variogram are calculated as follows:
yoi) = lE[(z( Xi )-z( Xj ) 2 ]

2
7 (h) = iE[(Y( Xi )-Y( Xj ) 2 ]

12
7 (h) = iE[(Z( Xi )-Z( Xj ))(Y( Xi )-Y( Xj ))]
where h = x; — Xj. A theoretical model is fitted to the experimental variograms by the
weighted least squares method assigning appropriate weights to different values with
different number of pairs. The above co-kriging system gives (n + m + 2) linear
equations with (n + m + 2) unknowns.

The variance of estimation error is given by:

n

<? = E ^ o + E ^ T k o + Mi
i=l k= l

CASE STUDY
The above mentioned technique was applied to the western Taurids area on the southern
coast of Turkey. The area comprises highly karstified mountains and plateaux. The
estimation of the spatial distribution of precipitation over this area is an important step
in many hydrological applications such as evaluation of the water balance or forecasting
karst flows using mathematical models. In the area, several important karst springs are
located and their subsurface drainage boundaries have not been defined clearly. The
errors in estimating the areal precipitation cause uncertainties in water balance
calculations and hydrological model calibrations. The accuracy of such models is based
upon the quality of the inputs. Precipitation is the easiest measurable input for these
models. However, the stations are not representative for the area, because the density of
the stations and the altitude of the stations are not satisfactory.
 Spatial distribution ofprecipitation in the western Taurids 205

a>
E

ie ter

Fig. 1 - Altitude distribution and precipitation stations in the western

Taurids.

The precipitation stations are located in the altitude range 0-1500 m (Fig. 1), and the
density of stations is not satisfactory to represent the distribution of precipitation in
higher parts of the area. The type of precipitation is mostly orographic, therefore, the
distribution of precipitation is highly affected by altitude.
In this study the annual precipitation and altitudes of 47 stations have been used to
estimate the spatial distribution of the precipitation. The spatial covariance structure of
the precipitation and altitude values are determined by semi-variograms. The
experimental semi-variograms of both variables and theoretical models are given in Figs

0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000
Seperation Distance (Metres)

Fig. 2 - Experimental and theoretical semi-variograms of altitude.

206 Lèvent Tezcan & Alparslan Arikan

1.20 T

0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000
Seperation Distance (Metres)

Fig. 3 - Experimental and theoretical semi-variograms of precipitation.

2 and 3. Theoretical models have been fitted by weighted least square technique. Both
of the semi-variograms have been expressed by spherical models with a range of 9 km,
so the spatial relation between the variables over this distance has not been taken into
account. The spatial correlation between two variables has been expressed by the cross-
variogram (Fig. 4). A negative relation between precipitation and altitude has been
observed in the region. A spherical theoretical model has been fitted to this variogram
with a range of 11 km.

Fig. 4 - Experimental and theoretical cross-variograms of precipitation

and altitude.
 Spatial distribution of precipitation in the western Taurids 207

After modelling the spatial covariance structure, the spatial distribution of precipitation
over the area is estimated by co-kriging method (Fig. 5). The annual precipitation in
coastal region is nearly 1000 mm, and this value increases to 1500 mm in the western
parts of the area with increasing altitude up to 1300 m as shown in Fig. 5. In the
southeastern part a similar relation can be seen. However, in the northern part, the
precipitation decreases to 300 mm with increasing altitude over 1500 m. Thus, most of
the precipitation falls below an altitude of 1500 m.

O 4000 3000 12000 16000 2000O

ineters
Fig. 5 - Spatial distribution of precipitation estimated by the co-kriging
method.

CONCLUSIONS
The distribution of precipitation is estimated by using the co-kriging interpolation
technique taking into consideration annual precipitation and altitude in the western
Taurids region of southern Turkey. The altitude-precipitation relation is modelled by a
cross-variogram over the area. The negative values of the cross-variogram indicate that
a positive increase in one of the variables corresponds, on average, to a decrease in the
other variable.

Precipitation is the dominant input over the western Taurids region. There are different
micro-climatological systems affecting the precipitation distribution over the area. Thus,
it is important to take account of these systems in karst hydrology applications in the
area. The method used in this study can be extended by using other variables causing the
existence of different micro-climatologic systems.

REFERENCE
Tabios, G. & Salas, J. D., 1985, A comparative analysis of techniques for spatial
interpolation of precipitation. Wat. Resour. Bull., 21(3), 365-380.
Hydrogeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990). -.--.
IAHS fubl. no. 207, 1993. ^ 209

EFFECT OF KARST SPRINGS ON FLOW DURATION CURVES OF RIVERS

ÛNAL ÔZΧ, AHMET ALKAN & ERGUN ÙNAL TATLIOGLU

Civil Engineering Department, Faculty of Engineering, Dokuz Eylul University, Bornova 35100, Izmir, Turkey

ABSTRACT

The contribution of spring effluents from karst formations significantly increase the
absolute and relative discharge of rivers and decrease their seasonal variation, in order
to alleviate the need for reservoir regulation. The quantitative effects of karst springs are
analysed and compared on three types of basins in Turkey: (a) a typical mediterranean
basin with significant spring effluents and with strong seasonal variations in moderate
annual precipitation; (b) a typical southwestern basin without significant springs and with
accentuated seasonal variation in moderate annual precipitation; (c) a typical northeastern
basin without significant springs and with moderate seasonal variation in high annual
precipitation. Preliminary comparisons show that karst springs exert a significant effect
in decreasing the seasonal variation of total river runoff in basins with strong seasonal
variation in precipitation, but not confined to the limits occurring in basins with moderate
seasonal variation in high annual precipitation.

INTRODUCTION
Karstified formations of soluble, mostly carbonate rocks cover almost one third of
Turkey, and are predominant along the entire southern coast and a large part of the
eastern regions (Eroskay & Giinay, 1980). There have been many studies of the
hydrology of these areas during the last two decades (Acatay, 1966 a,b; Dinçer & Payne,
1971; Çultu & Altinbilek, 1971; Ôzis & Ece, 1975; Ôzis & Keloglu, 1976; Ôzer, 1976;
Uslu, 1978; Yurtsever, 1978; ICOLD, 1976; Ôzis & Benzeden, 1979; Ôzis, 1979;
Giinay, 1980; Giinay & Karanjac, 1980; Yevjevich, 1980, 1981; Alparslan, 1980, 1982;
Harmancioglu, 1980, 1981, 1989; Ôzis et al., 1981; Ôzis & Yurtsever, 1982; Keloglu,
1984, 1985, 1987; Ôzis et al., 1985; Eroskay et al., 1986; Olcay, 1987; Olcay et al.,
1987; Baran et al., 1987; Benzeden & Ansoy, 1987; Bagah, 1988; Atis, 1988; Ôzis,
1989; Harmancioglu & Baran, 1989; Benzeden, 1989).

Karst spring effluents significantly contribute to river runoff in most of Turkey's river
basins; and exert a positive effect to decrease the required relative reservoir volumes for
flow regulation, because the propagation through underground karst formations leads to
smaller seasonal flow variations compared with direct surface runoff (Ôzis, 1989).

METHODOLOGY
The preliminary investigations were based upon the comparison of an average karst basin
with two non-karst basins with different meteorological and hydrological characteristics.

Selection of comparative hydraulicity

The comparisons were carried out for three years of representative hydraulicity: (a) a
year of almost average hydraulicity; (b) a dry year with about 95% probability of
exceedance; (c) a wet year with about 5% probability of exceedance. The selection will
be based on the probability distribution of annual hydraulicity coefficients H { of each
i.th year as:
210 Ûnal Ôzi§ et al.

Ha i = ^ (1)
' Qo.N
where Q0 ; is the average discharge of the i.th year and Q0 N the average discharge during
the entire period of N years under consideration (Ôzis, Ï971, 1974).

Dimensionless flow duration curves

The flow duration curves of each year will be transformed into a dimensionless form by
the use of the hydraulicity coefficient concept as:

V0,N
where H, ; is the daily hydraulicity coefficient of the t.th day of the i.th year, Q t j the
daily discharge of the t.th day of the i.th year, Q 0 N is the same as in formula (1). àince
usually the daily discharge sequences are skewed and a lognormal probability distribution
is in most cases the simplest, though not the best suitable for fitting the distribution of
stochastic elements, the comparisons will be carried out on In H t ; values:

In K^ ; = In (3)
Qo,N

Mathematical modelling of dimensionless flow duration curves

A simple exponential model of the type:

H,, = «,e-* W
is expected to fit a large part of the middle and lower portions of the dimensionless flow
duration curves, in order to permit comparisons as well as approximate, but quick
determination of the annual average diverted flow for a given diversion capacity.

BASIN SELECTION

Basin with significant karst spring effluents

The Mediterranean Lamas basin was selected as a typical karst basin, gauged at the
Kmlgeçit station (no. 1717) at an elevation 1070 m a.s.l., with drainage area 1055 km 2
(Figs 1 and 2). The average discharge is in the order of 7 m 3 /s, out of which one third
can be attributed to karst spring effluents and two thirds to direct surface runoff (EIE,
1964-1984; DEÛ, 1985; Benzeden & Ansoy, 1987).

The Lamas basin appears to be quite modest with regard to absolute values of karst
spring effluents, compared to the unusual basins like Manavgat, Kôprûçay (Baran et al.,
1987); but found to be adequate for these preliminary studies.

Non-karst basin with moderate annual precipitation

The Çine sub-basin of the Biiyuk Menderes basin in the Aegean region, was selected as
the non-karst basin with moderate annual precipitation. This is gauged at the Kayirh
 Effect ofkarst springs onflow duration curves of rivers 211

Fig. 1 - Location of the representative gauging stations at the three basins

(Kayirh at Çine; Kizilgeçit at Lamas; §im§irli at iyidere) in Turkey.

Fig. 2 - Drainage area of the Kizilgeçit gauging station in the Lamas

basin.

station (no. 701) (elevation 262 m a.s.l.; drainage area 948 km 2 ), with an average
discharge of 6.6 m3/s (EIE, 1964-1983). The drainage area as well as the average
discharge of Kayirh are of the same order as those of Kmlgeçit.

Non-karst basin with high annual precipitation

The iyidere basin in the eastern Black sea region was selected as the non-karst basin with
high annual precipitation. This basin is gauged at the §im§irli station (no. 2218)
(elevation 310 m; drainage area 855 km 2 ), and has an average discharge of 28.1 m3/ s
(EIE, 1964-1984).

The drainage area of §imsirli is of the same order as that at Kmlgeçit, but the average
discharge is about four-fold, due to the large difference in precipitation.
212 tlnal Ôzi$ et al.

Effect of precipitation
The precipitation over the Lamas basin can be characterized by the observations at the
Guzeloluk meteorological station, with an average annual value of 780 mm during 1966
1984; the large seasonal variations are shown as monthly average data (DEÙ, 1985) in
Fig. 3. The precipitation over the Çine basin can be characterized by the observations
at the Yatagan meteorological station, with an average annual value of 673 mm during
1950-1970; the large seasonal variations are shown as monthly average data (DMI, 1974)
in Fig. 3. The precipitation over the Iyidere basin can be characterized by the
observations at the Rize meteorological station, with an average annual value of 2357 mm
during 1929-1970; the moderate seasonal variations are shown as monthly data (DMI,
1974) in Fig. 3.
These data show that precipitation over Lamas and Çine basins are almost of the same
order of magnitude, whereas that over iyidere is three times larger; moreover, the
seasonal variations of precipitation over Lamas and Çine are quite large, whereas the
variations over Iyidere are quite moderate.

Monthly Precipitation
(mm/month)
300 •

200 -

100 -

^•Months

Fig. 3 - Monthly average precipitation at Guzeloluk for Kizilgeçit/Lamas

(full line), at Rize for §im§irli/iyidere (dashed line), at Yatagan for
Kayirh/Çine (dotted line).

SELECTION OF YEARS WITH REPRESENTATIVE HYDRAULICITIES

Observed data
Daily discharges at the three streamgauging stations Kizilgeçit, Kayirh, §im§irli are
available for the 21-year period 1963-1983 (Table 1). Those for Kayirh and §im§irli are
all direct observations (EIE 1964-1984) and those for Kizilgeçit are direct observations
for the period of 1967-1983 (EIE, 1968-1984), and obtained by regression on the
upstream Sariaydm station for the period 1963-1966 (DEÛ, 1985). The annual mean
discharges for each year from 1963 to 1983 are given in Table 1.

Annual hydraulicity coefficients

The annual mean discharges are transformed into annual hydraulicity coefficients
according to formula (1), through dividing by their respective long-term mean values,
given at the end of Table 1. These hydraulicity coefficients, with their mean value being
1.0, are given in Table 2.
 Effect ofkarst springs onflow duration curves of rivers 213
Table 1. Annual mean discharges at the Kayirh, Kizilgeçit and §imsirli
gauging stations 1963-1983.
Kayirh Kizilgeçit Çimçirli
Years (701) (1717) (2218)
rrf/s irf/s nWs
63 9.99 8.42 30.1
64 2.23 4.49 27.9
1965 8.11 6.83 28.5
66 8.66 7.34 28.2
67 7.76 7.37 28.4
68 8.35 9.17 32.6
69 9.70 13.49 21.2
1970 9.25 8.58 26.1
71 7.02 6.53 31.7
72 2.78 5.66 29.0
73 4.57 3.32 31.2
74 3.25 2.77 28.5
1975 4.48 7.35 26.0
76 3.44 6.37 27.5
77 3.46 5.96 27.2
78 6.91 6.99 29.4
79 5.51 5.55 28.1
1980 8.33 7.00 23.8
81 10.01 10.00 30.3
82 8.99 6.65 28.9
83 5.98 5.85 25.5

Qo,63/83 6.61 6.94 28.11

Table 2. Annual hydraulicity coefficients at Kayirh, Kizilgeçit and §im§irli.

Kayirh Kizilgeçit Çimsirli
Years (701) (1717) (2218)

63 i 1.51 1 ' 1.21 1.07

64 0.34 0.65 0.99
1965 1.23 0.98 1.01
66 1.31 1.06 1.04
67 1.17 1.06 1.01
68 1.26 1.32 1.16
69 1.47 1.94 0.76
1970 1.40 1.24 0.93
71 1.06 0.94 1.13
72 0.42 0.82 1.03
73 0.69 0.48 1.11
74 0.49 0.40 1.01
1975 0.68 1.06 0.93
76 0.52 0.92 0.98
77 0.52 0.86 0.97
78 1.05 1.01 1.05
79 0.83 0.80 1.00
1980 1.26 1.01 0.85
81 1.51 1.44 1.08
82 1.36 0.96 1.03
83 0.90 0.84 0.91
1^,63/83 1.00 1.00 1.00
214 Ûnal Ôzi$ et al.

Frequency distribution of annual hydraulicity coefficients

An earlier study (Ôzi§, 1971) showed that a simple normal probability distribution
function proved to be adequate in the frequency analysis of annual hydraulicity
coefficients of the southern Bûyûk Menderes (represented by the same Kayirli station),
of the eastern Mediterranean (represented by the Karageçit station on the Berdan River
close to Lamas), of the eastern Black Sea (represented by the Kanhgeçit/Maçka station
on the Degirmendere River close to iyidere) basins.

The frequency distribution of the annual hydraulicity coefficients for Kayirli, Kizilgeçit,
SJmsjrli are shown in Fig. 4 on a normal probability distribution coordinate system. The
values of the hydraulicity coefficients corresponding to 5 % and 95 % probabilities of non-
exceedance are given in Table 3; that for 50% is equal to unity according to the
definition of this parameter.

Probability of
non-exceedance
nl i i ' i I i i i ' i i i i
1 2 5 W 2030*0 50607080 90 35 9899

Fig. 4 - Normal probability distribution of annual hydraulicity coefficients

at Kizilgeçit (full line; circles), at Simsirli (dashed line; triangles), at
Kayirli (dotted line; squares).

Table 3. Annual hydraulicity coefficients with 5%, 50% and 95%

probabilities of nonexceedance at the Kayirli, Kizilgeçit and SJmsirli
gauging stations.
Probability of Kayirli Kizilgeçit §im§irli
non-exceedance (701) (1717) (2218)

5% 0.36 0.46 0.85

50% 1.00 1.00 1.00
95% 1.64 1.54 1.15
 Effect ofkarst springs onflow duration curves of rivers 215

Selection of representative years

The annual hydraulicity coefficients for the driest, the wettest, and the nearest-to-average
years during the 1963-1983 period, along with their respective year of occurrence, at
each gauging station are given in Table 4. These data correspond to a plotting frequency
position of 1/(21 + I) = 4.5%, 11/(21 + I) = 50%, and 21/(21 + I) = 95.5%, and should
display values close to those in Table 3, if the dry and wet-extremes fit the normal
probability distribution well.
This condition can be considered as almost satisfied for preliminary investigations, with
the exception of the wettest year at Kizilgeçit, displaying a hydraulicity roughly 25%
larger than that excepted for a probability of 95% non-exceedance. However, keeping
this discrepancy in mind in later discussions, the 1969 data are kept as representative of
the wet year at Kizilgeçit. Hence, the dimensionless flow duration curves of the years
given in Table 4 are analysed and compared for three typical hydraulicities at each of the
gauging stations.
Table 4. Annual hydraulicity coefficients for the driest, the wettest and
the nearest-to-average years during the 1963-1983 period at the Kayirli,
Kizilgeçit and §im§irli gauging stations.
Type of Kayirli Kizilgeçit Çimsirli
the year (701) (1717) (2218)

Dryest 0.34(1964) 0.40 (1974) 0.76(1969) '.

Average 1.05 (1978) 1.01 (1978) 1.00 (1979)
Wettest 1.51(1981) 1.94 (1969) 1.16(1968)

COMPARISON OF DIMENSIONLESS FLOW DURATION CURVES

The logarithmic transformation (formula (3)) of the daily hydraulicity coefficients,
forming the dimensionless flow duration curves for the driest years (Kayirli-1964,
Kizilgeçit-1974, §im§irli-1969) are given in Fig. 5; those for the nearest-to-average years
(Kayirh-1978, Kizilgeçit- 1978, §im§irli-1979) in Fig. 6; those for the wettest years
(Kayirh-1981, Kizilgeçit-1969, §im§irli-1968) in Fig. 7.

A comparison of these dimensionless flow duration curves show that, under almost
similar precipitation patterns, the seasonal variation at Kizilgeçit is lower than at Kayirli,
and even competes with the narrow limits at §im§irli, fed from a completely different
pattern of large precipitation.

Fig. 5 - Logarithmically transformed dimensionless flow duration curves

for the driest years: 1964 at Kayirli (dotted line), 1974 Kizilgeçit (full
line), 1969 at Sim§irli (dashed line).
216 Ûnal Ôzi$ et al.

t / T (%)

Fig. 6 - Logarithmically transformed dimensionless flow duration curves

for the nearest-to-average years: 1978 at Kayirh (dotted line), 1978 at
Kizilgeçit (full line), 1979 at §imsirli (dashed line).

t / T (%)

Fig. 7 - Logarithmically transformed dimensionless flow duration curves

for the wettest years: 1981 at Kayirh (dotted line), 1969 at Kmlgeçit (full
line), 1968 at §imsirli (dashed line).

In fact, the necessary relative volume for the regulation of the dry-season flows, as
shown in Fig. 8, is for the nearest-to-average years 0.47.V o at Kayirh, 0.24.Vo at
Kizilgeçit, 0.31.Vo at §imsirli. For the driest years these volumes increase to 0.69.V o at
Kayirh, to 0.62.Vo at Kmlgeçit, to 0.41.Vo at §imsirli; for the wettest years, they
decrease to 0.34. V0 at Kayirh, to 0.05. V0 at Kizilgeçit, to 0.26. VQ at §imsirli.

It can be concluded that the water regime at Kmlgeçit in the Lamas basin with relatively
significant karst springs requires about 50% less volume for eventual reservoir
regulation, compared to that at Kayirh in the Çine basin which is without significant
karst springs but has similar precipitation. This ratio is anticipated to be greater when
larger karst basins are considered. On the other hand, though an extreme in high
precipitation regions of Turkey, the water regime at §im§irli in the non-karst iyidere
basin, result in regulation volumes below the rate in the karstic Lamas basin for a dry
year, of the same order in an average year, and above in a wet year.

EXPONENTIAL MATHEMATICAL MODELLING OF DIMENSIONLESS FLOW

DURATION CURVES

The simple exponential model of the so-called Maillet-type (formula (4)) has found wide
application in modelling dry-season flows of rivers with significant karst spring effluents
 Effect ofkarst springs onflow duration curves of rivers 217
InH

V„ <rn / y ) 31,5.306.OoN ta3/s)

365
V, (m J /y) Ï (1-H,.).C.o , N -
t-1

Fig. 8 - Definition of the necessary relative volume (Vr) for the regulation
of the dry-season flow with regard to the long-term annual average flow
volume (VJ.

(Giinay, 1980; Gunay & Karanjac, 1980; Ôzis, et al., 1981, 1985; Keloglu 1984, 1985,
1987; Ati§, 1988). The same concept appeared to be useful in modelling the timely
unsequential flow duration curves, in order to easily determine the average diverted flow
rates and volumes for unregulated diversion-type water uses, as shown in Figs 9 and 10.
The fitting of the simple exponential model to the logarithmically transformed
dimensionless flow duration curves at Kizilgeçit for the entire duration, for the 90%,
80%, 70%, 60% dry-season parts of the year are shown in Fig. 11 for the driest, in

Fig. 9 - Average diverted flow with regard to diversion capacity, based

upon the flow duration curve.

Fig. 10 - Average diverted flow with regard to diversion capacity, based

on dimensionless flow duration curves of daily hydraulicity coefficients.
218 Ùnal Ôzi§ et al.

Fig. 12 for the nearest-to-average, in Fig. 13 for the wettest years. The convergence of
the parameters and the correlation coefficient R are given in Tables 5 , 6 , 7 respectively.
It can be concluded that after a duration of almost 20-25 %, the exponential model can
be considered satisfactory. This means that for a diversion capacity exceeded during 75-
90 days or more in a year, the model can be used to compute the average diverted flow.

CONCLUSION
A method has been developed for direct evaluation of the decreasing effect on the
seasonal variation of river flows, when karst spring effluents constitute a significant
contribution.

Fig. 11 - Fitting simple exponential model to the dimensionless flow

duration curve during the driest year at Kizilgeçit.

(ÎT1)(%)

Fig. 12 - Fitting simple exponential model to the dimensionless flow

duration curve during the nearest-to-average year at Kizilgeçit.

Fig. 13 - Fitting simple exponential model to the dimensionless flow

duration curve during the wettest year at Kizilgeçit.
 Effect ofkarst springs onflow duration curves of rivers 219

Table 5. Change in parameters of the exponential model for the driest year
at Kmlgeçit.
Part of
the duration a -pia 3 R

%100 0.58 -2.43 - 0.756

%90 0.45 -1.43 - 0.881
%80 0.41 -1.03 - 0.954
% 70 0.40 -0.96 - 0.960
%60 0.41 -1.06 - 0.960

Table 6. Change in parameters of the exponential model for the nearest-to-

average year at Kizilgeçit.
Part of
the duration a -pia3 R

%100 2.10 -4.93 - 0.960

%90 1.78 -4.26 - 0.974
% 80 1.56 -3.75 - 0.983
%70 1.43 -3.44 - 0.988
%60 1.41 -3.39 - 0.982

Table 7. Change in parameters of the exponential model for the wettest

year at Kizilgeçit.

Part of
the duration a -pio- 3 R

%100 4.42 -5.59 - 0.969

%90 3.84 -5.03 - 0.972
%80 3.25 -4.39 - 0.983
%70 3.08 -4.19 - 0.977
% 60 3.13 -4.23 - 0.966

Based upon the dimensionless flow duration curves of daily hydraulicity coefficients, i.e.
the ratio of daily discharge to long-term mean discharge, the application of this method
to three representative basins in Turkey yielded interesting results. The comparison of
a karst basin (Lamas) and a non-karst basin (Çine), both with large seasonal variations
in annual precipitation, showed that the above-mentioned decreasing effect is in the order
of 50%, being 10% in the driest, 50% in the nearest-to-average, 85% in the wettest year
of the common record period of 21 years. The comparison of the karst basin (Lamas)
with a non-karst basin with moderate seasonal variation in very high annual precipitation
(Iyidere) yielded similar results in a wet year, but the opposite in a dry year.

Furthermore, a large part in the middle and dry periods of the dimensionless flow
duration curve of the daily hydraulicity coefficients in karst basins can be fitted by a
simple exponential mathematical model, in order to obtain a straight line after
logarithmic transformation.
220 ÏJnal Ôzis et al.

Related analyses showed that for Lamas basin, this part is of the order of 20-25% of the
year. This corresponds to a diversion capacity exceeded during 75-90 days in a year and
the model can be used to compute the average diverted flow for diversion capacities up
to this value.

REFERENCES
Acatay, T., 1966a, Miimkiin oldugu nisbette radyoaktif maddeler kullanarak, Tûrkiye
dûdenlerinden dort tanesinin etiidu. Ankara, DSÏ Ara§tirma Dairesi, no. PR-357.
Acatay, T., 1966b, Beysehir gôlû - Manavgat irmagi ili§kisi. Ankara, DSÎ, Teknik
Bûlten, no. 9, 1-24.
Alparslan, A., 1980, Karst ortamda yeraltisuyu genel aki§ denklemlerinin belirlenmesi
için hidrolojik bir sistem yakla§imi. Izmir, Ege Ûniversitesi, In§aat Fakiiltesi,
Hidroloji ve Su Yapilan Doktora Tezi, no. 3.
Alparslan, A., 1982, An approach to karst hydrology using the relationships between
reservoir water level and spring discharge. Aachen-Essen, Bull. Int. Ass. Engng
Geol., no. 25, 111-115.
Ati§, I., 1988, Karst pinar katkili akarsularda iistel bo§ahm fonksiyonu esasina dayanan
matematik modeller yardimiyla yeralti ve yuzeyalti akis bilesenlerinin aynlarak
yiizeysel akipn belirlenmesi. Izmir, Dokuz Eylûl Universitesi, Fen Bilimleri
Enstitûsû, Hidroloji ve Su Kaynaklari Doktora Tezi, no. 9.
Bagali, K., 1988, Karst havza taçkmlarimn yiizeysel akis. bile§enlerinin analizi. Izmir,
Dokuz Eylul Ûniversitesi, Fen Bilimleri Enstitûsû, In§aat Mûhendisligi Hidrolik Ana
Bilim Dali Yuksek Lisans Tezi, no. 21.
Baran, T., Harmancioglu, N. & Ôzi§, U., 1987, Tûrkiye'nin akarsu havzalarmda karst
pinarkatkilan. Ankara, în§aat Mûhendisleri Odasi, Tûrkiye însaat Mûhendisligi, IX.
Teknik Kongresi, C.II, 299-311.
Benzeden, E. & Ansoy, Y., 1987, Lamas havzasi ôrneginde biriktirmesiz çevirme
santrallannda degerlendirilebilen debilerin çevirme kapasitesi ile degipmi. Ankara,
in§aat Mûhendisleri Odasi, Tûrkiye însaat Mûhendisligi IX. Teknik Kongresi, C.II,
65-79.
Benzeden, E., 1989, Karstik akarsu havzalarmdaki taçkinlar için matematik model
yakla§imlan. Izmir, DEÛ, Fen Bilimleri Enstitûsû Hidrolik Ana Bilim Dah
Aragtirma Projesi, no. 922-87.04.02.
Çultu, L. & Altinbilek, D., 1971, Statistical analysis of Manavgat River basin recharge
characteristics. Ankara, Orta Dogu Teknik Universitesi.
DEÛ (Ôzis, Benzeden, Arisoy), 1985, Mersin-Limonlu hauzasi Lamas hidroelektrik
santrallari (IIIA,B; IV) yapilabilirlik etudû hidroloji raporlari, C.I.: Genel su
potansiyeli. Izmir, Dokuz Eylûl Universitesi Mûhendislik. MimarlikFakiiltesi, Insaat
Mûhendislig Bôlûmû.
Dinçer, T. & Payne, B. R., 1971, An environmental isotope study of the southwestern
karst region of Turkey. J. Hydrol., 14, 233-257.
DMI, 1974, Ortalama ve ekstrem kiymetler meteoroloji bulteni. Ankara, Devlet
Meteoroloji Isjeri.
EÎE, 1964-1984, Akim neticeleri. Ankara, Elektrik isleri Etût îdaresi.
Eroskay, O. & Gunay, G., 1980, Tecto-genetic classification and hydrogeological
properties of the karst regions in Turkey. In: International Seminar on Karst
Hydrogeology, Oymapinar 1979 (ed. by G. Gunay), 1-14. DSI-UNDP Project.
Eroskay, O., Gunay, G. & Ankan, A., 1986, Oymapinar Baraji'mn karst ortamindaki
etkileri. Ankara, Tûrkiye Bilimsel ve Teknik Arastirma Kurumu, Proje no TBAG -
625.
Gunay, G. (éd.), 1980, Karst Hydrogeology. Ankara, DSI-UNDP Project (Oymapinar
1979, Proceedings).
Gunay, G. (éd.), 1986, Karst Water Resources (Proc. Ankara-Antalya Symp., July 1985).
IAHSPubl. no. 161.
 Effect ofkarst springs onflow duration curves of rivers 221

Gûnay, G. & Karanjac, J., 1980, Dumanh Spring, Turkey, the largest karstic spring in
the world. J. Hydrol. 45, 219-231.
Harrnancioglu, N., 1980, Hidrolojik sûreçlerde bilgi içeriginin entropi ile ôlçûlmesi.
Izmir, Ege Universitesi, în§aat Fakiiltesi Hidroloji ve Su Yapilan Doktora Tezi no.
4.
Harrnancioglu, N., 1981, Measuring the information content of hydrological processes
by the entropy concept. Izmir, Ege Universitesi, In§aat Fakiiltesi Dergisi, Ozel Sayi,
13-38.
Harrnancioglu, N. & Yevjevich, V., 1986, Transfer of hydrologie information along
rivers partially fed by karstified limestones. In: Karst Water Resources (Proc.
Ankara-Antalya Symp., July 1985), 115-131. IAHS Publ. no. 161.
Harrnancioglu, N. & Yevjevich, V., 1987, Transfer of hydrologie information along
river points. /. Hydrol., no. 91, 103-118.
Harrnancioglu, N., 1989, Baraj haznelerinin boyutlandirlmasi ve içletilmesi açisindan
karst pinar katkih akarsulann bilgi içeri gi. Izmir, DEU Aragtirma Fonu için, "Karst
Ortamin Baraj Haznelerine Giren Akimlara ve Hazne Kapasitelerine Etkileri" adh
ve 908.88.04/4 kod numarah Projesi Raporu. Ocak 1989.
Harrnancioglu, N. & Baran, T., 1989, Effect of recharge systems on hydrologie
information transfer along rivers. In: New Directions for Surface Water Modelling
(Proc. Baltimore Symp. May 1989) (ed. by M. L. Kavvas), 223-233. IAHS Publ. no.
181.
ICOLD Turkish National Committee, 1976, Investigation of leakages at Keban Dam,
Mexico. Douzième Congrès des Grands Barrages, Q. 45, R. 27.
Keloglu, N., 1984, Karst pinar katkih ayhk aki§lann iistel bo§ahm fonksiyonlu
matematik benzetim modelled. Izmir, Dokuz Eylûl Universitesi, Fen Bilimleri
Enstitiisii Hidroloji ve Su Kaynaklan Doktora Tezi, no. 8.
Keloglu, N., 1985, Mathematical simulation model with exponential deterministic
component for monthly river runoff receiving karst spring effluents. In: Karst Water
Resources (Proc. Ankara-Antalya Symp., July 1985), 135-139. IAHS Publ. no. 161.
Keloglu, N., 1987, Karst pinar katkih ayhk aki§larm iistel deterministik bile§enli
matematik benzetim modelled. Ankara, Devlet Su Isleri Teknik Btilteni, no. 57, 33-
43.
Olcay, M. S., 1987, Karst ortamin taskin zirve aki§lan dagihmi ûzerindeki etkileri.
Izmir, Dokuz Eylul Universitesi, Fen Bilimleri Enstitûsû, însaat Mûhendisligi
Hidrolik Ana Bilim Dab Yiiksek Lisans Tezi no. 17.
Olcay, S., Turkman, F. & Benzeden, E., 1987, Oymapinar Baraji ôrneginde ônemli karst
pinar katkih akarsularda çevirme sistemi kapasitesinin incelenmesi. Ankara, In§aat
Miihendisleri Odasi, Tûrkiye înçaat Mûhendisligi, IX Teknik Kongresi, C.II, 313-
319.
Ôzer, A., 1976, Yukan Dalaman akimlanmn stokastik analizi. Izmir, Ege Universitesi,
inçaat Mûhendisligi Bôlûmû, Hidroloji ve Su Yapilan Yiiksek Lisans Tezi no. 6.
Ozi§, U., 1971, Akislilik katsayilan vasitasiyla ortalama ve dû§ûk akimlann tayini,
gûvenilir su kuvveti potansiyelinin hesaplanmasi. Izmir, Tûrkiye Bilimsel ve Teknik
Ara§tirma Kurumu, MAG-206.
Ôzi§j Û . , 1974, Aki§lihk katsayilan vasitasiyla gûvenilir su kuvveti potansiyelinin tayini.
Izmir, Ege Universitesi Muhendislik Bilimleri Fakultesi, no. 1.
Ozi§, U . , 1979, Turkiye'de karst hidrolojisi çah§malanna toplubaki§. Istanbul, I. Ulusal
Hidroloji Kongresil, 49-63. (trans. N. Harrnancioglu) "An outlook to karst
hydrology studies in Turkey". Int. Karst Hydrogeology (ed. by G. Gûnay), 294-304.
Ôzi§, Û. & Ece, A. ; 1975, Indirektes matematisches Abflussmodell der karstischen
Sarikiz-quellen. Izmir, Ege Universitesi, Muhendislik Bilimleri Fakultesi Dergisi,
Y . 3 , no. 2, 1-30.
Ôzi§, Û . & Keloglu, N., 1976, Some features of mathematical analysis of karst runoff.
In: Karst Hydrology and Water Resources (Dubrovnik 1975, Proceedings) (ed. by
V. Yevjevich). 221-235. Water Resources Publications, Fort Collins.
Ôzi§, Û . & Benzeden, E., 1979, Theoretical considerations on simple linear correlation
222 Ûnal Ôzi§ et al.

applied to hydrologie processes. In: Modelling Hydrologie Processes (ed. by H.

Morel-Seytouxetal.). Third Int. Hydrol. Symp. 1977 Proc., Sec. 1-6, 85-99. Water
Resources Publications, Fort Collins.
Ôzi§, Û., Ûnal, E., Harmancioglu, N., & Benzeden, E., 1981, Underground storage
capacity of dams in karst regions, estimated through recession hydrographs. Bull.
Int. Ass. Engng. Geol., no. 24, 75-79.
Ozi§, 0 . & Yurtsever, Y., 1982, An outlook on tracer applications in karst hydrology
of Turkey. Beitr. Geol. der Schweiz- Hydrologie, 28, 149-160.
Ôzi§, U., Benzeden, E., Keloglu, N. & Harmancioglu, N., 1985, Assessment of karst
water resources. Vth World Congress on Water Resources Proceedings, vol. I, 95-
102. IWRA.
Ozi§; 0 . , 1989, Su kuvveti potansiyelinin geli§tirilmesinde karst ortamin etkileri.
Istanbul, "Miihendislik Jeolojisi Turk Milli Komitesi Bulteni" (1988 Erguvanh
Miihendislik Jeolojisi Sempozyumu Tebligleri), Y. II, n. II, 45-50.
Uslu, O., 1978, Karstik akimlarin parametrik çekirdek fonksiyonlu dogrusal sistemlerle
benzetimi. Izmir, Ege Ûniversitesi, In§aat Fakûltesi, Doçentlik Tezi.
Yevjevich, V., 1980, Investigations of karst hydrogeology, hydrology and water
resources in southern Turkey. Ankara, DSI, International Seminar on Karst
Hydrogeology, Oymapmar 1979, Proceedings (ed. by G. Giinay), 55-117.
Yevjevich, V., 1981, Karst waters of southern Turkey. Ankara DSI-UNDP Project. Final
Technical Report.
Yurtsever, Y., 1978, Tabii izotoplarm hidrolojide kullamlmasi esaslan ve Antalya civari
karstik bôlgede yeriistu-yeralti suyu ili çkilerinin tabii-izotoplarla araçtinlmasi
sonuçlan. Ankara, TBITAK, VI. Bilim Kongresi 1977, Miihendislik Arastirma Grubu
Tebligleri, no. 384/MAG-49, 1061-1078; Ankara, ElE, ElElBMteni, no. 75-76, 51-
64.
5 Modelling
Hydrogeologicd Processes in Karst Termites (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHSPubl. no. 207, 1993. ' 225

A RAINFALL-RUNOFF MODEL FOR LARGE KARSTIC AREAS

ALPARSLAN ARIKAN & LEVENT TEZCAN

International Research and Application Center for Karst Water Resources, Hacettepe University, 06532 Beytepe,
Ankara, Turkey

ABSTRACT

The enormous development in computer technology has made it possible to construct

more realistic physically-based mathematical models of hydrological processes. The
definition of the groundwater circulation system mainly based upon the spring
hydrographs in karst areas and, a rainfall-spring flow relationship can be used in
modelling the behaviour of karst aquifers in which the surface and subsurface catchment
boundaries do not coincide. In this study, surface and subsurface catchments are
simulated in two different catchment systems, one represents the surface catchment
system and simulates the peak flow, and the other simulates the contribution from
adjacent aquifer systems. Each catchment has two reservoirs in order to represent the
saturated and unsaturated zones. The parameters of the model are the catchment areas,
the minimum and maximum levels in the reservoirs, the initial heads and the discharge
coefficients that decrease exponentially with depth between maximum and minimum
values given as parameters. These parameters are adjusted by using generalized least
squares approximation in order to match the computed and observed flows, and applied
to the Manavgat River basin in southern Turkey. After calibration, a correlation
coefficient of 0.97 was obtained.

INTRODUCTION
The modelling of karst systems can be taken in the quantitative sense of reconstructing
the input-output of the system. The rapid and huge development of computer technology
allows the construction and use of more realistic mathematical models for catchment
processes.
The hydrograph of karst springs or streamflow in karst basins is the overall continuing
response of the karst system to the previous precipitation and evaporation events over the
catchment. In this study these events are modelled in two catchment systems, since the
surface catchment boundary does not coincide with the subsurface catchment boundary.
Thus, one of the catchment systems used in this model simulates the output of the surface
catchment, i.e. peak flow of the hydrograph, the other simulates the contribution from
adjacent catchments through the karstic features. Each system consists of two reservoir
systems (Fig. 1).

The first reservoir represents the behaviour of the unsaturated zone. In this zone, the rate
of infiltration is evaluated by taking into account soil moisture and evaporation and the
excess water transferred to the second reservoir system, representing the saturated zone.
In this zone, there may be a number of reservoirs connected with each other in a parallel
fashion or serially with different hydraulic characteristics. Reservoirs with rapid
discharge form peak flows, while those involving slower discharge rates contribute to the
baseflow. In this model, the saturated zone is handled as a single reservoir with variable
discharge coefficients decreases exponentially with depth. Thus, it can be thought that
this zone is simulated with an infinite number of reservoirs.

The parameters of the model are the areas of the catchments, the minimum levels of the
first reservoirs, the maximum levels of the second reservoirs, initial heads of both
reservoirs, the discharge coefficients of the first reservoirs and the maximum and
226 Alparslan Ankan & Lèvent Tezcan

minimum discharge coefficients of the second reservoirs. Model parameters are adjusted
during the calibration process in order to minimize the difference between observed and
calculated flows by using generalized least squares approximation. Calibration studies
improves the model quality. A better agreement with the observed flows is obtained than
the manual calibration of the system (Ankan, 1988).

STRUCTURE OF THE MODEL

The model simulates the response of the hydrological system to the precipitation
producing the streamflow hydrograph in karst basins. The hydrological system is
considered as two zones. At the top, the soil cover with vegetation constitutes the first
zone which is usually absent in karstic areas. The second zone is highly karstified and
the degree of karstification decreases with depth.

Due to the presence of intricate and unpredictable subsurface drainage systems, the
contribution from adjacent basins should also be considered in karst areas. In order to
simulate this contribution, a second catchment system with the same reservoir structure
connected hydraulically to the first is also included in the model (Fig. 1). The first
reservoir simulates the response of the system on effective precipitation taking into
account the real évapotranspiration. Due to the presence of karstic features, a direct
infiltration through the sinkholes and networks of joints and fissures happens instead of
overland flow.
(M D B) (ACS)
MAIN DRAINAGE BASIN SYS. ADJACENT CLOSED BASIN SYS

Ka(1)
Hm(1 Had)

Q1a

Km(2)
Ha(2)
Hm(2)

-Wi.l
QTm
7|vSI
'Trrr-TLm TLa

Symbols (m) and (a), which identify the main drainage basins system (MDB) and the
adjacent closed basins systems (ACB) respectively, are not considered within the
following abbreviations which have the same physical meaning for both systems.
P precipitation, at time T [mm];
ETR évapotranspiration from the first reservoirs [mm];
LMJN minimum water level required for flow to the second reservoirs [mm];
K(l) discharge coefficient of the first reservoirs;
Q(l) flow from the first reservoirs to the second reservoirs [mm];
H(2) head in the second reservoirs at time T [mm];
MAXL maximum level in the second reservoirs [mm];
KMIN discharge coefficient varying between KMAX and KMIN as a function of
H(2);
T lime [days];
QT calculated flow at time T [mm];
TL time lag between precipitation and the response of flow related to this event
[days];
QT + TL calculated flow taking tlte time lag into account [mm];
Q total outflow coming from bothsyslems calculated considering the areas of
the systems, converted into m s .
Fig. 1 - Structure of the model.
 A rainfall-runoff model for large karstic areas 227

The daily precipitation value is added to the head in the reservoir and real
évapotranspiration, calculated by Thornthwaite's method, is subtracted. If the water level
in this reservoir is below the soil moisture, the evaporation rate obtained by
Thornthwaite's method is adjusted by a factor calculated from an exponential function
of évapotranspiration with depth. This factor lies between 0.1 and 1, showing that
évapotranspiration continues under all conditions. The flow from the first reservoir to
the second zone is expressed as:

Qi = H, * K,

where H j is the head in the reservoir and Kj is the discharge coefficient of this zone. If
Kj is equal to 1, infiltration occurs directly through the sinkholes and all the excess water
of that day is delivered to the second reservoir. If Kj is less than 1, a portion of the
excess water is transferred gradually into the second reservoir and this causes a time lag
between the peaks of hyetograph and hydrograph.

The second reservoir simulates the karstified aquifer zone. The main hydraulic
characteristic of this zone is that the discharge coefficient decreases exponentially with
depth as the intensity of karstification. In previous studies, such as Bezes (1976), this
zone has been simulated with three or more discharge coefficients calculated from
recession curves by the Maillet equation. In this study, the discharge coefficient is
expressed with an exponential function of head in the reservoir:

K 2 = e x p [ - ( A * H 2 + B)]

where K 2 is the discharge coefficient in the second reservoir, A and B are constants
calculated from the boundary conditions, that, when the head is equal to the maximum
level (MAXL) in the reservoir, K 2 is equal to KMAX and when the head is zero it is set
to K M I N . Thus:

K2 = exp[(ln(KMIN) - ln(KMAX))/MAXL * H 2 + In (KMIN)]

The head in the second reservoir is calculated as the sum of the previous day's head and
flow coming from first reservoir (Qj). Evapotranspiration is ignored in this zone. Daily
discharge from this zone is calculated as follows:

Q 2 = H 2 * K2

The total flow computed from both catchment systems may be expressed as:

Q = Q s + Qa

where Q s is the daily discharge computed for the surface catchment system and Q a is that
for the adjacent catchment system.

CALIBRATION OF THE MODEL

The calibration of the model is the process by which model parameters are adjusted so
as to minimize the difference between simulated and observed discharges. However, it
should b e noted that the obtained parameters should be physically realistic. The more
nearly deterministic the model, the more physical meaning we may be able to give the
parameters.

The calibration could be performed manually or automatically. Manual calibration is the

process by which parameters are adjusted one at a time with a model run, and user
228 Alparslan Ankan & Lèvent Tezcan

examination of the results between each adjustment allows the user to control the physical
meaning of the parameters. Nevertheless, the goal of this calibration is to minimize the
same objective function that could be minimized automatically, then there is little
difference in principle between the two methods, although the manual calibration would
always be much less efficient in finding the optimum fit (Troutman, 1985).

The objective function used in the automatic calibration of the model parameters is:
E = S[QOBS - QSIM(KMAXS) KMIN5, MAXLS, KMAXa, KMINa) MAXLa, K l „ Kl a )]

where (s) stands for the surface catchment system and (a) for the adjacent catchments
system. The minimization of this function is performed by the algorithm of generalized
least squares technique developed by Lybenon (1985). The generalized least squares
technique is a weighted form of ordinary least squares which can eliminate the effect of
serially correlated errors (Williams & Yeh, 1983). The parameters calibrated in the
model are maximum levels in the second reservoirs and the discharge coefficients of the
first reservoirs of both catchments.

CASE STUDY
The model is applied to the Manavgat River basin (Fig. 2) in southern Turkey. This basin
is highly karstified and detailed geological and hydrogeological studies have been
performed during the last 20 years mostly because of the construction of the Oymapinar
and Manavgat dams. Based on the results of previous investigations, it is understood that
the water drains into the Manavgat River not only from its surface catchment, but also
from the closed surface basins to the northwest and especially to the east of the Manavgat
River (Yurtsever, 1979).

Fig. 2 - Location map of the study area.

 A rainfall-runoff model for large karstic areas 229

The surface catchment area of the Manavgat River at the Homa gauging station is about
900 km 2 . The recorded monthly and annual discharges are high for such a small drainage
area, and thus a considerable subsurface contribution must be coming from adjacent
basins (Gunay, 1981). It has been proven by repeated dye tests (Gûnay et al., 1978) that
the water of the Eynif Polje supports several springs observed along the Manavgat River
(Fig. 2). Various lithological units such as the Paleozoic flysch and schists, Eocene
flysch and Miocene molasses generally trend northwest-southeast, thus forming regional
or local hydrogeological barriers to the groundwater flow (Eroskay et al., 1986).

Precipitation data were obtained from the records of the three stations of Aydmkent,
Cevizli and Akseki. Previous multiple correlation analysis between precipitation and
Homa flows has shown that precipitation recorded at the Aydinkent station influences the
flow with a time lag of zero and two months, that is, precipitation that occurred in the
ith and the (i — 2)nd months influences the flow recorded in the ith month. This
conclusion shows that the precipitation occurring throughout the catchment area affects
the flow at the Homa gauging station within two months depending on the distance
between the stations. This effect shows itself as a piston flow.

The period covering the water years 1971-1977 was chosen for simulation. The data for
this period are in good accord with the general trend of the data for the long-term period.
Three precipitation stations are used to estimate the daily precipitation used as the input
for the model. However, all the stations are at lower altitudes so the precipitation data
do not represent the entire area. After calibration of the model parameters a good fit is
obtained between simulated and observed flows with a correlation coefficient of 0.97
which is more than satisfactory for karst areas. The fit can be seen on Fig. 3.

700
FLOW O OBSERVED
O CALCULATED
600

500

400

300

200

100

0
250 500 750 1000 1250 1500 1750 2000
DA'"
Fig. 3 - Observed and calculated hydrographs.

CONCLUSIONS
The model presented here has contributed valuable information to the previous
investigations on the determination of the dynamic structure of the karst groundwater
230 Alparslan Arikan & Lèvent Tezcan

circulation which will provide a good basis to the planners of hydrotechnical

constructions in karst regions. The main difference of this model from existing ones is
the derivation of the discharge coefficient as a function of depth which is physically more
meaningful in karst groundwater systems.

As in all models, there are deviations between the computed and observed values which
may be contributed to the assumptions made in the model and also to the errors involved
in the observed data. However, in this model, the calculated and observed hydrographs
show a good agreement. A correlation coefficient of 0.97 was obtained between the
observed and calculated discharges for the whole simulation period. The accuracy of the
model is improved by using automatic calibration of model parameters affecting the shape
of the hydrograph namely discharge coefficients of the reservoirs. The generalized least
squares technique was applied in order to minimize the differences between observed and
calculated data. A very good agreement was obtained between low-rate flows, but
deviations from peak flows cannot be prevented. The causes of these deviations could be
measurement errors or snowmelt contributions. Moreover, the number of the precipitation
stations are not satisfactory for entire representation of the precipitation distribution over
the area. Some new interpolation techniques such as kriging or co-kriging may improve
the model accuracy.

REFERENCES
Arikan, A., 1988, Modalp. A deterministic rainfall-runoff model for large karstic areas.
Hydrol. Sci. Bull., 33 (4), 401-414.
Bezes, C , 1976, Contribution à la modélisation des systèmes aquifères karstiques:
Etablissement du modèle Berner, son application à quatre systèmes karstiques du midi
de la France. CERGA, Montpellier, France.
Eroskay, S. O., Gunay, G. & Arikan, A., 1986, Effects of Oymapinar dam in the karst
medium. TBTAK TBAG-625 Project (unpublished and in Turkish).
Gûnay, G., 1981, Investigations of karst hydrology of the Manavgat basin and its close
vicinity (in Turkish). Affiliation thesis, Hacettepe University, Ankara.
Gûnay, G., Ataman, Y. & Dogan, L., 1978, Oymapinar project Asagi Eynif Polje Dye
Test: Final Report. Tech. Report, no. 17 of UNDP/DSI Project, Ankara.
Lybenon, M., 1985, A simple generalized least squares algorithm. Computers and
Geosciences, 11 (4), 501-508.
Sorooshian, S. & Gupta, V. K., 1985, The automatic calibration of conceptual catchment
models using derivative-based optimization algorithms. Wat. Resour. Res., 21 (4),
473.
Troutman, B. M., 1985, Errors and parameter estimation in precipitation runoff
modelling: 1. Theory. Wat. Resour. Res., 21 (8), 1195-1213.
Williams, B. J. & Yeh, W. W. G., 1983, Parameter estimation in rainfall-runoff models.
/. Hydrol., 63, 373-393.
Yurtsever, Y., 1979, Environmental isotopes as a tool in hydrogeological investigations
of southern karst regions of Turkey. Proc. International Seminar on Karst
Hydrogeology (Oymapinar-Antalya).
Hydrogeologicai Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHS Publ. no. 207, 1993. 231

ASSESSMENT OF MATHEMATICAL MODELS WITH EXPONENTIAL

FUNCTIONS DESCRIBING KARSTIC SPRING DISCHARGES

TURKAY BARAN & NÏLGÛN B. HARMANCIOÔLU

Dokuz Eyliil University, Faculty of Engineering and Architecture, Civil Engineering Department, Bornova
35100, Izmir, Turkey

ABSTRACT
The recession period streamflows of the Manavgat River, where karstic spring flows
contribute significantly to surface runoff, have been described in previous studies
basically by three mathematical models: simple exponential functions (single slope
recession curves), composite exponential functions with piece-wise components (multiple
slope recession curves), and continuous exponential functions to approximate the whole
period. The problem is how to select among these available models to better describe the
recession flows at different points along the river. A criterion based on the variance of
standard errors have been used earlier to test the goodness of fit of the three functional
approaches. The presented study compares these models on the basis of their
informativeness, using the entropy principle as a measure of the rate of information
conveyed by a particular model. Each model is assessed by defining the amount of
information it reveals about the true process. This is realized by computing the
information common (transinformation) to both the observed flows and the concurrent
model outputs. Then, the available models are ranked for performance in order of
decreasing transinformation with the best model showing the highest common
information.

INTRODUCTION
Among the various hydrological techniques used in estimation of karst runoff, the most
frequently applied method is recession hydrograph analysis. This method is based on the
assumption that the recession part of the runoff hydrograph represents the dry season
discharge of the karst aquifer, when the input precipitation ceases so that the continuous
discharge of the karstic system is observed as spring flows. Recession curves have often
been approximated by mathematical models in the form of exponential functions. These
models have been developed to varying degrees of complexity as required by particular
basin characteristics. In certain cases, simple exponential functions, as proposed first by
Maillet (1905), are found sufficient; whereas, in others, the combination of a number of
exponential functions with different slopes are selected to give a better description of the
recession curve.

The recession period streamflows of the Manavgat River, where karstic springs contribute
significantly to surface runoff have been modelled in previous studies by various
applications of recession hydrograph analysis. The method was used both with daily flows
(Karanjac, 1978; Karanjac & Altug, 1978, 1980) and monthly flows (Unal, 1980, 1981;
Ôzis. et al., 1980, 1981; Benzeden & Tatlioglu, 1985). A fairly detailed model (Keloglu,
1985) comprises reciprocal exponential functions, with a decreasing one for the dry
period and an increasing one for the wet period to describe the karst related part of the
deterministic component of river runoff, while the surface runoff portion is represented
by periodic functions in the wet period.

A recent study by Baran (1989) uses three types of recession curves to describe the dry
period flows at the Homa gauging station: (a) simple exponential functions (single slope
recession curves); (b) composite exponential functions with piece-wise components
(multiple slope recession curves); (c) continuous exponential functions to approximate
232 Turkay Baran & Nilgun B. Harmancioglu

the whole period. Each of these models has its merits with respect to representativeness
and easiness of application. The question is how to select the model which will best
describe the recession flows at Homa.

The presented study compares the three approaches used by Baran (1989), rating them
on the basis of their informativeness (or representativeness). The entropy principle is
used as a measure of the rate of information conveyed by a particular model. Each model
is assessed by defining the amount of information it reveals about the true process (or by
the amount of reduction of uncertainty in observed flows). This is realized by computing
the information common (transinformation) to the observed flows and the concurrent
model outputs. Then, the three models are ranked for performance in order of decreasing
transinformation with the best model showing the highest common information.

In essence, the study was carried out firstly to assess the goodness of fit of conventional
mathematical models with exponential functions in representing the recession period
streamflows; and secondly to test the applicability of the entropy concept in evaluating
the fitness of a model to observed phenomena.

APPLIED METHOD
Available models to represent the recession flow at Homa

At a point on a river fed by karstic spring discharges, the recession limb of the discharge
hydrograph is defined mathematically by Maillet (1905) and Tison (1953) to describe the
karstic aquifer discharge in the dry season, when the input precipitation is negligible.
Later, Meier (1980) has shown that these definitions are valid under certain assumptions
for linear underground reservoirs. Such mathematical expressions use the aquifer
discharge coefficient (a) as the basic parameter to represent the discharge of the karstic
aquifer.

Baran (1989) used similar models to investigate the dry period runoff observed at the
Homa and Sinanhoca streamgauging stations in the Manavgat basin. She developed three
types of functional expressions for the recession hydrographs of both stations and the
intermediate basin.

The first expression is the conventional simple exponential function which defines a
single slope recession curve:
Q(t) = Q0 e - ' (1)
Next, a composite exponential function with two discrete components is used to describe
the recession curves with two different slopes:

Q(t) = Q 0i e- a,t (t < t,) (2)

1
Q(t) = Q ^ (t > t,) (3)
Finally, a continuous exponential function of the form:

Q(t) = Q 0 2 e^ 1 = QoV"'"' (4)

is employed to approximate the total recession period. This last function preserves the
parameter for the second component of the piece-wise approximation given in (3) and
extends it in a continuous form to cover all time points t. In the above expressions, Q 0
(m3/s) represents the discharge at time t (day) when the recession starts, or is the aquifer
 Models with exponential functions describing karstic spring discharges 233

discharge coefficient in (1/day). Q(t), then, describes the spring discharge (m3/s) at any
day t. Figure 1 shows these three basic features, i.e. a, Q 0 , and Q(t) of the recession
curve.

When the recession curve described by the simple exponential function is plotted on semi-
logarithmic paper (Fig. 2), the aquifer discharge coefficient (a) is obtained as the slope
of the recession hydrograph:
a = tan 6» = [lnQ0-lnQ(t)]/(t-t^) (5)
Baran (1989) compared the three models using a criterion based on the variance of
errors. This criterion defines the sum of squares of errors as:

SSE = £ {Q(t) - Q(f)}2 (6)

where Q(t) describes the observed and Q(t) the estimated discharge at any time t. The
variance of errors is then:

s02 = SSFV[(T - g - k] (7)

where k represents the number of parameters specified for a model, and T (days), the
total length of the recession period. The application of this measure has shown in Baran's
study (1989) that the composite exponential function with piece-wise components gives
better estimates than the other two.

o t0 t

Fig. 1 The general form of the recession hydrograph.

ilnft(t)

In So

^ « = tg Q

Fig. 2 The effect of logarithmic transformation on the recession limb of

a hydrograph.
234 Turkay Baran & Nilgun B. Harmancioglu

Application of the entropy criterion in assessing models

The entropy concept can be applied to derive an objective criterion based on marginal
entropy, conditional entropy and transinformation to assess the goodness of fit of
mathematical models. Amorocho & Espildora (1973) initiated such an approach and
showed the limitations and the merits of using the entropy criterion in model evaluation.
They used the concept to assess the uncertainty of the Stanford Watershed model in
simulating streamflows from a basin in California for which historical records were
available. Later, Chapman (1986) extended the original use of the method by Amorocho
& Espildora to evaluate the reduction of uncertainty in hydrological data due to the
application of a model. He proposed a complementary approach to overcome the
limitations of the technique.
The entropy (uncertainty) of a discrete random variable X with N elementary events of
probabilities p n = p(xn), n = 1,..., N is defined in information theory as:

H(X) = -K I P(xn) log p(xn) (8)

with K = 1 if H(X) is expressed in napiers for logarithms to the base e. The entropy of
two independent variables X and Y is:
H(X,Y) = H(X) + H(Y) (9)
and of two dependent variables is:
N M
H(X,Y) = - K £ £ p(x n ,yj log p(x n ,yj (10)
n= 1 m= 1

with p(x n , ym) describing the joint probabilities of X and Y. The difference between (9)
and (10) gives:
T(X,Y) = H(X) + H(Y) - H(X,Y) (n)
which is defined as transinformation or information contained in both X and Y. If X and
Y are independent T(X,Y) = 0.
For continuous density functions, the discrete probabilities of p(x n ) are approximated by
f(x)ôx for small ôx, and the summations in (8) and (10) are replaced by integrals. In this
case, f(x) and f(x,y) represent continuous probability density functions. When
assumptions are made as to the particular type of function for f(x) and f(x,y), the
integrals are solved to obtain simplified expressions for entropy.

If f(x) and f(x,y) are selected to be normal distributions, with means nx, n standard
deviations <rx, ay, and the correlation coefficient p x y , equations (8) through (11) reduce
(Harmancioglu, 1981) to:
H(X;ôx) = ln(2ïï e<rx)/2 + ln(l/ôx) ( 12 )

H(X,Y;ÔX) = l/2 + ln[(2n e<rx)(l-pxy)]/2 + ln(l/ôx) (13)

T(X,Y) = -[ln(l-p xy )]/2 (W)

Similarly, if the choice is for a lognormal density function, the X and Y processes are
transformed by:
z = log x
(15)
w = log y
 Models with exponential functions describing karstic spring discharges 235

so that the variables z and w are normally distributed, with means \xz, jtw, standard
deviations az, a^,, and the correlation coefficient p2W. Then, equations (8) through (11)
reduce (Amorocho & Espildora, 1973; Chapman, 1986) to:
H(X;Ôx) = Mz + ln[(2nea^)/2 + ln(l/Ôx) ( 16 )

H(X, Y; ox) = Mz + ln[(2IIe a2x)(l - p2zv,W2 + ln(l/ôx) ( 17 )

T(X,Y) = -[ln(l- P L)]/2) ( 18 )

In the above expressions, the definition of entropies in the case of continuous probability
density functions involves the discretization interval ôx, which is a highly crucial value
in determining entropy. The condition p(xn) ~ f(x)ôx is valid only for sufficiently small
values of ôx; whereas, such a choice increases the required computer time and load
significantly. On the other hand, a relatively large ôx interval tends to reduce the
uncertainty, and hence the entropy of both the observed and predicted variable values.
A small ôx value presents a stricter test in evaluating what is uncertain and what is not
(Amorocho & Espildora, 1973). Chapman (1986) proposed the use of a class interval ôx
which is proportional to variable values to assess the relevant uncertainties more
realistically. Such an approach gives better credit to high and low levels of either
observed or predicted flows. Thus, the last term ln(ôx) in equations (12) through (18) is
replaced by ln(ôx/x) in Chapman's formulation. In this study, the computations are
carried out for both fixed ôx class intervals and proportional (varying) intervals in the
form of ôx/x where x represents the variable values.

ASSESSMENT OF RECESSION MODELS FOR HOMA BY ENTROPY MEASURES

The data set used in the study comprise daily streamflows at the Homa streamgauging
station observed between 1964 and 1982. The recession period of 153 days at Homa is
specified by Baran (1989) to extend from 1 June of each year to 31 October. The first
step of the analysis was to compute model outputs by using the parameters of the three
models, determined earlier by Baran for each year of the observation period. Figure 3
shows the observed daily means of recession flows versus those predicted by each of the
three models. Figure 4 gives similar curves for daily variances.

It may be observed from these figures that daily means are sufficiently well reproduced
by all models, except for about 20 days at the end of the recession period. However, each
model appears to be less successful in representing the variance characteristics of the
original data, especially at the beginning and the end of the recession period. The same
is true for skewness and kurtosis parameters not shown here.
Next, marginal and conditional entropies (uncertainties) of both the observed flows (Y)
and the predicted model outputs (X) are computed for 153 days of the recession period,
where each day is represented by 19 variable values for the observation period between
1964 and 1982. These computations eventually produce the variation of entropies
throughout the recession period.

The information common to both the model and the observed values, i.e. T(X,Y), is also
computed for each day to show the goodness of fit of each model in representing the
observed flows within the total period.
Figure 5 presents the daily marginal entropies H(Y) of the observed data for fixed class
intervals ôx of 1 m3/s and 2 m Vs. It may be observed from the figure that a change in
fixed ôx values does not alter the variation of H(Y); however, it affects the scale of
values so that a larger ôx results in smaller entropies, and vice versa. The same figure
 Tûrkay Baran & NilgUn B. Harmancwglu
~150 \
1/1 \
\
L\
£.'40

130 — Observed
••• Simple Exp
•— Composite Exp.
120 — Continuous

110

100

90

80

70

40,
50 100 150
t (day)
Fig. 3 The observed mean flows at Homa during the recession period,
together with those predicted by three types of models.

8 2
«„I8

Observed
Simple Exp.
Composite Exp
Continuous

100 150
t (day)

Fig. 4 The variances of observed and predicted flows at Homa during the
recession period.
 Models with exponential functions describing karstic spring discharges 237

3 _ Continuous

'o 50 100 150

t ( day )

Fig. 5 Marginal entropies H(Y) of observed flows for two class intervals
èx of 1 nr/s (lower solid line) and 2 m 3 /s (upper solid line), shown
together with reduction in uncertainties, H(X/Y), due to the application
of three types of models.

shows the amount of uncertainty still remaining in the recession process, H(Y/X), after
the three types of models are applied.

In all models, the uncertainties of flows approximately at the middle and the end
(between t = 120 and 145) of the recession period are more pronounced than in the case
of H(Y). The continuous exponential approximation shows this effect more clearly than
the other two which basically give similar curves.

The same situation is observed in Fig. 6, which describes T(X,Y) with respect to the
three types of models. According to this figure, the simple and composite exponential
functions seem to fit the observed recession process well, approximately between the 10th
and the 130th days. In all these computations, the normal probability density function is
assumed.

Next, the same application is realized by selecting ôx intervals proportional to flows in

the order of ôx/x = 0.05. Figures 7 and 8 present the results for normal distribution.
Proportional class intervals give credit to the variation of flows, so that the marginal
entropy of the process is reduced in comparison to that for fixed class intervals. In fact,
H(Y) appears to the constant at 3.0 ~ 3.2 throughout the recession period, excluding the
last 20-25 days. The models reduce this uncertainty to around 2.0 ~ 2.2, while failing
to handle it at the end of the recession period.

The transinformation in Fig. 8 confirms this result, further emphasizing the effectiveness
of the first two models for a longer duration than that shown by the continuous
approximation.
 Tiirkay Baran & Nilgun B. Harmancwglu
v. 1.2

/ ^ V ^ ^ ' ^ ' T ^ ' ^ fy <Cr

7r-
^ /
v
*r''»L

5
V
S
S
1
1
1

1
0.6

i
Simple Exp. 1
Composite Exp

0.2 - -
Continuous
i
$

50
u 150
V 1
t (day)
Fig. 6 Transinformation between observed flows and model outputs for
fixed <5x of 1 m 3 /s.

Observed
Simple Exp
Composite Exp
Continuous

3.5

1.5
100 150
t (day)
Fig. 7 Marginal entropies H(Y) of observed flows (solid line) and
conditional entropies H(Y/X) for three types of models, computed by the
assumptions: (a) proportional <5x/x = 0.05 class intervals; (b) normal
probability distribution function.
 Models with exponential functions describing karstic spring discharges 239

0.6

\\\
Simple Exp. •! J |
Composite Exp. jj Î
Continuous i) |

0.2

. , Ld
0 50 100 150
t(day)
Fig. 8 Transinformation between observed and predicted flows for ôx/x
= 0.05, assuming normal distribution.

Figures 9 and 10 present the results of similar computations for lognormal distribution
with ôx/x = 0.05. Two basic improvements are observed in this case. First the marginal
entropies H(Y) are further reduced, particularly at the beginning and the end of the
recession period. This indicates that the lognormal distribution gives a better fit to the
observed data, again especially for the early and late stages of recession. Second, the
variation in all entropies are more smoothed out than in the case of normal distribution.
On the other hand, the level of transinformation remains approximately the same as that
computed for normal distribution, although the smoothing effect of lognormal
distributions is still valid.

RESULTS
Figures 5-10 reveal two basic features with respect to the three types of models applied:
(a) the simple and composite exponential functions fit the observed data much better than
the continuous approximation; in essence, the two models give similar results so that, for
the sake of simplicity, one may be content with the simple exponential function; b) all
models seem to perform well during a large portion of the recession period; however,
they deviate significantly from the observed values for 10 to 20 days in the beginning and
20 to 30 days at the end of the recession period.

This may require a re-evaluation of the starting and ending dates of the recession period
at Homa. This means that the three models may be appropriately used to describe the
recession, but beyond this period, some other approaches will be needed to represent the
streamflows.
With respect to the application of entropy measures, two points are noteworthy. First,
the selection of proportional class intervals in the form of ôx/x gives more realistic
 Turkay Baran & Nilgun B. Harmancioglu
.2.5

Observed
Simple Exp
Composite Exp
Continuous

1.5

i
:C*A*0v:v
i-^t**J Nt^y^.-]
<CB ^*
0.5

50 150
t(day)
Fig. 9 Marginal entropies H(Y) of observed flows (solid line) and
conditional entropies H(Y/X) for three types of models, using ôx /x =
0.05 and lognormal probability distribution function.

>. 1.2

?K?0

i
0.6

0.4
•• •
—
—
Simple Exp
Composite Exp
Continuous I
0.2

;A/

50 100 150
t (day)
Fig. 10 Transinformation between observed and predicted flows for ôx/x
= 0.05, assuming lognormal distribution.
 Models with exponential functions describing karstic spring discharges 241

results in representing the uncertainty of variable values. Second, the use of the
lognormal distribution function produces greater reduction in the uncertainty of observed
flows. This distribution also smooths the highly variable flows and their consequent
uncertainties at the beginning and the end of the recession period, although such
uncertainties are not significantly reduced.
The above conclusions are drawn on the basis of the variation of marginal entropies,
conditional entropies, and transinformation. One may also compare the three models by
using the mean values of these entropy measures. To this end, three approaches can be
used: (a) comparison of mean marginal entropies of both the observed data, H(Y), and
the model outputs, H(X); (b) comparison of mean transinformation T(X,Y); (c)
comparison of T(X, Y)/H(Y) ratios which represent the percentage of the total uncertainty
of the process that is reduced due to the use of a particular model.

Table 1 shows these comparisons for normal and lognormal probability distribution
functions. According to the figures in Table 1, all three approaches used in comparing
the models indicate that the composite exponential function gives the best fit to observed
data, followed by the simple exponential, and then by the continuous approximation. This
is also the result obtained by the variance of standard error criterion in Baran's (1989)
study. Although the general result is the same, a major difference exists between the
variance of standard error and the entropy criteria. The former has to be evaluated for
each year of the historical record to arrive at an overall solution.

Table 1. Comparison of the three models by mean values of the entropy measures (in
napiers).
NORMAL PDF LOGNORMAL PDF
Model type H(Y) H(X) T(X,Y) T(X,Y) H(Y) H(X) T(X,Y) T(X,Y)
/H(Y) /H(Y)
Simple 3.123 3.081 0.985 0.315 1.781 1.754 1.005 0.564
exponential
functions
Composite 3.123 3.082 1.004 0.321 1.781 1.754 1.024 0.575
exponential
functions
Continuous 3.123 3.085 0.960 0.307 1.781 1.751 0.993 0.557
exponential
functions

Furthermore, Baran (1989) points out that difficulties are encountered in the numerical
computation of this criterion, so that certain data values had to be excluded from the
analysis to arrive at a reasonable value. In some years, such requirement has significantly
reduced the number of usable data. On the other hand, the entropy measures produce the
expected overall result in one step, using all the available data. In addition, they permit
the evaluation of model performance at different times along the recession curve. That
is, one can observe the variation of both the uncertainties in the original data and the
amounts of information produced by the models to reduce these uncertainties.
Consequently, it is possible to conclude as to when a model is applicable and when it is
not within the recession period.

Despite all these advantages of the entropy measures, one major difficulty is that
marginal and conditional entropies assume values that are relative to the axes. Only their
differences or transinformation have absolute values. This means that transinformation
should be preferred to the other two in evaluating model performance. If marginal and
conditional entropies are to be used for assessment, then their variation throughout the
242 Tûrkay Baran & Nilgtin B. Harmancioglu

recession period gives a better indication of model performance than their mean values
as given in Table 1. This particular feature of the entropy measures deserves further
investigation and provides the impetus for future studies on the subject.

CONCLUSION
Entropy-based measures are used to assess the performance of recession hydrograph
models with exponential functions. The application of the method is demonstrated in the
case of Homa recession flows. Consequently, the previously developed models are ranked
in order of decreasing fitness as composite exponential, simple exponential and
continuous approximation.

The study shows the applicability of entropy measures in assessing model performance,
pointing out to their advantages and limitations. The authors believe that these limitations
require further consideration so that entropy measures can be effectively and widely used
in practice.

ACKNOWLEDGEMENT
The authors gratefully acknowledge the support of Serpil Baran and Prof. Ertugrul
Benzeden for providing the required data and all the relevant information about their
previously developed models.

REFERENCES
Amorocho, J. & Espildora, B., 1973, Entropy in the assessment of uncertainty in
hydrologie systems and models. Wat. Resour. Res., 9 (6), 1511-1522.
Baran, S., 1989, Analysis of dry period runoff in karstic basins for the case of Manavgat
(in Turkish). Dokuz Eyliil University, Institute of Technical Sciences, Civil
Engineering Department, Izmir, MSc Thesis in Hydrology and Hydraulic Structures,
no. 22.
Benzeden, E. & Tathoglu, E., 1985, Hydraulicity and aquifer discharge coefficient in
river basins with significant karst spring effluents. In: Karst Water Resources (Proc.
Ankara-Antalya Symp., July 1985) 77-86. IAHS Publ. no. 161.
Chapman, T. G., 1986, Entropy as a measure of hydrologie data uncertainty and model
performance. J. Hydrol., 85, 111-126.
Harmancioglu, N., 1981, Measuring the information content of hydrological processes
by the entropy concept. Izmir, Ege University, Journal of Civil Engineering Faculty,
Special issue on the Centennial of Ataturk's birth, 13-40.
Karanjac, J., 1978, Oymapinar dam downstream springs. Ankara, DSÏ-VNDP Project,
Technical Report, no. 23.
Karanjac, J. & Altug, A., 1978, Oymapinar dam downstream springs. Hydrogeology and
hydraulics. Ankara, DSl-UNDP Project, Technical Report, no. 23.
Karanjac, J. & Altug, A., 1980, Karstic spring recession hydrograph and water
temperature analysis: Oymapinar dam project, Turkey. /. Hydrol., 45, 203-217.
Keloglu, N., 1985, Mathematical simulation model with exponential deterministic
component for monthly river runoff receiving karst spring effluents. In: Karst Water
Resources (Proc. Ankara-Antalya Symp., July 1985), 135-139. IAHS Publ. no. 161.
Maillet, E., 1905, Essais d'Hydraulique Fluviale. Paris.
Meier, R., 1980, Analyse und vorhersage von trockenwetterabflûssen eine anwendung
der systemhydrologie. Munchen, Versushsanstalt fiir Wasserbau der Technischen
Universitdt Mûnchen, Bericht Nr. 43.
Ôzi§, Û., Benzeden, E., Harmancioglu, N. & Ûnal, E., 1980, Contribution of soluble
 Models with exponential functions describing karstic spring discharges 243

rocks to effective storage capacity of reservoirs (in Turkish). Istanbul, National

Turkish Committee of International Association of Engineering Geology, 2nd
Symposium on Engineering Geology (preprint).
Ôzi§, Û . , Ûnal, E . , Harmancioglu, N. & Benzeden, E . , 1981, Underground storage
capacity of dams in karst regions, estimated through recession hydrographs. Aachen-
Essen, Bulletin of the International Association of Engineering Geology, no. 24,
75-79.
Tison, L. J., 1953, Cours d'hydraulique, I, Cours d'eau en mouvement permanent.
filtration. Gand.
Ûnal, E . , 1980, Analysis of the recession hydrograph at a river point fed by karstic
aquifer discharges: Homa/Manavgat (in Turkish). Izmir Ege University, Civil
Engineering Department, Applications in Karst Water Resources (1979/80), pp. 3c-l
to 3c-7.
Ûnal, E . , 1981, Analysis of the underground storage capacity of reservoirs in karstic
basins, using the recession hydrograph method (in Turkish). Izmir, Ege University,
MSc Thesis in Hydrology and Hydraulic Structures, no. 13.
Hydrogeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHSPubI.no. 207, 1993. 245

A CONCEPTUAL MODEL FOR THE BEY§EHIR LAKE KARST SYSTEM

MEHMET EKMEKÇI
International Research and Application Center for Karst Water Resources, Hacettepe University, 06532 Beytepe,
Ankara, Turkey

ABSTRACT
Beysehir Lake is situated in the Western Taurus karst region where there are many
hydrotechnical projects. A statistical approach was employed to analyse the
hydrometeorological data to investigate the interactions between Bey§ehir Lake and the
flow occurring in the adjacent basins. An accurate water budget computation was made
on a monthly basis over a period of 23 years in order to obtain an idea of the amount of
leakage from the lake through karstic sinkholes and channels. Results revealed that this
leakage is of the order of 5 m 3 /s, reaching up to 29 m3/s as the water level of the lake
rises during wet periods. Sinkholes were also found to function as springs during the wet
periods causing a flood. The relation between the water level of the lake and the amount
of leakage suggests that leakage increases significantly as the water level rises and no
leakage occurs when the water level decreases to a critical level which corresponds to the
zone of fluctuation where the sinkholes were concentrated. The apparent effect of leakage
was found to be very small due to the propagation effect within the huge karst aquifer
underlying the area between the lake and the Mediterranean Sea. Based on these results,
a conceptual model was developed to explain the Beysehir Lake karst system and its
effects on the major karstic springs in the adjacent basins.

INTRODUCTION
The Taurus karst region where the Beysehir Lake is situated, is of great importance from
the standpoint of water resources potential as well as the interesting karstic features
found on a large scale (Fig. 1).

The role of the lake has long been investigated by researchers using different techniques.
Two opposite views proposed as a result of these studies have not yet been verified. One
hypothesis suggests that the lake leaks a significant amount of water through sinkholes,
particularly through those located at the bottom of the lake which are a source for the
huge karst springs downstream in the Manavgat River and Koprucay basins. The other
hypothesis states that no significant leakage occurs from the lake but the main source for
the excess flow of the springs might be the northeast part of the region where Sugla Lake
is located (Fig. 1).
In this study, an attempt is made first to calculate the amount of leakage from the lake,
if any; secondly, to determine the impacts of the leakage on the discharge of the karst
springs in the adjacent basins; and finally, to propose a conceptual model to explain the
probable karst water system of the region based on the interpretation of the figures
obtained from calculations together with the field observations and the results of the
analyses of the hydrogeological structure in the region.

To achieve this aim, the water balance of the lake was calculated using accurately
determined data on a monthly basis. Results of the computations were then used in a
series of statistical analyses to establish the relationships between the water level in the
lake and the amount of leakage, and between the amount of leakage and the discharge of
the springs.
 Mehmet Ekmekçi

!
5 O 5 10 15 20 25 Km
I i i i i • '

SCALE : 1/500.000

HYDROGEOLOGICAL SETTINGS
The study area is located in the central part of the southern Taurus karst region where
the geology and subsequently the hydrogeology are very complex due mainly to the
tectonic movements that have taken place during almost all geological times until today.
Overthrusts of the ophiolithic nappes are the elements that control the hydrogeology of
the region. As hydrogeological units, the carbonate rocks of Mesozoic and Paleozoic age
are pervious and of these, the Jurassic-Cretaceous limestones and dolomitic limestones
compose the major karst aquifers.

Recession curve analyses of the huge karst springs which issue from these aquifers
suggest that the volume of the aquifers must be very large, with an extension in the order
of tens of kilometres (Giinay, 1981). Tertiary flysch and ophiolithic units are the
impervious lithologies in the region (Fig. 2).

HYDROMETEOROLOGICAL DATA ANALYSIS

A standard base period of 23 years for hydrometeorological data was selected from the
 A conceptual model for the Beysehir Lake karst system 247

_«<£ River
SCALE' 1/500000 J^ Formation Bounty
J O I 2 Km
' ' 1 1

Fig. 2 - Hydrogeological map of the study area.

hydrometeorological stations located in the study area. The areal mean precipitation and
evaporation were found by the polygon method. Discharge and water level data were
extracted from discharge data books published by the Electrical Power Resources Survey
and Development Administration and the State Hydraulic Works. Statistical analyses of
the data revealed that only a weak correlation exists between the water level in Bey §ehir
Lake and discharges of the Manavgat and Koprucay River basins.

The correlogram in Fig. 3 shows that the relation between the water level in the lake and
the discharge increases considerably during certain periods which correspond to rainy
seasons as indicated also by the graph of deviation from cumulative precipitation. This
may suggest that the aquifer lying between the Beysehir Lake and the basins to the south
is so large that it is affected by the cumulative precipitation but not by the individual
events. This confirms the conclusions revealed by the isotope studies stating that, on
average, the travel time of the karst springs is about 15 years (Yurtsever, 1979; Gunay
& Sipahi, 1979). Detailed results of the statistical analyses of the hydrometeorological
data can be obtained from Ekmekçi (1987).
248 Mehmet Ekmekçi

AMOUNT OF LEAKAGE FROM THE LAKE

The existence and amount of leakage from the Beysehir Lake were based on an accurately
computed water balance of the lake. Direct precipitation on the lake surface, the surface
and subsurface runoff from the drainage basin and groundwater flow from the adjacent
basins cause a rise in the water level of the lake and they are considered as the input
components of the system. Outflow from the lake through the natural outflow canal,
evaporation and water escaping through the sinkholes on the periphery of the lake, are
the output components of the system.

Fig. 3 - Correlation between water level and streamflow.

Water balance computations were based simply on the continuity equation. The observed
volume of the lake was found using the relationship between water elevation-volume-area
plotted on a graph. The difference between the observed and calculated volumes was
evaluated as leakage or an unrecorded gain according to its sign, minus or plus
respectively.

V0bServed ~ V computed > 0 gain (not recorded) by estavelles

V
obServed ~ V compuled < 0 leakage through sinkholes
Some examples of the relation between the observed and computed volumes of the lake
are given in Fig. 4.

- 4 0 0 | I . I j • • l 9 | l . •^ I I I j • I i ^ 1 I 1^1 l , . ^ , l ^1 I . ^.1 I ^1 l . ^i

\ 1964 | 1965 | 1971 | 1972 |

Fig. 4 - Results of the water budget computations for the lake.

 A conceptual model for the Beysehir Lake karst system 249

The maximum amount of leakage, about 28 m 3 /s, was found to occur in March 1975
when the water level in the lake reached a maximum, which is about 5 cm/s on the
average.

LEAKAGE RELATED TO ZONE OF KARSTIFICATION

Correlation analyses between the amount of leakage and the water level of the lake
revealed that the interaction between amount of leakage becomes stronger as the water
level exceeds a certain level corresponding to an elevation of 1123 m.

Meanwhile, when the water level is below this level, the correlation coefficient between
the mean precipitation and the water level increases significantly (Fig. 5). This can be
considered as strong evidence to suggest that no significant leakage occurs below this
certain level and especially from the bottom of the lake. Besides, very thick (about
250 m) silty sediments occur under the lake. Otherwise one may expect this shallow lake
(maximum depth is 10 m) to have emptied since the hydrological regime could not
provide such a rapid circulation.

YEARS
_,£_ K0PR0ÇAY STREAMFLOW — ^ MANAVGAT STREAMFLOW
..,,-- CUMMULATIVE DEVIATION
Fig. 5 - Correlogram showing the relation between precipitation and the
water level in the lake.

During the field studies, a large number of various sized sinkholes were located at the
west periphery of the lake, along a large-scale fault. The sinkholes have developed within
the zone where the water level fluctuates during wet and dry periods. It is known that a
fluctuating water level may promote the karstification process along a tectonic element
like a fault.

MODEL PROPOSED FOR THE REGIONAL KARST SYSTEM

Based on the results summarized in this paper, a conceptual model is proposed to explain
the regional hydrological regime. The term "conceptual" was used to indicate the
physical parameters and relations that cannot be measured quantitatively or verified by
means of the techniques used at present.
According to this model the Beysehir Lake leaks water through the sinkholes located at
its western periphery but not from the bottom. And the amount of leakage is not enough
to be the source for the excess discharge occurring in the adjacent basins. Incidentally,
the leaked water must be going to the Manavgat River basin but not to the Koprucay
250 Mehmet Ekmekçi

basin as the statistical analyses revealed. This can also be confirmed by the results of
detailed hydrogeological studies carried out by the State Hydraulic Works in 1983.
Water balance computations showed that excess water comes into the lake which is not
recorded by the gauging stations. This water gain is seen during a period after a heavy
rainfall. This gives the impression that the sinkholes which are located at relatively lower
part of the lake must be working as springs due to the pressure that occurs after a heavy
rainfall.
Figure 6 illustrates the model of the overall karst system of the region, with a sketch of
a cross section of the region.

Fig. 6 - A conceptual model for Bey§ehir Lake-Manavgat River karst

system.

CONCLUSIONS
Statistical analyses used with hydrometeorological data and interpreted together with the
hydrogeological structure of the region and field observations, contribute to the
explanation of the hydrological relationships between the Bey §ehir Lake-Manavgat River-
Koprucay River basin karst system.

According to the model proposed, the Beysehir Lake leaks only a small amount of water
compared to the excess discharge of the adjacent basins and this cannot constitute a
 A conceptual model for the Beysehir Lake karst system 251

unique source for this discharge though it may contribute a certain amount. The
Koprucay River basin must be a part of a partially, if not completely, different karst
system. The Manavgat River basin is fed mainly from the northeast part of the region
where the Sugla Lake, with a large number a sinkholes on its periphery, is situated.
Sugla Lake is recharged mainly from the Bey§ehir Lake through a natural surface canal
along which there are a number of sinkholes.

The above summarized results unfortunately will not be verified because (a) the area is
too large to investigate in detail, (b) the geology and hydrogeology of the region are so
complex that the available research techniques fail to give reliable results, and (c) some
of the features such as the Dumanli spring and other important sinkholes are not in their
natural situation any longer as they had been part of hydrotechnical projects.

REFERENCES
Gunay, G. & Sipahi, H., 1979, Egridir-Beysehir Gôlleri ile Akdeniz Arasindaki Alanda
Yapilan Karst Hidrojeolojisi Calismalari. Miih. Jeo. Semp., TJK, Ankara.
Gunay, G., 1981, Karst hydrogeological investigation of the Manavgat River basin.
Habilitation Thesis, Hacettepe University, Ankara (unpublished).
Ekmekçi, M., 1987, Beysehir Gôlûnûn Komsu Havza Akimlanna Olan Etkilerinin
Ara§tinlmasi. Yuk. Miih., Tezi, Hacettepe Ûniversitesi, Ankara.
Yurtsever, Y., 1979, Environmental isotopes as a tool in hydrogeological investigation
of southern karst regions of Turkey. In: Proc. Int. Symp. on Karst Hydrogeology-
Antalya. DSÎ-UNDP Project Publ. Ankara.
Hydrogeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990). _
IAHS Publ. no. 207, 1993. 253

NUMERICAL VERSUS STOCHASTIC MODELLING OF WATER BALANCE

AND MINIMUM DISCHARGE OF A KARST HYDROGEOLOGICAL SYSTEM

NEVEN KRESIC, NEBOJSA KUKURIC & MILENA ZLOKOLICA

University at Belgrade, School of Mining and Geology, Djusina 7, 11000 Belgrade, Yugoslavia

ABSTRACT
After detailed hydrogeological research on the catchment of the Grza karstic spring, two
models for determining its water balance and monthly minimum discharge were studied.
Autocorrelation and cross-correlation of system input and output were analysed, as well
as the first to ninth-order ARMAX model. On the basis of a four-year observation period
of spring discharge and a 30-year observation period of precipitation, the ARMAX model
was used for the simulation of 30 years of spring discharge. The fourth-order model was
shown to be the most reliable for the whole adjusting period of four years. On the basis
of 30 calculated values for each month the probability of minimum spring discharge was
analysed. Calibration of deterministic, i.e. finite difference model, was performed for
two assumed stages: quasi-stationary adjusting on the basis of inferred groundwater levels
along the system boundaries (three points were used for the control), and non-stationary
calibration according to the recorded recession curve. The assumed paths of increased
karst groundwater flow were given extremely high values of transmissivity introducing
modified finite difference model (basic rectangle elements of the grid were divided into
the smaller ones along the regional faults and neotectonically active zones).

INTRODUCTION
A complex hydrogeological study was made of the catchment of the Grza spring in
eastern Serbia, Yugoslavia, in order to estimate the minimum discharge of the system,
as well as the possibility of obtaining additional quantities of groundwater at the planned
intake during the dry season. The system's macro geometry (boundaries and regional base
of karstification) and micro geometry (preferential flow paths and regional porosity) were
assumed on the basis of hydrogeological mapping, groundwater tracing, neotectonic
analysis, remote sensing, speleological research and, indirectly, on the basis of
geophysical research in similar karst terrains in eastern Serbia.

The karst hydrogeological system of the Grza spring is developed in intensively karstified
Cretaceous and Jurassic limestones of the southwestern wing of the large Kucaj anticline.
Middle Paleozoic shales make the impervious base of the karst aquifer, while the barrier
on the west is overthrust Permian sandstones. Northern and southern boundaries toward
the catchments of Sisevac and Krivi Vir springs respectively are uncertain since all three
drainage areas are developed in the same formation.

The Grza karstic spring is located at the contact of Mesozoic limestones and overthrust
Permian limestones. This contact is very steep (around 60°) and two boreholes detected
the presence of karstification in limestones at depths more than 70 m below it. A several
hundred metres long cave behind the Grza spring has a siphonal lake with its base below
the level of the spring discharge.

STOCHASTIC MODELLING

Stochastic properties of the system

Stochastic properties of the spring discharge and its response to recharge by precipitation
254 Neven Kresic et al.

are analysed through autocorrelation and cross-correlation respectively (Fig. 1). The
independence of the time series was tested by modified Anderson's test (Prohaska, 1981)
for the length of correlograms being k = 100, i.e. not only for the first correlation
coefficient.
The autocorrelogram of the spring discharge indicates a highly dependent process,
specially in dry years. However, for the confidence limits of 95%, a cross-correlogram
shows that the process of discharge is independent on precipitation in dry years which
is surprising for an open karstified hydrogeological structure such as the Grza spring
catchment. More than 5% of coefficients, including that of lag 1, lie outside the
confidence limits in wet years indicating the presence of such dependence.
On the basis of these results one may conclude that small joints, fissures and clastic
material deposited in karst cavities play the most important role in the process of
accumulation of groundwater in the Grza aquifer. Their slow draining during the dry year
(when the effect of hydrostatic pressure of rapidly infiltrated precipitation is not so
distinct as in wet years) causes high interdependence of spring discharge rates.

From Fig. 1 it is also seen that the delay of the spring's response to recharge is
approximately 10 and 65 days in dry years, while in the wet year the delay is 6 and 52
days.

o so wo

0.4 ' (b)

Fig. 1 - Autocorrelogram of the Grza karstic spring discharge (a) and

cross-correlogram of the Grza spring discharge and the precipitation in its
catchment (b). Lag step (k) is given in days. 1 - Wet year; 2 - dry year.

ARMAX model
Different orders of the ARMAX model based on one input (gross precipitation) and one
output (the Grza spring discharge) have been tested for their reliability. The results are
shown in Fig. 2. Considerably higher values of R for the dry year indicate the strong
self-dependence of the process of the system's discharge, which was already concluded
on the basis of its autocorrelogram. For all four years with recorded data on daily spring
discharge, the ARMAX model of the fourth order (i.e. using data on the system's input
and output four days before) was shown to be the most appropriate, With decreasing
 Numerical versus stochastic modelling of water balance and minimum discharge 255

annual precipitation in the catchment, the reliability of the ARMAX model is higher
(Table 1).

Fig. 2 - Dependence of ARMAX model reliability (expressed through the

coefficient R) on the number of its order (n). 1 - Wet year; 2 - dry year.

Table 1. Dependence of ARMAX model reliability on the annual precipitation in the

catchment of the Grza karstic spring (for the value of R = 1 the dependence is
functional).
Year Deviation from the average annual Reliability of the model Model
precipitation (R) no.
1980 + 17% 0.59 I
1981 -3% 0.72 II
1982 -22% 0.92 III
1983 -24% 0.96 IV

The simulation of 30 years of the Grza karstic spring discharge was performed in order
to determine the probability of the minimum daily discharge for each month. For the
years having annual precipitation within the limits of ± 10% of the average, the ARMAX
model no. II was used. Model no. I served for the simulation of discharge in wet years
with precipitation exceeding the average more than 10%. Dry years with precipitation
10% or more less than the average were simulated by model IV.

800
\ 1 Q
\
(1/8]

1
°\ 600
AY
ar" l M I

^<
'SJO

a'

r-
50 SO 95 9699 9&5 99a 99.99
PROBABILITY ( % )

Fig. 3 - Theoretical probability distributions (log-Pearson type III) for the

Grza spring minimum discharges for the values calculated by the ARMAX
model.
256 Neven Kresic et al.

Thirty calculated values of the minimum daily discharge for each month were tested using
the HOMS programme, i.e. six common functions of the probability distribution. The
results for May (the month with the maximum recharge) and August (the month with the
most of the calculated absolute minimums) are shown in Fig. 3. The best fitting
probability function for all months, except for January and December, is log-Pearson
type III.

NUMERICAL MODELLING
Physical basis of the system and its deterministic modelling
Since the karstified media could by no means be treated as a continuum, the main
requirement for simulation of the system studied was to "imitate" its nature as closely
as possible with an improper tool. Namely, a finite difference approach was chosen for
its relatively simple calculation and flexible calibration possibilities according to the
modeller's knowledge of the hydrogeological characteristics of the system.

The flow domain was divided into a regular grid of rectangles (Fig. 4). The values of the
karst aquifer head along the boundaries with zero flux, both impervious and between the
neighbouring aquifers, were assumed. According to the predicted base of karstification,
based on geophysical research in similar terrains and surface mapping of discontinuities,
each element got the assigned transmissivity value.

The assumed preferential flow paths within the system (large fractures) have been divided
into smaller rectangles with extremely high values of transmissivity (up to 10 000 m 2 /s).
These paths have been changed during the model calibration, though not considerably.
Three control points with known (or, at least, most reliable) heads served for the
calibration: the level of spring discharge, the level of the siphonal lake in the cave and
the bottom of the deepest explored jama in the catchment.

0 1 ^ 2 MS • «
Fig. 4 - Schematic diagram of the grid used for the modelling of the Grza
karstic spring catchment by the finite difference method. 1 - Element with
known head value; 2 - element with given (assumed) head value; 3 -
element with highly exaggerated transmissivity value; 4 - element with
given (assumed) "normal" transmissivity value.

Calibration of the numerical model

The first calibration of the model was made on the assumption that, for one moment, the
system was in the steady state - the amount of discharged water equalled the system
recharge. This was necessary simply because of the fact that the model had to start
 Numerical versus stochastic modelling of water balance and minimum discharge 257

calculating. For such assumed quasi steady state flow, the boundary conditions were
given as the head values controlled by the predicted hydraulic gradient. The requirement
that head should be below the deepest explored dry jama and matched with the known
levels of spring discharge and siphonal lake in the cave (Fig. 5) was fulfilled by changing
the transmissivity values of the grid elements, as well as the number of preferential flow
paths. This phase of calibration gave very important information on the nature and the
role of particular fractures simulated by high transmissivity elements.

Fig. 5 - Schematic cross section showing the elements for the calibration
of the deterministic model in quasi steady state. 1 - given (assumed) head
value along the boundaries; 2 - level of spring discharge; 3 - level of
siphonal lake in the cave; 4 - level of the bottom of the dry jama; 5 -
unadjusted heads; 6 - adjusted head.

The second calibration of the model assumed non-steady flow during a period without any
recharge of the system. Since a very long recession period had not occurred during the
analysed years, the recession hydrographs were calculated by the method introduced by
Drogue (1972). In addition, the areas above the calculated recession curves were used
as the starting values for the final "day-for-day " modelling of the spring discharge during
one year.

The fitting of the calculated and modelled recession curve was performed basically by
changing the coefficient of specific storage of the porous media (i.e. limestones).
Surprisingly, the average value of 0.85 exp( —3) was found to be the most appropriate
for the model. Its similarity with the values commonly found for artesian intergranular
aquifers indicates that drainage under the pressure of joints, fissures and clastic material
filling the karst cavities was the most important process in the system's discharge. This
conclusion is supported by the considerably lower rate of effective precipitation obtained
by the model calibration than that shown in Fig. 6 (i.e. the areas above the recession
curves).

One might expect that similar shapes of these areas, i.e. similar intensity and amount of
precipitation would always cause similar effective infiltration in the same period of each
year, e.g. summer. However, when the model was applied for similar conditions of
summer storm events (Fig. 6c) in a year not used for parameter identification, it was
found to be inadequate. Consequently, it could not be used to simulate 30 years of spring
discharge since it requires determined effective precipitation for the input.

CONCLUSIONS

In the case when only short periods of discharge data are available, probabilities of the
minima (calculated by common statistical procedures which require at least 30 values for
correct interpretation) could be determined only by stochastic modelling. Primarily this
is because the ARMAX model operates with gross precipitation, while the deterministic
(numerical) model requires already prepared values of effective infiltration. However,
a strict comparative quantitative analysis of infiltration rates in karst areas may enable
a more reliable application of deterministic modelling.
258 Neven Kresic et al.

300 Qo (a)
1/» Q o
Qt
W
200- (1 + rtU)"

«0-

1
M j ' J ' A s '
°
(b)

Fig. 6 - Hydrographs of the Grza karstic spring discharge during the

summer periods. The areas above the calculated recession curves were
used as a first estimation of effective precipitation in deterministic
modelling.

Both types of model may be used for predicting the discharge after a long period without
precipitation. In the case of the ARMAX model this is particularly true for the system's
strong autoregressive nature. On the other hand, unsteady state calibration of the numeric
model can be highly reliable (in the case of the Grza karstic system the highest obtained
reliability measure was 0.92) and depends mostly on the modeller's patience. The
difference between predicted minimum discharge after three months without precipitation
is only 18%.

The numerical model is irreplaceable for predicting the effect of an artificial influence
on an aquifer. This is of extreme importance for e.g. estimating of the possibility of
additional extraction of groundwater in the dry season and its influence on the
environment (rate and consequences of head depletion, providing a certain minimum
surface flow downstream from the spring for biological reasons etc.).

A study of assumed preferential flow paths during the calibration of the deterministic
model gave important indirect information on the nature and hydraulic role of the system
fractures.

Finally, it is recommended that both models be used together whenever possible because
of their complementary rather than contradictory nature as was often considered in
practice. In addition, the development of a combined model in which the deterministic
part would use stochastic properties for transforming gross precipitation into effective
precipitation seems promising.
 Numerical versus stochastic modelling of water balance and minimum discharge 259

REFERENCES
Drogue, C , 1972, Analyse statistique des hydrogrammes de décrues des sources
karstiques. J. Hydrol., 15, 49-68.
Kresic, N. N., 1987, Kukuric, Models for defining of minimum monthly discharge of
karst springs (in Serbo-Croatian, English abstract). In: Proc. IX Yug. Symp. on
Hydrogeology and Engng Geol., vol. 1, 135-136.
Prohaska, S. A., 1981, stochastic model for long-term predicting of river flow (in Serbo-
Croatian, English abstract). Vode Vojvodine, 15-106.
Yevjevich, V., 1974, Stochastic Processes in Hydrology (translation into Serbo-Croat).
Sarajevo.
Zlokolica, M., Some possibilities of comparison of results of hydrogeological and
speleological explorations (in Serbo-Croatian, English abstract). Nas Krs, XII, 22,
47-54.
Hydrogeologictd Processes in Karst Terranes {Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHS Publ. no. 207, 1993. 261

DOUBLE POROSITY MODELS IN KARSTIFIED LIMESTONE AQUIFERS:

FIELD VALIDATION AND DATA PROVISION

MARTIN SAUTER
Geologisches Institut, Universitât Tubingen, Sigwartstrasse 10, D-7400 Tubingen, Germany

ABSTRACT

Modelling results and long-term field investigations from a karst aquifer (Upper Jurassic)
in southwest Germany are presented, providing further evidence and support for the
Double Continuum Ground Water Modelling approach in karstified limestone areas. With
water flowing at different rates through a karst aquifer, slowly through small fissures and
very rapidly through large fractures and solution enlarged channels, the simulation of the
flow and transport requires information on both systems and their interaction. This
information is provided by investigations on a catchment scale, an intermediate and on
a borehole scale. Time series analysis of rainfall events utilizing well hydrographs,
spring discharge, 1 8 0/ 1 6 0 ratios, temperature, electrical conductivity and turbidity of the
water allow the discrimination of not only event and pre-event water, but also the
quantitative distinction between the fast and the slow component of the storm derived
water. The latter is an essential calibration tool for the interchange between the two
systems in a double porosity model and the 1 8 0/ 1 6 0 ratio serves as a regional tracer, the
breakthrough of which could be modelled. Further transport parameters were provided
by two different kinds of tracer tests, with the input in sinkholes (information on fast
system), as well as in boreholes ("statistical" information on slower system). Hydraulic
parameters were derived from slug tests. The attempt to model the aquifer with a
standard porous medium model failed, primarily as a result of a different response from
the two systems. Whereas peak discharge generally occurs two to three days after the
storm event, the maximum in the well hydrographs lags behind by more than one month.
A satisfactory representation of both the flow and the transport processes could only be
achieved with a regional double porosity model, allowing flow in both the "fracture
porosity" and the "matrix-porosity".

INTRODUCTION
A growing concern with groundwater contamination through agricultural chemicals and
potential spills of hazardous materials calls for some method of prediction and possible
tool for remedial measures. This is especially applicable in the case of a karst aquifer,
which is known to be highly vulnerable to contamination. However, the lack of spatial
knowledge of the karst parameters, the relative unpredictability and the extreme
heterogeneity in aquifer properties has often discouraged researchers from attempting to
model such aquifers.

This paper attempts to demonstrate the applicability and practical use of a double porosity
model in a karst environment for the simulation of flow and regional transport. Issues
concerning the provision of the specific data input and aquifer parameters specifically
designed for a karst aquifer are also addressed. Special emphasis is laid on formulating
methods for the quantification of the input data, required for the model simulation.

AREA OF INVESTIGATION
For the investigations, the spring catchment of the Gallusquelle was selected. It is
situated in southwest Germany on the Swabian Alb, a small mountain range that stretches
262 Martin Sauter

Fig. 1 - Location map of the project area.

in an approximate southwest-northeast direction for roughly 200 km (Figs 1 and 2).

Morphologically, the project area dips gently from an escarpment (1000 m a.s.l.) in the
northwest down to about 600 m in the region of the spring.
The Gallusquelle groundwater basin forms a part of the catchment (450 km2) of the
Lauchert River, which is fed mainly by karst springs. The River Fehla represents the
boundary to the northeast. Approximately 60% of the catchment area is covered by
spruce, the remaining land is predominantly agricultural. The area is well suited for the
intended measuring programme. More than 25 years of continuous records of discharge,
water-level fluctuations (weekly readings), and climatic data were available, as well as
data from about 20 wells that allowed the construction of a water level map, even in
areas far removed from the discharge point. Fifteen tracer tests helped to delineate fairly
accurately the catchment boundaries which cover an area of approximately 45 km 2 .

Geology
Geologically, the area is composed exclusively of carbonate rocks of the Upper Jurassic
(Fig. 3). At the surface, predominantly the massive limestones of the Kimmeridge 2/3
(ki2/3) are exposed, which reach a maximum thickness between 90 and 140 m. They are
underlain by the Kimmeridge 1 (kil), a marly limestone sequence, with more or less
expressed bedding and a thickness of approximately 50 m. The lowest relevant geological
unit consists of well bedded limestones, the Oxford 2 (ox2). The whole stratigraphical
succession dips southeast. The southwest border of the catchment is formed by the
Hohenzollern Graben, which is tectonically still active. Another fault zone strikes north-
south and borders the project area in the southeast.

Hydrogeology
As shown in Fig. 3, the aquifer is formed by three geological units, the massive
 Double porosity models in karstifled limestone aquifers 263

• BOREHOLES _ CONTOURUNES IN M ABOVE S.L

• SPRING — CATCHMENT AREA
~ TRACERTESTS

Fig. 2 - Map of catchment area and monitored spring.

limestone (ki2/3) in the southeast, the marly limestone (kil) in the centre and the Oxford
2 in the northwest. The aquifer base does not follow any stratigraphical boundary. It can
be deduced from the geological logs of the boreholes, that the bottom of the active part
of the aquifer runs parallel to the water table, with saturated thicknesses estimated to
range between 0 and approximately 30 m, depending on the season, although the
limestone sequence may be in excess of 150 m.
An increased gradient of the piezometric surface can be observed in the centre of the
catchment (Fig. 2), where the water table cuts across the less permeable marly lime-
stones. Further upgradient, the aquifer comprises of Oxford limestones. The water table
constitutes the top of the aquifer. Unconfined conditions prevail in the entire catchment.
The unsaturated zone is highly karstifled and reaches thicknesses between 90 and 120 m.

The lateral boundaries of the catchment could be derived from numerous tracer tests; in
the lower course of the River Fehla, however, it could not be clarified, whether the river
is in direct contact with the aquifer or just perched above the water table. Closer to the
spring, annual water level fluctuations range between 5 and 15 m and.further upgradient,
264 Martin Sauter

Fig. 3 - Geological cross section of the karst aquifer.

between 10 and 30 m, reflecting the decrease in transmissivity and storage coefficient

away from the point of discharge.

MODELLING APPROACHES IN FRACTURED AND KARSTIFIED ROCKS

Two fundamental approaches to modelling karst water resource systems have been taken:
the black box model and distributed parameter model. As a result of the scarcity of
spatial data, the heterogeneity of the aquifer parameters and because of its relative
simplicity, a black box approach has been frequently preferred in the simulation of karst
aquifers. Despite its global approach, encouraging results have been obtained by
recession analysis (Atkinson, 1977), the identification of transfer/kernel functions
(Dreiss, 1989b; Aiguang et al., 1988; Avias & Joseph, 1984) and also through simple
regression analysis (Zaltsberg, 1984).

The inadequacies of the black box models become apparent when one attempts to model
spatially variable output phenomena, like characteristic water level fluctuations, that have
a definite physical basis. Frequently, geological information that could explain observed
differences which in many cases is of a spatial nature has to be ignored in such models.
Moreover, they fail to consider the different processes that determine flow and transport
in a karst aquifer, i.e. the mechanism of groundwater recharge, the influence of the
unsaturated zone and the phenomena in the aquifer itself. Furthermore, each of these
factors in turn has a different influence on fast and slow flow components.
A viable alternative is the distributed parameter modelling approach. Three major
methods have been used to describe the flow and transport through fractured porous
media. When the fractures are narrow, evenly distributed, and if there is a high degree
of connectivity an equivalent porous medium model can be applied (Pankow et al., 1986).
A recent development for this kind of model has been described by Neuman (1987). The
author simulates the flow in the fractured porous rocks with an equivalent porous medium
model, integrating the permeability as a stochastic variable.
 Double porosity models in karstified limestone aquifers 265

The second approach, the discrete fracture model implies that the effect of the matrix is
neglected. Flow is simulated by considering it as flow between two parallel plates and
it requires some detailed information on fracture apertures, density, orientation and
connectivity (Snow, 1965; Irmay, 1984; Romm, 1966). Especially as it is very difficult,
if not often impossible to obtain the required information, the fracture networks have
been statistically simulated (Long et al., 1985; Smith et al., 1987).

The modelling of flow and transport in fractured rocks could benefit greatly from studies
carried out by petroleum engineers in the field of reservoir engineering and from the
research efforts in the search for safe repositories for radioactive wastes. The preferred
approach in these areas of research has been the double porosity approach (Barenblatt et
al., 1960; Warren & Root, 1963; Duguid & Lee, 1977). The fractured medium and the
porous matrix blocks are modelled as two separate overlapping continua, each with its
own flow equation. The coupling of the two media is handled with a source/sink term in
each equation. The exchange of flow is controlled by the local difference in potentials.
The applicability of the distributed parameter approach to karst aquifers was described
in detail by Thrailkill (1986) and it's appropriateness to conduit-flow aquifers in
particular was examined. Kiraly (1984), for example, could successfully apply a finite
element model, representing the matrix blocks with a porous medium approach, with a
superimposed discrete regular network of line elements, simulating the fast system.

For the work described, a double porosity model was chosen. It is considered suitable
for the aquifer concerned, which is located at an intermediate position, possibly more
towards the diffuse end member of the conduit-diffuse-flow spectrum (White, 1977), and
where the physico-chemical characteristics of the spring water and tracer tests indicate
that conduit flow is still an important factor. The discrete fracture approach is
inapplicable, because the required data input cannot be obtained and the parallel plate
assumption does not hold.

MODEL INPUT DATA PROVISION

This section demonstrates how the data for the double porosity model were obtained. The
approach taken is in accordance with the concept forwarded by Smart & Friederich
(1986). The aquifer system is decompartmentalized into the processes of recharge,
storage and transmission, the influence of each of them is examined and the resulting
responses explained in terms of the hydraulics of the aquifer and spring water chemistry.
The aim is to quantify each of these phenomena, as well as to describe, where applicable,
their variability with time using time series analysis.

Recharge evaluation
Groundwater recharge was computed on a daily basis, applying a soil water balance
approach. Potential evaporation was calculated with the Haude (1955) method, which
requires air humidity and air temperature at 14:00 as input variables. The resulting
values are corrected for the varying transpiration by plants. The soil moisture balance
approach was carried out according to Uhlig (1959), a procedure similar to that of
Thornthwaite & Mather (1957), applying a field capacity (measured) of 70 mm.

However, due to the high percentage of woodland, interception becomes important, which
is not corrected for in the Haude/Uhlig method. Therefore, with the exception of drought
years, a systematic overestimation of annual recharge was obtained, even after varying
the field capacity within acceptable limits. A satisfactory agreement (error 10%) between
the computed and the measured (cumulative spring discharge, with AS = 0) annual
recharge (period 1965-1990) could be obtained by changing the Haude factors according
266 Martin Sauter

to Sokollek (1983) and using a canopy interception capacity of 4.7 mm; for rainfall
events below 12 mm, interception was taken as a percentage of 35% of the total daily
rainfall (Benecke, 1978). The storage in snowfall was accounted for and the release of
snowmelt followed a simple relationship, using a degree day factor of 6 mm.

Regarding the mode of entry, it can be shown that a high percentage (approximately
90%) of the recharge reaches the water table in the form of dispersed flow, and the
remaining portion via shaft flow, which is collected in surface depressions. This figure
can be substantiated by the variations in spring water chemistry examined below.
Recharge, slowly released from the epikarstic zone via subcutaneous drains (Williams,
1983; Smart & Friederich, 1986) falls into the category of dispersed flow. Neither
sinking streams nor impermeable cap rocks were observed.

Storage
Storage can be categorized into unsaturated and saturated storage, each of them can again
be subdivided into different reservoirs, the soil zone and the epikarst (subcutaneous
zone), the conduit storage and the water stored in microfractures and pores. It can be
assumed that the thin cover of orthic luvisols is unable to hold large quantities of water.
Even after minor rainfall events, the soil is at field capacity. Soil physical examination
(Hemme, 1970), suggests a field capacity varying between 30 and 70 mm. Only in
depressions, where brown earths prevail, does field capacity reach 180 mm.

The storage capacity of the subcutaneous zone is difficult to assess. Through repeated
borehole logging, temperature variations have been observed in the top 5 m of the
phreatic zone (100 m b.g.l.) that are closely related to the surface air temperature, even
after prolonged dry periods. The only possible mode of heat transfer could be by water,
temporarily stored in the unsaturated zone. These variations suggest that this type of
recharge must be an important quantity, otherwise, it would have been buffered by the
karst water stored in the saturated zone.
Ongoing research is attempting some quantitative estimate of this type of recharge. After
some recharge events, especially after prolonged dry periods,it was observed that
electrical conductivity of spring water increased for a short period followed by a rapid
decrease. According to Ashton (1966), the water with the higher conductivity represents
so-called deep phreatic water as a result of the longer reaction time between water and
rock. If however, this type of water is interpreted as water from the subcutaneous zone
(Williams, 1983), which is displaced by event water and discharged only via the conduit
system, some estimate of the stored quantity can be made.

Storage in the saturated zone can be subdivided into conduit and diffuse storage
(Atkinson & Smart, 1981; Shuster & White, 1971; White, 1977). Conduit storage could
be evaluated with the approach suggested by Williams (1983), which is a modification
of Ashton (1966). The values obtained this way varied between 0.01% and 0.02%.
Another method employed tracer tests, whereby the dye was injected into a doline
connected to preferential flow paths (Fig. 4). The volume of groundwater discharged
from the time of injection until the arrival of the dye, divided by the volume of the
saturated rock produced the figure of 0.0002 + 0.0001 for the conduit storage. The
storage of the whole aquifer system, which consists mainly of diffuse storage, could be
similarly evaluated by dividing the water volume discharged by the volume of the aquifer
drained. The drained rock volume could be calculated using the water level changes in
the boreholes. A fairly constant value of about 2% could be determined within the main
aquifer body. This value is somewhat lower at very high water levels.
 Double porosity models in karstified limestone aquifers 267

Tracertest Bitz
Tracer: Fluorescein, 10kg
g:9S20rn
Distance to Spring: :
Date of Input: 8.6.88.10:30
Input in Sinkhole
Average Discharge of Spring: 250 S/s
Average Velocity: 80 m/h

•a -
E

Fig. 4 - Tracer-test "Bitz", with the injection of the dye directly onto the
conduit system via the unsaturated zone.

Flow within the aquifer

On the scale from conduit to diffuse flow, the aquifer investigated can be placed in an
intermediate position (mixed-flow). Whereas the water quality parameters and the
discharge react "flashy", and the response to storms is still very rapid, the determining
part of the aquifer really is the matrix blocks, that store and release the water over
several months. Nevertheless, it is understood that the drainage is still via a tributary
network of conduits. As can be shown below, following storms, approximately 10% of
the total flow reaches the spring directly via the conduit network.
Hydraulic conductivity has been frequently determined in the context of water resources
developments. The values obtained range between 1*10"4 m/s and 2.6*10"3 m/s. Due to
the higher probability of success and the lower costs for shallower boreholes, all the
tested wells were drilled in valleys, and the values therefore constitute an overestimation,
taking the aquifer as a whole.
Alternative approaches were taken by Villinger (1977). Attributing the average flow rate
(500 1/s) to three areas of equal gradient (close to the spring, increased gradient in the
centre of the project area, area near the groundwater divide in the northwest) and
assuming a throughput area, Darcy's law can be used to calculate an average hydraulic
conductivity for each of the sub-areas. The values varied from 2*10"4 m/s in the centre,
to 11* 10"4 m/s near the spring.

The low conductivity end of the spectrum could only be evaluated from boreholes further
upgradient. Because the unsaturated thickness is very high (100 m) and the 3" casing
cannot accommodate powerful pumps, it was impossible to conduct pumping tests.
Instead, slug tests and constant rate injection tests were carried out. The injection tests
were evaluated according to Bourdet & Gringarten (1980), an analytical type curve
matching procedure, that assumes that the aquifer can be represented with the double
porosity approach and which allows the determination of skin and wellbore storage
effects.
The resulting hydraulic conductivity values ranged between 0.3*10"5 and 4*10"5 m/s
(whole system). No analytical method is available to evaluate slug tests where double
268 Martin Sauter

porosity effects have been observed (Fig. 5) and where the boundary conditions are
complicated by partial penetration (Dougherty & Babu, 1984; Barker & Black, 1983).
The tests were therefore simulated with the TRAFRAP-WT code (Huyakorn et al., 1983)
using a double porosity approach. This helped to account for geometrical factors and the
wellbore storage. Assuming that the fractures represent 0.1 vol% of the tested area,
hydraulic conductivities of 2-12* 10"3 for the fractures were obtained.
The matrix hydraulic conductivity varied between 10"5 m/s and 10"8 m/s. Taking into
account the relative volume percentages of the fractures and the matrix, values similar
to those evaluated according to the Bourdet & Gringarten (1980) method could be
obtained.

At2 Rmponss of MeSIx

AH ftasporoo of tra Fracture*

Slugtest, B8
Tests

0001 0.1 10 1000 100000

Time (s)

Fig. 5 - Recovery curve of a slug test, displaying double porosity effects.

Dynamic analysis of recharge and aquifer flow

In order to be able to describe the flow mechanism and the temporal distribution of the
slow and fast components of recharge and the flow within the aquifer, time series of a
number of recharge events have been analysed. The parameters shown in Fig. 6,
displaying the event of April 1989, have been measured. In the majority of events, it
could be observed that the peak discharge occurs two to three days after the rainfall
event, whereas the actual event water reaches the spring approximately 1.5 days later,
which implies that peak discharge consists mainly of pre-event water, as has also been
described by Williams (1983) and Dreiss (1989a). The maximum water level in the
aquifer was read about five weeks after the event. The highest values in spring water
turbidity were observed contemporaneously with peak discharge, i.e. the turbid water
consists mainly of pre-event water.
The groundwater hydrographs show a response that suggests that the two distinct flow
systems exist. Some intermediate maxima could be measured in both boreholes
simultaneously with peak discharge. These might reflect the maximum potential in the
fracture system, which dissipates quickly, due to the inherent high hydraulic
conductivity. The bulk of the recharge water eventually arrives, as described above, five
weeks later at the water table, delayed by the thick unsaturated zone (100 m) and the
 Double porosity models in karstified limestone aquifers 269

F
É 15 • - Effective Rainfall
al Station Herrmannsdorf
10
«-
to
oc
CD

0
1 April 1989
s ill. j
I.May 1989 U u n é 1989 I.July 1989
0.8

^ 0.7
Spring Discharge
k(D
1
0.6 Gallusquelle
E
(O
0.5

"O)5 0.4

Q 0.3

0.2
I.April 1989 1.May 1989 U u n e 1939 I.July 1989
8.1.)

661.5 : B7 B14 687.0

CO
F 686.0
661.0
CD 685.0
J! 660.5
684.0
CD Groundwater Hydrographs
CO
îS 660.0
I.April 1989
B7, PI 4 683.0
I.May 1989 U u l y 1989

3 0.8
Turbidity
rb dity (NT

0.6
Springwater Gallusquelle
0.4

0.2
3
1- 0.0 ^A^AAAyNl
I.April 1989 I.May 1989 U u n e 1989 U u l y 1989
^-^
fc 580
elect. Conductivity
f/>
3. 570 Springwater Gallusquelle
T3
560
o
550
LApfll 1989 I.May 1989 U u n e 1989 U u l y 1989

o 8.65 Temperature
8.60
Springwater Gallusquelle"
855
o.
8.50
E
I.April 1989 I.May 1989 U u n e 1989 U u l y 1989

Fig. 6 - Rainfall event of 1-2 April 1989; time series of hydraulic and
physico-chemical characteristics.

lower hydraulic conductivity of the aquifer matrix. The turbidity is probably the most
suitable parameter to distinguish between the fast and the slow event water component.
In order to keep the clay particles in suspension, a minimum of hydraulic energy and
turbulence is required, a criterion which allows the distinction between the two flow
components. It also enables the determination of the time when the fraction of the fast
water becomes zero.
An analysis of the electrical conductivity variations allows a semi-quantitative
determination of the total fraction of event water transmitted in the fast system. It is
assumed that fast percolating water does not take up any dissolved constituents on its 1.5
270 Martin Sauter

day passage through the unsaturated zone, whereas recharge water (slow component), in
contact with the carbonate rocks for more than a month, approaches the electrical
conductivity of the groundwater by the time it reaches the water table; the latter
statement could be verified by continuous measurements in a cave over a period of about
three months.
Assuming an electrical conductivity for the fast recharge of 50 jtS/cm, which is a rather
conservative estimate, and applying established component separation methods (Sklash
& Farvolden, 1979; Pinder & Jones, 1969), then the fast component amounts to
approximately 2.5% of the total event water.

Comparing the spring water temperature record with the electrical conductivity, the same
assumption about the slow water cannot be upheld. Even after a period of 2.5 months,
the spring water temperature still reveals some memory effect of the previous event.
Therefore, an attempt to separate the slow and the fast components was made, using a
straight line recession method (plot of temperature against log time, Fig. 7).

E 8.60 Temp.imputee, test comp., 1st event

o Temp.imputee, fast comp.. 2nd event
2
e - measured temperature (after correction)
8. - Residual (slow component)
E

Temperature variations
due to stow and fast components

I.April 1989 I.May 1989 I.June 1989 I.July 1989

Fig. 7 - Temperature changes attributed to the different flow components.

With the assumption that air temperature at the end of the rainfall event corresponds to
the input temperature, a figure of 12% for the fraction of the fast component can be
obtained, using the same component separation method as above (Fig. 8).
A common approach to distinguish between two components is the simple recession
analysis, assuming a distinct hydraulic response for each of the two systems (Maillet,
1965; Atkinson, 1977). Whereas the previously described methods quantify the event
water components directly, i.e. the spring water analysed consists of event water, the
recession analysis considers the change in discharge as a result of the pressure pulse
induced by the new recharge water, i.e. the change in discharge reflects the arrival of
the fast event water component. This feature is very useful, because it helps to describe
the time variation in the input to the groundwater system (Fig. 9). The fraction of the fast
component is evaluated at about 15%, assuming that the initial drop in the recession limb
describes the hydraulic characteristics of the fast system over the whole aquifer
thickness.
 Double porosity models in karstified limestone aquifers 271

Component Separation
based on temperature

Spring Discharge

Fast Comp. Event 1

a 0.5 Fast Comp, Event 2

S3

0.2 3u i M i H | < i i i u u u m i n u i i n 1111M1111[11111111111M i i i i n i i i i i u i i i j i i u i m

LApril 1989 I.May 1989 I.June 1989 I.July 1989

Fig. 8 - Component separation, based on the temperature as a tracer.

Component Separation
based on hydraulic characteristics

Spring Discharge
Fast Comp. Event 1
JE.
Fast Comp. Event 2
g, 0.5

I.April 1989 I.May 1989 I.June 1989 Uuiy 1989

Fig. 9 - Component separation, based on the principles of recession

analysis.

The above described methods would enable the quantification and the temporal
distribution of the input of the fast component in the model, if the recharge water
remained unchanged in temperature or chemical composition during its passage through
the unsaturated zone.

However, these methods were all based on non-conservative tracers, and although they
can describe the extremes and the time variation of the input, they are still somewhat
arbitrary. Therefore, one further parameter was analysed, the relative abundance of the
I8
0 isotope in groundwater and rainfall (Hess & White, 1974; Bakalowicz & Mangin,
1980). Spring water samples were taken automatically at 16-h intervals as well as daily
rainfall at five stations.

The results are plotted in Fig. 10, together with the rainfall input. The figure contains
two sets of information. Squares and crosses represent values of the spring water,
272 Martin Sauter

Oxygen Isotope Ratios in Spring Water

S -99
Spring Water Samples RaintaJI Input

"T* Ssmpte* tt&sn d u r i g wsSsr Efcraînscfeon MMJI'I of ben» rcprsflMftts n^csss

i. -loo oTOInlhatrInput
[j ScmptM not Wtwi dLrtiç tnster sbs^ccton
p BBiMiMa
> 20 mm
Error: 0.15%. e m s s i 15-20 mm
-10.1 W.-.-..;.-J i o - l 5 m m
5-10 mm
< 5mm

%
e

y -B
11
-15 \-
< il '1
tj^TmftLAg^.Sdayi
Il \
+
Spring WMar SeckgrounrJ Rsto

l l l i i l l l 11 11
6 L.
s K f n s d crwige f i rafio
duetoevent o!15/1 OAprl

Ill i n n i n Mm i l l 11111 r r i i m i m m i n u n Mi m i I m i m i l u r n m l m Mi n u t l l i M

1. April 89 1. May 89 1. June 89 1. July 89

Fig. 10 - Oxygen isotope variations in rainfall and in spring water.

whereas the bars indicate the deviation of the ô 18 0 in the rainfall from the long-term
spring average of -10.55%t>. The shading of the bars represents the weighting of the
input, corresponding to the rainfall height.

The spring data are represented by two different symbols, because it is suspected that the
water discharged during abstraction periods might be slightly different in its chemical
composition. The rainfall isotope data are shifted by 4.5 days to account for the observed
time lag.

The dotted line indicates an interpretation of the expected bimodal breakthrough of the
ô 18 0 as a response to the event of 1 April, the first peak indicating the arrival of the fast
component, the second, the delayed slow event water component. The rainfall input was
evaluated at -9.5%o. The spring water analyses display quite a pronounced scattering,
which is partly due to the analytical error of 0.15%o.

The largest variations, however, can be explained by the interferences of other rainfall
events: one event on the 15-16 April, with a fairly negative input of around — \3>%o and
an event at the end of April, with values reaching almost - 17%o. The arrival of the fast
component of the latter event water probably coincides with the slow component of the
event of 15-16 April, as indicated by the temperature residual in Fig. 7.

Applying the same component separation technique (mass balance of constituents) with
an input ô 18 0 of —9.5%o only about one third of the total event water (integral
underneath the discharge curve at AS = 0) could be recovered. The input ô 18 0 had to be
varied by up to — 10.2%o in order to balance both totals. In this case, the fast water
component amounts to about 12 %o.

This observation indicates that even the ô 18 0 is subject to change, in the present case
probably due to mixing processes in the epikarstic and the vadose zone. Previous (two
weeks) rainfall <5180 varied between — ll%o and —13%o. The same applies to component
 Double porosity models in karstified limestone aquifers 273

separation with electrical conductivity, that produced an underestimate of the fast

component fraction as a result of mixing with higher conductive waters of the epikarstic
zone.
The temperature of the spring water probably behaves like a more conservative tracer,
as far as the fast component is concerned, because the temperature stored within the
epikarst adjusts itself more to the ambient surface temperature. In sum, a proportion of
approximately 10% for the fast component can be assumed with some confidence. The
isotope data do not only provide an estimate of the fractions of the two recharge
components, but also provide the output function for the regional transport of dissolved
constituents.

Temporal distribution of groundwater recharge

The input to the system at the water level can be derived from hydraulic considerations
for both components. The pressure pulse due to the fast water, analysed above (Fig. 10)
is closely related to the spring discharge. The arrival of the slow component, delayed by
the flow through the thick unsaturated zone can be described by comparing spring
discharge and groundwater hydrographs (Fig. 11).

r S/-

Distribution of Seepage
H II I H | I III H H
LApril 19 1.May 1989 I.June 1989 I.July 1989

Fig. 11 - Evaluation of the temporal distribution of the slow event water

components (seepage water).

The rate of seepage water (slow component) is higher than the discharge during periods
of rising water level and lower when the potentials drop. At times of no head change,
both quantities are equal. Together with the relative fractions of the two components and
the distribution with time, the recharge input and the inputs of dissolved constituents can
be described (Fig. 12).

MODELLING REGIONAL FLOW AND TRANSPORT

After evaluation of the hydraulic and transport input and output functions for the two
flow components and estimation of the aquifer parameters, modelling of the aquifer
system can be attempted.
274 Martin Sauter

Recharge Distribution

RecfW98 to M ^ t ( Bock»

R$c£iy yu to CondL^s

i aaiiiBB -PL
Day since Recharge Evenî

Fig. 12 - Distribution of recharge input in the model.

Flow modelling
The model selected is the one described by Teutsch (1989). It is a one-dimensional finite
difference double porosity model that allows flow in both continua. The code was
extended to allow for the necessary selective recharge into both continua, because only
then, could the distinct response of the two systems (different recession characteristics,
the distinct bimodal output function, intermediate maximum in the hydrographs) be
explained.
This type of recharge distribution also corresponds to the physical process, that water is
collected at the surface in depressions, providing a concentrated fast input, that exceeds
the proportional area of fractures and conduits at the surface. The model was also
modified to allow regional and time variable input and it consists of two times 22 cells.
The simulation run extended over a period of 25 years, and the spring flow and the
groundwater levels could be modelled reasonably well (Figs 13 and 14).
Only for the last two years, for which continuous records exist and are displayed, are
there some differences in the measured and modelled flow that are mainly a result of the
recharge input, that was calculated based on the rainfall of one station, assumed to be
representative for the whole catchment and due to the uncertainty in the release of
snowmelt from store. The differences between the modelled and the simulated
hydrographs can be explained by the homogeneity of the storage coefficient in the
vertical dimension that had to be assumed in the model. At the level of the observed
kinks in the field hydrograph, a change in storage coefficient is suspected.

The parameters with which the fit was achieved are:

Storage coefficient (matrix): 0.01 to 0.03

Storage coefficient (conduits): 0.0004
Hydraulic conductivity (matrix): 1*10"4 to 5*10"3 m/s
Hydraulic conductivity (conduits): 1*10"2
The lower values are generally further upgradient. The somewhat higher values for the
hydraulic conductivity compared with the field values, are a result of the fact that a
 Double porosity models in karstified limestone aquifers 275
2.8

2.6 Comparison of Model and Field Data

2.* Spring Discharps
Modal
23.

2.0

E 1-8

a 1.4

12
&
1.0

0.8

0.6

0.4

0.2

0.0
1987 1988 1989 'l990
Fig. 13 - Comparison of modelled and measured spring discharge.
constant saturated thickness of 30 m had to be assumed in the evaluation of a regional
average hydraulic conductivity. The exchange parameter does not have a physical
meaning in the model; the relative magnitude of the hydraulic conductivity of the matrix
blocks and the exchange term can only be fixed by running the transport model (see
below).

Comparison of Model and Field Data

Measured Hydrograph
Model Hydrograph

690 _ Borehole B14

T l s l 0 | N | 0 | j | F | M I A l M | J | J | A I S I O | N | 0 | J | F I M | A I M I J I J | A I S I O I N I D | J | F I M I A | M | J | J | A
f
1987 ' 1988 1989 1990
Fig. 14 - Comparison of modelled and measured water level fluctuations.
276 Martin Sauter

At low flow conditions, the observed regional steep gradient could only be maintained,
if the hydraulic conductivity was reduced to one fifth of the above value at a level of
about 10 m above the base of the model aquifer. This measure also prevented some model
nodes from drying up, which, although physically realistic, would produce unrealistic
results in a one-dimensional model.

Transport modelling
With the flow calculated by the model for the period from April to July 1989, and
assuming the recharge distribution as a tracer input function, the relative change in
concentration of a regional tracer was simulated with a random walk approach (Teutsch,
1989). The breakthrough curve obtained compares reasonably well with the observed
change in the ô 18 0 ratio (Fig. 15).
The porosities used corresponded with the respective storage coefficients of the flow
model. A dispersivity of 15 m was used for the transport in the matrix continuum, a
value which could be obtained from the tracer tests. The transport model also allowed
the calibration of the hydraulic conductivity of the matrix blocks, because if the
matrix/fracture exchange term chosen is too high, the tracer left the system even before
the beginning of May, which does not correspond to what has been observed. With the
same model, it was also possible to simulate the tracer test "Bitz", for which the bulk of
the tracer was injected directly into the conduits via a sinkhole.

1 .April 1989 1 .May 1989 1 .June 1989 1 July 1989

Fig. 15 - Simulated breakthrough of a regional tracer.

The "B8" test (Fig. 16), however, was somewhat problematic to simulate. Due to the fact
that the dye was injected directly into the groundwater and because the tracer probably
travelled for only part of the time within the matrix before reaching a major conduit, the
model described is unsuitable for the simulation of this kind of transport. Matrix
diffusion processes, which the model did not include, are also likely to determine the
shape of the breakthrough curve.
 Double porosity models in karstified limestone aquifers 277

Fig. 16 - Tracer test with input directly into the groundwater.

CONCLUSION
The presented results demonstrate that even a karst aquifer with its extreme flow systems
can be successfully modelled with a one-dimensional double porosity model, which can
therefore provide a valuable tool in the understanding of the flow and transport
processes. Some of the discrepancies between modelled and measured data can probably
be overcome by including a second vertical dimension.

However, the data input requires substantial time and efforts, and the continuous data
records necessary to calibrate the models are not always available. In most karst
groundwater basins, large numbers of boreholes, somewhat removed from the spring are
not available either.

ACKNOWLEDGMENTS
The work was undertaken with the financial support of the Federal State of Baden-
Wflrttemberg, within the research programme "Geowissenschaftliche Umweltforschung".
The author likes to thank Prof. Einsele for his scientific and administrative support in the
project. Thanks are also due to the "Landesanstalt fur Umweltschutz, Karlsruhe" (tracer
test "Bitz"), and to P. Merkel (tracer test "B8") for letting me use their tracer test data
and to my wife Susan for correcting the English.

REFERENCES
Aiguang, C , Hanchen, L., Juanming, X. & Jinli, S., 1988, Stochastic modelling of the
karst spring of Xin'an Shanxi Province. In: Karst Hydrogeology and Karst
Environment Protection (Proc. 21st Congress of the Int. Assoc, of Hydrogeologists,
Guilin, China, October 1988), vol. 2, 621-628. IAHS Publ. no. 176.
Ashton, K., 1966, The analysis of flow data from karst drainage systems. Trans. Cave
Res. Group GB, 7, 161-203.
Atkinson, T. C , 1977, Diffuse flow and conduit flow in a limestone terrain in the
Mendip Hills, Somerset. J. Hydrol., 19, 323-349.
278 Martin Sauter

Atkinson, T. C. & Smart, P. L., 1981, Artificial tracers in hydrogeology. In: A Survey
of British Hydrogeology 1980, 173-190. The Royal Society, London.
Avias, J. & Joseph, C , 1984, Modélisation des systèmes karstiques. In: Hydrogeology
ofKarstic Terranes (éd. by A. Burger & L. Dubertret). International Contributions
to Hydrogeology, Heise, Hannover.
Bakalowicz, M. & Mangin, A., 1980, L'aquifère karstique, sa définition, ses
caractéristiques et son identification. Mém. H. Sér. Soc. Géol, France, 11, 71-79.
Barenblatt, G. E., Zheltov, I. P. & Kochina, I. N., 1960, Basic concepts in the theory
of homogeneous liquids in fissured rocks. J. Appl. Math, and Mechanics (USSR), 24,
1286-1303.
Barker, J. A. & Black, J. H., 1983, Slug tests in fissured aquifers. Wat. Resour. Res.,
19, 1558-1564.
Benecke, P., 1978, Der Wasserumsatz eines Buchen- und eines Fichtenwaldôkosystems
im Hochsolling (Methoden und Ergebnisse). Universitât Gôttingen, Gôttingen.
Bourdet, D. & Gringarten, A. C , 1980, Determination of fissure volume and block size
in fractured reservoirs by type curve analysis. Presented at the SPE-AIME 55th
Annual Technical Conference and Exhibition, Dallas, Texas. Society of Petroleum
P Engineers, SPE-9293.
Dougherty, D. E. & Babu, D. K., 1984, Flow to a partially penetrating well in a double
porosity reservoir. Wat. Resour. Res., 20, 1116-1122.
Dreiss, S. J., 1989a, Regional scale transport in a karst aquifer: 1. Component
separation of spring flow hydrographs. Wat. Resour. Res., 25, 117-125.
Dreiss, S. J., 1989b, Regional scale transport in a karst aquifer: 2. Linear systems and
time moment analysis. Wat. Resour. Res., 25, 126-134.
Duguid, J. O. & Lee, P. C. Y., 1977, Flow in fractured porous media. Wat. Resour.
Res., 13, 558-566.
Haude, W., 1955, Zur Bestimmung der Verdunstung auf môglichst einfache Weise. Mitt.
d. Dt. Wetterdienstes, 11, Bd2, Bad Kissingen.
Hess, J. W. & White, W. B., 1974, Hydrograph analysis of carbonate aquifers. Res.
Publ. no. 83, Inst. Res. Land and Wat. Resour. Pennsylvania State University.
Hemme, H., 1970, Die Stellung der "lassivierten" Terra fusca in der Bodengesellschaft
der Schwâbischen Alb. PhD Thesis, Univ. Hohenheim.
Huyakorn, P. S., Lester, B. H. & Faust, C. R., 1983, Finite element techniques for
modelling groundwater flow in fractured aquifers. Wat. Resour. Res., 19, 1019-
1035.
Irmay, 1984, Theoretical models of flow through porous media. RILEM Symp. on the
Transfer of Water in Porous Media (Paris).
Kiraly, L., 1984, La régularisation de l'Areuse (Jura Suisse): simulée par modèle
mathématique. In: Hydrogeology of Karstic Terranes (ed. by A. Burger & L.
Dubertret). International Contributions to Hydrogeology, Heise, Hannover.
Long, J. C. S., Gilmour, P. & Witherspoon, P. A., 1985, A model for steady fluid flow
in random three-dimensional networks of disc-shaped fractures. Wat. Resour. Res.,
21, 645-658.
Maillet, E., 1965, Essais Hydraulique Souterraine et Fluviale. Hermann, Paris.
Neuman, S. P., 1987, Stochastic continuum representation of fractured rock permeability
as an alternative to the REV and fracture network concepts. In: Groundwater Flow
and Quality Modeling (ed. by E. Custodio, A. Gurgui, & J. P. Lobo-Ferreira), 331-
362. Reidel, Dordrecht.
Pankow, J. F., Johnson, R. L., Hewetson, J. P. & Cherry, J. A., 1986, An evaluation
of contaminant migration patterns at two waste disposal sites on fractured porous
media in terms of the equivalent porous medium (EPM) model. /. Contam. Hydrol.,
1 (1), 65-76.
Pinder, G. F. & Jones, J. F., 1969, Determination of the groundwater component of peak
discharge from the chemistry of total runoff water. Wat. Resour. Res., 5, 438-445.
Romm, E. S., 1966, Fluid flow in fractured rocks. Nedra, Moscow.
Shuster, E. T. & White, W. B., 1971, Seasonal fluctuations in the chemistry of limestone
 Double porosity models in karstified limestone aquifers 279

springs; a possible means for characterising carbonate aquifers. /. Hydrol., 14, 93-
128.
Smart, P. L. & Friederich, H., 1986, Water movement and storage in the unsaturated
zone of a maturely karstified carbonate aquifer: Mendip Hills, England. In: Proc.
Environmental Problems in Karst Terranes and their Solutions Conference. National
Water Well Association, Bowling Green, Kentucky.
Sklash, M. G. & Farvolden, R. N., 1979, The role of groundwater in storm runoff. /.
Hydrol, 43, 45-65.
Smith, L., Schwartz, F. & Mase, C , 1987, Applications of stochastic methods for the
simulation of solute transport in discrete and continuum models of fractured rock
systems. In: Proc. Conference on Geostatistical, Sensitivity and Uncertainty Methods
for Groundwater Flow and Radionuclide Transport Modelling, 425-440. San
Francisco.
Snow, D. T., 1965, A parallel plate model of fractured permeable media. PhD thesis,
Univ. of California, Berkeley.
Sokollek, V., 1983, Der Einfluss der Bodennutzung auf den Wasserhaushalt kleiner
Einzugsgebiete in unteren Mittelgebirgslagen. PhD thesis, Giessen.
Teutsch, G., 1989, Groundwater models in karstified terrains: two practical examples
from the Swabian Alb, S. Germany. Proc. 4th Conference - Solving Groundwater
Problems with Models (Indianapolis, USA, February, 1989), 929-953.
Thornthwaite, C. W. & Mather, J. R., 1957, Instructions and tables for computing
potential évapotranspiration and the water balance. Publ. Climatol., Drexel Inst.
Technol, 10.
Thrailkill, J., 1986, Models and methods for shallow conduit-flow carbonate aquifers.
Proc. Environmental Problems in Karst Terranes and their Solutions Conference.
National Water Well Association, Bowling Green, Kentucky.
Uhlig, S., 1959, Wasserhaushaltsbetrachtungen nach Thornthwaite. Z. Acker- und
Pflanzenbau, 109, 384-407. Berlin.
Villinger, E., 1977, Uber Potentialverteilung und Strômungssysteme im Karstwasser der
Schwabischen Alb (Ob. Jura, SW-Deutschland). Geolog. Jahrbuch, C18.
Warren, J. E. & Root, P. J., 1963, The behaviour of naturally fractured reservoirs. Soc.
Petrol. Eng. J., 3, 245-255.
White, W. B., 1977, Conceptual models for karst aquifers revisited. In: Hydrologie
Problems in Karst Regions (ed. by R. R. Dilamarter & C. S. Csallany), 76-87.
Western Kentucky Univ. Press, Bowling Green, Kentucky.
Williams, P. W., 1983, The role of the subcutaneous zone in karst hydrology. J.
Hydrol, 61, 45-67.
Zaltsberg, E. A., 1984, Forecast of karst water levels. In: Hydrogeology of Karstic
Terranes (ed. by A. Burger & L. Dubertret). International Contributions to
Hydrogeology, Heise, Hannover.
Rydroeeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHS Publ. no. 207, 1993. 281

AN EXTENDED DOUBLE-POROSITY CONCEPT AS A PRACTICAL

MODELLING APPROACH FOR A KARSTIFIED TERRAIN

GEORG TEUTSCH
Institut fur Wasserbau, Universitat Stuttgart, Pf 801140, D-7000 Stuttgart 80, Germany

ABSTRACT

The aim of this study was to investigate various modelling approaches applicable to the
simulation of groundwater flow and transport in the karstified limestone aquifer of the
Swabian Alb in southern Germany. Results from two karst areas showing significantly
different flow characteristics are presented. At the first site near Heidenheim, the
hydrogeological data indicated homogeneous aquifer conditions. A three-dimensional
numerical model with a minimum mesh size of 87 m was used to simulate the
groundwater flow pattern over a period of six years. The good agreement between
predicted and measured water levels demonstrates the ability of standard porous media
models to simulate the regional flow pattern in moderately karstified terrains. At the
second site near Reutlingen, the karstification of the Jurassic limestone aquifer is more
advanced resulting in more heterogeneous hydraulic conditions. From tracer tests,
groundwater flow velocities of up to and above 100 m/h are reported. However, stable
isotope investigations show that usually less than a quarter of the total flow contributes
to what is considered the fast flowing conduit type system. A double-porosity flow and
transport groundwater model was therefore developed to simulate the spring discharge,
the groundwater level fluctuations and some of the tracer breakthrough curves. One
porosity represents the moderately karstified aquifer zones with low conductivities and
high storativities. The other porosity represents the highly karstified areas yielding high
flow velocities but hardly any water level fluctuations. The double-porosity model was
able to adequately reproduce all observed characteristics of the highly karstified system
at the Reutlingen site.

INTRODUCTION
Karstified limestone terrains are in general very prolific sources of groundwater. Their flow
characteristics have been studied in much detail at numerous locations. The complexity of
the flow system is primarily determined by the degree of karstification. Depending on
lithology, type and time of aquifer exposure to erosion and the prevailing structural
features, a fissure/conduit type or a diffusive fracture flow regime may dominate.

A classification scheme for carbonate aquifers, based on the type of flow, recharge, and
storage was developed by Smart & Hobbs (1986). Thrailkill (1986) provides an overview
of different (single "porosity") modelling concepts and their applicability to karstified
aquifers. He distinguishes between a granular (porous) aquifer type and three karst aquifer
types with flow conditions ranging from laminar Darcy to turbulent pipe-flow.

However, hardly any practical application of these concepts within a quantitative karst
water resources study has been reported so far. The aim of this paper is to investigate in
theory and practice various modelling concepts for the simulation of groundwater flow
and transport in karstified terrains.

FRACTURE FLOW MODEL CONCEPTS

A variety of modelling concepts has been proposed for the simulation of the hydraulic
282 Georg Teutsch

behaviour of fractured rocks. The concepts are based on either a discrete or continuous
representation of the fracture flow system, together with equations describing the
exchange between the fractures and the rock matrix.
In the discrete fracture approach the location and the geometry of the fractures is assumed
to be either exactly known, or at least describable through a statistical model. Several
analytical and numerical studies have been conducted employing the discrete fracture
approach, including early laboratory and theoretical works by Wittke & Louis (1968),
Gringarten (1971), Grisak & Pickens (1980), Smith & Schwartz (1984). Most of these
studies dealt with crystalline rocks or some other low-permeability system. It was
therefore in general assumed that groundwater flow occurs in the fracture system only,
the advective flow within the porous matrix being negligible.

The major drawback of the discrete fracture approach is the necessity to describe the
geometry and hydraulic properties of all individual fractures within the flow domain. For
most regional aquifer systems this is not feasible in practice. So far, only a few very
intensively investigated radioactive waste disposal sites of very limited areal extent have
been characterized using the discrete fracture modelling approach. Bearing in mind that
fracture permeability is proportional to the square of the fracture width, no reliable
regional rock permeability assessment is to be expected using indirect underground
investigation techniques available today (e.g. geophysics). In the continuum approach it
is assumed that the fractured medium can be described by one or more porosity
equivalents. Often a double-porosity approach is employed, one porosity representing the
low-permeability porous or fractured rock matrix, the other porosity representing the
higher permeability fracture system.

The double-porosity model concept was introduced by Barenblatt et al. (1960). It was
extensively used and further developed in oil industry studies including works by Warren
& Root (1963), Kazemi (1969), De Swaan (1976), Streltsova-Adams (1978), and van
Golf-Raacht (1982). At first, the magnitude of the cross-flow between the matrix and the
fracture system was assumed to be directly proportional to the lumped pressure difference
between the two systems. The implicit assumption of an instantaneous pressure adjustment
was later refined in models which represent also the matrix-block storativity. Since most
of the studies dealt with systems of comparatively low matrix permeability, advection was
generally assumed to take place in the fracture system only.

SINGLE-"POROSITY"

porous continuum discrete fractures

or
fracture continuum

DOUBLE-"POROSITY"

Pores * Fractures Fiss-Jres » Fractures

1) porous continuum 1 ) discrete fractures
2) discrete fractures 2) discrete fractures
or or
1) porous continuum 1) fracture continuum
2) fracture continuum 2) fracture continuum

Fig. 1 - Schematic representation of single- and double-porosity aquifer

systems.
 An extended double porosity concept as a practical modelling approach 283

Figure 1 shows four pictograms (simplified physical representations) of single- and

double-porosity systems. A possible range of model representations for fractured or
karstified aquifers is shown in Fig. 2.

In karstified aquifers, groundwater flow is restricted mostly to fractures and fissures or

conduits. The rock matrix, usually carbonates or evaporites, is in general impermeable.
Therefore, the effective hydraulic components within such a system may be represented
through fracture networks with varying width, length and orientation. A simple realization
of such a system is shown in the lower right pictogram of Fig. 1. In an extended double-
porosity approach the discrete fracture networks are modelled as two overlapping
continua, where advection and dispersion may occur in both systems. One continuum
represents the small aperture low-conductivity fracture system and the other continuum
represents the high-conductivity large conduits.

In Fig. 3 a one-dimensional double-continuum or extended double-porosity system is

shown. The upper block with the steep hydraulic gradient represents the fracture system
whereas the lower block with the shallow hydraulic gradient represents the conduit
system. Both systems are connected through hydraulic resistances which are determined
by the exchange coefficient.

The respective three-dimensional double-continuum partial-differential flow equations are:

9 ,, a dh v d ,, a dh*
—(k ) -(k^d-) =\ — a 0 (h a - h b ) (1)
9x^ xx
dx ' ây-^'ây
b
b dh1
f(kxb, 9hJ
dx dy (fcyy-jj-) +
— k
dz(v x x 3 dz
—)
cb3h
<*o(ha - h b ) (2)
dx
where :
x,y,z = Cartesian coordinates,
hydraulic head [L],
k k k = hydraulic conductivity [L/T],
transmissivity [L 2 /T],

SINGLE POROSITY DOUBLE TOROSITY- FRACTURE NETWORKS

ONE COMTMtiUM

l-ICONVECTIOH • tXSTCRSION] 2-(CONVECTKW • U 8 K M I O N ] 1-fCOMVECTION • USP£ll8ION/OIFFuatO»|

K£J1£L MMUMiMTfi
OF DHÇMTE FRACTURES

flCONVECTKM • 0 O K M 9 0 N ] «•(CONVECTION • DSWSRKOjJ

( inumuuntmiMtmim > < iMmuMuamjumtm I

DETAIL AND SCALE OF INVESTiSATION

STOCHASTIC
REV MO0SLS
PRACTICAL APPLICABILITY OF
CONCEPT mSCSETE
FRACTURES
REPRESENTATION OF HETER08EMEITIE8

Fig. 2 - Model conceptualization of fractured and karstified aquifer

systems.
 Georg Teutsch
Groundwater Recharge

Fracture-System

Fissure-System

Fig. 3 - One-dimensional double-continuum model.

= specific storage [1/L],

S = storage coefficient [-],
«o = specific exchange coefficient [1/(LT)],
a = index of fracture-flow system (porosity 1),
b = index of conduit-flow system (porosity 2).

CASE STUDIES
Two areas located on the Swabian Alb aquifer outcrop, showing significant differences
in the flow characteristics were selected for a comparative study. With respect to the
prevailing groundwater flow conditions, these areas are believed to represent opposite
extremes of the existing karst aquifer on the Swabian Alb.

Figure 4 presents a generalized hydrogeological northwest-southeast section through the

Swabian Alb outcrop area. The term "Deep Karst" refers to that zone of the aquifer
where the piezometric levels do not reach the aquifer bottom, the head gradient being
controlled by the topographical level of the existing rivers. In the "Shallow Karst" the
aquifer piezometric levels are controlled by the topographical position of the underlying
aquiclude (oxl-marl beds). It should be noted that the underground water divide forms
the boundary between the "Deep Karst" in the southeast and the "Shallow Karst"in the
northwest.

"Deep Karst" at Heidenheim

The study area is located in the so called "Deep Karst" north of Heidenheim. Due to a
limited shallow karstification the vast majority of the groundwater flow is believed to
occur in a network of mostly narrow fractures. The available hydrogeological data
indicate fairly homogeneous aquifer conditions, continuous at a scale larger than what
may be considered a REV (Representative Elementary Volume), probably in the range
of 10 to 100 m.
 An extended double porosity concept as a practical modelling approach 285
m
Swabian Alb

20 JO 40 km

Fig. 4 - Generalized hydrogeological section through the Swabian Alb

Jurassic aquifer (Villinger, 1977).

The study area' comprises about 60 km2 of karst aquifer outcrop, together with
Quaternary sediments in the River Brenz valley. Elevations vary between 450 m in the
valley and 650 m in the surrounding hills.

A sequence of upper Jurassic limestone and dolomitic limestone with interbedded marl
beds forms the karst aquifer system. The aquifer thickness varies between 40 m along the
axes of a northeast-southwest striking structural anticline in the north, and up to 170 m
in the southern part of the study area. In general, the limestone formations exhibit no
primary porosity (sedimentary), but a well-developed secondary (structural) and tertiary
porosity (solution), especially in the shallow parts of the aquifer. The limestone aquifer
is unconfined outside the Brenz valley, receiving recharge from precipitation that
percolates through an unsaturated zone, which reaches 70 m near the catchment
boundaries.

The Pleistocene sand, silt, and gravel sediments of the Brenz valley build up to
thicknesses between 40 and 50 m. The valley sediments consist of several layers of
coarse elastics, exhibiting high permeabilities and high storativity. These are interbedded
with finer low permeability silt and clay layers.
The general flow direction in the karst aquifer follows the structural and topographical
slopes with dip towards the southwest on the eastern side of the Brenz valley, and dip
towards the southeast on the western side of the valley. The groundwater level
fluctuations range between less than 1 rn near the river and more than 10 m near the
catchment boundary in the northeast.

Groundwater recharge was calculated using a simple soil moisture-field capacity water
balance. The calculation was performed based on daily values of precipitation, air
temperature, and relative humidity. Results were calibrated against the occurrence of
groundwater level peaks and regional water balance data. During the six-year study
period from 1980 to 1985, average groundwater recharge amounted to about 37% (258
mm/year) of precipitation. The high recharge percentage is due to the sparse soil cover
and the lack of surface water flow, which is typical of karst terrains.

Flow model
A three-dimensional single-porosity flow model covering the whole study area was
employed to simulate the flow in the karst and valley aquifer systems. Using a three-
dimensional instead of a two-dimensional horizontal model enabled the inclusion of the
286 Georg Teutsch

vertical flow occurring in and around the valley alluvium, as well as the observed
vertical permeability variations in the karst aquifer.
The model calculates the movement of groundwater of constant density through a single-
porosity media, as described by the partial differential equations (1) or (2) omitting the
exchange term on the right-hand side. Groundwater flow is simulated utilizing a block-
centred finite-difference approach. The finite-difference equations are solved by a
successive over-relaxation technique. Horizontal grid spacing varied from 87 m around the
"7. Fuss" water works to 1050 m in the eastern and south-western parts of the modelled area.
The vertical discretization was chosen such that layers 1 to 3 comprise both the limestone
and the alluvial aquifer, whereas, layers 4 to 8 comprise the limestone aquifer only. The
entire grid is composed of 7812 cells arranged in 39 rows, 35 columns, and 8 layers.

The model boundaries were chosen to coincide in the north, east, and west with the
catchment boundaries. No flow conditions were assumed along these boundaries. The
absence of a convenient boundary in the southeast required that an arbitrary location be
selected for the model boundary. This boundary was located along an imaginary line
formed by three observation wells. It was incorporated in the model as a time variant
head boundary, the heads being determined through linear interpolation from three
observation wells.

Initial conditions were adjusted using a dynamic water balance procedure (Rushton,
1979). The six-year monthly averages of recharge and abstraction were used for a 15
years model run in order to put the model into a balanced initial state, prior to the
historical period. For aquifers exhibiting high groundwater level fluctuations, using
steady-state initial heads instead, may cause considerable water balance errors within the
first phase of the historical simulation period.

The calibration of the model consisted of matching six years of weekly groundwater level
measurements at 26 observation wells, located in the karst and the valley aquifer. This
resulted in 300 model time steps each 7 days long. Longer time steps could not be used
due to the rapid response of the karst system to precipitation events. The observation well
network is presented in Fig. 5. Twelve wells are completed in the alluvial aquifer only,
five are completed in the alluvium but reach the underlying karst aquifer, and nine are
completed in the karst aquifer only. The majority of the observation wells are located
around the "7. Fuss" water works, where information requirements are highest.

The transmissivity values for the karst aquifer range between 0.000 35 and 0.0013 m 2 /s.
Specific yield values for the karst aquifer ranged between 0.003 and 0.04. Aquifer
parameters that were varied in the simulation were limited to the aquifer hydraulic
conductivity and river leakage. Boundary conditions that were varied included the
recharge rates and the recharge distribution. Parameters were adjusted within reasonable
ranges until a satisfactory match was obtained between observed and modelled hydraulic
heads. Subsequently, a series of sensitivity simulations were conducted to assess the
significance of each individual parameter.

The final hydraulic conductivity distribution shows a dramatic increase of the values from
the catchment boundaries towards the Brenz valley. This can be explained by the
increased chemical dissolution potential in areas of flow concentration. Another factor
might be the increased porosity due to structural lineaments located along the river
valley. In the vertical direction the hydraulic conductivity decreases by approximately
a factor of 2.

The groundwater flow velocities as predicted by the finite-difference model varied

between 0.5 m/day and 8 m/day. These values were in good agreement with velocities
obtained from short distance tracer tests conducted by underground injection. The flow
 An extended double porosity concept as a practical modelling approach 287

Fig. 5 - Observation well network.

velocities also agree with the calculated average groundwater residence times, as
determined from tritium measurements. Together with the overall good match between
observed and modelled hydraulic heads throughout the study area, this demonstrated the
ability of the standard porous media model to simulate the regional groundwater flow in
a moderately karstified aquifer.

Figure 6 shows the three-dimensional flow pattern as it appears at one time-step (23
March 1983) during the six-year simulation period.

"Shallow Karst" at Reutlingen

The study area is located in the so called "Shallow Karst" south of Reutlingen. In the
"Shallow Karst" the flow regime is characterized by slopes of the piezometric surface too
steep to be maintained by the high flow velocities observed in artificial tracer tests. A
considerable part of the flow is believed to occur in fissures or conduits enlarged by
solution and to be temporarily non-Darcian after recharge events. Based on the available
field evidence, showing high fluctuations of the piezometric levels together with very
high groundwater flow velocities, physical reality was approximated using the concept
of two overlapping fracture system continua. The high-conductivity porosity represents
the more karstified areas (fissure and conduits), yielding high groundwater flow
velocities but hardly any water level fluctuations. The low-conductivity porosity
represents the moderately karstified aquifer zones (fractures) which may yield
considerable water level fluctuations.
288 Georg Teutsch

_g3Q Heldsnholm-Ma^ell ° Run 8 8 , SCop 1BS CB303233, Loyor 1

tf*° ,ff

Fig. 6 - Simulated hydraulic heads - 23 March 1983.

The study area comprises the Echaz (110 km2) and Reissenbach (2.5 km2) catchment
areas located side by side along the northern escarpment line of the Swabian Alb
limestone plateau. Figure 7 shows the general outline of the catchment areas.

Fig. 7 - General outline of the Echaz and Reissenbach catchments with

tracer test locations (Villinger, 1977).
 An extended double porosity concept as a practical modelling approach 289

The karstified upper Jurassic limestone, dolomite, and marl beds form the aquifer of
almost 170 m thickness. In general, groundwater flows in a north and northwest direction
towards the escarpment line. However, the flow directions are not known in much detail,
due to the very limited amount of hydraulic head information. The aquifer is unconfined
and exhibits groundwater level fluctuations of up to 15 m during recharge events. The
fast and intensive response to recharge events may be expressed by the wide range of
spring discharges. The Echaz springs, draining the larger catchment show a high-
flow/low-flow ratio of 22 to 1. The Reissenbach spring, draining the smaller catchment
shows a flow ratio reaching even 73 to 1.

Daily groundwater recharge was calculated using the same procedure as for the
Heidenheim study area. Through calibration, a field capacity value of 40 mm was
obtained. The resulting average groundwater recharge amounted to more than 50% (470
mm/year) of precipitation.

More than a dozen successful tracer tests were conducted in the Echaz catchment area
over the last 25 years. In most cases, a tracer dye, usually fluorescein, was flushed into
sinkholes or quarries. Subsequently, the tracer concentration was observed at the
catchment outlet (spring or river) over a period of time. From breakthrough curves,
maximum tracer velocities between 17 m/h and more than 200 m/h were determined.
Tracer recovery varied between 1% and 38%.

Data available for model calibration included spring flow measurements, approximate
hydraulic head data, tracer breakthrough curves, and stable isotope hydrograph separation
data.

The stable isotope hydrograph separation technique is based on the occurrence of natural
isotope variations in rainfall water, which is traced through the underground to
distinguish between pre-event and event water. This method was extensively used to
determine the percentage of groundwater flowing in the fast fissure- and the slower
fracture-system.

The results showed that during peak discharge only about 30% of total spring flow
originated from the precipitation event itself, the rest comprising groundwater which was
recharged prior to the event. Since groundwater levels are highest in both systems during
peak-discharge, the 30% is likely to be equal to the maximum flow capacity of the
fissure-system.

The following assumptions were made during model calibration:

— The hydraulic head variation is caused primarily by the low-conductivity porosity.

— The high groundwater flow velocities observed during the tracer tests occur in the
high-conductivity porosity.
— Recharge enters the low-conductivity system only, the relative number of enlarged
fissures reaching the ground surface being small.
— The maximum percentage of event-water, determined using the stable isotope
hydrograph separation method, is equal to the flow occurring in the high-conductivity
system.

The simulated Echaz spring discharge, applying the extended double-porosity flow model
is shown in Fig. 8. It should be noted that the fissure-system flow percentage increases
during low flow conditions, the fracture-system being drained by the fissures.

To model the tracer breakthrough curves the extended double-porosity flow model was
290 Georg Teutsch

Fig. 8 - Simulation of the spring flow at the Echaz spring using an

extended double-porosity model.

extended to a transport model. The transport in both "porosities" is described by the

partial-differential equations:

_d_ ( D a 3 C '
.— + « " ( C 8 - Cb) = — (3)
dx R a 9x R. > dx dt

_d_ ( p b a c b vab a c b a*(Ca pb\ _ 9C (4)

dx p b dx R,b dx
R;
where:
D = hydrodynamic dispersion coefficient [L2/T],
C = concentration [M/L 3 ],
Rd = retardation coefficient [-],
va = interstitial velocity [L/T],
m = saturated thickness [L],
ne = effective porosity [-],
a = exchange coefficient [1/T],
a = a(h a - h b )/(n e * m).

Convection, as well as longitudinal dispersion is assumed to occur in both porosities.

Mathematically, the formulation is similar to a two-layer model. The transport equations
were solved using a random-walk particle tracking technique.

Figure 9 shows the observed and simulated tracer breakthrough curves for one of nine
tracer tests conducted within the Echaz catchment during low flow conditions. Based on
the concentrations observed at the Echaz springs a mean tracer flow velocity of
approximately 100 m/h was determined. Such high groundwater flow velocities,
indicating a fissure-type flow regime, were repeatedly observed in the study area. For
this tracer experiment the overall tracer recovery was 31 %, accumulated within the 7-day
observation period.

The transport simulation, which is based on the calibrated extended double-porosity flow
model was carried out using a 30-minute time step. To achieve an optimum match
 An extended double porosity concept as a practical modelling approach 291

ECHAZQUELLEN
C [NPART/QJ
BOD.O

SOO.O

400.0
o = simulated
— = observed
30D.O
3600m
ISO m/h
I00m/h
saa.a 10 m
ISO m

-lOO.O

.o
[Std.]
.O 24.0 4B.O 72.0 gS.D 1SO.O 144.O
RUN-3 E-B310D1 POS-2DO OIV-24

Fig. 9 - Observed and simulated tracer breakthrough curves for tracer test
no. 7 (Kleinengstingen).

between the observed and simulated tracer concentration data, 27% of the tracer mass
had to be entered directly into the fast flowing porosity (fissure-system) system, the
remaining 73% being entered into the slower porosity (fracture-system). According to
the model, only 4% of the tracer mass was exchanged from the slow to the fast flowing
system during the observation period of seven days. The rest of the tracer mass caused
a low concentration tailing over a period of almost three months. The small cross-flow
was due to the minor pressure differences which occurred in the two porosities during
low flow conditions (low groundwater heads in both systems). Low concentration
tailings, as predicted by the model, were repeatedly observed for a long time after the
main tracer breakthrough but have not been quantitatively monitored in the study area.
The tracer flow velocities determined in the study area are positively correlated with the
spring discharges observed during the tracer tests. Other factors influencing the shape
of the tracer breakthrough curve include; the distance between the location of input and
the observation point, as well as the size of the catchment area. Pronounced bimodal
breakthrough curves are predicted by the extended double-porosity model for short
distance tracer tests conducted in small catchments during high spring flow conditions.
These types of curves were repeatedly observed in karstified terrains and should be
regarded as a clear indication for the existence of a multi-"porosity" system.

CONCLUSIONS
The regional groundwater flow in the moderately karstified aquifer at Heidenheim could
be adequately described by a standard porous flow numerical model. The three-
dimensional model demonstrated clearly the hydraulic significance of the alluvial aquifer
system for the regional piezometric levels in the karst aquifer. The existing model,
calibrated for a period of six years, will be used for the simulation and prediction of
aquifer responses to water resources development measures. However, additional closely
monitored tracer tests are required in order to provide the data required for aquifer
contamination risk assessments.
292 Georg Teutsch

An extended double-porosity approach comprising two overlapping fracture continua was

employed to simulate the flow regime in the well-developed karst aquifer system south
of Reutlingen. The model adequately describes the observed field data. These included;
total spring flow together with the pre-event and event water distribution from isotope
measurements, the piezometric level fluctuations, and the groundwater flow velocities
observed in artificial tracer tests. Model simulations showed clearly that the groundwater
flow velocities obtained from an artificial point tracer input are not comparable with the
flow velocities obtained from areal natural tracer inputs (stable isotopes). Further field
investigations, including short distance well to well tracer tests are planned to provide
a better insight into the karst flow regime.

REFERENCES
Barenblatt, G. I., et al., 1960, Basic concepts in the theory of seepage of homogeneous
liquids in fissured rocks (in Russian). Prikl. Mat. Mekh., 24 (5), 852-864.
De Swaan, H., 1976, Analytical solution for the determination of naturally fractured
reservoir parameters by well testing. Soc. Pet. Eng. J., 16, 117-122.
Gringarten, A. C , 1971, Unsteady-state pressure distributions created by a well with a
single horizontal structure, partial penetration, or restricted entry. PhD thesis,
Stanford University.
Grisak, G. E. & Pickens, J. F., 1980, Solute transport through fractured media: 1. The
effect of matrix diffusion. Wat. Resour. Res., 16 (4), 719-730.
Kazemi, H., 1969, Numerical simulation of water-oil flow in naturally fractured
reservoirs. Soc. Pet. Eng. J., 9, 317-325.
Rushton, K. R., 1979, Starting conditions for aquifer simulations. Ground Water, 11, 37-
42.
Smart, P. L. & Hobbs, S. L., 1986, Characterisation of carbonate aquifers: A
conceptual base. In: Proceedings of the Environmental Problems in Karst Terranes
and Their Solutions Conference (Bowling Green, Kentucky), 1-14. Published by
NWWA.
Smith, L. & Schwartz, F., 1984, An analysis of the influence of fracture geometry on
mass transport in fractured media. Wat. Resour. Res., 20 (9), 1241- 1252.
Streltsova-Adams, T. D., 1978, Well hydraulics in heterogeneous aquifer formations.
In: Advances in Hydrosciences (ed. by V. T. Chow), vol. 11, 357-423.
Thrailkill, J., 1986, Characterisation of carbonate aquifers: A conceptual base. In:
Proceedings of the Environmental Problems in Karst Terranes and Their Solutions
Conference (Bowling Green, Kentucky), 17-31. Published by NWWA.
Van Golf-Raacht, 1982, Fundamentals of Fractured Reservoir Engineering. Elsevier,
Amsterdam.
Villinger, E., 1977, Uber Potentialverteilung und Strômungssteme im Karstwasser der
Schwâbischen Alb. Geol. Jb., C18, 3-93.
Warren, J. E. & Root, P. J., 1963, The behaviour of naturally fractured reservoirs. Soc.
Pet. Eng. J., 3, 245-255.
Wittke, E. & Louis, C , 1968, Modellversuche zur Durchstrômung kluftiger Medien.
In: Aktuelle Problème der Geomechanik und deren theoretische Anwendung (ed. by
L. Miiller). Springer Verlag.
6 Regional Systems
%,
Hydrogeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHS Publ. no. 207, 1993. 295

RESULTS FROM STUDIES IN SEVERAL KARST FORMATIONS IN

SOUTHERN CATALONIA (SPAIN)

EMILIO CUSTODIO, ALFONS BAYO, MANEL PASCUAL & XAVIER BOSCH

Polytechnic University of Catalonia, Department of Engineering and International Course on Groundwater
Hydrology, Gran Capita sin, Môdul D-2, 08034 Barcelona, Spain

ABSTRACT
Catalonia, in northeast Spain, has several important permeable carbonate formations that
contain karstic features. They provide important water resources and are subject to
intensive exploitation. The paper considers the results of studies carried out during the
last 10 years in several areas: the Garraf and Vandellos coastal massifs of Jurassic and
Cretaceous age, the southern pre-littoral range of Triassic and Eocene age, and the deep-
seated Cretaceous carbonates of the Plana de la Galera area, near the coast. Studies in
the coastal formations point to the importance of ion exchange and organic matter
oxidation in the generation of carbonate dissolution capacity, that may overcome the
effect of fresh-water and salt-water mixing in the enhancement of the corrosiveness.
Local tectonic disturbances and possibly the dissolution of gypsum are also important in
other areas, and control the development of major flow paths. Sedimentary discontinuities
explain the presence of buried karsts that if non-obliterated may constitute major regional
aquifers.

INTRODUCTION
Catalonia is situated along the Mediterranean coast in northeast Spain; it has a length of
350 km. Numerous, important carbonate formations are found, corresponding to the
Pyrenees, Catalanides and Maestrat mountain ranges (Triassic, Jurassic, Cretaceous and
Eocene) and Miocene deposits in the tectonic depressions (Fig. 1). Most formations have
a thickness of several hundred metres. The limestones and dolomites are frequently
between marls and even gypsum and anhydrite layers (Solé-Sebarfs, 1958; Anadon et al.,
1979; AIH, 1989). Some areas present conspicuous karst features and cavities but others
are dominated by fracture permeability characteristics. Non-obliterated old sedimentary
discontinuities subjected to karst processes under subaerial environment have good water
yielding capacity.
Hydrogeological studies have been carried out since 1965 by the Eastern Pyrenees Water
Authority and the Public Works Geological Service office in Barcelona, both presently
incorporated into the Water Authority of Catalonia, with the cooperation of the
Polytechnic University of Catalonia and the International Course on Groundwater
Hydrology (REPO, 1970; PHPO, 1985). Most of these studies have been directed to solve
real supply problems or to provide the basis for groundwater use planning, but recently
more research oriented hydrogeological works have been done by university teams.

The water demands of the relative dense population of the coastal Catalan areas,
supporting the large conurbation of Barcelona (about 4 million), important industrial
belts, popular tourist resorts and intensively irrigated agricultural areas, have resulted
in the over exploitation of aquifers as a source of fresh water. There are numerous
examples of large cumulative drawdowns and sea-water intrusion. Also, badly planned,
uncontrolled and even criminal pollution activities are common near the large urban,
industrial and tourist (mainly weekend) areas. Thus groundwater contamination problems
also need to be considered (Custodio, 1982).

The complicated geology and tectonics and the mountainous character of the area, with
296 Emilio Custodio et al.

Fig. 1 - Location map of Catalonia with main areas and mountain ranges.
The lower figure is a simplified map indicating the main carbonate
formations of hydrogeological interest in southern Catalonia. T = Triassic;
J = Jurassic; C = Cretaceous; E = Eocene.

ranges parallel and close to the sea, allows only for small aquifers, thus aggravating local
problems. The present paper deals with some examples taken from southern Catalonia,
with relatively low rainfall (500-550 mm/year) and moderate recharge (10-50 mm/year
under most common circumstances). They do not include the metropolitan area of
Barcelona, although the Garraf massif is just at the southwest edge. Carbonate aquifers
are mostly of the open type, after the classification of Bayô (Bayô, 1982; Bayô et al.,
1986).

Further to more classical studies on hydrogeology and groundwater resources, some

research has been done in the southern Garraf and Vandellôs massifs. These studies were
in parallel to research carried out on Mallorca Island (Price, 1988; Herman et al., 1985),
and all were part of a cooperation between the US Geological Survey (Reston) and the
Polytechnic University of Catalonia (Back et al., 1989). One of the goals was to extend
studies done in Florida and Yucatan (Back &Hanshaw, 1970a,b, 1982, 1983; Back et al.,
 Results from several karst formations in southern Catalonia 297

1979, 1981, 1984, 1986; Hanshaw & Back, 1979, 1980a,b) and elsewhere (Cotecchia,
1977) to other areas in order to understand the role of mixing fresh water and marine
water on the enhancement of karstic processes in carbonate coastal aquifers, following
the original ideas of ion strength changes (Wigley & Plummer, 1976; Plummer, 1975;
Sanford & Konikow, 1989; Herman, 1986).

These studies have been based mainly on chemical equilibria studies and mass balances
(Plummer & Back, 1980), but the framework was obtained by more conventional
chemical studies and water salinity logging, with the aid of environmental isotopes. The
existence of vertical flows and water renovation inside the boreholes has been checked
by means of radioactive tracers ( m I ) using known simplified techniques (Baonza et al.,
1979; Custodio & Llamas, 1976, Sect. 12; Custodio, 1981; Halevy et al., 1966; Plata,
1982).

Processes in the saturated zone that has been taken into account, further to those leading
to ion strength effects, are cation exchange, redox processes and dolomite
formation/dissolution (Custodio & Bayô, 1989; Custodio, 1986). Cation exchange is a
common process if negative charges exist on mineral surfaces and edges (Sayles &
Mangelsdorf, 1977; Mercado, 1985; Magaritz & Luzier, 1985) and dolomite-related
processes have also received attention (Hanshaw et al., 1971; Magaritz et al., 1980;
Shatkay & Magaritz, 1987; Ward & Halley, 1985).

THE GARRAF MASSIF

The Garraf massif is just to the southwest of Barcelona and constitutes a sparsely
populated area due to its abrupt relief and lack of suitable soil, except along the coast
when some flat land is available, and along the contact with the Penedés graben. From
northeast to southwest carbonate formations (dolomites and limestones) with clay and
gypsum layers, dipping to the southeast, are successively covered by Jurassic dolomites
and then by Cretaceous limestones and dolomitic limestones, with calcarenite and
conglomerate Miocene sediments filling some depressions (IGME, 1982). The massif is
the result of a south-north regional overthrust, probably above Paleozoic schists, and
consequently is intensively faulted and has horizontal displacements along vertical planes
normal to the coastline.

The dominant bare carbonates, especially in the eastern part, present conspicuous karstic
features, both in surface and in depth, but sinkholes, some of them more than 100 m
deep, generally do not reach sea level. But some submarine caves exist, developed mainly
along the horizontal displacement faults, that actually are the main water drains of the
massif (Custodio, 1975).
The largest submarine spring is La Falconera (Astier, 1970, 1971) that discharges
brackish water with a mean fresh-water content of about 400 1 s"1 (Fig. 2). Sea water
enters the La Falconera main duct through the small network of fissures that link the cave
with the sea bottom. The greater specific gravity of sea water accounts for the small salt-
water flow that converts the fresh water into brackish water (Batista et al., 1983;
Custodio & Llamas, 1976, sect. 13; Custodio & Bruggeman, 1986), and precludes the
use of these water resources near the coast to solve existing water supply problems.
Several attempts have failed.

A large sanitary landfill was constructed in 1973 in the eastern part of the massif, at
relatively high elevation (Fig. 3), on Jurassic dolomites (CAPO, 1973). At the start of
the operation, a part of the leachates not collected by the bottom drainage system
percolated through the fissures and karst-enlarged fractures and moved laterally towards
the La Falconera and nearby areas, about 4 km away (Petit, 1981).
298 Emilio Custodio et al.

Fig. 2 - Cross section through La Falconera submarine spring (modified

from Astier, 1970) indicating fresh-water flow and salt-water flow
originating offshore due to the higher sea-water specific weight (Custodio
& Bruggeman, 1986). Surveys made by speleologists do not show a clear
salinity vertical stratification in the water; it may be due to turbulent flow,
at least in some sections. The map shows the assumed groundwater flow
paths in the active saturated zone (after Soler, 1984).

The first symptoms appeared after eight months in the main spring in the form of a
hydrogen sulphide smell, dark colour of the water (probably due to elevated sulphur and
tiny iron sulphide particles) and loss of dissolved oxygen, as well as an increase in
bicarbonate and hardness in local well water. It was shown that water discharging in La
Falconera was depleted in sulphate with respect to the mixture of fresh water and sea
water (Fig. 4) as a consequence of the strongly reducing underground environment
created by infiltrating leachates (Custodio & Galofré, 1976; Custodio, 1981; Hoyos-
Limôn & Palomero, 1975; Cebollada & Urcola, 1978).

Karst caves and sinkholes near the landfill were soon invaded by methane gas, and
studies for its recovery for energy purposes were commissioned. Leachate collected by
 Results from several karst formations in southern Catalonia 299

1 Water table elevation,m

x $ 3 Bedrock above sea level

Bedrock above 50m elevation

f* Quarry

'.\ _ ~ — __ ~~""~C" S - •—fcj—JtiiiC^- / / / / j

•3 — _

' * ? \ « VC^'GARRA*
J_?
La Falconera

VAU.CARCA

Aigua Dotç

Fig. 3 - The elevation of the groundwater table in the northeast of the

Garraf area. The bedrock elevation (Lower Triassic formations acting as an
impervious base) is indicated. Anisotropic permeability deviates
groundwater flow westwards. Intensive quarrying activities produce local
piezometric depressions in low permeability areas due to porosity increase
(CAPO, 1973; Custodio & Galofré, 1976).
the landfill drains showed a total chemical oxygen demand equivalent to about 130 g L 1
of 0 2 (Custodio, 1981). The depletion of sulphate in the La Falconera needs an
infiltration of about 1 1 s'1 of leachate, in spite of the valley floor on which the landfill
is placed being previously lined with cement and a layer of low permeability soily
material, a costly and rather ineffective exercise. Observations show that contaminants
move at a low angle to groundwater head contours due to anisotropic regional
permeability. This is a consequence of the normal (gravity) faulting orientation of the
carbonate materials.

In the Garraf carbonate formations, after heavy rains - mainly in early autumn - data
from the Catalonia Water Authority monitoring groundwater network show fast
groundwater head increases, with rapid descending limbs (Fig. 5). Although no detailed
studies have been carried out, there is evidence to confirm that this is due to the
formation of perched aquifers that also collect in the long screened monitoring boreholes.
Small inflows in the upper part can produce large water level increases if the lower
section is open to permanently saturated, low permeability carbonates. This is more
likely far from the sea, where karstic development is less important below the water table
due the high elevation. This also explains the large differences between the monitoring
boreholes (Llamas, 1968; Nifierola & Fuster, 1969; PHPO, 1985; REPO, 1970).

The productivity of the wells varies widely as can be expected in fissured and karstified
rocks. Most of the wells are in or close to the coastal plains. A plot of well productivity
versus position of the well bottom relative to sea level (Fig. 6) shows that most of the
wells penetrate just a few metres below sea level in order to get the required yield
(generally a few 1 s"1). Some wells unable to get enough water at this penetration were
drilled deeper, but generally did not attain the desired results (Pascual, 1990; Pascual et
300 Emilio Custodio et al.
1 1

A-,
/ . /
* * / ' / •

/•
~f/ /
v* •

•i f •

— /'/•'
/

1000 2000 3000 UNO MOO SOOO 7000 «OQO

moj/l CI

0.5 î ?S 5 K) ÎO Î3 « M 60 *J 60 90 5 ! n Î 9 S9.S

F r e q u e n c y , V.

Fig. 4 - Chemical composition changes induced by the incorporation of

leachate from the Bègues controlled landfill of Barcelona, about 4 km far.
The upper figure refers to the Falconera submarine spring water, sampled
inside the gallery by divers. The theoretical mixing line is that observed
before the landfill and correspond to a simple fresh and marine water
mixing. The actual mixing line is a mean position of observed S0 4 2 " vs. CI"
contents in 1974 and 1975, after the landfill site was in operation. There is
a loss of S0 4 2 " (the intensity varies according to the moment of sampling),
equivalent to the introduction of about 6 tons of soluble organic matter. The
lower figure shows the corresponding increase in HC0 3 " in some shallow
wells at nearby Garraf village. Line A is the distribution before the landfill
operation and B afterwards (after Custodio & Galofré, 1976; Hoyos-Limôn
&Palomero, 1975).

al., 1986b; PHPO, 1985). This shows a preferential permeability zone around the present
sea level. Drillers' well logs frequently mention some karst features at this depth. Below
it the carbonate rock seem little affected by general dissolution, in part related to
Quaternary sea level changes in the Mediterranean sea (Riba, 1981; Fairbridge, 1972).
Present intensive groundwater exploitation in these coastal areas has resulted in sea-water
intrusion (Fig. 7) and consequent groundwater quality degradation (Carbonell et al.,
1988; Pascual, 1990; Font & Rafaelli, 1984; Pascualetal., 1986a; Alom&Soler, 1989).
Wells obtain a mixture of locally recharged fresh water with inflowing sea water. The
landward flow of fresh water and salt water from the coast toward the wells explains why
 Results from several karst formations in southern Catalonia 301
i — i — i 1

«7-4-28
Olivelki
Altitude 99Um

\ *N \ ^ irv | \
N

vj v
L JJ Ï
f\.
M J c 1*^
?
J,
J
H
«7-4.21

JV > ~>
Uu ^
/, 1 1
W^^vJ
^r
Altitude 32.2m
M
r- u
E. 100
,
£ c
(7

1 1
-U i ^l l
a s i ill M LI g>
U i

Fig. 5 - Hydrograms of two wells in the central Garraf massif, one close to
the sea (Sitges) and the other in a valley far from the sea (Olivella). Data
from the Catalonia Water Authority groundwater monitoring network.
Effective rainfall values (thick bars = recharge; lines = runoff) from daily
water balances in the soil (Pascual, 1990) are shown below.

wells with water levels permanently below sea level (Fig. 8) yield brackish water instead
of salt water (Pascual & Custodio, 1990). Both fresh water and salt water have water
heads below sea level.

Assuming the wells yield a mixture of local and marine water without large chemical
changes, it is possible to obtain the chemical composition of the fresh water component
if chloride is assumed to be a conservative ion, since the composition of the sea water
is known. This shows a reasonably homogeneous fresh water source all over the southern
sector (Pascual, 1990) except in the area downstream of the Foix-Castellet dam (Fig. 9),
a reservoir which leaks water into the Cretaceous limestones in the Foix River valley
(GEOTECNIA, 1980; Ninerola & Fuster, 1969; REPO, 1970; Llamas & Doménech,
1967, 1969). Foix River water is allochthonous to this area and has a lower chloride
content and a characteristic high sulphate content. Also the environmental isotope content
of the water is lighter and affected by evaporation effects.

CHEMICAL PROCESSES IN THE MIXING ZONE IN SOUTHERN GARRAF

CARBONATE FORMATIONS

A detailed hydrochemical study has been carried out in the coastal Cretaceous carbonate
formations in the southern part of the Garraf massif to understand the rock-water
interactions inside the fresh-water/salt-water mixing zone of the aquifer. This mixing
zone is maintained by groundwater abstractions away from the coastline.

A borehole that penetrates the upper half of the mixing zone and placed between the
pumping wells and the coast have been periodically sampled, as well as another well
inland from the pumping wells, and used as a reference to find the chemical changes in
fresh-water recharge with time (Pascual, 1990). This fresh-water well does not represent
the actual fresh-water mixing with salt water, but since the area has a rather
homogeneous landscape the assumption of representativeness seems acceptable.
302 Emilio Custodio et al.

1
\ \
VA \

1
>o
•\\

0 \
Illll

POSIT ETRATK N
A Co -20 m °'.\
B Cc =-20 m
C Ir - 20m
D I - 20m
-1

-7
90 99

frequency. %

Fig. 6 - Probability distribution of well specific discharge according to

situation and penetration below sea level (after PHPO, 1985; Pascual,
1990). The wide range (large standard deviation) is a consequence of the
wide variation in karst development.

Samples were taken with a grab sampler that was open when lowered to the desired depth
and then closed by a sliding weight. The boreholes are cased with PVC. It is assumed
that PVC does not induce chemical changes in the water. To know the representativeness
of the samples the existence of vertical flows inside the bore was studied by introducing
a radioactive tracer (131I) and measuring the dilution rate and also by repeated well
logging with a submersible electrical conductivity cell. No bulk vertical flows exist, but
between the permeable sections - mainly karstified fissures and layers - there is some
flow to adjust the salinity vertical distribution along the water column in order to
compensate for hydraulic potential deviations (Custodio et al., 1989).

Permeable sections correspond to the fast changes in water electrical conductivity, and
the sections of the salinity log presenting small changes indicate low permeability rock
between specially permeable zones (Fig. 10).
To get a time series of samples that can be compared, the low vertical electrical salinity
change section was selected - at 23, 40 and 58 m depth - in spite of the possibility of
 Results from several karst formations in southern Catalonia 303

Fig. 7 - Simplified distribution of groundwater salinity in exploitation wells

in the Garraf massif coastal plains (modified from Carbonell et al., 1988;
and Soler & Trilla, 1979). The salinity penetrations correspond to the areas
with more developed karst features.

chemical changes due to the sluggish renovation of the water (Pascual & Custodio, 1987;
Bosch et al., 1990; Custodio et al., 1990).

Conspicuous changes in groundwater temperature have been found in spite of the relative
large depth of the samples (Fig. 11). This is explained as the effect of preferential paths
for a fraction of rainwater recharge, especially of relatively warm early autumn rainfall,
when the soil is less vegetated and dry.

The study of chemical equilibria between water and rock, and also with gas, and the mass
balances (Figs 12 and 13) show that (Pascual, 1990; Pascual & Custodio, 1990):

Fig. 8 - Schematic cross section to show groundwater flow in the coastal

carbonate aquifers due to well pumping and the mixing of fresh and saline
water. Boreholes S.01 and S.02 are those mentioned for the detailed
studies. Bedrock is non-karstified carbonates or obliterated karst, the
boundary being irregular and transitional.
304 Emilio Custodio et al.

-Miocene caicareratts = PIio Quaternary cover

Fig. 9 - Chemical analysis of groundwater in the Garraf coastal area. The

vertical logarithmic columns at the left-hand side shows the difference
between water derived from the Foix River and locally recharged fresh
water. The right-hand map shows the modified Stiff diagrams of the fresh
water component of pumped groundwater. A large fraction of S 0 4
downstream from the Foix Reservoir means there is a possibility of
receiving recharge from leakage (Pascual, 1990).

— Ca-Na ion exchange plays a role in the chemical behaviour and is due to the fact that
in the area there is a continuous trend toward groundwater salinity increase. Ca 2 +
increases in water. Exchange positions are thought to be in clays derived from
impure limestones and marls.

— There is calcite precipitation in the zone close to the water table and also some C 0 2
exchange with the unsaturated zone gas, mainly dissolution of it.
- Carbonate dissolution occurs in the lower half of the borehole, inside the upper half
of the mixing zone.

- Conditions exist for dolomitization and this is probably happening, although it is not
easily distinguishable from possible Ca-Mg ion exchange.

Some environmental isotope studies ( 1 8 0, 2H in water; 13C in dissolved inorganic carbon)

do not contradict these conclusions but the number of data available is not adequate for
conclusions to be drawn from them with confidence. The unsaturated gas isotopic
composition is unknown, thus preventing water-gas exchange studies.

Cation exchange processes seem the rule in the Garraf carbonate coastal aquifers due to
the continuous trend of increasing salinity (Fig. 14) as shown by PHPO (1985), Carbonell
et al. (1988) and Pascual (1990).

SEA-WATER INTRUSION IN THE VANDELLOS CARBONATE MASSIF

The Vandellôs coastal carbonate massif, south of Tarragona, is formed mainly by Jurassic
 Results from several karst formations in southern Catalonia 305

Fig. 10 - Characteristics of borehole S.01. A - lithology; karstified and

highly fissured parts are indicated as indentations of the column and a
change in the brick-like symbol for carbonates; below 61 m the borehole is
uncased and infilled with debris. B - Gamma-ray log; high radioactivity
deflections towards the right. C - Radioactive tracer (1-131) dilution test
results (the one made in July 1986) after tracing the whole water column
inside the PVC casing; figures show elapsed time after first log (0),
immediately after the extraction of the tracer injection hose. D - The two
extreme water salinity (electrical conductivity) logs from a series of 10; the
shape is conserved in spite of large seasonal salinity variations. E -
Explanation.

dolomites and dolomitic limestones (Anadônetal., 1979; Bayé, 1985; Bayôetal., 1976).
The Vandellôs Nuclear Site (CNV) is on an alluvial fan at the foot of the mountain range.

The salinization at two deep wells in this carbonate formation (Cadaloques), fairly far
from the coast, promoted a study that included the drilling and testing of four deep
exploratory wells and several surveys to get chemical and environmental isotopic samples
(Custodio et al., 1984). The pumping tests and salinity changes at different well
discharges indicate that permeability is higher in the section close to the water level
except at one of wells, close to the shoreline, in which permeability was high along the
whole penetrated depth.
The iron casings of the wells precluded detailed chemical studies due to the induced pH
changes (Bosch & Custodio, 1987a), but electrical conductivity logs (Fig. 15), chloride
determinations and some isotope analyses are representative of the groundwater flow
conditions. The sluggish flow of groundwater due to the small head gradient explains the
slow re-establishment of salinity stratification after completion of drilling or pumping
tests. The fresh-water/salt-water mixing zone thickness is fairly well defined up to a
suspected fault running parallel to the coast, where permeability apparently decreases
(Fig. 16).
Calculation of fresh-water flow towards the sea can be done by means of common
approximate formulae, such as (Custodio & Llamas, 1976, sect. 13; Custodio &
Bruggeman, 1986):
306 Ermlio Custodio et at
2qoX-WX 2
k(l+a) a

in which:
z = depth of the interface (taken as the mid-point of the mixing zone) below local
mean sea level;
q0 = fresh-water discharge into the sea per unit coast length;
X = effective distance to the coast of the point of measurement of z;
W = mean groundwater recharge rate;
k = mean aquifer permeability;
a = relative specific mass change = (7S - Tf)/Yf;
7 = specific mass; s: salt water; f: fresh water.
f

S.01 B 23m If
• i
on- : 1
— • 40 m
a 5Jm • I f
: 1
A..
... \ S04 •* 27m
l Ut
7.5-
•'•iâ,.^....
\ ••"•4
if I/
.**\- ""^...
-Ait
^4-Â 4 \
If
if
7.0' M
/
7 V
V
as- "I "1
I I v—i ' T ""I M—y if"—'"i"" 1 1 B~— i — i — i — i — i r~T—«—r—

Fig. 11 - Evolution of field-measured pH and temperature of groundwater

samples at three depths (water table at 21 m) in the studied (S.01) PVC-
cased borehole in the Calafell area (southern Garraf massif). Borehole S.04
is outside the area of sea-water intrusion and shows a similar behaviour.
Changes are explained as the effect of preferential path recharge of
aggressive warm water from late summer rainfall events (Pascual &
Custodio, 1990).
 Resultsfromseveral karst formations in southern Catalonia 307

12 3 4 5 6 7 8 9 10 11 12 ' 2 3 4 5 6 7 8 9 1011 12

1986 198~

Fig. 12- Rock-groundwater exchanges in borehole S. 01 at 23 m depth (2 m

below the water table). Positive (and negative) values indicate Ca 2+
increase (and decrease) by exchange with Na + , calcite dissolution (and
precipitation), dolomite formation (and dissolution) and C 0 2 inflow from
the unsaturated zone (and degassing). After Pascual (1990) and Pascual &
Custodio(1990).

which is valid when salt water can be assumed steady (with a head equal to mean sea
level), the aquifer reasonably homogeneous, and for sites not too close to the coastal
outflow. According to the Badon Ghijben-Herzberg formula, the fresh-water thickness
is @ = z + zla = z(l + 1/a). Then:
q0 = [@2 k/2(l + a) X] + (WX/2)
For wells close to the coast the second right-hand term can be neglected. Some
approximation for k was derived from pumping tests and also from radioactive tracer
( I ) dilution tests. Results show that q 0 values thus obtained are clearly less than those
expected by considering that recharge in the massif flows at right angles to the coast line.
Also, water environmental isotope data (Bosch & Custodio, 1987b) show that recharge
in the more rainy high altitude areas is not found in the wells at lower altitudes and that
this water is similar to low elevation recharge (Fig. 17).

High altitude recharge moves westwards at a low angle to the coast, instead of normal
to it due to the high permeability anisotropy introduced by the open fractures subparallel
to the coast, and discharges in a series of submarine, near-shore springs, well known to
the local fishermen. This can explain that coarse, carbonate-dominated alluvial fan
308 Erriïlio Custodio et al.

EXCHANGE

'A j
*3 EXCHANGE
13
Co"- No*

12
AOm

10

» \
8

0
if
1/J
S

'
2

o pffl
rkyfff

1
-2
CALCiTE
- 3

-4

-9

p
-2

•3

-«
DOLOMITE "\
-9

1 2E3 4 3 6 7 0 S K) D 1! 1 3 S 0 i 9 T H S 1 @l
1986 fit?

Fig. 13 - Rock-groundwater exchanges in borehole S.01 at 40 and 58 m

depth. Positive (and negative) values indicate Ca 2+ increase (and decrease)
by exchange with Na + , calcite dissolution (and precipitation), dolomite
formation (and dissolution) and C 0 2 inflow from the unsaturated zone (and
degassing). After Pascual (1990) and Pascual & Custodio (1990).

sediments in the Vandellôs nuclear site contain diluted marine water and that temperature
vertical gradients are reversed, i.e. temperature decrease with depth, due to the cooling
effect of deep groundwater flow in the underlying carbonates (CIHS, 1990), recharged
at high altitude (Fig. 18).

DETAILED STUDIES IN THE VANDELLOS AREA

To get a better understanding of the hydrogeological processes in the Vandellôs area, two
research studies have been carried out, one in the unsaturated zone and another in
dominantly carbonate sand deposits near the mouth of a gully. The first was aimed at
 Results from several karst formations in southern Catalonia 309
no of 30
samples 20
1 i
10 8 1 1 • •
1 1 I 111....
! • ^y
-4
s J
H -8
22 .. 1
i
18
. / \
14 / 1
^
*a
o 10 - / \
* 6 . .-' • \
ë 2 — \
0 —*
-4 -7f—
z /
-8
\ 1
t

N
« , -* >-
|HCO3
i- 0 _^ ._ . . . _ ;-^

CI- (g/l)
Fig. 14 - Deviations of actual groundwater composition from theoretical
(closed system mixing) composition for well water in the Garraf coastal
area (spring 1984 survey). The theoretical composition is calculated by
adding fractions of local fresh water and sea water according to the CI"
content of actual sea water. The excess in Ca 2+ is in large part matched by
the deficit in Na + and Mg 2+ resulting from cation exchange reactions. The
deficit in S0 4 2 " is not fully explained (after Custodio et al., 1986). Studies
by Carbonell et al. (1988) show the same pattern.

investigating carbonate dissolution in the vadose zone, as suggested by Wood (1985), and
the second to evaluate the influence of chemical modifying phenomena, as commented
before.

In the carbonates surrounding the Vandellôs nuclear site three places were selected to
drill boreholes in the thick unsaturated zone and to install inside them a bundle of small
diameter copper tubes. They are tubes open at different depths and isolated ones from the
other in order to get point samples from soil gas (Bosch et al., 1990). The maximum
depth sampled was 22 m. The C0 2 content was monitored and once the oxygen content
and 13C in C0 2 . Results (Fig. 19) show that C 0 2 partial pressure tends to increase with
depth, at the time that 13C becomes heavier and 0 2 partial pressure tends to decrease. In
part this is due to the oxidation of down-dragged organic matter and to the effect of
carbon isotopic exchange between C0 2 gas and solid carbonate by means of reactions in
the unsaturated zone. This leads to carbonate dissolution in this unsaturated zone. Two
of the sites behave this way, with one showing a reverse trend, but this is in an alluvial
fan and piedmont carbonate deposits.
The studies in the Cala Jostell sandy sediments were aimed at the understanding of
chemical changes inside the fresh-water/salt-water mixing zone under long-term steady-
state conditions. These sediments are deposited near the mouth of the Bassa Nova creek
and consist mainly of carbonate particles (limestone and dolomite) with some feldspar
particles. A series of six shallow bores were fitted with a PVC pipe open only at a given
depth. They cover the full thickness of the mixing zone. The area is rich in organic mater
and part of the year is under marshy conditions.
310 Emïlio Custodio et al.

WATER ELECTRICAL CONDUCTIVITY, «S/cm

Fig. 15 - Well water electrical conductivity logs from the Cadaloques

observation wells, in the Vandellôs massif. Only extreme values are shown.
The saturated thickness is in Jurassic dolomites except for the first 20
saturated m in well C, in which a thick Plio-Quaternary valley fill exists.
Variations in the fresh and upper mixed zone is due to sluggish renovation
of water after well construction and testing. The left-hand boundary is a
good representation or undisturbed conditions, except for small vertical
flows due to salinity adjustments to get a stable distribution. Black bars
indicate sections along which no radioactive tracer dilution has been seen
(no flow but not necessarily impermeable) (Custodio et al., 1989). See Fig.
20.

A sand and gravel bar separates this plain from the beach. In this case water-carbonate
rock equilibria is influenced by sulphate reducing conditions, with ion exchange
processes acting when there are salinity modifications due to hydrodynamic changes. Ion
strength effects on activity coefficients play a minor role in this case. There is a clear
increase in HC0 3 " with a decrease in S0 4 2 ', accompanied by increases in all major
cations in diluted water, but in the more saline water Mg 2 + , Na + and K + decrease in
exchange of Ca 2+ (Fig. 20). C0 2 dissolution from organic matter oxidation is a major
process that affects carbonate rock-water chemical equilibria.
 Resultsfromseveral karst formations in southern Catalonia 311

distanc* to th* shoreline, m

Fig. 16 - Situation map of Cadaloques-Coll de Balaguer area and the

Vandellôs Nuclear Site (CNV) in the Vandellôs massif, and schematic cross
section with results of sea-water intrusion according to evidence from
observation boreholes and pumping wells.

CARBONATE AQUIFERS OF THE SOUTHERN PRE-LITTORAL RANGE

The southern sector of the pre-littoral range, a part of the Catalanides (Catalan Coastal
Ranges) is locally formed by a series of folded and fractured Triassic and Eocene
sediments with thick carbonate layers, tectonically pushed in a southeast-northwest
direction. Schists and granites overthrust the northeast boundary. The southeast boundary
is a big fault that constitutes the limit with the deep Penedés graben (Bayô et al., 1979;
REPO, 1973). Three main carbonate formations can be distinguished from bottom to top
(Fig. 21):
312 Emilio Custodio et al.

Lower Muschelkalk dolomites, largely a confined aquifer, poorly known, resting on

low permeability Bundsanstein clays, sandstones and conglomerates.
Upper Muschelkalk limestones and dolomites, largely outcropping in the southeast
part, and resting on clayish formations with anhydrites.

Middle Eocene alveoline limestones, occupying the northwest part, largely

unconfined and above the local water table in anticlinal folds. They rest on Keuper
and lower Eocene clayish formations, sometimes with thick anhydrite layers, and
toward the northwest deep below grey Oligocène marls of the Ebre basin
morphological depression (Bayô & Custodio, 1976).

. La Puntaire-La PorteHada (700m)

Mota del Grèvol (600m)

Coll de Balaguer IfflOm)

Cadaloques
El Torn (150m)

s»p y

ay /
,*/ y -20
/ ^/
#>
•<* / •30

6. y* 1? A y i

/ 6 y
40

-9 -8
'/V -7 -6 -5
S'fe srnow.7..
o high altitude spring Observation wells
A CD pumping wells A B C D
O «• X A

Fig. 17 - Schematic cross section of the Vandellôs massif through the

Cadaloques-Coll de Balaguer area (modified from Pascual et al., 1986a)
indicating groundwater recharge. High altitude recharge flows at an angle
to the cross section due to anisotropic permeability. The ô2H vs. ô 18 0 plot
(using salinity corrected values) shows that well water is isotopically
heavier than high altitude recharge, and heavier the closer to the coast. This
indicates that recharge flows out of the section towards the west (see Fig.
20). d is the deuterium excess and the lines are parallel to the world
meteoric line of slope 8.
 Results from several karst formations in southern Catalonia 313

Karstic features are frequent towards the west, but tectonically induced fractures,
possibly with some dissolution enlargement, dominate in the south and eastern parts,
where most of the springs appear. Anhydrite displacement and deformation of its cover
due to dissolution surely plays a major role in carbonate rock fracturing and in
introducing discontinuities in the aquitard layers.

EC T EC T

Fig. 18 - Two examples of water salinity (EC = electrical conductivity in

mS cm"1) and temperature (T in °C) logs in boreholes in the coarse,
carbonate-dominated, clastic sediments (S) at the Vandellôs Nuclear Site,
in the coastal Vandellôs massif. Heterogeneity and preferentially permeable
zones due to dissolution produce the step-like form of the curves. The great
thickness of the mixing zone is due to the tidal effect, high dispersivity and
some upwards fresh-water flow from the underlying carbonate rocks (J).
Temperature decreases with depth indicating the possibility of the upward
flow of cooler (higher altitude of recharge), fresh groundwater from the
lower carbonates. The two cross sections summarize available information.
The depth to the carbonate rock (the top indicated by the irregular line) at
the coast is about 130 m according to a deep borehole.
314 Emilio Custodio et al.

5"c,cOiqastPDBlV.. '/. vol. CO? gas V. vol.0; gas

22 -20 -18 -16 05 1.0 13 «9.5 200 205 21.0

8 /A
<
Q.

516

20

24
™ -,6
m
o
i -18

f "^
o
Fig 19 - Results of
Co sampling
-22. gases at different depths in the vadose zone of
the Vandellôs massif sea side. Numbers indicate the site. There are four
sampling dates for C0 2 gas (the limits of composition is indicated), the first
one also with 0 2 gas measurements and the fourth one with 13 C-C0 2
determinations. Data from Bosch et al. (1990).

10 Co" Mg" NatK" CI" SO; HCOJ

9.

8.

7. /
/
6. \ '
\• il
5
\ • if
\ •
k. \ '•
/ \
I-1
3. \P2
V. /
7 N \ / / \p,\ li
\\
n
1.
X •\ li;
1
^^^ n,,-i ^ = > 7:
•-,s
/:
-2.
/P4
-3.

-4.

Fig. 20 - Deviations of actual values with respect to theoretical mixing

values for point groundwater samples in Cala Jostell wetland on carbonate
materials. End members for mixing calculations are the freshest water (PI
sample for the first survey) and the saltiest (P5 sample for the second
survey). The depth increases from PI to P5. Carbonate dissolution is the
joint result of C 0 2 addition by organic matter oxidation in the soil, S0 4 2 "
reduction and cation exchange. In this case, cation exchange is zero in the
long term (stable mean salinity) but it has a seasonal effect. Between the
first and the second survey salinity increased (Bosch et al., 1990).
 Resultsfromseveral karst formations in southern Catalonia 315

CENTRAL
M-Q Miocene-Quaternary
E-0 Eocene-Oligocene
E Eocene limestones
T Triassic
P Schists & granite
X Travertine
Stain spring
Piezometrk contour, m

t
<^
Main flow direction
Contact between units
*•** Overthrust
•— Riult
=»-«= Cross-section

M-Q
PENEDES GRABEN

Triassiez B - M W M 2 * M 3 « K
Jurassic = J
I Bedrock ( P ) g3]Low permeability I I Permeable
t x Highly permeable breccia

Fig. 21 -Location map, groundwater head contours, main groundwater flow

direction and hydrogeological cross sections of the Carme-Capellades-
Mediona area of the southern pre-littoral range. The multilayer carbonate
aquifer present a tectonic disturbance by overthrusting and faulting that
allows deep aquifers to discharge through springs and form large patches of
travertines (Custodio & Bayô, 1983).

This is a dominantly rural and forest area, with rainfall higher than in the surrounding
plains where the population is concentrated and creates a water demand that clashes with
the traditional rights of municipalities in which the springs appear. The main springs
have been monitored and sampled over a period of five years to understand changes in
316 Emilio Custodio et al.
ca** «g** Na* cr 804 COjH*

Fig. 22 - Mean chemical characteristics of water from the main springs

showing the effect of turnover time (increase in dissolved salts) and time
fluctuations of the major ions due to mixing of different aquifer
contributions (Custodio & Bay6, 1983).

major chemistry and water isotopic composition ( 1 8 0, 2 H, 3 H), as well as a few flowing
boreholes in the overthrust area (Custodio & Bay<5, 1983).

Chemical results show that sulphate is a dominant anion. Its proportion increases
downflow, as also does chloride, in that case at low concentrations. This is the effect of
progressively longer flow paths and residence times, with the progressive incorporation
of water from deep aquifers that leaks upward by the tectonically disturbed aquitards. But
ion concentrations also fluctuate with time (Fig. 22). For the lowest altitude springs this
is interpreted as variable contributions from various aquifers, each with different
residence times and consequently different water chemistry, although conserving the same
chemical facies.

Spring water is saturated with respect to calcite at the relatively high C 0 2 partial
pressure found in the aquifer. Near the discharge point, C 0 2 degassing to the atmosphere
produces calcite oversaturation and the formation of travertines (calcite tufa) if water
flows slowly. Actually the area is rich in travertine deposits (REPO, 1973). Tritium
concentration in spring-waters shows a decreasing trend in agreement with the decreasing
concentration in rain-water (Fig. 23). The large observed fluctuations are the result of
mixing of waters of different depths in the springs, as commented before. The deeper the
origin of the water the lower the concentration due to a greater component of pre-nuclear
recharge.

Trends in tritium concentration can be studied to yield mean turnover times by means of
a lumped parameter model of the exponential (well mixed) type (Maloszewski & Zuber,
1982; Zuber, 1985). The use of such a model is justified by the mixing at the outflow
 Resultsfromseveral karst formations in southern Catalonia 317
50

40 \V / t-WW1

30
. ®^L
vv
Tritium.TU
(3) '

20

10 £ k2 /^
» v^^^-^-
•—v\ v
V \ ^ \
LU P^\\

77 1978 1979 1980 1981 1982

")n r\ d
. ~^
' .
»-*
^
\^
£/ * Ç\h&tf>
• • ^
79 latio ' 1962

<3\ @ 5"b SMOW'/..

.'all bona

Fig. 23 - Tritium content in spring water. There is an exponential decrease

which can be adjusted to a well-mixed (exponential) reservoir model
solution to yield mean turnover times of 6 to 25 years according to the flow
. path length (springs 1 to 5). à I 8 0 values and the mean value distribution are
also shown. There are variations according to recharge altitude in the south-
north direction. Springs in Capellades are a mixture of contributions.
Tritium and ô 18 0 time variations reflect changes in the proportion of the
different contributions (Custodio & Bayô, 1983).

point. A simplified solution allows for a fast determination of turnover times (Custodio
& Bayô, 1983).
The trend of minimum tritium values, assumed to represent the behaviour of the deepest
aquifer contribution (with some unavoidable mixing with water from the aquifers above)
yield larger turnover times. The mean turnover time multiplied by the spring discharge
yield the associated aquifer storage volume. The values thus obtained are larger than the
Eocene limestone open space, thus indicating contributions from the Triassic carbonate
aquifer, even where they do not outcrop and are deeply seated. Spring water temperature
agrees with this conclusion and shows that fissures in the aquitards convey significant
quantities of water. 1 8 0 and 2 H isotope values, although not easy to interpret due to some
non-systematic analytical deviations in the laboratory, indicate that the lower altitude
springs also discharge lower altitude water.
318 Emilio Custodio et al.

THE PLANA DE LA GALERA AREA

The Plana de la Galera is a plain that extends from the lower reaches of the Ebre River,
in southernmost Catalonia, to the boundary with Valencia County, and occupies a trench
between the high Ports de Beceit range (the union of the Maestrat-Iberica ranges with the
Catalanides) and the coastal Montsià ranges. It is a northeast-southwest oriented graben
(Fig. 24), between the thick carbonate formations of the ranges. The bottom comprises
the same materials, buried below continentally dominated sediments and extensive
alluvial fans extending from the Ports de Beceit mountains (Bayô et al., 1976; CIHS,
1988; Bayô, 1984; AIH, 1989; SGOPU, 1974).

A rural and traditionally dry farming area for olive and carub trees, presently is being
transformed into irrigated lands for fruit trees. Since water supply from the Ebre River

{Alluvial fans

( Miocene-quaternary
segments
i trustified surface
\ upper cretaceous
'reactivated flow
{ Albtan

0 10km
I I 1 Active zone . Freshwater
jjllll 2 Sluggish flow zone. Saline water
gpgg 3 Build-up of hydrocarbons
SES3 4 Bedrock and impervious basement
ES23 5 Miocene-Quaternary sediments

| P»lo*ok<irst.Confined flew""""' ËiiiSiii Sraben fillings

E M Hard-ground • c a r b o n a t e rock
1 Low permeability ESSS Marls

Fig. 24 - Schematic hydrogeological behaviour of La Galera Plain. The

cross section shows groundwater flow and the effect of low permeability
Miocene-Quaternary deposits. The detail shows the role of the palaeokarst
below the graben filling as a drain. Shallow wells has a small specific
discharge, but the deep ones, if they intersect open palaeokarst, can have
a high specific discharge. The block diagram shows the existence of
different palaeokarst surfaces in ancient subaerial erosion surfaces (A and
B) with possible open spaces for groundwater flow (modified from Bayô et
al., 1986).
 Results from several karst formations in southern Catalonia 319

by means of an almost completed large canal failed due to administrative and social
problems, farmers are getting their water from wells that penetrate deeply the carbonate
rock base of the depression. Sedimentary discontinuities in the carbonate rock sequence,
formerly subjected to subaerial weathering processes, conserve open karst features that
constitute water transmitting layers capable of feeding large discharge wells (CIHS,
1988).

Water heads are only slightly above sea level in spite of the relative large distance from
the sea and the existence of the Montsià range in between (SGC, 1986; Touris &
Custodio, 1983). The situation is one of concern since the large investments in

Fig. 25 - Sulphate and environmental isotopic composition of groundwater

from the Plana de la Galera and Baix Ebre (modified from Tourfs &
Custodio, 1983; SGC, 1986). Sulphate points to a local source of recharge
for the plain except along the La Senia River since all water from the valley
springs infiltrates in the plain. Isotopic analysis differentiate allochthonous
Ebre River water (lighter and with a smaller deuterium excess due to its
high altitude headwaters and Atlantic origin), water from the Ports de
Beseit (light) and locally recharged water (heavier). Anomalous S. Carles
(5) groundwater is due to sea water contamination. Sample 11 is possibly
an analytical error.
320 Emilio Custodio et al.

agriculture and food industries, and also municipal water supplies, depend on poorly
known groundwater resources, their recharge source being unclear. The natural discharge
is to the sea and to the large Ebre River (SGOPU, 1975; Baytf et al., 1986), and thus also
poorly known.

There are four main sources of recharge:

— Rainwater recharge, directly on the plain. Semiarid conditions make this quantity
rather small, about 20 to 50 mm/year from 500 to 550 mm/year mean precipitation.
The contribution of return irrigation flows is probably small, except locally.
— Infiltration of runoff from the high elevation areas in the alluvial fans and along the
gully channels. Most alluvial fans are now dissected by creeks and then this recharge
is significant only along the creeks and locally.
— Infiltration of surface water. The only river is the La Senia, just in the southwest
boundary of the area. Most of the normal flow infiltrates the river channel, which
displays closed depressions that reflect collapses in the karstified limestones below.
But this barely affects the rest of the plain. Some springs in the mountains sustained
small flows that fully infiltrated, but now most of them are captured for irrigation
purposes.

— Underground transfer of water from the mountain ranges, mainly from the high Ports
de Beseit. It seems unlikely that this transfer is important due to the deep trench
separating the range from the plain carbonate base, which is filled with low
permeability materials. Recharge in the mountains flow laterally to the deep valleys
crossing them, mainly the La Senia River, where groundwater discharge in springs
and the river conveys the flows outside the ranges, towards the plain, where it
infiltrates.

To solve the questions raised by the last point, a preliminary chemical and isotopic
survey was carried out (SGC, 1986; Tourfs & Custodio, 1983). The results (Fig. 25)
point to the lack of significant underground transfer of water, a rather common
circumstance in other systems like this in carbonate formations in southern Catalonia, as
the Penedés graben with respect to the southern pre-littoral range before commented or
with respect to the southeast boundary of the Garraf massif (PHPO, 1985).

CONCLUSIONS
This study of the hydrogeological behaviour and exploitation problems of the southern
Catalonia carbonate aquifers has make successful use of chemical data. Hydrogeochemical
studies, both simple surveys, repeated surveys or frequent monitoring of selected points,
yield very useful results if the study techniques are selected in order to fit the problem
to be solved. Best results are obtained when combined with environmental isotope
studies, an extension of hydrogeochemical studies, and also tracer tests in boreholes to
know the existence of vertical flows and what the samples represent.

Detailed studies have aimed at a better knowledge of water-carbonate rock interactions

in coastal aquifers. Further to the influence of ion strength effect on the activity
coefficients to explain calcite dissolution or precipitation in the fresh-water/salt-water
mixing zone, other chemical influences play their role and even dominate the chemical
processes leading to karstification, karst obliteration, dolomitization and
dedolomitization. Such are redox processes and also ion exchange in areas in which
salinity evolution with time has a definite trend.
 Results from several karst formations in southern Catalonia 321

ACKNOWLEDGEMENTS
A large part of the data come from internal reports of the former Eastern Pyrenees Water
Authority and the Public Works Geological Service, where three of the authors work or
have worked, and are complemented by studies done by the International Course on
Groundwater Hydrology and for the doctoral theses of the third and fourth authors. Also
the owners of the Electronuclear Plants of Vandellôs (I and II) helped in studies and
financed some drilling and surveys. Part of the research studies have been done under
project 8309/007 of the extinguished American-Spanish Joint Committee for Science and
Technical Co-operation. Dr William Back, of the US Geological Survey co-operated with
these studies, as well as Dr M. Ramon Llamas, of the Madrid's Complutense University.

REFERENCES

AIH, 1989, Carte et Notice Explicative Feuille B 6 Madrid. In: Carte Hydrogeologique
Internationale de l'Europe 1:1 500 000. Assoc. Int. Hydrogéologues, Bundesanstalt
fur Geowissenchaften und Rohstoffe, Hannover, et UNESCO, Paris.
Alom, A. & Soler, X., 1989, Sfntesis de la circulation y dinâmica del acuffero cârstico
de Garraf (Barcelona). Tecnologia del Agua, Barcelona, 55, 35-41.
Anadôn, P., Colombo, F., Esteban, M., Marzo, M., Robles, S., Santanach, P. & Solé-
Sugranes, LI., 1979, Evolution tectonoestratigrafica de los Catalânides. Acta Geol.
Hisp. "Homenaje a Lluis Sole i Sabaris", 14, 242-270.
Astier, LI., 1970, Exploration de la surgencia submarina de la Falconera. Geo. y Bio.
Karst, VII (27), 706-709.
Astier, LI., 1971, Surgencias marinas en el karst litoral de Garraf. Geo. y Bio. Karst,
VIII (29), 755-758.
Back, W. & Hanshaw, B. B., 1970a, Comparison of chemical hydrogeology of the
carbonate peninsulas of Florida and Yucatan. /. Hydrol, 10, 330-368.
Back, W. & Hanshaw, B. B., 1970b, Major geochemical processes in the evolution of
carbonate-aquifers systems. /. Hydrol., 43, 287-312.
Back, W. & Hanshaw, B. B., 1982, Geochemical significance of brackish-water springs
in limestone of coastal regions. Ill Semana de Hidrogeologia, Faculdade de Ciências
de Lisboa-Portugal.
Back, W . & Hanshaw, B. B., 1983, Effect of sea-level fluctuations on porosity and
minéralogie changes in coastal aquifers. In: Paleoclimate and Mineral Deposits (
ed. by T. N. Cronnin et al.), 6-7. US Geological Survey Circular 882.
Back, W., Hanshaw, B. B., Pyle, T. E., Plummer, L. N. & Weidie, A. E., 1979,
Geochemical significance of groundwater discharge and carbonate solution to the
formation of Caleta Xel Ha, Quintana Roo, Mexico. Wat. Resour. Res., 15 (6),
1521-1535.
Back, W., Hanshaw, B. B. & van Driel, J. N., 1984, Role of groundwater in shaping the
eastern coastline of the Yucatan Peninsula, Mexico. In: Groundwater as a
Geomorphic Agent (ed. by R. G. La Fleur), 281-293. Allen & Unwin, Inc.
Back, W., Hanshaw, B. B., van Driel, J. N., Ward, W. & Wexler, E. J., 1981,
Chemical characterization of caves: cove, caleta, and karst creation in Quintana Roo.
Geol. Soc. Am. Abstracts, 94th Annual Meeting, 13 (7), 400.
Back, W., Hanshaw, B. B., Hernan, J. S. & van Driel, J. N., 1986, Differential
dissolution of a Pleistocene reef in the ground-water mixing zone of coastal Yucatan,
Mexico. Geology, 14, 137-140.
Back, W . , Bosch, X., Custodio, E. & Pascual, M., 1989, Geochemical behaviour in
freshwater-saltwater mixing zone in carbonate coastal aquifers in Catalonia, Spain.
28th International Geological Congress 1989, Abstracts (Washington, DC), 1.69-
1.70.
Baonza, E., Plata, A. & Piles, E., 1979, Aplicaciôn de la téenica del pozo tfnico
mediante el marcado de toda la columna piezométrica. In: Isotope Hydrology, 690-
322 Emilio Custodio et al.

711. International Atomic Energy Agency, Vienna.

Batista, E., Bayd, A., Custodio, E. & Doménech, J., 1983, Some aspects of sea water
intrusion in Catalonia (Spain). Geologia Applicata e Idrogeologia, Bari, XVIII, part
II, 299-314.
Bayô, A., 1982, La exploraciôn hidrogeolôgica de acufferos en rocas carbonatadas desde
la optica de la exploraciôn de recursos y utilizaciôn del almacenamiento. Reunion
Monograflca sobre el Karst (Larra-82), 177-215.
Bayé, A., 1984, Estudio hidrogeolôgico para mejora del abastecimiento de agua a Santa
Barbara y La Galera. Servicio Geologico de Obras Piiblicas y Urbanismo, Barcelona
(internal).
Bayé, A., 1985, Historia natural dels Països Catalans. Gran Enciclopèdia Catalana.
Barcelona, III, 17-56.
Bayô, A. & Custodio, E., 1976, Hidrologia subterrânea de la cuenca de la riera de
Carme y ârea de Capellades. Hidrologeologia y Recursos Hidrâulicos, Madrid, I,
108-128.
Bayô, A., Berga, L. & Fayas, J. A., 1976, Avance del conocimiento de la hidrogeologia
del "Baix Ebre" (valle inferior y delta del rio Ebre). Hidrogeologia y Recursos
Hidrâulicos, Madrid, I, 73-89.
Bayô, A., Doménech, J. & Custodio, E., 1976, Estudio geologico del macizo de
Vandellôs para définir sus caracten'sticas hidrogeolôgicas. Hidrogeologia y Recursos
Hidrâulicos, Madrid, I, 89-106.
Bayô, A., Custodio, E. & Fayas, J. A., 1979, Particularidades del funcionamiento del
sistema acuffero multicapa de un sector de la Cordillera Prelitoral Catalana entre los
rios Anoia, Carme y Foix (Barcelona). Hidrogeologia y Recursos Hidrâulicos,
Madrid, IV, 41-58.
Bayô, A., Castiella, L, Custodio, E., Ninerola, J. & Virgôs, L., 1986, Ensayo sobre las
diversas tipologias de acufferos en rocas carbonatadas de Espana. Técnicas de
estudio y explotacion. Jornadas sobre el Karst en Euskadi, San Sebastian, 255-340.
Bosch, X. & Custodio, E., 1987a, Alteraciôn de la alcalinidad en muestras de agua
procedentes de piezometros entubados con hierro en un acuffero carbonatado
afectado por intrusion marina: efecto de la corrosion del hierro sobre la
concentraciôn del ion H C 0 3 \ Hidrogeologia y Recursos Hidrâulicos, Madrid, XII,
197-206.
Bosch, X. & Custodio, E., 1987b, Interpretaciôn de datos de isôtopos ambientales
estables (D y 0-18) de muestras de aguas subterrâneas del acuffero carbonatado
costero del macizo de Vandellôs en la zona del Coll de Balaguer: T.M. de Vandellôs
(Tarragona). Il Jornadas de Geoquimica, Soria, Colegio de Geôlogos de Espafla, 67-
70.
Bosch, X., Custodio, E. & Pascual, M., 1990, Geochemical behaviour in the freshwater-
saltwater mixing zone and vadose zone in carbonate coastal aquifers in Catalonia,
Spain. In: Proceedings of the Int. Ass. Hydrogeologists Congress, Washington.
Heisse (in press).
CAPO, 1973, Estudio de las afecciones hidrolôgicas del vertedero controlado de basuras
en el macizo de Garraf, 3 vols. Comisaria de Aguas del Pirineo Oriental y Excmo.
Ayuntamiento de Barcelona (internal).
Carbonell, M., Koch, M., Teran, W., Pacual, J. M. & Custodio, E., 1988, Estudio
hidrogeolôgico del sector Moja-Canyelles-Pla de Jorba-Castellet, del Macizo
calcâreo del Garraf (Barcelona). Tecnologia de la Intrusion en Acufferos Costeros.
Instituto Geologico y Minero de Espana, AlmMecar, Granada, III, 27-42.
Cebollada, P. & Urcola, J., 1978, Estudio de la contaminaciôn de las aguas de la
poblacion de Garraf. Curso Internacional de Hidrologia Subterrânea. Barcelona.
CIHS, 1988, Sfntesis geolôgica e hidrogeolôgica del Maestrazgo. Curso Internacional de
Hidrologia Subterrânea, Barcelona (internal).
CIHS, 1990, Estudio de los resultados de la red de control de la CNV-II. Curso
Internacional de Hidrologia Subterrânea, Barcelona (internal).
Cotecchia, V., 1977, Studies and investigations on Apullian groundwaters and intruding
 Results from several karst formations in southern Catalonia 323

sea waters (Salento Peninsula). Quaderni dell Istituto di Ricerche, Roma.

Custodio, E., 1975, Metodologia y resultados del estudio hidrogeolôgico del macizo
kârstico de Garraf. Barcelona. Bolettn Geolôgico y Minero, Madrid, 76-1, 31-44.
Custodio, E., 1981, Técnicas radioisotôpicas artificiales para estudios de la
contaminaciôn del agua subterrânea: Anâlisis y Evoluciôn de la Contaminaciôn de
la Aguas Subterrâneas en Espaiïa. Curso Internacional de Hidrologia Subterrânea.
Barcelona, 2, 713-740.
Custodio, E., 1981, Contaminaciôn de aguas subterrâneas en relaciôn con los vertidos
de basuras urbanas de Barcelona: Jornadas sobre Anâlisis y Evoluciôn de la
Contaminaciôn de Aguas Subterrâneas en Espana. Curso Internacional de Hidrologia
Subterrânea, Barcelona, 1, 223-247.
Custodio, E., 1982, Calidad del agua y protecciôn de los acufferos kârsticos ante la
contaminaciôn. Reunion Monogrâfica sobre el Karst de Larra-82, Isaba (Navarra).
Diputaciôn Forai de Navarra, 291-327'.
Custodio, E., 1986, Hidrogeoqufmica del karst. Jornadas sobre el Karst en Euskadi-86,
San Sebastian, 2, 131-179.
Custodio, E. & Llamas, M. R., 1976, Hidrologia Subterrânea, 2 vols. Omega,
Barcelona. .
Custodio, E. & Galofré, A., 1976, Evoluciôn en la calidad del agua en la surgencia
litoral de la Falconera en relaciôn con un gran vertedero de basuras en el macizo de
Garraf. In: II Asamblea Nacional de Geodesiay Geoflsica (Barcelona), 2131-2173.
Instituto Geogrâfico Nacional, Madrid.
Custodio, E. & Bayô, A., 1983, Nuevas contribuciones al conocimiento geohidrolôgico
del acuffero de Carme-Capellades-San Quintfn (Barcelona). In: VAsamblea Nacional
de Geodesia y Geoflsica (Madrid 1983), 4 (III), 1795-1816. Comisiôn Nacional de
Geodesia y Geoffsica.
Custodio, E. & Bruggeman, G. A., 1986, Sea Water Problems in Coastal Areas:
Groundwater. Studies and Reports in Hydrology no. 45, UNESCO Press, Paris.
Custodio, E. & Bayô, A., 1989, Procesos fisico-quimicos de disoluciôn-precipitaciôn de
carbonatos: algunos resultados en acufferos carbonatados espanoles. In: El Karst en
Espana, 321-332. Sociedad Espafiola de Geomorfologia, Madrid, Publ. no. 4.
Custodio, E., Bayô, A. & Galofré, A., 1984, Informe sobre el plan de control
hidrologico de los pozos del sector de Cadaloques (Vandellôs, Tarragona). Curso
Internacional de Hidrologia Subterrânea, Barcelona (internal).
Custodio, E., Pascual, J. M., Bosch, X. & Bayô, A., 1986, Sea water in coastal
carbonate formations in Catalonia, Spain. In: Proceedings 9th Salt Water Intrusion
Meeting (Delft), 147-164.
Custodio, E., Pascual, J. M., Bayô, A. & Bosch, X., 1989, Processes in the mixing zone
in carbonate formations: Central and Southern Catalonia. Natuurwet Tijdschr., 70,
263-277.
Custodio, E., Bosch, X. & Pascual, J. M., 1990, Geochemistry in the saturated and
vadose zone in carbonate coastal aquifers in Catalonia, Spain. Water & Wastewater
(Barcelona Meeting).
Fairbridge, R. W., 1972, Quaternary sedimentation in the Mediterranean region
controlled by tectonics, palaeoclimates and sea level. In: The Mediterranean Sea (ed.
by D. I. Stanley), 99-113. Dowden, Hutchinson & Ross, Strondsbourg.
Font, X. & Rafaelli, S., 1984, Actualizaciôn del conocimiento de los acufferos del llano
de Sitges-Sant Père de Ribes. Curso Internacional de Hidrologia Subterrânea,
Barcelona (internal).
GEOTECNIA, 1980, Estudio de la influencia de las pérdidas de agua del embalse de Foix
en la recarga de los acufferos miocénicos y cretâcicos explotados mediante pozos
para el abastecimiento a nûcleos urbanos y para riego. Excma. Diputaciôn Provincial
de Barcelona, Servei del Medi Ambient, Geotécnica Geôlogos Consultores,
Barcelona (internal).
Halevy, E. et al., 1966, Borehole dilution techniques: a critical review. In: Isotopes in
Hydrology, 531-564. International Atomic Energy Agency, Vienna.
324 Emilio Custodio et al.

Hanshaw, B. B. & Back, W., 1979, Major geochemical processes in the evolution of
carbonate aquifer systems. /. Hydro!., 43, 287-312.
Hanshaw, B. B. & Back, W., 1980a, Chemical reactions in the salt-water mixing zone
of carbonate aquifers. Geol. Soc. Am. Abstracts, 93rd Annual Meeting, 12 (7), 441-
442.
Hanshaw, B. B. & Back, W., 1980b, Chemical mass-wasting of the northern Yucatan
peninsula by groundwater dissolution. Geology, 8, 222-224.
Hanshaw, B. B., Back, W. & Deike, R. G., 1971, A geochemical hypothesis for
dolomitization by ground water. Economic Geol., 66, 710-724.
Herman, J. S., 1986, Diagenesis and mass transfer reactions in the groundwater mixing
zone. Fifth International Symposium on Water-Rock Interactions, Extended Abstracts
(Reykjavik, Iceland), 266-269.
Herman, J. S., Back, W. & Pomar, L., 1985, Geochemistry of groundwater in the
mixing zone along the east coast of Mallorca, Spain. In: Karst Water Resources
(Proc. Ankara Symp., July 1985), 467-479. IAHS Publ. no. 161.
Hoyos-Limôn, A. & Palomero, A., 1975, Estudio de la contaminaciôn de las aguas de
la poblaciôn de Garraf. Curso Internacional de Hidrologia Subterrânea, Barcelona
(internal).
IGME, 1982, Mapa Geologico de Espafla: E 1:50 000, Hoja 447. Villanueva y Geltre,
Instituto Geologico y Minero de Espana, Madrid.
Llamas, M. R., 1968, Informe hidrogeolôgico sobre las filtraciones del embalse de Foix
y laposible utilizaciôn de aguas subterrâneas para resolver el déficit de agua de la
zona costera comprendida entre Sitges y Calafell. Servicio Geologico de Obras
Pûblicas, Barcelona.
Llamas, M. R. & Doménech, J., 1967, Study of seepage from a reservoir situated on
calcareous land. Mem. Int. Assoc. Hydrogeologists Congress, Istanbul, VIII, 208-
216.
Llamas, M. R. & Doménech, J., 1969, Estudio de las filtraciones del embalse de Foix
y la utilizaciôn conjunta de aguas superficiales y subterrâneas para el abastecimiento
y riego de la franja costera desde Calafell a Sitges (Prov. de Barcelona). Documentos
de Investigaciôn Hidrolôgica, Barcelona, 5, 157-169.
Magaritz, M., Goldengerg, G. L., Kafri, V. & Arad, A., 1980, Dolomite formation in
the sea water-freshwater interface. Nature, 287 (5783), 622-624.
Magaritz, M. & Luzier, J. E., 1985, Water-rock interactions and seawater-freshwater
mixing effects in the coastal dune aquifer, Coos Bay, Oregon. Geochim.
Cosmochimica Acta, 49, 2515-2525.
Maloszewski, P. & Zuber, A., 1982, Determining the turnover time of groundwater
systems with the aid of environmental tracers: I. Models and their applicability. /.
Hydrol, 57, 207-231.
Mercado, A., 1985, The use of hydrogeochemical patterns in carbonate sand and
sandstone aquifers to identify intrusion and flushing of saline water. Groundwater,
23 (5), 635-645.
Nifierola, J. M. & Fuster, J., 1969, Informe hidrogeolôgico sobre las calizas del embalse
de Foix. Curso Internacional de Hidrologia Subterrânea: Barcelona (internal).
Pascual, J. M., 1990, Hidrogeoqufmica del macizo carbonatado de Garraf: anâlisis de los
procesos relacionados con la mezcla de aguas subterrâneas dulces y saladas en el
literal de Calafell-Bellvei (Tarragona). Doctoral Thesis, Dept Ground Engineering,
Polytechnic University of Catalonia.
Pascual, J. M. & Custodio, E., 1987, Procesos hidrogeoqufmicos en la zona de mezcla
agua dulce-agua salada en el literal del extremo meridional del macizo carbonatado
de Garraf (Tarragona). Hidrogeologiay Recursos Hidrdulicos, Madrid, XI, 477-492.
Pascual, J. M., Bosch, X., Custodio, E., Iribar, V. & Bayô, A., 1986a, Intrusion marina
en calizas costeras: aplicaciôn a Garraf Sur y Vandellôs (Catalunya). Jornadas sobre
el Karst en Euskadi, San Sebastian, 287-299.
Pascual, J. M., Bosch, X., Custodio, E. & Bayô, A., 1986b, Los acutferos mixtes
abiertos costeros: ejemplo de aplicaciôn de la tipologia de acufferos en rocas
 Results from several karst formations in southern Catalonia 325

carbonatadas: aplicaciôn a los acufferos de Garraf y de Vandellôs (Catalunya). i l

Simposio sobre el Agua en Andalucia, Granada, 2, 403-412.
Pascual, J. M. & Custodio, E., 1990, Geochemical observations in a continuously
seawater intruded area: Garraf, Catalonia (Spain). 11th Salt Water Intrusion
Meeting, Gdansk, 1990.
Petit, E., 1981, Un caso excepcional de contaminaciôn en zonas kârsticas: Garraf.
Jornadas sobre Anâlisis y Evoluciôn de la Contaminaciôn de las Aguas Subterrûneas
en Espafla, Barcelona, I, 277-279.
PHPO, 1985, Documentos para el Plan Hidrolôgico del Pirineo Oriental. Confederation
Hidrogrâfica del Pirineo Oriental, Barcelona. Vols.: Garraf; Penedés; Camp de
Tarragona; Serres de Cardô-Vandellôs (restricted circulation).
Plata, A., 1982, Single Well Techniques Using Radioactive Tracers. Centra de Estudios
y Experimentation de Obras Pûblicas, Servicio de Aplicaciônes Nucleares, Madrid.
Réf. I-1B11/85.
Plummer, L. N., 1975, Mixing of sea water with calcium carbonate ground water. Geol.
Soc. Am. Mem. 142, 219-235.
Plummer, L. N. & Back, W., 1980, The mass balance approach: application to
interpreting, the chemical evolution of hydrologie systems. Am. J. Sci., 280, 130-
142.
Price, R. M., 1988, Geochemical investigation of salt water intrusion along the coast of
Mallorca, Spain. Department of Environmental Sciences, University of Virginia,
Thesis for the Degree of Master Science.
REPO, 1970, Estudio de los Recursos Hidrâulicos Totales del Pirineo Oriental.
Comisaria de Aguas del Pirineo Oriental y Servicio Geologico de Obras Peblicas.
Barcelona. Vols., Garraf-Penedés-Gayâ; Prelitoral de Igualada; Tarragona-Reus-
Valls (restricted circulation).
REPO, 1973, Recursos de agua subterrânea en la Cuenca de la Riera de Carme y area de
Capellades, alrededores de Igualada (Barcelona). Comisaria de Aguas del Pirineo
Oriental and Servicio Geolôgico de Obras Pûblicas, Barcelona, 3 vols (internal).
Riba, O., 1981, Canvis de nivell i de salinitat de la Miditerrània Occidental durant el
Nebgen i el Quaternari. Treb. Inst. Cat. His. Nat. Barcelona, 9, 45-62.
Sanford, W. D. & Konikow, L. F., 1989, Simulation of calcite dissolution and porosity
changes in saltwater mixing zones in coastal aquifers. Wat. Resour. Res., 25 (4),
655-667.
Sayles, F. L. & Mangelsdorf, P. C , 1977, The equilibrium of clay minerals with sea
water, exchange reactions. Geochim. Cosmochimica Acta, 41, 951-960.
SGC, 1986, Recursos hfdrics subterrànis a la Vessant Catalana de l'Ebre. Informes no.
2, Servei Geologic de Catalunya, Barcelona.
SGOPU, 1974, Informe geolôgico sobre el pozo experimental y sondeos realizados en el
rio Gaià. Servicio Geologico de Obras Pûblicas, Barcelona-Madrid (internal).
SGOPU, 1975, Estudio hidrogeolôgico para abastecimiento de Amposta. Servicio
Geologico de Obras Pûblicas, Barcelona-Madrid (internal).
Shatkay, M. & Magaritz, M., 1987, Dolomitization and sulfate reduction in the mixing
zone between brine and meteoric water in the newly exposed shores of the Dead Sea.
Geochim. Cosmochimica Acta, 51, 1135-1141.
Solé-Sabarfs, LI., 1958, Geografia Fùica. Geografia de Catalunya, vol. 1. Aedos,
Barcelona.
Soler, X., 1984, Dinâmica de la circulation y control estructural en el acuffero kârstico
del macizo de Garraf (sector SW). Tecniterrae, "S 235".
Soler, X. & Trilla, J., 1979, Contaminaciôn marina y continental en un karst costero
(terminaciôn SO. del Macizo de Garraf. Barcelona). Tecniterrae, "S 170".
Soler, X. & Trilla, J., 1984, Litofacies del mioceno y geometria del zocalo en la
subdepresion de Vilanova-San Pedro de Ribas y su interes hidrogeolôgico.
Tecniterrae, "S 235".
Tourfs, R. & Custodio, E., 1983, Nuevos conocimientos hidrogeologicos del Bajo Ebre.
V Asamblea Nacional de Geodesia y Geoflsica (Madrid 1983). Comisiôn Nacional
326 Emilio Custodio et al.

de Geodesia y Geoffsica, 4 (III), 1819-1834.

Ward, W. C. & Halley, R. B., 1985, Dolomitization in a mixing zone of near-sea water
composition, late Pleistocene, northeastern Yucatan Peninsula. / . Sed. Petrol., 55
(3), 407-420.
Wigley, T. M. L. & Plummer, L. N., 1976, Mixing of carbonate waters. Geochim.
Cosmochimica Acta, 40, 989-995.
Wood, W. W., 1985, Origin of caves and other solution openings in the unsaturated
(vadose) zone of carbonate rocks: a model for C0 2 generation. Geology, 13, 822-
824.
Zuber, A., 1985, Mathematical models for the interpretation of environmental
radioisotopes in groundwater systems. In: Handbook of Environmental Isotope
Geochemistry: the Terrestrial Environment (ed. by P. Fritz & J. C. Fontes), vol. 2,
1-59. Elsevier.
Hydrogeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990). „ __
IAHS Publ. no. 207, 1993. 327

USE OF THE PHYSICO-CHEMICAL VARIATIONS OF KARST WATERS TO

UNDERSTAND HOW AN AQUIFER WORKS: EXAMPLE FROM THE
FONTAINE DE VAUCLUSE SYSTEM (SOUTHEASTERN FRANCE)

BERNARD BLAVOUX & JACQUES MUDRY

Hydrogeology Laboratory, 33 Rue Louis Pasteur, F-84000 Avignon, France

ABSTRACT
The Fontaine de Vaucluse karst system has been studied by spatial sampling of several
peripheral springs and by time sampling on the main springs of the system. Spatial data -
— on temperature, 18-oxygen and chloride content - of the catchment area and its
neighbourhood allows us to compute its average altitude, 850 m, a number that conforms
to geological and morphological data (the altitude range from 85 to 1909 m), and also the
average altitude of the intake area feeding the boreholes in use, which are situated on the
foot of the limestone massif. This was done in order to protect their quality. The good
correlation between 18-oxygen and chloride is indicative of the meteoric origin of both
tracers, but each of them shows a logarithmic fitting with the altitude, probably due to
the évapotranspiration phenomenon that affects the lower part of the intake area more
than the higher one. Between July 1981 and January 1982, during a low water period,
253 samples were collected for a daily survey of the Fontaine de Vaucluse spring, one
of the major springs in the world (average discharge 21 m 3 /s, minimum discharge 4 m 3 /s,
maximum discharge > 100 m 3 /s). The variation of different chemical ratios show that
several ions are related to the rainfall episodes, without any increase of the discharge
(chloride and manganese showing short and intense peaks), and that others are related to
the increase of the discharge (magnesium and silica showing a long-period decrease).
Both phenomena: immediate seepage, mobilizing atmospheric and pédologie tracers, and
deferred seepage, diluting the carbonate aquifer tracers, are superimposed on the flow
coming from the reserve. This study displays several phenomena:
— exclusive flow from the waterlogged reserves of the system (lack of seepage),
— flow from the same reserves, but mixed with immediate seepage,
— high discharge with a "reserve" chemistry ("piston effect"),
— dilution of the reserve water by quick and slow seepage water.
The examination of the flux (concentration x discharge) vs. discharge shows two great
families of water, with rather a steady composition:
— the permanent reserves segment, whose slope is the average concentration of that
unit,
— the mixing of that component with seepage water, that provides the underground
runoff.

INTRODUCTION
In southeastern France, on the rim of the Alps, the Vaucluse Plateau is a 1100 km 2
karstic massif in which the most important spring in France is located: the Fontaine de
Vaucluse (Fig. 1). One of the main geological features of this karst massif is the 1500 m
thick lower Cretaceous limestone series, in which sinkholes of 667 m, 640 m, 610 m,
575 m have been sunk. A 307-m-deep dewatered sinkhole (224 m under the sea level) has
been sunk under the spring.

The whole area has a mediterranean climate, but the range of altitudes induces variable
effective rainfalls: the altitude of the spring is 100 m, the highest mountain of the intake
area (Mount Ventoux) is 1909 m. The seepage (Fig. 2) is 120 mm in the plain while it
is 1300 mm on the mountains (Puig, 1987; Blavoux et al., 1988). The size of the intake
328 Bernard Blavoux & Jacques Mudry

Fig. 1 Geology of the Fontaine de Vaucluse karst system.

area, the effective rainfall and the exceptional thickness of the unsaturated zone (800 m)
allows the discharge to be maintained above 4 m 3 /s, to exceed 100 m3/s during flood
periods, and to reach an average value of 21 m3/s (Evin et al., 1968; Puig, 1987).

The system was studied using physico-chemistry and 18 0 data, by two complementary
methods: simultaneous sampling of several springs in the neighbourhood of the system,
in order to situate the main spring in its context, and temporal sampling of the Fontaine
de Vaucluse, Groseau and Aurel springs during a six-month dry period (July, 1981 to
January, 1982; Mudry, 1987).

EXPLORATION OF THE INTAKE AREA

Delimiting the intake area was tried after releasing dye in six caves or sinking streams
of the Plateau. Unfortunately, all these experiments were done in the same geological
context: the southern part of the Ventoux-Lure range. These dye tracings show the
connections, but do not allow the determination of the limits of the area, which were
plotted with geological and morphological data (Puig, 1987).

Some simultaneous chemical sampling was done, especially during dry periods. The
sampled springs were the outlets of the system, and also springs of the intake area and
its neighbourhood. Measurements of temperature, conductivity, pH and major ions were
undertaken at each spring (Malzieu, 1987).

The temperature of the water shows a decrease of 0.32°C for every 100 m increase of
altitude, when it is 0.28°C in the air. The correlation with the average altitude of the
intake area is slightly better (-0.938, Fig. 3), than the one with the altitude of the spring
(—0.931); this confirms the difference in the gradients.
Physico-chemical variations ofkarst waters to understand how an aquifer works 329

Tvater height (mm)

2000-,
tjijjjij real evapo-transpiration
JH effective rainfall
È
1000-

U.tf Ë
0 200 600 1000 1400 1800
altitude (m)

F i g . 2 H y d r o l o g i c a l b u d g e t for d i f f e r e n t a l t i t u d e b e l t s o f t h e F o n t a i n e d e
Vaucluse karst intake area.

température (° C) 18-oxygen (delta per 1000)

16
delta = 0.0 097 *t 2 +O.0 178*t 10.27
8
14 r = 0.774 (NS)
-8 •
12 '' e
10
-9 -
8
•
6
10 - 8
4 t = -0.00 636 Zi + 16.58
2 r = 0.931*
11 - 1
. J 1 . 1 •' 1 i 1 r -~ i — i —

0 1 000 2 000 2 6 4 8 10 12 14 16
average altitude of the intake area (m) température (° C)
chloride content (mg.l-1) chloride content (mg.1-1)
8 "
CI = 0.0 623 * 10 ( 0 0 7 1 ' " *
r = 0.910 (NS) /
6- e
e /

4-

2-
» "•'
0-
0 1 000 2 000 2 6 4 8 10 12 14 16
average altitude of the intake area (m) température (° C)
-f •oxygen (delta per 1 000) 18-oxygen (delta per 1 000)

*'•,• «'
-8 .
»
11 'a
-9 -
•

e
10 - 10
-delta = 2.5 428 * Z i (tU8481) delta » 0.38 797 *C1- 10.472
r = 0.943 *
11 - r = 0.950 * - 11
0 1000 2 000 0 2 4 6 !
average altitude of the intake area (m) chloride content (mg.1-1)
Fig. 3 Computation of the average altitude of the Fontaine de Vaucluse
intake area.
330 Bernard Blavoux & Jacques Mudry

t = -0.00 636 * Z(i) + 16.58 (n = 1 0 , r = - 0 . 9 3 1 , significant)

The improved correlation with the altitude of the intake area means that the reserves are
situated higher than the spring, and that a quick transit between reserve and spring does
not allow the water to balance its temperature with the one of the spring.

The chloride content shows a decrease with the altitude of the intake area Z(i), but the
correlation is not linear (Fig. 3), it is:

CI = 2320.7 * Z(i)(- 077 595> (n = 10, r = -0.936, significant)

The 18-oxygen content shows the same phenomenon, the equation is:

-Ô = 2.54 * Z(i) ( - 018 48I) (n = 10, r = -0.950, significant)

Both curves produce a linear relation between b and CI:

b = 0.386 * CI - 10.451 (n = 1 0 , r = 0.951, significant)

The curvature of the CI and of the b vs. altitude is due to the évapotranspiration
phenomenon, that is much more important (referring to the rainfall) at the lower altitude
belts than at the higher ones (Fig. 2).

With a planimetric method (surface of the different altitude belts of the intake area), Puig
(1987) found an average weighted altitude of 870 m for the Vaucluse system. Table 1
shows the three computed altitudes for the same system (from temperature, chloride and
18-oxygen content).
The difference between the three computed altitudes and the one obtained from the
planimetric method (870 m) is not very important for the Vaucluse intake area, the three
gradients giving equivalent computations, confirming that the intake of such a system
must include the rainiest parts of the massif, because of the very intense
évapotranspiration on the lower part (Fig. 2).

For the boreholes, the three computed altitudes are different, but these boreholes have
a very low tritium content: this may mean that the C1/ 18 0 ratios of the rain have been
different in the past compared with today. All the same it shows that these boreholes are
fed by an area higher than the plain, which is situated on the limestone massif.

STUDYING THE BEHAVIOUR OF THE SYSTEM

Chemograms

With a daily study of the spring, the chemograms show two kinds of reactions to the
rainfalls episodes (Fig. 4):
- The ions coming from the intake area (in the rain, the soil and the unsaturated zone;
chloride, sodium, manganese and nitric nitrogen) show very brief peaks, generally
one day after the rainfall, without any connection with the discharge at the beginning
of the rain.

- The ions dissolved in the limestone aquifer (silica, magnesium) show an important
decrease a fortnight after the first inflation of the discharge; at the same time there
is a dilution of the intake area tracers (when the superficial stock of these ions is
exhausted).
 Physico-chemical variations ofkarst waters to understand how an aquifer works 331

Table 1. computed altitude of intake areas, from temperature, chloride and 18-oxygen
content of karst waters (springs and boreholes).
Measured at springs: Computed altitude of the intake
area:
temperature CI 5 18 0 temperature CI S 18 0

Fontaine de Vaucluse 11.5 4.2 -8.87 828 762 850

Caromb 1 (borehole) 13 3.64 -8.25 624 851 602
Caromb 2 (borehole) 14 4.02 -8.05 489 788 538
Caromb 3 (borehole) - - -8.20 - - 585

This behaviour means that the immediate seepage, rich in chloride, sodium, manganese
and nitric nitrogen, is superimposed on the outflow of the saturated part of the aquifer,
rich in silica and manganese, before the dilution (due to the massive delayed seepage)
reaches the spring. The peaks of surface tracers are due to the leaching of the soil, where
there is a concentration due to évapotranspiration (Blavoux & Mudry, 1986).

8 f3Q&S&**'
ifflca 7 '
(ppm)

mangsneic
4^
magnesium
<PPtt (ppOO)

chloride
(ppm)

2
-* ?&^*SlG>/*jt S/\-
sodium
(ppm)

nitric
nitrogen
0.8 (ppm of N)

rainfalls
(ram)

ir-?
Fig. 4 Behaviour of the chemical content of the Fontaine de Vaucluse
spring during an autumn flood period.
332 Bernard Blavoux & Jacques Mudry

Figure 4 shows four phenomena:

— exclusive flow, without any recent seepage water, from the waterlogged reserves,

— flow from these reserves, but exaggerated with immediate seepage,

— "piston effect": increase of the discharge, but with "reserve" characteristic

chemistry,
— dilution of the reserve water with massive seepage.

Curves of flux vs. discharge

Using only tracers characteristic of the saturated zone, for instance silica, we can observe
(Fig. 5) the behaviour of that unit (Blavoux & Mudry, 1983; Mudry & Blavoux, 1988,
Mudry, 1987). The points show two linear fittings:

— The first straight line beginning at the origin, fits the low water plots, but also a
very high one. The slope of that straight line is the average silica concentration of
the waterlogged supplies of the system. The points situated near the origin are those
of the late recession. The high one is displaying the "piston effect": the seepage
waters, at the beginning of the flood, are pushing out of the channels well
mineralized water from the dewatered zone.

— The second one, whose slope is weaker, fits the decay of the discharge, after the
flood peak. The slope is lower because the concentration is lower: that unit is a
mixture of both water from the dewatered zone and from seepage.

discharge {m)
Fig. 5 Flux (discharge X concentration of silica) vs. discharge for
Fontaine de Vaucluse 1981-1982.

CONCLUSION
The use of the physico-chemical content of karst waters allows as much to delimit the
intake area so as to understand the working of the system.
Limiting the intake area may require the study of temperature, chloride and 18-oxygen
content. These three tracers give a similar average altitude for the intake area, but only
 Physico-chemical variations ofkarst waters to understand how an aquifer works 333

if the outlet of the aquifer is a spring. But, in order to use this method, we must be sure
of a climatic homogeneity; we cannot compare recent waters with low tritium content
ones, such as Caromb's boreholes. The rainfalls that have provided this aquifer did not
probably fit the same gradients as the actual ones.

Concerning the working of karst aquifers, the chemical content allows us to distinguish
the supply water, rich in long-lasting tracers such as silica, magnesium, from short
lasting seepage waters, which are bound to the rainfall episodes and rich in meteoric
tracers such as chloride, sodium, manganese and nitric nitrogen. At the beginning of the
rainfalls, particularly in the mediterranean area, high peaks of these tracers show the
leaching of the pédologie zone until that soil is poor in these tracers and then show a
decrease at the spring.

ACKNOWLEDGEMENTS

Mrs Liza Sandell helped the authors correct this text in Dakar.

REFERENCES
Blavoux, B. & Mudry, J., 1983, Séparation des composantes de l'écoulement d'un
exutoire karstique à l'aide des méthodes physico-chimiques. Hydrogéol. Géol. Ing.,
BRGM, Orléans, 4, 269-278.
Blavoux, B. & Mudry, J., 1986, Influence des pluies estivales sur la qualité des réserves
de l'aquifère karstique sous climat méditerranéen: le rôle du sol et de l'épikarst dans
la concentration des chlorures. Bull. Soc. Géol. Fr., Paris, 8, t 2, 4, 667-674.
Blavoux, B., Mudry, J. & Puig, J. M., 1992, Rôle du contexte géologique et climatique
dans la genèse et le fonctionnement du karst de Vaucluse (Sud-Est de la France). In:
Hydrogeology of Selected Karst Regions (ed. by W. Back & H. Paloc) (à paraître
pour AIH).
Evin, J., Flandrin, J., Margrita, & Paloc, H., 1968, Contribution des analyses
chimiques, des colorations et des analyses isotopiques à l'étude des circulations
souterraines en réservoirs karstiques: Exemple fourni par la Fontaine de Vaucluse
et le sondage de Salignan (France). AIH Congress (Istanbul, 1967) - 8, 298-308.
Malzieu, E., 1987, Ressource en eau entre Rhône et Ventoux: Hydrogéologie et
hydrochimie du bassin de Bedoin-Mormoiron, de la plaine des Sorgues et du massif
de Pernes-les-Fontaines (Vaucluse). Thèse 3° cycle Avignon.
Mudry, J., 1987, Apport du traçage physico-chimique naturel à la connaissance
hydrocinématique des aquifères carbonates. Ann. Scientif. Univ. Besançon, Géologie
- Mémoire H.S. no. 4. Thèse Besançon.
Mudry, J. & Blavoux, B., 1988, Quelques propriétés des courbes flux chimique - débit
dans les aquifères karstiques. 4° Coll. Hydrol. Pays Cale. Mil. Fiss.- Besançon-
Neuchâtel, sep 88, - Ann. Scientif. Univ. Besançon, Géologie Mémoire HS no. 6,
91-98.
Puig, J. M., 1987, Le système karstique de la Fontaine de Vaucluse. Thèse 3° cycle,
Avignon.
Hydrogeoloeical Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990). „ „ _.
IÀHS Publ.no. 207, 1993. * ' ^ 335

INTERNATIONAL COOPERATION IN RESEARCH ON KARST

ENVIRONMENTS

BRIAN FINLAYSON
Department of Geography, University of Melbourne, Parkville, Victoria 3052, Australia

UGO SAURO
Department of Geography, University ofPadova, Via del Santo 26, 35123 Padova, Italy

ABSTRACT
This paper provides an overview of international organizations who have substantive
interests in the field of karst research. Details are provided of the current activities of
the IGU Study Group "Environmental Changes in Karst Areas". The role of IGCP Project
299 "Geology, Climate, Hydrology and Karst Formation" is reviewed in relation to the
present needs for international research and action in this field. Proposals are made for
future international cooperation in karst research, specifically with reference to the
establishment of an integrated monitoring network along the lines of the IGBP
"geosphere-biosphere observatories" and to the establishment of a UNESCO convention
for the protection of significant karst sites.

INTRODUCTION
International cooperation has long been a feature of both karst research and caving in
general. In this paper we briefly outline the international groups which presently exist
in the field karst research and deal with the problems of coordination between different
commissions, working groups and projects. The future trends of karst research will
continue, as in the past, to be determined by individual workers or research groups aided,
of course, by the existence of international forums for the communication of research
ideas and results. The priorities now for international groups should be the fostering of
lines of communication through workshops, conferences and study groups; facilitating
the international exchange of data through the establishment of computer data bases in
agreed common formats; establishing an international inventory of karst resources which
could serve as the basis of a scheme for listing of threatened, rare or fragile karst areas;
and the establishment of an international scheme for such listing, probably through the
auspices of UNESCO.

An important opportunity for cooperation between the major international organizations

with interests in karst (see Table 1) is now given by projects such as the new
International Geological Correlation Program (IGCP) Project 299 "Geology, Climate,
Hydrology and Karst Formation" recently approved by the IGCP Board and the
International Geosphere-Biosphere Programme (IGBP) "A Study of Global Change". With
reference to the Study of Global Change, karst research now has the opportunity to
demonstrate that karst sediments, speleothems and landforms may give fundamental data
for the reconstruction of the evolution of terrestrial geoecosystems in recent geological
time. It is also appropriate to consider the establishment of a permanent monitoring
network of a large number of karst environments and waters. This could be done
cooperatively through the development of an "interdisciplinary integrated observatory
network" as suggested in the IGCP project. These and other issues are elaborated below.

International organizations
The main international organizations either wholly devoted to karst activities or which
336 Brian Finlayson & Ugo Sauro

have an important component of karst activities are listed in Table 1. This list is not
meant to be exhaustive but meant to give an overview of the international scene. In
addition there are many regional, national and local organizations such as the national
speleological societies and many specialized institutes in different countries.

Table 1. Major international organizations with interests in karst research.

Union Internationale de Spéléologie (UIS)
Five internal departments with commissions and groups:
(A) Protection and Management
Commission for protection, exploitation and tourism
Commission for cave management
(B) Research
Commission for physics-chemistry and hydrology of karst
Commission for palaeokarst and speleochronology
Commission for speleotherapy
Working group on karst and glaciers
Working group on artificial caves
Working group on hydrothermal karst
Working -group on volcanic caves
(C) Documentation
Commission for bibliography
Commission for large caves
Commission for the atlas of karst regions
Commission for information
Commission for the history of speleology
(D) Exploration
Commission for cave rescue
Commission on materials and techniques
Commission for cave diving
(E) Education
Commission for speleological education
Working group on speleological education in schools
International Geographical Union (IGU)
Study group on environmental changes in karst areas (S.88.6)
includes some national working groups (Japanese, Cuban, Polish, Czechoslovakian,...)
International Association of Hydrogeologists (IAH)
Karst Commission
International Geological Correlation Programme (IGCP) (Co-sponsored by the International
Union of Geological Sciences (IUGS) and UNESCO)
Project 299: Geology, climate, hydrology and karst formation
Project 287: Correlation of Mesozoic/Tertiary bauxites and related paleokarst phenomena in
the Tethyan realm
International Association of Geomorphologists (IAG)
At the second international conference held in Frankfurt in September 1989 a special symposium
and excursion was held on karst planation in tropical and middle latitude environments.

The IGU Study Group on Environmental Changes in Karst Areas

This study group was established at the Sydney meeting of the IGU in 1988; It is based
on an appreciation of the uniqueness of karst ecosystems and of the fact that in many
areas of the World this fact is not recognised and taken account of in their management.
The objectives of the study group are: to enhance the study of environmental changes in
karst areas throughout the World; to analyse the human impact within the framework of
natural karst ecosystems; to elaborate the best techniques for monitoring the changes and
collating the data; to classify karst areas for their susceptibility to human impact at local,
national and global scales.
 International cooperation in research on hirst environments 337

The group is currently preparing a publication on environmental changes in karst terrains

which will be published as a Catena Supplement, hopefully in time for the Washington
meeting of the IGU. A provisional list of chapters has been agreed and these are listed
below together with the authors' names:
Part A: A background to environmental changes in karst terrains
1. Environmental degradation and rocky "desertification" of karstlands: an introductory review
(Williams)
2. The global occurrence of karst and its relation to population distribution (Pfeffer)
3. Natural environmental change in karst: the Quaternary record (Ford, Lauritzen)

Part B: Human impacts on karsts around the world

4. Environmental change and human impact on karst in the mediterranean basin (Gams, Nicod, Sauro)
5. Environmental change and human impact on karst in southern China (Yuan, Sweeting)
6. Environmental change and human impact on karst in northern China (Song, Zhang)
7. Environmental change and human impact on karst in the Caribbean (Day, Troester)
8. Environmental change and human impact on raised coral reefs (Hori, Kawano, Miura)
9. Environmental change and human impact on karst in the United States (White, Quinlan)
10. Environmental change and human impact on karst in the western USSR (Andraithchuk,
Klimchouk)
11. Environmental change and human impact on karst in continental Europe (Pulina, Panos)
12. Environmental change and human impact on karst in British Isles (Goldie)
13. Environmental change and human impact on karst in sub-Saharan Africa (Brook)
14. Environmental change and human impact on karst in arid and semi-arid Australia (Gillieson)
15. Environmental change and human impact on karst in areas with sporadic or continuous permafrost
(Ford)
16. Environmental change and human impact in high mountain karsts (Kunaver)
Part C: General issues of human impact on karst terrains: principles and case studies

17. Impact of acid rain on karst environments (Trudgill)

18. Degradation of karst water resources through human impact (Crawford, Smith)
19. Environmental impacts of karst hydro-electric developments (Lauritzen)
20. Impact of quarrying on karst resources (Gunn)
21. Impact of tourism on karst resources (Huppert, Burri)
22. Human impacts on karst soils (Urushibara-Yoshino, Gillieson)

In addition, Prof. Gams (Jugoslavia) has proposed a project entitled "The Global Karst
Environment and Its Fragility" which would aim to classify karst areas on the basis of
their fragility and susceptibility to human impact. This project has been discussed by the
group and while it is very important it would be difficult to develop in the lifetime of the
Study Group. The Study Group could consider the methodology for such a program and
perhaps prepare special manuals as discussed below.

A proposal from Kevin Kiernan (Australia) to coordinate a volume on the conservation

of karst phenomena has been accepted by the group and it is hoped that this will contain
proposals for an international register of heritage areas in karst. (This topic is raised
again later in this paper.) Kazuko Urushibara-Yoshino (Japan) is planning to coordinate
a project on medium and small scale mapping in karst areas, and to prepare an annotated
bibliography on geomorphological and environmental mapping of karst areas in the
World.

A further proposal has come from Michel Bakalowicz (France) for the preparation of
special manuals on karst research methodology. Together with M. Pulina (Poland), he
plans to prepare a manual on karst water chemistry. Similar basic research manuals could
be prepared on topics such as: karst soils; karst morphometry; large scale mapping of
karst areas; small scale mapping of karst areas; cave environments; karst nomenclature.
These manuals should provide the necessary methodology for collecting the most
338 Brian Finlayson & Vgo Sauro

significant environmental data and for the production of maps, files of data for statistical
analysis, etc. These ideas are still being developed within the group.

IGCP Project 299: Geology, Climate, Hydrology and Karst Formation

The initiative for this project came from the 18th meeting of the Karst Commission of
the IAH in Guilin in 1988. Following further discussion at the 19th meeting in Florida
in 1989, the project was accepted by the IGCP Board in February 1990. The general
objectives of the IGCP is to establish correlations between different natural features
needed to understand the patterns, characters and sequence in time of geological
environments through to the present time. The first meeting of the international working
group of the project is the present one (Antalya, Turkey in October 1990).

Professor Yuan Daoxian has proposed a "Registration Form for a Typical Karst Area"
as a basis for the collection of information internationally within this project. The
information which this form requires seems to us inappropriate at the scale of the project.
Also, the underlying philosophy of this form, the "karst feature complex" is poorly
defined but appears to be based on the assumption that there are unique complexes of
karst features associated with particular climates; an unwarranted assumption in this
context. Three suggestions are given below which could form the basis of a revised
registration project.

(i) It could lead to the production of a finalized database of all the characteristics
useful for geological correlations, such as interrelations between geology and the
karst environment; stratigraphy, sedimentology and geochronology of paleokarst
and other types of karst sediments; landform "stratigraphy" ; present denudational
rates in relation to climate and geodynamic evolution; and present rate of
landform development.
(ii) It could be seen as a data base which provides the opportunity for large scale
(global) comparative work. Comparable work is currently going on in hydrology.
The Centre for Environmental Applied Hydrology at the University of Melbourne
now has a global database of monthly and annual river discharges which is used
for studies in comparative hydrology. There is also the Flow Regimes from
Experimental Network Data (FREND) project in western Europe and its broader
extension as the Flow Regimes from International Experimental Network Data
(FRIEND) project, now UNESCO Project IH 5.5. These hydrological databases
have the advantage that the data consist of quantitative information on
streamflows that can be analysed by statistical and numerical procedures. It is
difficult to see how the essentially descriptive information in the proposed karst
registration could be used in this way. If it were limited to, say, spring
discharges and water chemistry it would be more amenable to comparative
analyses.
(iii) A third possibility is that it could be seen as a karst cataloguing system. As such
it would uniquely identify each karst area and its general characteristics (location,
climate(s), controlling government(s), etc.); contain a bibliography of
publications on each area; and be associated with local karst registration systems
already in operation such as the Australian Karst Index. Probably the project
which fits better with IGCP is (i), but also (ii) and (iii) may be promoted better
if in cooperation with other interested organizations

Perspectives on the future of international cooperation in karst research

Future international cooperation in karst research should be viewed at three levels:
scientific collaboration in the establishment of databases and monitoring networks and
 International cooperation in research on karst environments 339

the exchange of scientific information, including international conferences and meetings;

fostering and developing educational activities in karst areas and on karst environments;
and the more politically oriented activities designed to conserve and protect karst
resources and establish a system of international compliance with agreed standards. A
better understanding of interactions within karst systems both in natural conditions and
in different human impact situations should be one of the main aims of karst research.

Table 2. Some international karst research meetings.

Subject of meeting Location/year Organization group
Environmental changes and anthropogenic Brno, Czechoslovakia; IGU Study Group
impact in karst 1990
Excursion in conjunction with the Brno Hungary; 1990 IGU Study Group
meeting
Environmental changes in karst areas Padova, Italy; 1991 IGU Study Group
Modular excursion with thematic symposia in Central and southern IGU Study Group
conjunction with the Padova Symposium Italy; 1991
Symposium and field seminar Guilin, China; 1991 IGCP Project 299
Symposium and field seminar Canberra, Australia; IGCP Project 299
1991
Karst environmental problems of the Washington DC, USA; IGU
Appalachians 1992
North American correlation seminar and field Washington DC, USA; IGCP Project 299
excursions 1992
11th International Congress China; 1993 ISU
International Meeting China; 1993 IGU Karst
Commission (?)
Third International Conference on Hamilton, Canada; IAG
Geomorphology 1993
International Meeting Poland; 1994 IGU Karst
Commission (?)
Environmental change in permafrost karst Yakutsk, USSR; 1995 IGU Karst
Commission and
Frost Action
Commission (?)
International congress To be decided; 1996 IGU

Scientific Collaboration: An important aspect of scientific collaboration is the provision

of international forums for the exchange of ideas and information; a role partly filled by
the international journals but also importantly by meetings and conferences. Face to face
contact provides for communication of a type and quality which cannot be achieved by
the written word.
In Table 2 some international meetings on karst are listed. This list is not exhaustive but
even so it illustrates a growing problem for international collaboration; that of too many
international meetings. Too many meetings could be as detrimental to international
communication and cooperation as too few. Few people in full time employment as
scientists, academics, environmental managers etc. can afford either the time or expense
of attending so many meetings. It would be preferable if the various groups responsible
for organising international meetings on karst research were to coordinate their activities
and perhaps run joint meetings to improve the level of cross-disciplinary interaction.
International communication between research groups working in different countries and
in different languages will be facilitated by the preparation and dissemination of manuals
340 Brian Finlayson & Ugo Sauro

of practice for the study of various aspects of karst systems so as to achieve

standardization in terminology and in the reporting of results. It is now possible to
exchange data directly by computer and to this end a uniform set of formats for the
storage and retrieval of information would allow more efficient use of the available
technology. International programmes, such as karst spring water analysis using agreed
standard methods and units for reporting and the establishment of a database of spring
water quality parameters should be developed based on these standard methods.

The IGU Study Group is currently considering the establishment of an international list
of scientists working on karst. If such a list were to be compiled on one of the widely
used desk-top computer database systems (such as HyperCard on the Macintosh) it could
be widely disseminated and used for the mailing of conference information etc.

International Monitoring and Management: International programmes are already in

existence for the protection of sensitive areas. UNESCO have for some time been
operating a system of World heritage listing. This programme covers all types of
environments, both natural and artificial, and may not be providing the protection
necessary for specific types of sites. This has already been recognised as it applies to
wetlands with the establishment, also under the auspices of UNESCO, of the RAMSAR
convention. The main purposes of this convention are the conservation of selected
wetlands and to promote education in the understanding of the relation between man and
his environment. Sensitive karst areas will be better protected if a similar system is
established for karst areas of global importance. Karst environments offer ideal
conditions to promote interdisciplinary research and to introduce young people to the
environmental problems of the conservation of natural resources and interrelations
between atmosphere, vegetation, land surface, soil, subterranean waters and cavities.
While karst areas can now be nominated in the present world heritage listing they have
to compete for attention with a wide variety of different types of sites.

The establishment of a UNESCO convention for the protection of karst areas

(International Register of Karst Geoecosystems - IRKG) would require that the necessary
data be collected to demonstrate to bureaucrats with no specific understanding of karst
areas that such a protection scheme is necessary. One of the more important aims for the
future is the collection and preparation of such data. Scientists in the field of karst
studies would also need to be active in setting up monitoring networks for sensitive karst
areas and in the international coordination of such monitoring. In this context it is useful
to consider the proposals being made through the IGBP for "geosphere-biosphere
observatories" and "data management and information systems". Large karst springs
reflect environmental changes through quantity and regime of flow and water quality and
could be important elements in a geosphere-biosphere monitoring programme. Other
significant sites in this programme could be large caves and surface karst depressions.
Strategies for the better conservation and planning of karst resources will need to be
developed if this international monitoring is to be successful.

These aims, especially as they apply to setting up international monitoring programmes,

will be more easily achieved if the various international karst groups mentioned in Table
1 act in concert. A registration project could be developed so as to be an inventory of
karst resources which could be used as the basis for protecting endangered karst areas
and features. This could be developed as a cooperative project between the major
international organizations working in this field, UIS, IGU, IAH, IAHS, IGCP and IAG.
In this way it will be more easy to lobby UNESCO to set up an IRKG Convention for
karst areas.
Hydrogeological Processes in Karst Terrants (Proceedings of the Antaîya Symposium and Field Seminar, October 1990).
IAHS Publ. no. 207, 1993. 341

THE KARST AQUIFERS OF TUSCANY (ITALY)

PAOLO FORTI
Department of Geological Sciences, Italian Institute of Speleology, Via Zamboni 67, 40127 Bologna, Italy

LUIGI MICHELI
Department of Land Management of the Region of Tuscany, Italy

LEONARDO PICCINI & GIOVANNI PRANZINI

Department of Earth Sciences, The University of Florence, Via Pirali, 50125 Florence, Italy

ABSTRACT
The overexploitation of the alluvial aquifers in Tuscany and their increasing
contamination, especially by chemical agents, has led people to turn to the region's karst
aquifers, which are less exploited and contaminated. Hydrogeological studies of the
regions containing karst aquifers are necessary to insure their proper exploitation. Such
studies have reached different stages in the various areas. Three karst aquifers, chosen
both for their importance and because hydrogeological studies of these aquifers have
reached different stages, are described. The Apuan Alps, in part because they are the
most famous speleological area of Tuscany, have been studied for many years, and their
hydrogeology is known in detail. In the Colle Val D'Eisa region the general underground
circulation patterns followed by the waters are known, but precise estimates of the
groundwater resources and studies for their protection have yet to be made. A karst
aquifer discovered under the lagoon in the region of Orbetello was examined for the first
time in the course of this study; its catchment area was determined, and a preliminary
calculation of its potential was made.

INTRODUCTION
The most heavily exploited aquifers in Tuscany are those in the Quaternary alluvial
deposits of the internal and coastal plains. In many areas they are beset by problems of
both quality and quantity: overexploitation (exacerbated, over the past three years, by a
drastic drop in precipitation) has reduced the reserves, while contamination, both induced
(intrusion of salt water into the coastal groundwaters) and direct (pollution by industry
and agriculture), has made the water unfit for use. Given this situation, the lightly
exploited karst aquifers have become quite important. Clearly, the exploitation of these
aquifers must be based on a thorough knowledge of their hydrogeology. Specifically,
hydrogeological balances must be made to estimate the exploitable reserves, and, at the
same time, to plan the steps necessary to protect them from contamination.

Though carbonate outcrops are not very large in Tuscany, in some areas they contain
significant hydrological reserves. The principal areas in which carbonate rocks are
exposed, according to the survey of karst areas carried out by the Department of Land
Management of the Region of Tuscany (Forti et al., 1989), are shown in Fig. 1. The
importance of these areas as aquifers depends upon their surface area, rate of recharge,
and, more importantly, degree of karstification.

The most important karst aquifers are in the formations of the Tuscan series with ages
ranging from Triassic to Jurassic. In particular, the Triassic Calcare Cavernoso and
Liassic Calcare Massiccio Formations are the thickest and most highly karstified. At
present, our knowledge of the hydrogeology of these systems varies considerably from
area to area. In some areas, detailed studies have been made to estimate both reserves
and vulnerability to contamination, while in other areas, the underground hydrogeological
342 Paolo Forti et al.

Fig. 1 - Carbonate outcrops in Tuscany. 1: Moderately karstified areas.

2: Heavily karstified areas. 3: Areas discussed in this paper - (a) Apuan
Alps; (b) Colle Val d'Eisa area; (c) Orbetello area.

resources have yet to be determined. In the following sections, three areas containing
karst aquifers that have been studied to differing extents, will be discussed.

THE APUAN ALPS

The major hydrological reserves associated with karst aquifers in Tuscany are
concentrated in the Apuan Alps. For this reason, and because Tuscany's most famous
caverns are in the Apuan area, many studies to unravel the area's complex hydrogeology
have been carried out over the years; by now the hydrogeology of the Apuans is
essentially understood. The Apuan Alps form a large tectonic window in which a
metamorphic series outcrops. The rocks ranging from Middle Triassic to early Cretaceous
in age are for the most part carbonates (dolomites, marbles, and siliceous limestones),
and in some places reach a thickness of 1000 m. The complex tectonic phases that caused
the uplift of the Apuan Massif produced intense fracturing, which, combined with the
well developed surface and hypogene karst of the carbonates, has produced excellent
aquifers with very high permeabilities. The high rainfall in the area, which averages 2500
mm/year, and easily surpasses 3000 mm/year in the higher parts of the chain, contributes
to make these aquifers extremely rich. As can be seen in Fig. 2, which shows the springs
with the highest discharges, karst springs are numerous, and border the massif. The
subsurface catchments of the individual springs were determined through hydrological
balances and dye tests. In some cases, the subsurface catchments were found to be
considerably larger than the hydrographie basins (Piccini & Pranzini, 1990, Fig. 3). The
 The karst aquifers of Tuscany (Italy) 343

Fig. 2 - Hydrogeological sketch of the karst aquifers of the Apuan Alps.

1: Carbonate rocks, highly permeable due to fracturing and karst. 2:
Impermeable or slightly permeable rocks. 3: Hydrogeological divides,
determined in part from dye tests. 4: Paths followed by groundwater
determined through dye tests. 5: Non exploited karst springs with
discharges greater than 1000 1/s, between 1000 and 100 1/s, and between
100 and 10 1/s. 6: Exploited karst springs with discharges between 1000
and 100 1/s, and between 100 and 10 1/s.

largest spring in the Apuan Alps, located inland from the city of Massa, has a discharge
of 1500 1/s, and supplies the Frigido River. On the opposite flank of the chain, the
largest spring is the Pollaccia, which was a discharge of about 1000 1/s. There are at
least 10 springs with discharges between 400 and 100 1/s, and many others have
discharges greater than 10 1/s; the total average discharge is therefore about 6 m 3 /s.
Given that the total area covered by the karst aquifers that supply the above-mentioned
springs is 170 km2, the specific discharge is about 0.035 nr/s/km 2 , equivalent to an
infiltration of 1100 mm of water per year.

Since the average precipitation over this area, which is concentrated in the higher parts
of the massif, is about 2500 mm, we can calculate that there is an infiltration rate of
about 45%. Considering the steep slopes in this region, this value is quite high, and

Fig. 3 - Hydrogeological section across the Apuan Alps. 1: Marbles and

Dolomites, highly permeable due to fracturing and karst. 2: Calcareous-
siliceous rocks, moderately or slightly permeable. 3: Crystalline
basement, impermeable. 4: Caves. 5: Karst circulation. 6: Superficial
divides. 7: Subsurface divides.
344 Paolo Forti et al.

shows how well developed the surface karst features are. Even though there are large
amounts of excellent water whose salinity, despite its coming from a carbonate aquifer,
is low, little more than 15% of it is put to use. The Apuan Alps is therefore a region
whose hydrological resources can be further exploited, even though care must be taken
to avoid damaging its delicate fluvial ecosystems, which are important enough for the
region to have been declared the Apuan National Park in 1988.
The hydraulic characteristics of these aquifers do, however, limit the degree to which
they can be exploited. The well developed hypogene karstification, combined with a
phreatic network which is probably well channelled, conspire to make the discharges of
the majority of these springs, and in particular the larger ones, quite variable. The
minimum discharges are in general about 10% of the average discharges, and about 2-3%
of the peak discharges. This results in less water being available in the summer months,
just when the demand is highest. Another limitation is the high vulnerability of these
springs to contamination, due to the intense karst of the region. The situation is
complicated by the extensive quarrying of the renowned Apuan marble, which has been
going on for over 2000 years, and which has produced high risk conditions. A map based
on a number of parameters including lithology, degree of karstification, slope, etc., has
been drawn to show the differing vulnerabilities of the karst aquifers (Civita et al.,
1990). Just two classes are shown in Fig. 4, one can see that because the marbles are the
most intensely karstified rocks, quarrying is concentrated in the areas at greatest risk.
There is therefore a fundamental incompatibility between quarrying and the exploitation
of the hydrological resources, which can only be resolved by limiting or rigidly
regulating the former, while protecting the environment in the most vulnerable areas.

THE COLLE VAL D'ELSA REGION

There are some important exposures of the carbonates of the Tuscan series in central
Tuscany, in structural highs produced by the extensive tectonic phase that began in the

Fig. 4 - Map showing the vulnerability to contamination of the karst

aquifers of the Apuan Alps. 1: Moderately vulnerable areas. 2: Highly
vulnerable areas. 3: Areas where marble is quarried. 4: Hydrogeological
divides.
 The karst aquifers of Tuscany (Italy) 345

w E

Q ' ^ 2 IS33
m F^g 4
/ '/ ';

• *
' il L

Fig. 5 - Schematic hydrogeological cross section (not to scale) of the

Colle Val D'Eisa area (from Casagli et al., 1990). 1: Calcare Cavernoso
(Triassic), highly permeable due to fracturing and karstification. 2:
Travertine complex (Quaternary), permeable due to primary porosity. 3:
Recent alluvial deposits, permeable due to primary porosity. 4: Clays and
sands (Pliocene), generally impermeable. 5: Shaly-marly-arenaceous
complex (Cretaceous), practically impermeable. The arrows indicate the
flow path of the water that feeds the springs of Colle Val D'Eisa (see
Fig. 6).

Miocene and is still under way; the structural lows contain Miocene lacustrine or
Pliocene marine sediments. Figure 5 is a schematic cross section of the region, in which
the highs are formed by carbonate rocks rendered permeable by fracturing and kargt,
while the basins are filled with unconsolidated sediments (clays and sands) whose
permeabilities are generally low. The highs are therefore source regions for deep
aquifers, whose waters rise along faults, producing springs whose discharges can be quite
high. In many cases the circulation is very deep and the waters become highly saline and
quite thermal (reaching temperatures of 70°C); one must keep in mind that the thermal
gradient in central and southern Tuscany is high because there are granitic stocks at
shallow depths. Indeed, Tuscany was the first place in which geothermal energy was used
to produce electricity. In other cases, the waters have low temperatures and salinities,
and can therefore be used as drinking water and for industrial and agricultural needs.

/ \ \
•H
|p?*siii|a,
\fe^
/"-COlLETfAI. D'ELSA

Bsë'T^"^' H
i- ,.,X \a» \
•^ V v i \
r <"1S*^\ //

0 i 2Km
gli." liia^sffi'M'lliiiiâiiiài

Fig. 6 - Hydrogeological map of the Colle Val D'Eisa region. 1:

Travertines and alluvial deposits (Quaternary), permeable due to primary
porosity. 2: Clays and sands (Pliocene), generally impermeable. 3:
Calcare Cavernoso (Triassic) and Conglomerates derived from the Calcare
Cavernoso (Miocene), highly permeable due to fracturing and karst. 4:
Hypothermal springs (20-22 °C); the larger dot indicates the Vene degli
Onci, a spring with an average discharge of 1.5 m 3 /s.
346 Paolo Forti et al.

There are some large springs in the Colle Val D'Eisa area (Fig. 6) whose hydrologie
reservoir is in the Calcare Cavernoso. This formation outcrops as a breccia, consisting
of limestone and dolomitic fragments cemented by a calcite cement, whose origin is
controversial. However the Calcare Cavernoso formed, its very high permeability has
allowed intense groundwater circulation, which has produced karst that is deep and well
developed, even though it is only rarely traceable. Rain water infiltrates the elevated
regions of Monte Maggio and Poggio del Comune, and rises along faults. During the
Quaternary, the waters from the carbonate rocks produced significant travertine deposits.
The studies carried out in this area have shown that the average infiltration in the
carbonate terrains is equivalent to about 45 % of the annual precipitation. Because of the
high permeability of the Calcare Cavernoso, the underground circulation is well
channelled, and almost all of the water flows from a few springs. In addition, the
elevation of the mouths of the springs is constant, indicating that the carbonates form a
continuous aquifer with a single piezometric surface. Not all the water rising from the
carbonate reservoir reaches the surface; in some cases there are local thermal and salinity
anomalies in the superficial aquifers, which indicate that significant amounts of thermo-
mineral water are dispersed into these aquifers (Casagli et al., 1990, Fig. 7). Some
thermo-mineral springs are used as spas, while many other low temperature thermal
springs could be used as low enthalpy geothermal resources, for example, to heat houses
or greenhouses.
The waters of other springs that are normal, though fairly hard, are used for municipal
supplies, and, more rarely for agriculture. Since the demand for water for these uses
increases yearly, while the reduction in rainfall that has occurred in the past years has
reduced the amount of water available, new sources of water must be found. In the Colle
Val D'Eisa area, these sources essentially lie in the aquifers of the Calcare Cavernoso.
A well 180 m deep has already been drilled in the vicinity of Monteriggioni (7 km

'?/•

£.' ' I-
i-*
y

Fig. 7 - Thermal anomaly in the groundwater in the travertines and

alluvial deposits, due to the intrusion of thermo-mineral water rising from
the Calcare Cavernoso. 1: Springs. 2: Wells for measuring the
temperature and chemistry of the superficial waters. 3: Isophreatic
contours. 4: Flow lines of the superficial groundwater. 5: Temperatures
of the water in the wells.
 The karst aquifers of Tuscany (Italy) 347

southeast of Colle Val D'Eisa). It produces 30 1/s, which demonstrates the aquifer is
fairly rich. However, before planning to exploit the aquifer more intensely, one must
examine the hydrological balance of the entire system to avoid overexploiting it. It is in
fact clear that, since the aquifer is supplied exclusively through rain water infiltration,
one cannot pump more than the average yearly supply. Overexploitation through wells
would, in the long run, reduce the discharges of the springs.
Another problem that must be examined is the vulnerability to contamination of this
groundwater; indeed, the high circulation rates and small capacity of the carbonates for
purifying the water makes these wells and springs vulnerable to pollution sources at
considerable distances. With regards to possible pollution sources, there are few
potentially dangerous industries in the area. The greatest danger is posed by agriculture,
from the fertilizers and herbicides used. An example of this problem can be seen in the
area of Rosia, near Siena, where considerable amounts of herbicides are used in the
cultivation of a small alluvial plain. In some areas, the alluvial aquifer already contains
concentrations of atrazine in excess of those allowed by law for drinking waters. Since
there are three wells at the edge of the plain that draw water from the Calcare Cavernoso
of Monte Maggio for Siena's municipal aqueduct, there is a real danger that the
contamination of the superficial aquifer will spread to the deeper aquifer, rendering its
water unfit to drink.

THE REGION OF ORBETELLO

This part of the Tyrrhenian coast has an intense seasonal tourist trade; the considerable
increase in the population during the summer creates serious water supply problems. A
significant confined aquifer has been found below the Orbetello Lagoon (Fig. 8). Its
piezometric surface is above sea level, and, as the low salinity of the water in the lagoon
indicates, it in part supplies the lagoon. The extreme hardness of the water renders it

Fig. 8 - Geological map of the Orbetello area. 1: Pliocene and Quaternary

deposits: clays and sands. 2: Calcare Cavernoso Formation (Triassic):
dolomitic limestones brecciated and karstified. 3: Verrucano Formation
(Triassic): phyllites and quartz-rich conglomerates. 4: Cross section
shown in Fig. 9.
348 Paolo Forti et al.

unfit for drinking without treatment. It does however appear suitable for irrigation, fish
breeding (its temperature, close to 20°C, favours this activity), and for use in the
lavatories of the many resorts along the coast. As drinking water presently serves these
needs, this would considerably reduce the strain on the drinking water supply. In order
to manage this aquifer correctly, without creating problems such as salt water intrusion
or subsidence, a careful hydrological study of the region must be made.
Tectonically, the area consists of a depression, that contains the Orbetello Lagoon,
between two structural highs, M. Argentario and Poggio dei Venti. These highs are
almost entirely made up of Calcare Cavernoso, which is more than 400 m thick here, and
is exposed over an area of more than 67 km2. The Calcare Cavernoso is intensely
karstified in these mountains (Mori, 1923; 1932), and there are many dolines, some with
diameters greater than 500 m. At M. Argentario, the Verrucano, a formation underlying
the Calcare Cavernoso that consists of impermeable phyllites and quartz-rich
conglomerates, is also exposed.

The sediments of the Orbetello Lagoon, which date to the Middle Pleistocene, are over
100 m thick, and consist of alternating fine sands and clays. Because of the presence of
significant superficial ore deposits, whose genesis is related to the uprising of granitic
stocks, the area has been thoroughly explored (RIMIN, 1980), and a precise geological
cross section can be drawn (Fig. 9).

Fig. 9 - Hydrogeological cross section of the Orbetello area. 1: Lagoonal

deposits: clays and fine sands, slightly permeable. 2: Calcare Cavernoso
Formation: dolomitic limestones, extremely permeable due to
karstification. 3: Verrucano Formation: phyllites and quartz-rich
conglomerates, practically impermeable. 4: Pyrite ore deposit. 5: Faults.
6: Water flow.

A sizable pyrite deposit has been found under Poggio Mortaio. Mining it has proven
difficult because of the large amounts of water that filter into the shafts. Water under
pressure has also been found in the lagoonal sands on the eastern side of the depression.
Below the Orbetello Lagoon there are, therefore, sizable aquifers; the most important lies
within the Calcare Cavernoso, which is isolated from the sands by an impermeable
reddish brown clay layer. The water comes from the karstified carbonates of Poggio dei
Venti, where the Verrucano serves as the impermeable base of the aquifer. The lack of
significant springs around Poggio dei Venti indicates that most of the infiltrating water
goes to feed the aquifer below the lagoon. It is also possible that some of the water feeds
the deep circulation that emerges as a copious thermal spring 15 km southwest of the
hills.
An initial analysis of the hydrogeological basin leads to the calculation that the Calcare
Cavernoso exposed at Poggio dei Venti covers an area of 35 km 2 . The records taken at
the Orbetello meteorological station over the past 20 years indicate the average annual
precipitation is 716 mm; with an infiltration coefficient of 40%, 10 million m^of water
 The karst aquifers of Tuscany (Italy) 349

would go to supply the aquifer below the Orbetello Lagoon each year. The area, because
of its high degree of karstification, appears vulnerable to contamination. However,
because the karst area is wooded and there are no potential sources of contamination, it
is not presently endangered.

ACKNOWLEDGEMENT
We thank RIMIN for making their geological data available.

REFERENCES
Casagli, N., Ceccarelli, A., Duchi, V. & Pranzini, G., 1990, Low temperature thermal
springs of Colle Val d'Eisa (Italy). XXII Congress of IAH (Lausanne, August-
September 1990).
Civita, M., Forti, P., Micheli L., Piccini L. & Pranzini, G., 1990, Carta della
vulnérabilité all'inquinamento degli acquiferi carsici delle Alpi Apuane. Congr. on
Protection and Management of Groundwater (Manaro sul Panaro, Modena,
September 1990).
Forti, P., Micheli, L., Piccini, L., Pranzini, G., Sassoli, U. & Trevisani, M., 1989, The
study of karst regions in Tuscany. In: Proceedings of 10th International Congress
of Speleology (Budapest), vol. II.
Mori, A., 1923, I fenomeni carsici dell'Argentario. AM VII Congr. Geogr. Ital., II.
Mori, A., 1932, I fenomeni carsici dell'Orbetellano e del Capialbese. Mem. Reg. Soc.
Geogr. Ital., XVII.
Piccini, L. & Pranzini, G., 1990, Idrogeologia e carsismo del bacino del Fiume Frigido
(Alpi Apuane). Atti Soc. Tosc. Sc. Nat., série A, XCVI.
RIMIN s.p.a, 1980, Convenzione Ministeriale Industria e Commercio. Ricerche
minerarie area M. Argentario.
Hydrogeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990). ^_,
IAHS Publ. no. 207, 1993. 351

HISTORICAL DEVELOPMENT OF CONCEPTS OF REGIONAL

GROUNDWATER FLOW IN THE FLORIDAN AQUIFER SYSTEM,
SOUTHEASTERN UNITED STATES

RICHARD H. JOHNSTON
US Geological Survey, 108 Tolomoto Trail, St Simons Island, Georgia 31522, USA

ABSTRACT

The Floridan aquifer system, a major source of groundwater in the southern United
States, is a thick sequence of highly permeable Tertiary limestone and dolomite. An
important contribution to understanding the Floridan flow system was made in the early
1930s by Stringfield (Stringfield, 1936) who mapped the Floridan potentiometric surface
and identified areas of recharge and discharge. Later Stringfield & LeGrand (1966) noted
that the present distribution of karst features, permeability, and pattern of groundwater
circulation can be associated with the erosion during Pleistocene time of Miocene beds
that had overlam the older rocks of the Floridan. From the 1940s through the 1970s,
many investigators contributed to an understanding of the Floridan flow system. Kohout
(1965) proposed a hypothesis of cyclic flow in which cold sea water flowed inland
through the lower part of the Floridan, was heated, then moved upward and mixed with
seaward-flowing fresh water in the upper part of the Floridan. Hanshaw et al. (1971)
proposed a hypothesis by which extensive dolomitization can occur as a result of
circulation in the brackish and saline zones of the Floridan. Computer simulations made
by several hydrologists during the 1970s and 1980s provided insight into the Floridan
flow system, especially the aquifer system's response to development. Bush & Johnston
(1988) designed a coarse-mesh, finite-difference model for the entire Floridan aquifer
system which provided (for the first time) regional flow rates and hydrological budgets
for the Floridan.

INTRODUCTION
The Floridan aquifer system is one of the major sources of groundwater supplies in the
United States. More than 120 m3/s is withdrawn from the aquifer system for all uses,
and, in many areas, the Floridan is the sole source of fresh water. The Floridan is made
up of a thick sequence of mostly Tertiary carbonate rocks that contain a regional
groundwater flow system. More than one-half of the natural discharge from this system
is through 300 major springs with a combined flow of about 360 m'/s.

The Floridan aquifer system is located in the southeastern part of the United States and
underlies an area of about 260 000 km2 - including all of Florida and parts of the States
of Georgia, Alabama, and South Carolina. The high average rainfall (about 1350
mm/year) and generally flat topography combine to provide abundant recharge (up to 500
mm/year) to the aquifer system.

The carbonate rocks (limestone and dolomite) that compose the Floridan generally range
in age from late Paleocene to early Miocene. Initially, these rocks were referred to as
"the principal artesian formations" by Stringfield (1936) and later he and other workers
used the term "principal artesian aquifer". In the early 1950s, however, Parker (in Parker
et al., 1955) proposed the term "Floridan aquifer" for these rocks and this name has
since been widely used in hydrological literature. Recently, Miller of the US Geological
Survey, provided the first detailed hydrogeological description of the Floridan throughout
its four-State area of occurrence (Miller, 1986). He defined the lateral and vertical extent
of the Floridan on the basis of permeability and noted that the top and base of the system
locally cross both formation and age boundaries. Accordingly, the stratigraphie units that
352 Richard H. Johnston

make up the Floridan vary from place to place. Because the Floridan generally consists
of an upper and lower aquifer separated by rocks of low permeability, Miller (1986)
proposed the term "Floridan aquifer system".

Large transmissivity values (greater than 10 000 m2/day) are characteristic of most of the
Floridan aquifer system. Water-bearing openings vary from intergranular spaces between
limestone particles and fossil fragments to networks of solution-widened joints to large
cavernous openings in some areas. These openings range in size from a few millimetres
to tens of metres. Diffuse, or Darcian, flow predominates where the small openings are
present, whereas conduit flow occurs where there are large cavernous openings,
especially in the vicinity of large springs.

DEVELOPMENT OF CONCEPTS OF REGIONAL GROUNDWATER FLOW

Hypotheses, early 1900s to the 1970s

Speculations about the location of recharge areas of the geological formations composing
the Floridan appeared in reports of the US Geological Survey in the early 1900s. Maison
& Sanford (1913) provided the first comprehensive description of the hydrogeology and
groundwater resources of Florida. They speculated that the principal catchment area of
the Tertiary carbonate rocks "occupies a belt in the central portion of the Florida
Peninsula and extends northward into Georgia".

However, the existence of a regional flow system within these carbonate rocks was not
recognized until the mid-1930s when Stringfield published his classic paper "Artesian
water in the Florida Peninsula" (Stringfield, 1936). Stringfield mapped the potentiometric
surface (Fig. 1) and areas of artesian flow for the "principal artesian formations" in
peninsular Florida on the basis of measurements, made mostly during 1934, of water
levels or shut-in pressure in wells. Stringfield explained that "the principal areas of
recharge are in the central, north-central, and northwestern parts of the peninsula".
Recharge areas included "areas in which the artesian formations are at or near the
surface, areas where sinkholes extend to the artesian formations, and areas where
drainage wells penetrate the artesian formations". Stringfield (1936, p. 10) also mapped
the areas of artesian flow on the Florida Peninsula in 1934. Not only was this map of
practical interest for water-supply development, but it indicated areas where upward
discharge from confined parts of the Floridan was occurring.

Following the development by Theis of the equation for non-steady flow to wells in 1935,
Cooper & Warren (1945) applied the Theis equation to aid in understanding the response
of the Floridan flow system to large withdrawals in the coastal area from Jacksonville,
Florida to Savannah, Georgia. By 1943, extensive cones of depression had developed
around four industrial pumping centres in this coastal area. Cooper & Warren (1945)
correctly forecast that water levels in the Floridan would stabilize within a few months
after each increase in withdrawal rate in the Jacksonville area and within 1 to 10 years
at Savannah. Proximity to recharge areas and larger transmissivity of the aquifer system
near Jacksonville are the reasons that a steady-state condition is reached more quickly
at Jacksonville than at Savannah following an increase in pumping.

During the 1950s, 1960s and 1970s, the Floridan was used as a natural laboratory by
many hydrogeologists and geochemists interested in carbonate aquifer systems. Their
investigations involved the development of permeability, complexities of the regional
flow system, and relations between geochemical processes and regional groundwater
circulation. Hanshaw et al. (1971) proposed a hypothesis by which extensive
dolomitization can occur in the Floridan as a result of groundwater circulation. They
noted that groundwater in the brackish and saline zones of the Floridan generally has
 History of concepts of regional groundwater flow in the Floridan aquifer system 353
87° 86" 85° 84= 83° 82° 81° 80°

Fig. 1 - Potentiometric surface of the principal artesian formations, 1934,

(modified from Stringfield, p. 12, 1936).

magnesium to calcium ratios (in milliequivalents) greater than one and, thus, can
dolomitize limestone. Sea-level fluctuation and climatic change cause the position of the
brackish-water zone to move and, thus, provide an opportunity for large volumes of rock
to be dolomitized. Later, Thayer & Miller (1984) concluded that dolomitization has been
a major diagenetic process affecting the permeability of the Floridan aquifer system.
Locally, micrite beds have been recrystallized into a highly permeable mosaic of dolomite
crystals. Conversely, in places, finely crystalline dolomite has filled the openings in
loosely packed, pelletai limestone, thereby creating a less porous and less permeable
rock.

On the basis of the temperature anomalies in some deep wells, a hypothesis of cyclic
circulation in the Floridan due to geothermal heating was proposed by Kohout (1965) for
part of south Florida. He observed a reverse geothermal gradient - that is, groundwater
temperatures decreased with depth. Kohout proposed a convection cell as shown in Fig.
2 consisting of: (1) cold sea water flowing inland in the lower part of the Floridan, which
crops out on the sea floor; (2) heating of the sea water, which then moves upward
 Richard H. Johnston

Gulf J; Big Cypress

of g Swamp Everglades
Mexico 2 SURFICIAL A Q U I F E R . ^

UPPER CONFINING UNIT

z
ce
ffi<UJ

D uOu. /^

a. ^
2 ">
o
u. 32? LOWER CONFINING UNIT.

EXPLANATION
TOP OF SEAWATER-LIKE SALINITY
WATER TEMPERATURE—in degrees Celsius
FLOW LINE
FRESH WATER AREA OF
SALT WATER SCHEMATIC
SECTION
0 3 0 KILOMETERS Naples - v - - 7 ™ »
i
, • j Lauderdale

Fig. 2 - Theory of cyclic circulation in the Floridan aquifer system due to

geothermal heating (modified from Meyer, Fig. 24, 1989).

through vertical openings (sinkholes or solution pipes) in the overlying confining units;
and (3) mixing of the rising sea water with seaward-flowing fresh water in the upper part
of the Floridan. Some workers have questioned Kohout's hypothesis because of the scanty
temperature data. However, recently collected data on hydraulic heads, water chemistry,
and natural isotopes support Kohout's hypothesis. Meyer (1989) described this evidence
based on new data from a deep well west of Miami, Florida. Meyer reported: (1) the
existence of an inland hydraulic gradient in the Lower Floridan, (2) decreased carbon*14
activity and increased water temperature inland within the lower part of the Floridan, and
(3) the local occurrence of warm, salty groundwater with anomalously high carbon-14
activity in the upper part of the Floridan. Therefore, Kohout's circulation theory is
preferable to an earlier explanation that considered salt water in the Lower Floridan to
be either connate water or residual sea water from a former high stand of sea level.

Stringfield & LeGrand (1966) concluded that the present distribution of solution openings
and, therefore, increased permeability in the Floridan was developed following erosion
of overlying Miocene age clay beds during the Pleistocene epoch. They also suggested
that the present pattern of groundwater circulation was developed chiefly during low
stands of sea level during the Pleistocene. Recently, Bush & Johnston (1988, p. 2)
presented a map showing the regional distribution of transmissivity of the Upper Floridan
aquifer based on aquifer tests and computer simulation. The transmissivity varies by more
than three orders of magnitude, due primarily to the degree of development of solution
features in the aquifer, and, to a lesser extent, the aquifer thickness. Comparison of this
transmissivity map with a map showing the occurrence of confined and unconfined
conditions for the Upper Floridan aquifer (Bush & Johnston, 1988, p. 1) indicates that
all the areas of very large transmissivity (greater than 100 000 m2/day) are located where
the aquifer is either unconfined or semiconfined. In these areas, the upper part of the
Floridan contains numerous caves, sinkholes, pipes, and other types of solution openings
that account for the very large transmissivity values. Small values of transmissivity (less
than about 5000 m2/day) occur where the aquifer is confined by thick beds of clay and
 History of concepts of regional groundwater flow in the Floridan aquifer system 355

contains thick beds of less-permeable limestone (generally micrite) or in updip areas

where the aquifer is thinnest.

QUANTIFICATION OF FLOW RATES USING COMPUTER SIMULATION, 1970s

AND 1980s
Although recharge areas were known and much of the natural discharge (which occurs
mostly as spring flow and discharge to streams) had been measured, flow rates through
the Floridan were not quantified on an areal basis until the advent of digital computer
models. Computer simulations of parts of the Floridan flow system were made by several
hydrogeologists during the 1970s and 1980s using various finite-difference and finite-
element models. Application of such models makes a major assumption that flow in the
Floridan behaves as flow in porous medium. Aquifer-test analyses indicate that this
assumption is probably valid on a scale of a few hundred metres or more in those parts
of the Floridan where development of solution cavities is not extensive (Bush & Johnston,
1988, p . C7). However, in the karst areas of central and northwestern Florida near major
springs, conduit (non-Darcian) flow probably occurs on a local scale. Even so, at the
scale of a coarse mesh digital model (with grid-block spacing of 5-10 km or more), flow
in the karst-springs area realistically can be treated as porous-media flow.

During the early 1980s, Bush designed and calibrated a coarse-mesh, finite-difference
model for the entire Floridan aquifer system (Bush & Johnston, 1988). Because his
calibrated model output closely matched measured heads and the amounts and distribution
of natural discharge to springs and rivers, Bush's model is considered to be an accurate
representation of the flow system in the Floridan. This model provided flow rates and
hydrological budgets (for the first time) of both predevelopment and present-day
conditions in the Floridan.
Simulation indicated that prior to development total flow through the Floridan was about
610 m 3 /s (Fig. 3). The average recharge rate to the Floridan was about 110 mm/year over
the area where recharge occurred. Recharge was highest (250-500 mm/year) in
unconfined and semiconfined areas near major springs. The proximity of high-recharge
to high-discharge areas implies a vigorous and well-developed shallow flow systems in
the unconfined and semiconfined parts of the Upper Floridan. Aquifer discharge to
springs, streams and lakes, nearly all of which occurred primarily in unconfined areas,
accounted for 88% of the discharge. Based on simulation, diffuse upward leakage, which
occurred primarily in confined areas, accounted for the remaining 12% of the

PREDEVELOPMENT DEVELOPMENT CONDITIONS

(1980)
Natural and
Induced
Recharge Pumpage
Natural Recharge 610 680 120

\ t
AQUIFER Natural Natural
Discharge - * • Discharge
^ SYSTEM - 610 560

Negligible change in storage.

Fig. 3 - Groundwater budgets for the Floridan aquifer system (all values
in m3/s) (modified from Bush & Johnston, Fig. 23, 34, 1988).
356 Richard H. Johnston

predevelopment discharge. Simulation indicates that development has affected the

hydrologie budget of the Floridan aquifer system (Fig. 3). By 1980 withdrawals from the
Floridan totalled about 120 m 3 /s. The pumpage was supplied primarily by diversion of
natural outflow and by induced recharge rather than by loss of water from aquifer
storage. Groundwater development also has resulted in an increase in the size of the
recharge area and the corresponding decrease in the size of the discharge area (Bush &
Johnston, 1988, p. 36).
Simulation indicates that about 75% of all 1980 discharge left the aquifer system as
spring flow or as discharge to streams and lakes. The remaining one-fourth of the 1980
discharge was divided between pumpage (17%) and diffuse upward leakage (8%). As in
predevelopment time, most recharge and nearly all spring flow and discharge to streams
and lakes in 1980 occurred in the areas where the Floridan is unconfined or thinly
covered by clayey confining beds.

CONCLUSION
Understanding of regional groundwater flow in the Floridan aquifer system has evolved
from the reconnaissance studies of the early 1900s to the computer simulations of the
1980s. It is noteworthy that the hypotheses and conclusions of some early workers have
been confirmed by recent hydrogeological investigations. Stringfield's conclusion that the
removal of Miocene clay beds from above the rocks of the Floridan by erosion during
Pleistocene time is largely responsible for the present distribution of karst features, and
associated high permeability and pattern of circulation is supported by recent
hydrogeological mapping and computer simulation. Kohout's hypothesis of cyclic
circulation in the Floridan due to geothermal heating is supported by recent data on
hydraulic heads, groundwater temperatures, and natural isotopes.

REFERENCES
Bush, P. W. & Johnston, R. H., 1988, Ground-water hydraulics, regional flow, and
ground-water development of the Floridan aquifer system in Florida and in parts of
Georgia, South Carolina, and Alabama. US Geol. Survey Prof. Pap. 1403-C.
Cooper, H. H. & Warren, M. A., 1945, The perennial yield of artesial water in the
coastal area of Georgia and northwestern Florida. Economic Geol., 40 (4), 263-282.
Hanshaw, B. B., Back, W. & Deike, R. G., 1971, A geochemical hypothesis for
dolomitization by ground-water. Economic Geol., 66, 710-724.
Kohout, F. A., 1965, A hypothesis concerning cyclic flow of salt water related to
geothermal heating in the Floridan aquifer. New York Acad. Sci. Trans., Series 2,
28 (2), 249-271.
Matson, G. C. & Sanford, S., 1913, Geology and ground waters of Florida. US Geol.
Survey Wat. Supply Pap. 319.
Meyer, F. W., 1989, Hydrogeology, ground-water movement, and subsurface storage in
the Floridan aquifer system in southern Florida. US Geol. Survey Prof. Pap.
1403-G.
Miller, J. A., 1986, Hydrogeologic framework of the Floridan aquifer system in Florida
and in parts of Georgia, Alabama, and South Carolina. US Geol. Survey Prof. Pap.
1403-B.
Parker, G. G., Ferguson, G. E., Love, S. K. et al., 1955, Water resources of
southeastern Florida, with special reference to geology and ground water of the
Miami area. US Geol. Survey Wat. Supply Pap. 1255.
Stringfield, V. T., 1936, Artesian water in the Florida peninsula. US Geol. Survey Wat.
Supply Pap. 773-C. 115-195.
 History of concepts of regional groundwater flow in the Floridan aquifer system 357

Stringfield, V. T. & LeGrand, H. E., 1966, Hydrology of limestone terranes in the

Coastal Plain of the Southeastern United States. Geol. Survey Am. Special Paper 93.
Thayer, P. A. & Miller, J. A., 1984, Petrology of lower and middle Eocene carbonate
rocks, Floridan aquifer, central Florida. Gulf Coast Ass. Geol. Soc. Trans., 34, 421-
434.
7 Tracing
Hydrogeotogical Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990).
IAHS Publ. no. 207, 1993. 361

PARAMETERS OF CARBONATE ROCK AQUIFERS FROM TRACER

METHODS

A. PLATA
Section of Isotope Hydrology, Division of Physical and Chemical Sciences, International Atomic Energy Agency,
PO Box 100, A-1400 Vienna, Austria

ABSTRACT
The determination of hydraulic parameters of karstic and, in general, carbonate rock
aquifers requires the use of direct methods, especially in the case of local studies. Tracer
methods based on the use of both artificial and environmental tracers have proved to be
of great help in many cases for this purpose. The paper describes the use of radioactive
and saline tracers for the investigation of the groundwater flow through tests in
boreholes. Horizontal and vertical flows are measured. Information on the following
parameters carl be obtained: local transmissibility (semi-quantitative) if no vertical flow
is present, vertical distribution of fractures in the aquifer, depth of the incoming water
(which can help for the interpretation of chemical and isotopic data of the collected
samples), dynamic response of the aquifer to precipitation (by repetition of tracer tests
in different conditions), identification of karstified or fractured stretches, etc.
Experimental results are presented which were obtained from two Brazilian aquifers. A
second point analysed in the paper is the use of environmental tritium for the
determination of the turnover time of groundwater by using the "well-mixing" model.
The suitability of this model is discussed for four karstic aquifers on the basis of more
than 300 tritium concentration data. Finally, the paper describes an example of the use
of artificial tritium for the determination of the relationship between primary and
secondary porosity of a karstic system by means of an interconnection experiment. The
results obtained were compared with the ones derived from the seasonal variations of the
stable isotope content (D and 18 0) interpreted on the basis of a "well-mixing" model with
sinusoidal input function.

INTRODUCTION
Tracer techniques including both artificial and environmental tracers have proved to be,
in many cases, useful tools for the investigation of karstic aquifers. The combined
application of these tracers can provide information on some basic characteristics of the
groundwater system, such as localization of recharge and discharge areas, calculation of
subsurface water storage and the determination of the residence or turnover time of the
groundwater.

When dealing with the investigation of local problems, for example dam leakage and
mine flooding studies, detailed and even exhaustive information on the characteristics of
the local groundwater flow is usually needed. Average values of hydraulic parameters and
general information obtained at a regional scale have a limited application in such cases
due to the large heterogeneity frequently shown by this kind of aquifer. Artificial tracer
techniques play a dominant role in such studies although the information provided by
environmental tracers can also be very important.

In the same way, the information and values of hydraulic parameters obtained from point
or local tests cannot be representative for the whole aquifer unless the appropriate
distribution and density of the point data are available. This is the case, for example, of
pumping tests, tracer dilution tests and, in some cases, interconnection experiments under
pumping or natural flow conditions.
362 A. Plata

TRACER TESTS IN BOREHOLES

Description of the techniques
Horizontal flow Artificial tracers have been used for the measurement of flows in
boreholes under natural conditions (Kocherin, 1916; Ogilvi, 1958; Matseev, 1965;
Grinbaum, 1965; Guizerix et al., 1967; Borowczyk et al., 1965; Halevy et al., 1967;
Drost, 1986). Most of the applications have been in granular media and only a few in
carbonate rock aquifers. Due to the characteristics of groundwater flow in this kind of
aquifer (concentrated flow), point dilution techniques are not appropriate for the
determination of the filtration velocity. For many years the author has used the technique
illustrated in Fig. 1 which is based on the labelling of the whole water column using a
radioactive (131I) or a saline (NaCl) tracer (Baonza et al., 1970; Plata, 1983). The use
of a radioactive tracer is advantageous because no significant change in the water density
is created. A plastic tube is inserted as far as the bottom of the borehole and is filled
with the tracer solution. The tube is open at both ends. When this tube is extracted at a
constant velocity, the water column is labelled, more or less, homogeneously. Tracer
concentration profiles are measured afterwards.
TRACER INJECTION

r
WATER TABLE
H

PLASTIC TUBE

-I'V
BORED CASING

vl

•I I;-
:••".!

•••-I
i •••

Fig. 1 - Illustration of the method for labelling the whole water column,

The previous method provides a rapid general view of the flows existing in the borehole.
When vertical flow does not exist, concentration profiles like the ones shown in Figs 2
and 3 are obtained. Fractured or karstified stretches are clearly identified and the
groundwater velocity at the permeable horizons can roughly be estimated by using the
equation:
1.81 d
(1)
a t 1/10
where Vf = filtration velocity, d = diameter of the well, a = correction factor due to
the distortion of the flow lines caused by the well, t 1/10 = the time needed for the
concentration of the tracer to be reduced to 10% of its initial value.
 Parameters of carbonate rock aquifers from tracer methods 363
TRACER CONCENTRATION [CPS] otsertvATioNs
, "P . . . . T " rJGfï-

TRACER CONCENTRATION (CPS)

OBSERVATIONS
0 WOO 20C0

30
<

Ï
Q=^t FLOW
2
V=44 em/d

1
IMPERMEABLE

o t»6.» k /nj
o t - 2 3 , 3 11 \H>

Fig. 2 - Tracer test performed in a borehole of a karst aquifer of Minas

Gérais (Brazil).

Fig. 3 - Tracer tests performed at different boreholes in a karst aquifer of

Minas Gérais (Brazil).
364 A. Plata

Obviously, no precise values of Vf can be expected from this method due to the vertical
mixing of the tracer and the uncertainty in the correct value of the coefficient a. This
coefficient can vary between 1 for open fractures and 2 or more for sand-filled fractures
depending on the borehole characteristics. In the case of karstified stretches, the diameter
of the borehole can also be very different to the nominal value due to the presence of
caves. Calliper profiles can help for the interpretation of the tests when dealing with
casing-free boreholes.
A better approach can be obtained by using the system which is illustrated in Fig. 4. The
permeable stretch, as identified through the previous method, is now isolated by two
inflatable packers. The tracer is injected and the concentration is kept homogeneous by
means of an electrical pump which recycles the water inside this stretch. If V is the
volume of water and Q the flow-rate entering the borehole between the two packers, the
tracer concentration Ct decreases with time according to the following equation:

Ct = C0 exp(-Q.t/V) (2)

where C0 = initial concentration. From this equation, the value of Q can be derived. The
quotient between Q and the cross-section of the borehole S gives the apparent average
velocity of the groundwater flow by assuming that a = 1. The main advantage of this
equipment is the possibility of adjusting the distance between the two packers according
to the length of the permeable stretch to be investigated.

PACKER
TRACER INJECTOR

PRESSURIZED AIR DIFFUSER

DETECTOR C M .

ELECTRICAL CONDUCTOR
STEEL CABLE

PLASTIC TUBE
PLASTIC TUBE

DETECTOR G.M.

PRESSURIZED AIR

Fig. 4 - Probe suggested for the measurement of groundwater flow

velocity in fractured or karst rocks.

Vertical flow Vertical flows in boreholes drilled in karst aquifers are very frequent. In
the majority of cases, when the borehole crosses two or more fractured or karstified
stretches, a vertical flow is created inside the borehole due to the differences in the
hydraulic heads of the water. The measurement of these vertical flows can provide very
useful information. The measurement is carried out by point injections of the tracer
 Parameters of carbonate rock aquifers from tracer methods 365

according to the techniques which are illustrated in Figs 5 and 6. In the case of Fig. 5A,
the tracer produces two peaks, one at each detector, from which the velocity is obtained.
A saline tracer can be used. In the case of the techniques illustrated in Figs 5B and C,
a radioactive tracer is necessary. A first peak is measured when the tracer passes in front
of the detector circulating inside the plastic tube and a second peak when the tracer
reaches the detector as a result of the motion derived from the vertical flow.

In the case of vertical velocities higher than about 2 cm/s, the use of commercially
available flow-meters is possible. Obviously, the resolution given by these devices is
much higher than in the case of tracer techniques and their use is recommended. The
precise location of the permeable horizons (inflow and outflow zones) is obtained from
the variation of the flow velocity along the water column. Constant velocity stretches
obviously correspond to impermeable strata.
The method shown in Fig. 6 allows the measurement of vertical flows using a saline
tracer and a single conductivity probe. The tracer is injected by means of a plastic tube
provided with an electromagnetic valve at its end. The point injection is controlled by
the opening of this valve. The time between this opening and the peak detected
afterwards gives the velocity of the vertical flow.

When the velocity of the vertical flow is lower than about 0.5 m/min, the tracer can be
injected at an appropriate depth by using a plastic tube and concentration profiles can be
measured afterwards by moving the detector along the water column. The successive
positions of the tracer cloud give the water velocity (Fig. 7).

Results and possibilities

The measurement of horizontal and vertical flows in boreholes can provide the following
information:

Vertical distribution of fractures The combination of horizontal and vertical flow tests
makes it possible to locate precisely fractures or karstified stretches tapped by the
borehole. If the number of investigated boreholes in a specific area is large enough, the
characteristics of the karstification can be obtained. Diagrams representing the vertical
distribution of the fractures which are associated with a measurable groundwater flow can

-DETECTOR

A B C
Fig. 5 - Illustration of techniques for the measurement of vertical flows
in boreholes.
366 A. Plata

SALINE TRACER

CONDUCTIVITY-METER

HATER TABLE

PLASTIC TUBE

CONDUCTIVITY PROBE

ELECTROMAGNETIC VALVE

Fig. 6 - Illustration of the method for the measurement of vertical flows

at boreholes using a saline tracer and a single conductivity electrode.
be constructed. An example is shown in Fig. 8 for a Brazilian aquifer located on the right
bank of the San Francisco River, where 92 wells were studied. It was found that most
of the permeable horizons (level of karstification) were located at the uppermost layer
of the aquifer. This information helped in the planning of the drilling of new wells.
Natural gamma profiles can also help in this investigation. Lower gamma counting-rates
are obtained at depths corresponding to fractures or caves. The relative importance of
these fractures can be evaluated if the probe is calibrated appropriately.
TRACER CONCENTRATION [CPSJ
&

583 l/h _,
3

527 t/h
I .

Fig. 7 - Measurement of the vertical flow inside a borehole in Minas

Gérais (Brazil) by point injection of a tracer at 50 m depth.
 Parameters of carbonate rock aquifers from tracer methods 367
PERCENT OF FRACTURES
0 10 20 30

2C-

3 4C-

NUMBER OF WELL = 92
SC-

8C- POOR STATISTIC

Fig. 8 - Vertical distribution of permeable horizons in a Brazilian karst

aquifer.

Identification of recharge and discharge areas Typically, vertical descending flows are
found in recharge areas and vertical ascending flows in discharge areas. This was the
case for the Jaiba aquifer (Brazil) shown in Fig. 9. Vertical ascending flows were more
frequent along the rivers where the discharge takes place. On the other hand, very
important information can be obtained if the flow-tests are repeated at different periods
of the year. The flow velocity and even the flow pattern of many boreholes can change
drastically from rainy periods to dry periods. The results obtained give the relationship
between groundwater and precipitation on a local scale.

Interpretation of chemical and isotopic data The knowledge of the flow pattern inside
boreholes is of great help, in many cases, for the interpretation of chemical and isotopic

H s HORIZONTAL FLOW
D = DESCENDING VERTICAL FLOW
A = ASCENDING VERTICAL FLOW

Fig. 9 - Distribution of borehole flow in the Jaiba aquifer, Brazil.

 368 A. Plata

data. Many results considered anomalous can be explained on the basis of the depth from
which the water comes. Vertical flows are especially important in this respect. The
example shown in Fig. 10 is for the Jaiba aquifer. A logarithmic correlation was found
for many wells between natural tritium content and depth. From the tritium content, the
percentage of pre-thermonuclear water (recharge prior to 1953) could be estimated on the
basis of the "good-mixing" model (see below).

DEPTH (m)

ol 1 1 1—lioo
2 3 4 5
NATURAL LOGARITHM OF DEPTH (m)

Fig. 10 - Variation with depth of the tritium concentration (Jaiba aquifer,

Minas Gérais, Brazil).

Flow velocity and correlation with other hydraulic parameters In the absence of
vertical flow, the tracer test provides a rough estimate of the flow velocity as shown in
Figs 2 and 3. Due to the dispersivity associated with the flow, the tracer injected in the
borehole is usually swept away from a stretch which is larger than the real thickness of
the permeable horizon. This phenomenon represents an additional difficulty for the
estimation of the transmissibility from the results of the tracer test. TransmissibilLty
values that are too high are frequently obtained.

Figure 11 represents an attempt to correlate the transmissibilities obtained from tracer

and pumping tests with the specific yield in the case of the Brazilian aquifer. The
deviation of many points can be explained because these tests were carried out at
different times.
The main advantages of the tracer test are its simplicity and low cost in comparison with
pumping tests. It can be considered a complementary tool from which additional
information can be derived. However, a qualitative correlation is usually observed.

THE USE OF NATURAL TRITIUM

General

The use of environmental tritium for tracing large-scale water movement in karst systems
has become a well established technique. The half-life of this radionuclide (12.43 years)
is very appropriate for flow dynamics studies in most of the cases. Natural tritium is very
often used for the identification of modern recharge, exclusively, in a qualitative way.
This information is, of course, important; but experience has shown that, in many cases,
a semi-quantitative approach for the determination of some hydraulic parameters of
 Parameters of carbonate rock aquifers from tracer methods 369
i r r— i — i
o
TRANSMISS
IBILITY OBTAINED BY PUMPING TEST %
TRANSMISS
IBILITY OBTAINED BV TRACER TEST

© /%
Q 10 1

m
o y
u. ,no ' .*]
<
f ®

r O®
f @
y£
•y*
y ®
—0

• * - \
5
10-6 10- lO"4 10-3 10-2 10" 1 10° 10 1

TRANSMISSIBILITÏ (i^/s)

Fig. 11 - Relationship between specific yield in wells and transmissibility

data from pumping and tracer tests.

groundwater flow is possible through the formulation of mathematical mixing models

which account for the relationship between the input and output functions of the system.

Conceptual models for carbonate rock aquifers as well as specific models for the
interpretation of environmental tritium data have been described elsewhere (Campana &
Mahin, 1985; Simpson & Duckstein, 1976; Maloszewski & Zuber, 1982; Przewlocki &
Yurtsever, 1974; Simpson, 1975; Yurtsever & Payne, 1985; Zuber, 1986). Due to the
characteristics of water circulation in karstic systems, the "good-mixing" model seems
to be the most appropriate. This model shows an exponential distribution of the transit
time and the tritium concentration in the well-mixed reservoir is given by the well known
equation:
t
1 + XT
C(t) = 1 ( C 0 (t') exp ( t - t ' ) dt' (3)

where C 0 (t') = the known tritium input function; t = present time (sampling time); X =
decay constant of tritium (0.05576 year"1); r — turnover time given by the relationship
between the volume of the reservoir and the inflow (volumetric flow-rate); T can also be
expressed in the form T = p.b/R with p = porosity, b = thickness of the aquifer, and
R = recharge in mm/year.

More complicated multi-compartmental mixing models for constant or time-variant flow

have been used for the simulation of tritium and stable isotope concentrations in karstic
systems (Simpson, 1975; Yurtsever & Payne, 1985). The main difficulty for the
implementation of these models is the correct definition of the mixing cells and the
interrelationships between them.

Experimental results
Experimental data of tritium concentrations from four aquifers in different regions have
been interpreted on the basis of the "good-mixing" model of equation (3). The results are
370 A. Plata

shown in Figs 12 and 13. The input functions C 0 (t') have been calculated from the known
tritium concentrations in precipitation of neighbouring stations included in the IAEA-
WMO network (International Atomic Energy Agency, 1981). In the case of the two
Brazilian aquifers (Fig. 12), the data corresponding to the Rio de Janeiro and Brasilia
stations were considered, while for Jamaica and the Dominican Republic the calculations
were based on data from Barbados and Puerto Rico.

The tritium concentrations for the periods where no data are available were calculated
according to the model given in Table 1 which has been obtained using data from 54
Northern Hemisphere stations of the IAEA-WMO network. This model is not very
different to the Ottawa model frequently used for similar calculations.

The tritium concentration given by equation (3) was calculated for different values of the
turnover time ranging from 2 to 500 years. The experimental values of the tritium
concentration found in the aquifer are shown at the right-hand side of the figure. As we
can see, all the points are located below the maximal concentration calculated with the
model and some of these points are very close to this maximal concentration. The data
correspond to water samples collected in the years 1977 (Bahia) and 1978 (Minas Gérais).
Similar results were obtained in the case of the examples from Jamaica and the
Dominican Republic shown in Fig. 13. The distribution of the tritium concentration
measured in groundwater samples taken from different aquifers is expressed in these
cases as the percentage of the maximal concentration given by equation (3). In the case
of Jamaica, the values correspond to water samples collected by Gonfiantini (1971-1973).
The values for the Dominican Republic correspond to samples collected during the years
1987 and 1988.

The previous results show that "good-mixing" processes seem to govern the maximal
concentrations of tritium at the four aquifers studied and that piston flows can be
discarded. Particularly in the case of Jamaica, it was expected that high tritium
concentrations derived from the thermonuclear tritium peak of 1963 (only 8 to 10 years
before sampling) would be found. However, it was not the case because of the presence
of well-mixed reservoirs.

I 1 "1
MINAS
• BAHJA -
GERAIS
^£em a
a
BAHIA
# K
a n = 13
a
«a e

fl
a
S
a
//^MINAS GE =*AIS a» a
a
V
aea a
a
99
a
•
a
a
999
- 9 - s
n = 48
X
4 aa
9
1 i i^"^ a»
10 100
TURNOVER TIME (YEARS)

Fig. 12 - Theoretical tritium concentrations obtained with "good-mixing"

model and comparison with the experimental data.
 Parameters of carbonate rock aquifers from tracer methods 371

JAMAICA
128 samples

JZL
DOMINICAN REPUBLIC
130 samples

PERCENT OF MAXIMAL CONCENTRATION

Fig. 13 - Distribution of experimental tritium data obtained in karst

systems from the Dominican Republic and Jamaica. Values are expressed
as the percent of the maximal tritium concentration calculated by assuming
the "good-mixing" model.

Table 1. Relative weighted mean tritium concentration for precipitation in the Northern
Hemisphere. Data taken from 54 stations of the IAEA-WMO network.

Year Relative Number of a(%) Year Relative Number a(%)

T cone. data T cone. of data
(%o)

1953 9.1 1 1971 61.9 29 7.7

1954 99.2 1 - 1972 44.6 39 8.9
1955 14.2 1 - 1973 37.0 38 9.0
1956 63.4 1 - 1974 42.5 34 13.8
1957 40.7 1 - 1975 32.6 30 12.4
1958 150.1 3 20.2 1976 25.2 38 8.6
1959 154.0 3 23.2 1977 26.9 41 6.2
1960 52.5 5 5.6 1978 28.9 41 5.7
1961 46.7 9 14.8 1979 19.8 39 6.8
1962 309.2 20 80.8 1980 17.0 38 7.3
1963 1000.0 26 - 1981 17.7 40 6.4
1964 581.6 33 93.4 1982 12.5 10 2.6
1965 269.1 32 39.6 1983 10.4 10 2.8
1966 164.7 38 25.3 1984 9.3 11 3.0
1967 100.7 33 16.1 1985 8.9 10 2.6
1968 74.8 39 9.1 1986 8.3 9 2.3
1969 75.0 32 6.1 1987 7.0 5 2.5
1970 67.6 44 9.6 1988 5.8 1 -
372 A. Plata

An important question which arises from the previous considerations refers to the validity
of the turnover time values obtained from the tritium data using this simple model. In the
opinion of the author, the answer to this question depends mainly on the characteristics
of the aquifer. Each tritium concentration value can be considered representative of a
small area of the aquifer. If we are dealing with a free-recharged aquifer where the water
stored is only coming from local precipitation, the turnover times calculated with the
"good-mixing" model can be meaningful. This will not be the case when the karstic
aquifer is confined within the sampling area and the water stored in it comes mainly from
lateral flow. The dispersion model could be more appropriate in this case. In general, the
probability of obtaining true values of the turnover time is higher for high values of the
tritium concentration because they represent, most probably, free-recharge conditions.

In parallel to the large variability of hydraulic parameters in karst systems derived

mainly from the flow concentration, chemical and isotopic data can also show large
variations even over short distances. This fact has to be taken into account for the
interpretation of tritium concentration data.

In the majority of cases, a high density of data is required to obtain reliable values of the
turnover time of water. Information on the flow-pattern inside the well or, at least, on
the depth of the permeable horizons or fracture systems intercepted by each well, will
help for this interpretation. It is not worth applying complicated interpretative models
when only a small number of data is available. Figure 14 shows the variation of the
tritium concentration for the aquifer of Minas Gérais. Four different zones could be
defined. The derived average turnover time for each area is indicated.

The verification of the fulfilment of the "good-mixing" model can be carried out through
the long-term observation of the variation of tritium concentration in the aquifer or in the
discharge from springs. This long-term observation has not been often done.

Fig. 14 - Distribution of tritium concentration in the aquifer of Minas

Gérais. The standard deviation of the experimental data and the calculated
turnover time are indicated for each zone.
 Parameters of carbonate rock aquifers from tracer methods 373

In the case of the aquifer of Minas Gérais (Fig. 14), three wells were sampled for tritium
in April 1977 and in August 1978. The respective values obtained were 7.7 TU, 13.7 TU
and 8.6 TU for 1977 and 7.1 TU, 11.6 TU and 6.6 TU for 1978. Very similar values of
the turnover time were calculated for each set of two values of tritium concentration.

Future possibilities
The results obtained at most of the stations of the IAEA-WMO network show that the
tritium concentrations in the precipitation have dropped close to the values existing prior
to 1953. In other words, the thermonuclear tritium has almost finished and only tritium
produced by secular sources such as nuclear reactions at high atmosphere and solar
emission remains. Contamination from manmade sources has been observed in some
stations during recent years. The tritium concentrations derived from these secular
sources are somewhat higher than expected and range from 2 TU for tropical islands to
25 TU for the central zones of continents in the Northern Hemisphere (Gonfiantini et al.,
1990).

Table 2 contains the tritium concentrations (yearly weighted mean values) measured in
recent precipitation at different IAEA-WMO network stations. Only the last years' data
with more or less constant values (no thermonuclear tritium) have been considered. If no
significant tritium release is made into the atmosphere in future, these or lower tritium
concentrations can be expected in the near future. It is important to stress that these
concentrations can be measured with good precision using the available measuring
techniques.

Figure 15 shows the tritium concentrations which can be expected in the future in well-
mixed karst reservoirs assuming a constant yearly mean concentration for the years after

Table 2. Tritium concentrations measured in the last years (before 1990) precipitation of
some IAEA-WMO stations. The values represent yearly weighted mean concentrations.
The figures in brackets are the number of years averaged.

COUNTRY STATION T CONC. COUNTRY STATION T CONC.

(TU) (TU)

Halley Bay 14.7 (4) JAPAN Tokyo 11.7(5)

Antarctica NEW ZEALAND Kaitaia 2.5 (5)
ARGENTINA Argentine Is. 7.3(3) POLAND Krakow 21.9 (2)
AUSTRALIA Adelaide 5.3 (2) - Cook Is.
•• Brisbane 4.0 (4) - Pacific O. 1.7(3)
" Darwin 3.2(4) SPAIN Madrid 14.1 (3)
" Melbourne 6.1(4) " Barcelona 19.0 (4)
Alice Springs 5.0 (2) SOUTH AFRICA Pretoria 6.5(1)
AUSTRIA Vienna 20.0 (6) TURKEY Ankara 15.8 (3)
COLOMBIA Bogata 5.1 (5) U.S.A. Alburqueque 17.1 (4)
EGYPT Alexandria 10.6 (2) " Anchorage 13.1 (2)
ETHIOPIA Addis Ababa 12.6 (4) Boston Maks 21.9 (4)
" Diego Garcia
GERMANY Berlin 22.9 (2) " Island 2.6 (4)
" Stuttgart 30.9 (3) " Lincoln ND. 1 5.5 (3)
INDONESIA Djakarta 3.2(5) Miami PI. 4.3(6)
IRELAND Valentia 7.5(6) " Waco Tex. 7.3(4)
ITALY Genoa 10.5 (4) CROATIA Zagreb 20.7 (4)
374 A. Plata

1990 of 2.5 TU (left scale) and 25 TU (right scale) for different values of the turnover
time. Assuming a lowest detection limit of 0.5 TU, it is clear that natural tritium will
continue to be an important tool for the investigation of karst systems even in the case
of tropical islands. The lowest curve of Fig. 15 would be valid for the time when the
influence of the thermonuclear tritium from the period 1953-1963 can be neglected. In
this case, the term C 0 (t') accounting for the input function in equation (3) takes the
constant value C0 and the concentration in the aquifer would be given by the expression:

C(t) = _ E ° (4)
1 + Xt

T 1 — I — i i i l 1 1 1 1—i—i i • I 1 1 1—i 50

TURNOVER TIME (YEARS)

Fig. 15 - Prediction of future tritium concentration in karst systems on the

basis of the "good-mixing" model.

REMARKS ON ARTIFICIAL TRACING

Introduction

Many experiments of artificial tracing of karst systems have been reported during the last
decades aimed at obtaining information on relevant dynamic parameters (Burdon et al.,
1963; Leontiadis & Dimitroulas, 1973; Maloszewski & Zuber, 1984; Atkinson & Smith,
1974; Batsche etal., 1970; Burdon, 1963; Calmels etal., 1977; Quinlan & Rowe, 1977;
Smart, 1976, 1982; Zojer & Stichler, 1989). Different chemical and radioactive tracers
have been used for this purpose.
Among the first, fluorescent dyes have achieved a position of prominence and they are
at present the most commonly used. In the case of radioactive tracers, artificial tritium
plays the most important role. The main advantages of this tracer are:
(i) Isotopic labelling of the water molecule.
(ii) Possibility of labelling very large volumes of water using reasonable tracer
amounts. For example, with 100 Ci of tritium and electrolytic enrichment of the
water samples, a volume of about 6 x 109 m3 of water can be labelled (limit of
detection considered is 5 TU equivalent to 0.016 ^Ci/m 3 ).

(iii) Low radiotoxicity: maximal permissible limit of intake for radiation workers is
109 Bq/year equivalent to 27 mCi/year; and a tenth of this value for members of
the public.
 Parameters of carbonate rock aquifers from tracer methods 375

In the opinion of the author, when quantitative information is needed, artificial tritium
is practically the only choice. In most cases, the combination of environmental (chemical
and isotopic tracers) and artificial tracer methods provides the best results. The following
information can be sought using these experiments:

(i) Determination of storage capacities.

(ii) Localization of recharge areas (origin of the water).

(iii) Determination of flow parameters, such as transit time and dispersivity.

Experiment
The experiment was carried out in connection with the investigation of the origin of the
water discharging at the Reocin mine (Spain) in a gallery located about 340 m below the
surface. A more or less constant flow rate of about 1200 1/s was emerging through a
single fracture of the calcareous rock (Gargasian dolomite and limestone) (Plata, 1989).
Deuterium and oxygen-18 analyses showed that about 72% of this water was coming from
the River Saja which is crossing the same calcareous formation at the outcropping area
located about 8 km from the mine (Fig. 16). The remaining 28% was coming from local
precipitation. The following mean annual &D and ô 18 0 values were obtained from one of
the three sampling programme which were performed:
Local groundwater not related to the mine: ÔD = —31.9%o, <5i80 = —5.62%o
River Saja: ÔD = -45.7%o, <5180 = -7.38%o
Mine discharge: ÔD = -41.1%», t>180 = -6.89%o
The more negative ô-values for the River Saja are because the water of this river comes
from high altitudes (about 1500 m mean altitude compared with 400 m for the mine site).
The seasonal variations of t>180 values for the River Saja and mine discharges are shown
in Fig. 17. In the case of the river, composite water samples representing each of the
indicated periods of time were measured. The dotted line represents the estimated
seasonal variation. Obviously, the probable error of the seasonal variations obtained for
the mine discharge is very large due to the small difference between the maximum and
minimum o180-values (the analytical error is about 0.5%o).

In principle, this seasonal variation can be treated as a sinusoidal function and, assuming
a well-mixed reservoir, the turnover time of the groundwater system can be calculated
by the equation (Oeschger & Gugelmann, 1974):

| | CENOMANIAN
FAULT
g £ g £ f MINERALIZATION

T EXPLOITATION WELL

Fig. 16 - Area of the Reocin mine, Spain.

376 A. Plata

Fig. 17 - Seasonal variations of ô 1 8 0 in the mine and River Saja.

AC (5)
C„ — sin2irt - 2 T cos2xt
(l/r)+4ir 2 r T

where AC = maximum deviation of ô 1 8 0 with respect to C m (peak of the sinusoidal

function); T = turnover time = V/Q with Q expressed in m 3 /year.

Due to the mixing process, the sinusoidal input function is smoothed in the reservoir and
the parameter AC at the outlet of the system (mine discharge) is transformed into A C
(Fig. 17). The difference in time between the two peaks is explained by the transit time
of the water from the river to the mine as well as by the smoothing effect of the
successive waves of tracer, as shown in Fig. 18. The difference between the mean values
C m and C m ' is explained by the mixing of the water from the River Laja with
groundwater coming from local precipitation which has an average ô 1 8 0 value of
-5.62%o.

The time corresponding to the peak of the tracer wave at the outlet of the system can be
calculated by making the first derivative of equation (5) equal to zero. The following
equation is obtained:

0.2 0.4 0.6

FRACTION OF TIME UNIT (YEAR)

Fig. 18 - Variation of the output function with the turnover time for the
"good-mixing" model with sinusoidal input function.
 Parameters of carbonate rock aquifers from tracer methods 317

1 • 1
-—arctan-— (6)
2TT 2TT

By substituting this value of t max in equation (5), the concentration (or ô-value)
corresponding to this peak is obtained, from which the value of AC can be derived.
Figure 19 shows the variation of the relationship AC/AC as a function of the turnover
time. In the present case, the approximative values AC = 0.41%o and A C = 0.109%o
were obtained. These values give a turnover time of r = 0.56 year.

As the flow-rate discharging in the mine is 1.2 m 3 /s, equivalent to 3.78 X 107 m3/year,
a volume of the groundwater reservoir participating in the process of 0.56 X 3.78 X 10 7
= 21.2 x 10s m5 is obtained. However, this volume cannot be considered valid because
the smoothing of the tracer wave is partially produced by the mixing of waters with very
different residence times, as will be shown later.

r= V/Q
Fig. 19 - Variation of the parameter AC/AC as a function of the turnover
time for "good-mixing" model with sinusoidal input function.

\ ?o statistical error

S- 12.3*^'days

60 80 100
TIME AFTER INJECTION IDAYSI

Fig. 20 - Breakthrough curve of the tritium injected at the infiltration zone

of the River Saja.
378 A. Plata

On the other hand, an interconnection experiment using artificial tritium was carried out
in order to verify the origin of the water discharged in the mine. An amount of 3.5 Ci
of this isotope was injected in a sinkhole near the River Saja in the zone of the Gargasian
calcareous formation outcrop. The tritium reached the mine about 20 days after the
injection and the breakthrough curve shown in Fig. 20 was measured. An average tracer
transit time of 62 days was obtained. The volume of water which emerged in the mine
during this period was about 6.4 x 106 m3, which represents only 30.2% of the volume
calculated on the basis of the ô 18 0 variations. In principle, one can suppose that the
tritium which emerged in the mine has travelled through the main fractures or solution
channels (primary porosity) and its residence time in the reservoir was not long enough
to achieve an equilibrium with the more stagnant water stored in the small fissures and
cracks (secondary porosity). The situation is different for stable isotopes because, in this
case, we are dealing with a continuous tracer injection throughout the year.

The integration of the tritium breakthrough curve gives an area of 12.3 /xCi m"3 day"1.
From this area, an activity of 1.27 Ci is obtained as the total tritium activity emerged in
the mine. This value represents only 36.4% of the injected value. The only reasonable
explanation is that the rest of the activity (63.6%) has been transferred into the water
which is travelling through the secondary porosity.

According to these results, the water flow emerging in the mine can be separated into the
following components:
(a) Water coming from local precipitation with an average o 18 0-value of — 5.62%o. Only
small seasonal variations of this value can be expected. This component comprises
28% of the total flow-rate.

(b) Water coming from the River Saja through the primary porosity. According to the
tritium experiment, a turnover time of 6.4 X 106/3.78 X 107 = 0.17 years is
estimated for this component. The <5180 output function for this water can be
calculated from equation (5) using the values AC = 0.41%o and C m = —7.38%o, as
well as the previous turnover time.

(c) Water coming from the River Saja through the secondary porosity. Due to the long
residence time of this water, confirmed by the lack of a clear tail for the tritium
breakthrough curve, a constant o180-value of — 7.38%o (mean annual value for the
river) can be accepted for this component.

Therefore, the ô 18 0 variation of the mine discharge as a function of time can be

calculated by the expression:

(ô 18 0) minc = 0.28 X 5.62 + 0.72 [7.38f + (1 - f)(o 18 0) rivcr ]

where f = fraction of flow-rate coming from the River Saja through the secondary
porosity (third component); (ô 18 0) river = ô180-value for the second component at the
discharge point.

The best fit to the o180-values observed at the mine discharge (Fig. 17) corresponds to
a fraction f = 0.3. The smoothing effect of the ô 18 0 wave is very well reproduced. Using
this value of f, the respective contributions to the mine discharge of the above three
components would be 28%, 50.4% and 21.6%. Using the tritium recovery data, these
values would be 28%, 72 X 0.364 = 26.2% and 72 X 0.636 = 45.8%, respectively.

The discrepancy can be explained by the fact that the tracer distribution between the
primary and secondary porosity was not the same in the case of the point injection (in
time and space) of tritium and the continuous total scale injection of stable isotopes.
 Parameters of carbonate rock aquifers from tracer methods 379

Obviously, the results provided by the stable isotopes are more representative. No
alternative method is available to prove the validity of these results. But they show the
limitations of artificial tracer experiments for the determination of quantitative values of
aquifer parameters.

REFERENCES
Atkinson, T. C. & Smith, D. I., 1974, Rapid groundwater flow in fissures in the chalk:
an example from south Hampshire. Quart. J. Engng Geol., 7, 197-205.
Baonza, E., Plata, A. & Piles, E., 1970, Aplicaciôn de la técnica del pozo unico
mediante el marcado de toda la columna piezométrica. In: Isotope Hydrology (Proc.
Symp. Vienna, 1970), 659-711. IAEA, Vienna.
Batsche, H. et al., 1970, Kombinierte Karstwasseruntersuchungen im Gebiet der
Donauversickering (Baden-Wûttenberg) inden Jahren 1967-1969, Steirische Beitrâge
zur Hydro'géologie, 55-165.
Borowczyk, M., et al., 1965, Laboratory investigations on the determination of filtration
velocity by means of radioisotopes. Atomkernenergie, 10, 51-56.
Burdon, B. J., Erickson, E., Payne, B. R., Papadimitropoulos, T. & Papakins, N., 1963,
The use of tritium in tracing karst groundwater in Greece. In: Radioisotopes in
Hydrology (Proc. Symp. Tokyo, 1963), 309-320. IAEA, Vienna.
Burdon, D. J., 1963, The use of tritium in tracing karst groundwater in Greece. In:
International Association of Scientific Hydrology Symposium, March 1963, 309-320.
Calmels, A., et al., 1977, Multitraçages dans les massifs karstiques du Vercors et de la
Chartreuse (France). In: Underground Water Tracing (Ljubljana: Institute of Karst
Research) (ed. by P. Gospodaic & P. Habic), 239-254.
Campana, M. E. & Mahin, D. A., 1985, Model-derived estimates of groundwater mean
ages, recharge rates, effective porosities and storage in a limestone aquifer. /.
Hydrol.
Drost, W., 1986, Groundwater measurement by means of radioactive tracers.
Atomkernergie, 48 (2).
Gonfiantini, R., 1971-1973, International Atomic Energy Agency, Vienna (unpublished
data).
Gonfiantini, R., Araguas, L. &Rozanski, K., 1990, Tritium in precipitation: comparison
between the years of high tritium (1963-67) and recent years. Bull. Rad. Protection,
13(1), Jan-March.
Grinbaum, 1.1., 1965, GeoflziceskieMetody OpredelenijaFiltracionnych Svojstv Gornyh
Porod. Nedra, Moscow.
Guizerix et al., 1967, Appareil par la mesure des vitesses relatives des eaux souterraines
par la méthode de dilution ponctuelle. In: Isotopes in Hydrology (Proc. Symp.
Vienna, 1966). IAEA, Vienna.
Halevy, E., et al., 1967, Borehole dilution techniques: a critical review. In: Isotopes in
Hydrology (Proc. Symp. Vienna, 1966). IAEA, Vienna.
International Atomic Energy Agency, 1981, Statistical treatment of environmental isotope
data in precipitation. Tech. Report Series no. 206, IAEA, Vienna.
Kocherin, O. J., 1916, Material po vodnym izyskanijam Krium, idrometriceskoij otdel.
Izd. otd. zem ulcsenij, vyp. 1. Simferopol.
Leontiadis, I. & Dimitroulas, C , 1973, The use of radioisotopes in tracing karst
groundwater in Greece: IV. Investigation of the possible interconnection between the
sinkholes of Milea and Tala Lake with various springs of the near area. Report Demo
73/4E, Demokritos Nuclear Research Centre, Athens.
Maloszewski, P. & Zuber, A., 1984, Interpretation of artificial and environmental
tracers in fissured rocks with a porous matrix. In: Isotope Hydrology 1983 (Proc.
Symp. Vienna, 1983), 635-651. IAEA, Vienna.
Maloszewski, P. & Zuber, A., 1982, Determining the turnover time of groundwater
382 Luigi Tulipano et al.

presence of different carbonate faciès, results in a very variable degree of permeability

in the different zones of the aquifer and, within the same zones, at different depths. The
Murgia, as a whole, represents a considerable hydrogeological unit where groundwaters
flow according to preferential water levels, irregularly distributed in accordance with the
strong degree of anisotropy characteristic of the aquifer (Grassi et al., 1977).
The aquifer is coastal in type and the sea level is the base level of the groundwater
circulation. Close to the coastline, fresh water almost certainly floats on intruding sea
water. The only discharge of groundwater is through springs scattered along the coastline
and offshore (Tadolini et al., 1983). Carbonate outcrops in the Murgia allow the
infiltration of a certain amount of precipitation. The amounts vary according to the
location and depend also upon the morphological conditions, the hydrological properties
of the outcropping rocks, the type and the thickness of the loam covering and the
presence and type of vegetation.

A preliminary analysis of such characteristics allows the determination in the innermost

zones, at altitudes between 300 and 600 m a.s.L, of the areas that mainly feed the
hydrogeological system. Groundwaters from the feeding areas flow to the sea and partly
feed the contiguous hydrogeological system of the Salentine Peninsula, starting from
hydraulic heads even exceeding 200 m a.s.l. The hydraulic gradient is very variable from
zone to zone, following the irregular distribution of the permeability of the aquifer
(Grassi & Tulipano, 1983).

STABLE ISOTOPIC COMPOSITION OF GROUNDWATERS

6D%o and ô180%o values in groundwaters sampled in the Murgia region are shown in
Fig. 1. This figure also shows the trends of the Meteoric Water Line (M.W.L.) (Craig,
1961) and of the Mediterranean Water Line (Mediterranean W.L.) (Gat & Carmi, 1970),
representing respectively the locus of the isotopic composition of fresh-water sources on
a worldwide scale and of precipitation in the eastern Mediterranean Sea area.

-44 -II -7.6 -12 -4M -6.4

-6.0 -5.6
6*0 % . SMOW
Fig. 1 - Stable isotope composition of groundwaters sampled in the
Murgia hydrogeological unit.
 The multitracing approach in studies ofkarstic amd coastal aquifers 383

Isotopic data of groundwaters in Fig. 1 are grouped according to three roughly parallel
alignments. These groups coincide with a geographical zonation of the territory.
Notwithstanding the apparent uniformity of the Murgia region, the three zones A, B and
C (Fig. 2), relating to the three groups of groundwaters, actually show different morpho-
altimetric and climatic characteristics. Moving from zone A to zone C, in decreasing
latitude, one notices a progressive diminution in the average topographic altitudes, an
increase in the average annual temperature and climatic variations due to a stronger sea
influence.

Fig. 2 - Outline of the connections between hydrogeological circuits

intercepted by the sampling wells and the feeding areas located at different
altitudes (m a.s.l). The zonation comes from the groupings recognized in
Fig. 1.

Values aligned in an isotopic diagram could be considered as a mixing product between

extreme terms. However, a careful assessment of the chemical composition of waters, as
well as of their geographical position, leads to the exclusion of the idea that alignments
are the result of mixing, as this is possible only in a few isolated cases. The distribution
along a single alignment may then be related to a change in altitude of the feeding area.
In this case it is possible to attribute to each well intercepting the examined water a
connection with the most probable feeding area. In practice it confirms that there is a
spatial selection of precipitation besides the seasonal selection already stated, that is
confirmed here by the depleted isotopic values in groundwater samples, which are
characteristic of winter precipitation.

The notable differences among the isotopic values in general and particularly among
those regarding geographically close sampling points, confirms that the aquifer is not
homogeneous (Alaimo et al., 1988) and that local characteristics of feeding waters are
likely to be preserved at great distance by means of preferential pathways that do not
connect with other circuits. Figure 2 represents only a scheme of the connections between
the hydrogeological circuit intercepted by each sampling well and its probable feeding
area. However such direct connections represent the minimum length of the path followed
by groundwaters.

CHEMICAL INDICATORS OF MATURITY

Previous studies dealt with the chemical composition of groundwaters of the Murgia
region, and with the main phenomena regulating the variation of these chemical
characteristics as a result of water-rock interactions (Alaimo et al., 1988) and mixing
with sea water (Fidelibus & Tulipano, 1986). Here, only aspects of groundwater
384 Luigi Tulipano et al.

°26
A52
10-
A108
o20
"A "% °« m
-e- 8 33 3^1
%
JE "• oM ^
• 7
82 D 37 "
t O67
re 73 A A
a "
69
6
32

5
— • 1 1 1

0.01 0.02 0.03 0.04

4
Sr" (meq/l)
Fig. 3 - (Ca 2+ + Mg 2+ ) vs. Sr 2+ content of sampled waters; symbols
refer to the groupings of Fig. 1.

chemistry that are likely to provide parameters of its evolution according to the different
residence times in the aquifer, are examined.
Groundwaters flowing in a carbonate aquifer undergo an evolution mostly resulting from
rock dissolution and from carbonate precipitation. The dissolution of a rocky matrix,
such as the one previously described, causes the groundwaters to become rich in Ca, Mg
and, among the minor constituents, Sr ions. The ratio between the Sr content and the sum
of Ca and Mg shown in Fig. 3 demonstrates that the Sr content is very variable under
equal (Ca + Mg) contents.

Carbonate precipitation processes likely to occur when groundwaters are moving from
feeding areas to the discharge point and in which surely Ca plays the main role, tend to
limit the increase of (Ca + Mg), even in the case of numerous cycles of dissolution and
re-precipitation. Conversely, Sr which is in small quantities in the rocky matrix and is
not found in precipitated calcite, increases in concentration after several cycles of
dissolution/re-precipitation (Plummer et al., 1976). The comparison between the Sr
concentration and the (Ca + Mg) content may indicate, for each sampled water, the
quantity of dissolution and possible re-precipitation.

Figure 3 includes sample no. 17: this sample shows the minimum value of (Ca + Mg)
sum, and the lowest value of Sr. Its chemical composition can be considered as the
starting point for the evolution towards the chemical composition shown by the other
groundwater samples.
The evolution can be estimated in the diagram of Fig. 4 showing the representative points
of the examined samples in accordance with the percent variation of Sr content and of the
(Ca + Mg) content as to the ones of sample no. 17. The same diagram shows the lines
corresponding to different A(Sr)%/Ad(Ca + Mg)% ratios; this allows the determination
of waters which, as the ratio increases, have undergone a growing number of dissolution-
precipitation cycles. Such waters, then, according to this evolution, reach higher stages
of "maturity". Such stages can be related to the residence time of the fresh water in the
 The multitracing approach in studies ofkarstic amd coastal aquifers 385

0 20 40 60 80 100
A (Ca~+ Mg~) % (meq/l!
Fig. 4 - Percent variation of both (Ca 2+ + Mg 2+ ) and Sr 2+ content of
sampled waters in relation to those of a sample considered to be the
starting point of groundwater evolution. Symbols refer to the groupings
of Fig. 1.

aquifer, since a longer period of water-rock contact causes higher stages of maturity to
be reached.

For waters mostly occurring in a dissolution process of a calcium-magnesium matrix, the

increase in salinity must be correlated to the (Ca + Mg) content. Figure 5 shows the
trend of (Ca + Mg) content with respect to total salt content, expressed in meq/l. If we
refer only to the waters that, according to previous data, have been classified as a
product of prevalent dissolution, the increase of the (Ca + Mg) sum provides an
evaluation, in relative terms, of the residence time of these groundwaters in the aquifer.
Such information makes it feasible to differentiate the hydrogeological circuits,
previously outlined using isotopic data, with reference to their relative velocity. These
fast circuits originate mainly from two main feeding areas (Fig. 6).

TEMPERATURE OF GROUNDWATER AS AN INDICATOR OF MOBILITY

Previous work, both in the present region and in other hydrogeological situations, showed
the great potential of methodologies based on space reconstruction and on the
interpretation of the interference brought about by groundwater circulation on the
geothermal gradient. Especially in a strongly anisotropic aquifer, the reconstruction of
isogeothermal surfaces allows the main feeding areas and the preferential flow paths to
be recognized (Tulipano, 1988). Past investigations in the Murgia region support this,
even though the scarcity of wells available for thermal loggings allowed no detailed
reconstruction (Cotecchia et al., 1978).

As an example, Fig. 7 shows the reconstruction of isothermal curves for a horizontal

section (600 m b.s.l.), together with a vertical section showing the distribution of
temperature resulting from groundwater circulation in the Murgia unit.
386 Luigi Tulipano et al.
10

72 O"

\
33<#>
o"
8-
55 O77 :s
12
o O 61
59, O «
47 O 73
69
.,32 0 R<2
O 2<R<5

o 5<R<10

O R>10

—J— I i 1

10 14 18 22
Z (Car + An) (meq/l)

Fig. 5 - (Ca 2+ + Mg 2+ ) vs. total salt content; symbols refer to the R

values of Fig. 4.

The global amount of annual precipitation feeding the aquifer is characterized by a

temperature corresponding to the average of atmospheric temperature during the recharge
period. During the percolation and the following filtration, waters tend to reach a thermal
equilibrium with the rocks they pass through, and in the mean time they modify the
geothermal gradient which would exist in the absence of groundwater circulation. In fact,
a heat transfer from rock to water occurs.

Then, the longer the residence time of the water in the aquifer, the higher the
temperature will be regarding the one peculiar to the feeding waters. Main discharge
directions coincide with the minimum increase in temperature directions, that is along
lines of minimum thermal gradient.

WATER SJtmKG

Fig. 6 - Hydrogeological connections (outlined in Fig. 2) differentiated

with reference to their relative velocity and to the maturity of sampled
groundwaters.
 The multitracing approach in studies ofkarstic amd coastal aquifers 387

sE
AC P I A T1C * _^——
\
/>?SL—r-—
• ; i»-r- •

« -^—
«5'
15

0 10 20 km
* 1 Tsrinte

m • Mo» • . i l l is ni A U SEA

Vy/^7~.Z-r— 1
0 is- r^T - p - l d — 1 ,. "
m
400
At\T l^XX^N "
\ \ \ \ \ w i>5 is ; • \
too
« 11 17 \ \ " ,' / ,'
1200 - X
X '' !«' /
1600

2000 0 5 10 km 17

Fig. 7 - An example of the reconstruction of isothermal curves for the

Murgia unit: horizontal section (600 m b.s.l.) and vertical section.

DETERMINATION OF GROUNDWATER CIRCULATION BY MEANS OF A

MULTITRACING APPROACH

Besides hydrogeological data obtained by traditional geological surveys, previous

paragraphs showed how other methodologies may provide valuable additional
information. The combination of all such information made it possible to determine a
complete and detailed hydrogeological picture concerning the actual flow paths of the
groundwaters, how each part of the aquifer is connected with the main feeding areas, and
the relationship between the Murgia hydrogeological system and the adjacent systems.

Such a combination is evident in the compiled map shown in Fig. 8 representing an

example of the effectiveness of multitracing methodology. Even though the map has been
drawn up from a limited number of observation points, it makes it possible to verify the
congruence of information that can be obtained by different ways. By increasing the
observation points it will be easy to obtain more detailed information

The synthetic elements used for the reconstruction of the hydrogeological conditions are:
the trend of isolines of 5D%o, establishing the connections between the recharge areas
and the discharge zones, the trend of isothermal curves, showing preferential paths and
final information about the degree of maturity.

The most important conclusion likely to be derived from the map is that the main
recharge areas are located along the ridge of the Murgian hills. From the first area,
northwest of the hydrogeological system, originates a fast flow bound to the coast which
reaches a great depth beneath the zone of well no. 20 and feeds by a preferential circuit
at least one of the Trani springs. A second very slow flow originates from the same area
and flows to an adjacent hydrogeological unit. Two fast flows separated by a zone with
a slower locally fed flow, originate from recharge areas along the ridge of the Murgian
388 Luigi Tulipano et al.

Fig. 8 - Map showing main and minor hydrogeological pathways outlined

on the basis of all previous information, including ôD%o isolines and
isothermal curves (200 m b.s.l.).

hills, at an altitude ranging from 400 to 500 m. The influence of the southward flow
reaches the coast, where it contributes directly to two big concentrated springs of the
Torre Canne group. The considerable contribution from the Murgia hydrogeological unit
to the Salentine one, bordering the southeast, is shown by the fast circuit with a trend
parallel to the coastline. Waters which normally belong to slow circuits flow to the
Ionian Sea; they are fed by the southeast zone of the Murgia, starting from an altitude
below 400 m. The presence of a big concentration of coastal springs of the Taranto Gulf
leads to the hypothesis of a swift hydrogeological circuit. However, no proof is available
because there is a lack of observation points upstream of the spring group.

The proposed representation shows the complexity of the system, but then it is well
definable on a regional scale. It represents a starting point for reconstruction on a local
scale, which can be obtained by the adoption of the same multitracing methodology.

REFERENCES
Alaimo, R., Aureli Grifeo, A., Fidelibus, M. D. & Tulipano, L., 1988, Chemical and
isotopic methodologies in the studies on origin and evolution of groundwaters
flowing in the coastal carbonatic and karstic aquifer of Apulia (southern Italy). 10th
Salt Water Intrusion Meeting (Ghent, Belgium).
Cotecchia, V., Tadolini, T. & Tulipano, L., 1978, Ground water temperature in the
Murgia karst aquifer (Puglia, southern Italy). Int. Symp. on Karst Hydrology
(Budapest).
Cotecchia, V. & Tulipano, L., 1982, Caratteristiche fisiche del territorio della provincia
di Bari ed alcuni aspetti dell'impatto geo-ambientale prodotto dalla urbanizzazione
(in Italian). Convegno di studi sull'area metropolitana di Bari (Bari, Italy).
Craig, H., 1961, Isotopic variations in meteoric waters. Science, 133.
Fidelibus, M. D. & Tulipano, L., 1985, Determination of the isotopic characteristics of
water feeding the coastal aquifers of the Murgia and Salento (Apulia - southern Italy)
deduced from an examination of the isotope contents of groundwaters mixed with
waters of marine origin. V Int. Symp. on Groundwater (Taormina, Italy).
Fidelibus, M. D. & Tulipano, L., 1986, Mixing phenomena owing to sea water intrusion
for the interpretation of chemical and isotopic data of discharge waters in the
Apulian coastal carbonate aquifer (southern Italy). 9th Salt Water Intrusion Meeting
(Delft, The Netherlands).
 The multitracing approach in studies ofkarstic amd coastal aquifers 389

Gat, J. R. & Carmi, I., 1970, Evolution of the isotopic composition of atmospheric
waters in the Mediterranean Sea area. / . Geophys. Res., 75.
Grassi, D., Tadolini, T., Tazioli, G. S. & Tulipano, L., 1977, Ricerche sull'anisotropia
dei caratteri idrogeologici delle rocce carbonatiche mesozoiche della Murgia nord-
occidentale (in Italian). Geol. Appl. e Idrogeol., XII, Bari.
Grassi, D. & Tulipano, L., 1983, Connessioni tra assetto morfo- strutturale della Murgia
(Puglia) e caratteri idrogeologici della falda profonda verificati anche mediante
l'analisi della temperatura delle acque sotterranee (in Italian). Geol. Appl. e
Idrogeol., XVIII, Bari.
Plummer, L. N., Vacher, H. L., MacKenzie, F. T., Bricker, 0 . P. &Land, L. S., 1976,
Hydrogeochemistry of Bermuda: A case history of groundwater diagenesis of
biocalcarenites. Geol. Soc. Am. Bull., 87.
Tadolini, T., Tinelli, R. & Tulipano, L., 1983, Discharge conditions and variations of
the main hydrological parameters of some coastal apulian springs relating to sea
water influence on groundwater. 8th Salt Water Intrusion Meeting (Bari, Italy).
Tulipano, L., 1988, Temperature logs interpretation for identification of preferential
flow pathways in the coastal carbonate and karstic aquifer of the Salento Peninsula
(southern Italy). In: Karst Hydrogeology and Karst Environment Protection (Proc.
21st IAH Congress, Guilin, China). IAHS Publ. no. 176.
Hydrogeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990). .n „
IAHS Publ. no. 207, 1993. 391

RELATIONS ENTRE REPONSES IMPULSIONNELLES ET CONDITIONS

HYDRODYNAMIQUES DANS LE CADRE DE TRAÇAGES ARTIFICIELS DES
AQUIFERES KARSTIQUES: APPLICATIONS SUR COLONNE DE
LABORATOIRE ET SUR UN SYSTEME KARSTIQUE A DOUBLE ENTREE
DANS LE CAUSSE DE GRAMAT (LOT, FRANCE)

M. DZIKOWSKI, E. CARLIER, N. CRAMPON

Laboratoire d'Hydrogéologie (SN5), Université des Sciences et Techniques de Lille, 59655 Villeneuve d'Ascq,
France

G. DE MARSILY
Laboratoire de Géologie Appliquée, Université Pierre et Marie Curie, 75252 Paris Cedex 05, France

RESUME
Cet article se propose d'étudier les problèmes de transport d'un traceur conservatif en
milieux poreux ou karstiques dans lesquels les conditions hydrodynamiques sont
susceptibles de varier au cours du temps. Il arrive en effet souvent, et en particulier dans
les systèmes karstiques, que le débit d'écoulement se mettre à varier au cours
d'expériences de traçages artificiels. Des relations théoriques sont ici établies entre la
réponse impulsionnelle à l'injection instantanée d'un traceur et le débit d'écoulement
constant à l'intérieur d'un tube. Extrapolées pour des conditions de débit variable ces
relations aboutissent à une "méthode de transformation" des réponses impulsionnelles en
fonction des conditions hydrodynamiques. Cette approche est une étape vers l'analyse des
systèmes non linéaires. La méthode est appliquée et validée sur des résultats obtenus en
colonne de laboratoire puis utilisée afin d'étudier les transferts de type conduit karstique
sur un système du Causse de Gramat où des expériences de traçages artificiels ont été
réalisées en périodes hydrologiques différentes.

ABSTRACT
The purpose of this paper is to study the problem of the transfer of a conservative tracer
in a karst or porous aquifer where the hydrodynamic regime can vary, as it is often the
case of tracer tests in karst terranes. Theoretical relationships established between the
impulse responses to an instantaneous injection of tracer and the constant flow rate in a
pipe are extrapolated for impulse responses defined in an unsteady state. This model is
called "method of transformation". This approach is a step toward the analysis of non
linear systems. For different scales of transport the impulse responses are defined from
short time injection. The validity of these relationships has been quantitatively assessed
by tracer tests in columns in the laboratory at different constant flow rates and for a
variable flow rate. The method of transformation is used for the study of transfer in
conduit flow on a karstic system in the "Causse de Gramat" where tracer experiments
have been undertaken for different hydrodynamic conditions.

INTRODUCTION
Cet article se propose d'étudier le problème de la caractérisation du transport d'un
traceur conservatif dans un système où les conditions hydrodynamiques sont susceptibles
de varier au cours du temps. Il arrive en effet souvent, dans les systèmes karstiques par
exemple, que le débit du karst se mette à varier pendant l'essai, empêchant généralement
toute interprétation quantitative de la "réponse" du karst à l'injection du traceur.

A partir de modèles théoriques simples, nous avons mis au point une méthode de calcul
392 M. Dzikowski et al.

de la "réponse" d'un système à une variation de débit. Cette méthode sera d'abord
comparée à des résultats obtenus en laboratoire, puis appliquée sur un système à double
entrée (Thémines-Théminettes-Besaces) situé dans le Causse de Gramat (Lot), où deux
campagne de traçages ont été réalisées, la première en période de basses eaux (Mai
1987), la seconde en période de hautes eaux (Février 1988).

L'interprétation de traçages artificiels en aquifère dans le cas où la structure des terrains

concernés est inconnue ou lorsqu'il est difficile d'appréhender les paramètres du transport
nécessite une approche dite "approche système" du phénomène ou encore "système
traçage" (Mangin, 1975; Shirley & Dreiss, 1989). Ce type d'approche repose sur des
concepts de stationnarité et de linéarité des phénomènes et nécessite donc des conditions
d'applications particulières (Margarita et al., 1984; Duffy & Gelhar, 1985).

Lorsque le traceur est injecté de façon brève et ponctuelle à l'entrée d'un système, la
courbe d'évolution des concentrations ou des flux en fonction du temps à la sortie
représente la durée des temps de séjour (DTS) du traceur dans le système. A partir de
cette DTS, il est possible de déterminer la réponse impulsionnelle (réponse à une
injection unitaire), qui est caractéristique du système étudié. Cette fonction dépend donc
du système physique (Walliser, 1976; Lepiller & Mondain, 1986), dans le cas particulier
des traçages, le système est défini par l'espace occupé par la masse de traceur lors de son
transfert (Molinari, 1976). La variation des conditions hydrodynamiques d'un traçage à
l'autre, peut modifier cet espace, et joue un rôle essentiel dans la durée des temps de
transfert (Smart, 1988). Lorsque le système est sous pression, cet espace n'est pas
modifié (Stanton & Smart, 1981).

Certains modèles de transfert sont basés sur la "définition" de la réponse impulsionnelle

ou de la fonction de transfert (transformée de Laplace de cette première), ces modèles
induisent un échange de matière entre un liquide en écoulement et une phase stationnaire
(Villermaux & Swaaij, 1969; Zuber, 1986). La démarche entreprise ici n'est pas de
caractériser une réponse impulsionnelle, mais d'étudier les variations de celle-ci en
fonction de conditions hydrodynamiques variables.

DETERMINATION DE LA REPONSE IMPULSIONNELLE D'UN SYSTEME-

TRAÇAGE

La réponse impulsionnelle est dérivée de la fonction concentration-temps à la sortie du

système (pour une injection brève et ponctuelle) par les relations suivantes:

h(t) = Q(t)C(t)/JQ(t)C(t)dt (1)

-00

où Q(t) est le débit au temps t, C(t) la concentration au temps t, et h(t) le valeur de la

réponse impulsionnelle au temps t.

Lorsque le débit entre le temps d'apparition et le temps final de la DTS est constant
l'équation (1) devient:
-f 00

h(t) = C(t)/ [ C(t)dt (2)

-<x> „
Ces équations permettent d'écrire la relation:
+ 00

f h(t)dt = 1 (3)
 Relations entre réponses impulsionnelles et conditions hydrodynamiques 393

REPONSES IMPULSIONNELLES ET DEBIT VARIABLE

Les problèmes liés à la variation du débit
Lorsque le débit varie au cours du traçage, l'état du système est influencé par cette
variation. Dans la suite de l'exposé, le système physique (espace occupé par le traceur
au cours de son transfert) est supposé ne pas varier en fonction du débit. Seul le débit
ou la variation du débit influe sur la réponse impulsionnelle.

L'analyse des réponses impulsionnelles dans un tube en fonction du débit constant

d'écoulement a permis d'établir des relations entre les signaux synthétisées par la Fig. 1
(Dzikowski étal., 1991).

i i A
h(t)

h(n At'
s
v temps temps
>

Fig. 1 - Transformation de t et du signal h(t) au débit Q en t' et h'(t') au

débit Q'.

Les relations permettant de transformer la réponse impulsionnelle 1 et d'obtenir la

réponse impulsionnelle 2 sont:

t' = Q t/Q'
(4)
h'(t') = (Q'h(t))/Q

L'établissement de relations entre réponses impulsionnelles à débits variables impose la

discrétisation du temps à partir du temps d'injection en pas égaux à At. Cette
discrétisation du temps implique dans le cas de la Fig. 1:

h(t) At = h'(t') At' (4')

Cette relation vérifie:

S)h/(tOAt/ = 1

en effet:

£h(t)At = £hV)At' = £^h(t)At'

-OO -00 -Co Ne

Ce qui impose:
At' = (Q/Q') At
394 M. Dzikowski et al.

RELATION ENTRE REPONSE IMPULSIONNELLE A DEBIT CONSTANT ET

REPONSE IMPULSIONNELLE A DEBITS VARIABLES (CAS DU TRANSFERT DE
MASSE)
Hypothèses de base

Dans le cas d'un système piston considérons le signal de sortie suivant (Fig. 2). Le
traceur étant injecté au temps t = 0 de façon brève et ponctuelle nous définissons:

h(t)

temps
t
Fig. 2 - Sortie du signal d'entrée au temps t.

Qt = XS
où Q est le débit constant, S la surface moyenne d'écoulement, X la distance (entrée-
sortie), et t le temps convectif. Imaginons, dans un deuxième temps, la même expérience
mais avec une variation du débit Q' en fonction du temps à partir du temps d'injection
t = 0. Si t' est le temps convectif de sortie du système il est possible d'écrire:
t'

XS = [Q'(t)dt

Cette relation implique:

t'

Qt = [Q/(t)dt (5)

Il est donc possible de déduire t' (temps de transfert prenant en compte la variation du
débit) à partir du temps de transfert t au débit Q et de la variation du débit Q' en
fonction du temps.

Réciproquement à partir de t', Q'(t) et Q, il est possible de retrouver le temps de

transfert t au débit constant Q.
Dans le cas d'un système piston-dispersion, Il est toujours possible de calculer un débit
constant équivalent Qm tel que:
t'

Q'(t)dt/t' (6)

D'après l'équation (4) h'(t') se déduit de h(t) par la relation:

h'(t') = (Qm/Q) h(t) (7)

En considérant que la réponse impulsionnelle est une suite de créneaux, il est ainsi
possible de construire la réponse pour toute variation de débit.
Remarques: Ces hypothèses impliquent que la quantité de matière représentée par un
signal h(t) caractérisé par son temps de transfert t occupe la même surface d'écoulement
quel que soit le débit .
 Relations entre réponses impulsionnelles et conditions hydrodynamiques 395

Les problèmes liés à la discrétisation du temps

Les méthodes de transformation des réponses impulsionnelles à débits variables
proposées, respectent la relation suivante:
+ 00

[ h(t)dt = 1
-co

L'utilisation de la relation (7) basée sur le temps de transfert convectif de chaque signal
pose le problème du chevauchement des signaux lorsque le débit va augmenter (Fig. 3)
et de la disjonction des signaux lorsque le débit va décroître entre le temps d'apparition
et de disparition du nuage (Fig. 4).

Q(t)

temps temps

h'(t')
h(t)

^Z
i
temps V

Fig. 3 - Chevauchement des signaux après transformation au débit Q'(t).

temps temps

h'(t')
h(t)

Y7
" '/
0
temps
0
Ï^YTTA temp
M 12 t'1 t'2

Fig. 4 - Disjonction des signaux après transformation au débit Q'(t).

De plus dans ces conditions:

[ h'CtOdt' * 1

La méthode de calcul est donc basée sur les temps convectifs d'arrivée et de sortie de
chaque signal afin de juxtaposer les signaux constituant la réponse impulsionnelle.

En résumé, la méthode de transformation des réponses impulsionnelles proposée repose

sur le principe d'invariance du système physique en fonction de la variation du débit
entre traçages différents.

Après avoir établi les relations entre réponses impulsionnelles dans un tube, la méthode
de transformation est appliquée à des réponses impulsionnelles issues de traçages
effectués sur colonnes de laboratoire.
396 M. Dzïkowski et al.

APPLICATION SUR COLONNE EXPERIMENTALE

Dispositif expérimental

Afin de tester les transformations à débit variable, un dispositif expérimental a été mis
au point permettant de contrôler la variation du débit au cours des expériences de
traçage. L'injection d'uranine est effectuée en "volume" par l'intermédiaire d'une
seringue dans un tube souple à l'entrée du dispositif. Le transfert s'effectue à l'intérieur
de deux tubes de 3 cm de diamètre pour une longueur respective de 210 et 187 cm. Ces
tubes sont disposés en série et réunis par un manchon d'une longueur de 4 cm et d'un
diamètre de 0.9 cm.

Les échantillons récupérés à la sortie sont analysés au fluorimètre Turner. Les réponses
impulsionnelles de trois traçages ont été calculées, deux à débit constant (Fig. 5) et un,
à débit variable (Fig. 6). Les caractéristiques de ces traçages sont synthétisés dans le
Tableau 1.

12 14 16 18 20 22 24 26 28 30 32 34 36 38 40

Fig. 5 - Réponses impulsionnelles en fonction des débits constants 640 et

148 ml/min.

h(t) 10e3 m n - 1 Q(t) en ml/mn

12 13 14 15 1 (

Fig. 6 - Réponse impulsionnelle h(t) obtenue pour une variation du débit

Q en fonction du temps.

Tableau 1 Caractéristiques des traçages en colonne.

No. Débit Temps Temps Temps Mass Concentration
traçage (ml/min) d'arrivée modal moyen restituée maximale
(min) (min) (min) (g) (g/1)
1 148 13.67 17.8 21.57 4452 3494
2 640 2.9 4 4.82 5067 4757
3 variable 2.33 3.76 4.98 5659 4334
 Relations entre réponses impulsionnelles et conditions hydrodynamiques 397

Application de la méthode de transformation

La réponse impulsionnelle obtenue pour un débit de 148 ml/min a été transformée pour
un débit de 640ml/min et comparée à la solution expérimentale (Fig. 7). Les différences
entre les valeurs calculées et réelles sont représentés dans le Tableau 2.

600 - h(t) 10e3 mn- 1

500 -

400 -
* o
*
300 " t- h(t) (0=640 ml/mn]
t • h(t) calculé
200 - o

100 -
+ ' * • • . .
n , iemps en mn
0 -— — . * - • — I > — i — ' — i — ' — i — ' — i ' i " — r - * - * — i — • — i

Fig. 7 - Comparaison de la réponse impulsionnelle avec la solution réelle.

Tableau 2. Différence entre réponse impulsionnelle expérimentale (Q = 640 ml/min) et

transformée.
h(t) pïc
% de masse moy h(W1 % de masse
(min) (min'1) restituée au (min) (min" ) restituée au
temps moyen
Q = 640 ml/min 4.05 0.501 38 4.79 0.309 66
Calculée avec Q 4.05 0.476 35 4.80 0.312 62
= 148 ml/min

Cette même réponse a ensuite été transformée de façon à prendre en compte la variation
du débit au cours du traçage n° 3. Les résultats obtenus sont comparés à la solution réelle
(Fig. 8 et Tableau 3).

500 h(t) 10e3 mn-1

400 -

300 -
• h(t) (débit variable)
200 -
+ h(t) calculé
100 -
•-••••••. temps en mn
, 1 , • ' . I I J WW.TT T » ».
0 -)0"*«-»0T»«-!

Fig. 8 - Comparaison de la réponse transformée avec la solution réelle.

Tableau 3. Différence entre réponse impuisionneiie expérimentale (débit variable) et

transformée.
h(t) *pic h(t p i ) % de masse tmoy h(tmoy) % de masse
1
(min) (min ') restituée au (min) (min' ) restituée au
tpic temps moyen
Débit variable 3.7 0.382 33.8 4.98 0.166 71
Calculée avec Q 3.7 0.394 32.6 4.76 0.197 66
= 640 ml/min
398 M. Dzikowski et al.

CONCLUSION
Les deux méthodes proposées pour transformer la réponse impulsionnelle d'un système
aquifère à un traçage quand le débit varie, permettant ainsi de reconstituer ce qu'aurait
été cette réponse à un débit constant, ou à un débit différent, semble donner de bons
résultats. Elles découlent de l'hypothèse de linéarité et d'invariance du système étudié,et
de la prédominance du transport par advection sur le transport par diffusion. Elles
permettent d'interpréter des essais sur colonne réalisées en laboratoire.

APPLICATION SUR UN SYSTEME KARSTIQUE

Les campagnes de traçages
En 1987, 1988, afin de vérifier l'existence d'un pouvoir épurateur du karst vis à vis des
eaux qui le traversent, une étude menée par le BRGM (SRG Midi-Pyrenées, Département
Eau, Environnement, Energie), le CENG de Grenoble (section d'application des
traceurs), le SRAE Midi-Pyrenées, la DDAF du Lot et le groupe spéléologique de Gramat
sur le réseau Karstique de l'Ouyse a fait l'objet de campagnes d'expérimentations sur le
terrain.

La réalisation de traçages à débit constant a déterminé le choix des périodes

d'expérimentation, la deuxième motivation était d'effectuer ces traçages pour des régimes
hydrodynamiques bien différences. Deux campagnes de mesures se sont donc déroulées,
la première en période d'étiage (Mai 1987) et la deuxième, en période de hautes eaux
(Février 1988).

Le choix du système karstique de l'Ouysse est du à la "bonne" connaissance des

écoulements de cette région. Ces séries d'expérimentations avaient pour but de préciser
les conditions régissant les écoulements souterrains qui affectent ce bassin
hydrogéologique.

Dans le cadre de cette étude, nous avons appliqué le modèle sur les données de traçages
effectués les systèmes de l'Ouysse amont et du Frances (Fig. 9). Les injections ont été
effectuées le plus brièvement possible dans deux pertes se situant sur le Frances (perte
de Théminettes) et sur l'Ouysse amont (perte de Thémines), la sortie se situe au niveau
du gouffre de Besaces (Fig. 10).

La campagne en période d'étiage (Mai 1987): 4000 g de Rhodamine Wt liquide ont été
injectés dans la perte terminale de Thémines le 13 Mai 1987 à llh30.

6000 g d'Amino G prédilués dans un bidon de 30 1 ont été injectés à 10 m de la perte de

Théminettes le 13 Mai 1987 à llh.
La campagne en période de hautes eaux (Février 1988): 20 000 g d'Amino G prédilués
ont été injectés au droit de la perte terminale de Thémines le 17 Février 1988.
Le 17 Février 1988, 10 000 g de Jaune Sulfacide dilués sont injectés à 10 m de la perte
terminale de Théminettes.

Résultats
Les résultats de ces séries de traçages sont présentés sous formes d'évolution des
concentrations en fonction du temps à la sortie du système à partir du temps d'injection.
Celles-ci sont présentées en relation avec la somme des débits moyens moyens journaliers
Relations entre réponses impulsionnelles et conditions hydrodynamiques 399

SYSTEME KARSTIQUE DE
L'OUYSSE
10 km
fCombe Nègre] IMeyraRuet.

g Mamo-calcaires du
Kimméridgien Calcaires du Dogger et du Malm
-_-_-_-_i Mamo-calcaires du
Lias moyen et supérieur Lias inférieur carbonate
Trias et Paléozoïque
Limite occidentale de la zone d'infiltration du système karstique
+ + Limite souterraine approximative du système
Limite précise du système ( zones de ruissellement et d'infiltration)
-> Traçage ® Emergence O Perte T Gouffre
Fig. 9 - Le système karstique de l'Ouysse (d'après Beaudoing et al.,
1989).

(Station de jaugeage GOAF) (Station de jaugeage DDAp)

THEMINETTES
(perte)

4= m

Fig. 10 - Schéma du système Thémines-Théminettes-Besaces.

400 M. Dzikowski et al.

de l'Ouysse amont et du Frances pour les deux périodes d'expérimentation (Figs l i a

14). Les caractéristiques de ces différents traçages sont synthétisés dans le Tableau 4.

90 110 130 150 170 1 90

Fig. 11 - Traçage Thémines-Besaces (Mai 1987).

60 - Cen 10e-9 kg/1 Q en l/s - 100<

50 - - 900
-800
40 -
-700
30 -
-600

s -500
10 -
-400
ten h
-l
i i i ' i • i -300
90 140 190 • 240 290 340

Fig. 12 - Traçage Théminettes-Besaces (Mai 1987).

20 25 30 35 40 45 50 55

Fig. 13 - Traçage Thémines-Besaces (Février 1988).

100-1 C en 10e-9 kg/1 Q en l/s

25 35 45 55 65 75 85 95 105 115 125 135 145

Fig. 14 - Traçage Théminettes-Besaces (Février 1988).

 Relations entre réponses impulsionnelles et conditions hydrodynamiques 401

Tableau 4. Caractéristique des traçages du système Thémines-Théminettes-Besaces.

Traçage Dates Temps moyen Temps moyen Taux de
arithmétique géométrique restitution
Thémines Mai-87 130,30 h 128,33 h 71.0%
Besaces Fév-88 30,95 h 29,80 h 73.0%
Théminettes Mai-7 178,60 h 169,82 h 90.7%
Besaces Fév-88 53,17 h 48,55 h 73.0%

Détermination des réponses impulsionnelles

Les réponses impulsionnelles de ces différents traçages sont établies à partir des masses
de traceur injecté. Ces réponses tiennent compte par conséquent du taux de restitution de
chaque expérience. Les débits utilisés pour calculer les flux de traceur à la sortie du
système sont représentés par la somme des débits moyens calculés au entrées (Thémines
et Théminettes). Les réponses impulsionnelles ainsi définies sont représentées sous
formes de courbes où est associé la moyenne des débits moyens journaliers en fonction
du temps, à partir du temps d'injection (Figs 15 à 18).

2300 -i moyenne des h(t) 10e3h-1 - 120

débits en l/s A - 100
2200 -
-80
2100 -
-60
2000 -
-40
1900 -
"~""~ temps en h
1600- i 1 i ' = — i 1 1 1 1 -0
30 60
Fig. 15 - Réponse impulsionnelle et moyenne des débits pour le système
Thémines-Besaces (Février 1988).

225 250

Fig. 16 - Réponse impulsionnelle et moyenne des débits pour le système

Thémines-Besaces (Mai 1987).

Fig. 17 - Réponse impulsionnelle et moyenne des débits pour le système

Théminettes-Besaces (Février 1988).
402 M. Dzïkowski et al.

Fig. 18 - Réponse impulsionnelle et moyenne des débits pour le système

Théminettes-Besaces (Mai 1987).

APPLICATION DE LA METHODE DE TRANSFORMATION DES REPONSES

IMPULSIONNELLES

Les conditions d'applications

La variation des débits au cours de la campagne de traçages en période de basses eaux

est prise en compte afin d'établir par transformation (débit variable - débit constant) les
réponses impulsionnelles correspondant aux conditions hydrodynamiques des expériences
réalisées au cours de la période de hautes eaux. Durant cette période la variation des
débits moyens journaliers à partir du temps d'injection est négligeable vis à vis des temps
de transfert des traceurs dans les différents systèmes. Les réponses impulsionnelles en
période de basses eaux (Mai 1987) sont donc transformées pour des dédits moyens
constants correspondant aux temps de transfert des traceurs durant la période de hautes
eaux (Février 1988).

Fig. 19 - Comparaison de la réponse impulsionnelle de Mai 1987

transformée pour un débit de 2180 1/s avec la réponse de Février 1988
(Thémines-Besaces).

h(t)10e3 h-1

- h(t) Février 1988

• h(l) transformée

temps en h

Fig. 20 - comparaison de la réponse impulsionnelle de Mai 1987

transformée pour un débit de 1950 1/s avec la réponse d Février 1988
(Théminettes-Besaces).
 Relations entre réponses impulsionnelles et conditions hydrodynamiques 403

Comparaison des résultats

Les réponses impulsionnelles obtenues en basses eaux pour les systèmes Thémines-
Besaces et Théminettes-Besaces sont transformées respectivement pour des débits
constants de 2180 1/s et 1950 1/s. Ces réponses sont comparées au résultats obtenues en
périodes de hautes eaux (Figs 19 et 20).

CONCLUSIONS
Les temps de transfert des traçages semblent être liés aux conditions hydrodynamiques
à l'entrée du système. En effet les débits de 2180 et 1950 1/s correspondent aux valeurs
moyennes des débits des traçages effectués en février 1988. La bimodalité des courbes
de la période de basses eaux n'apparait pas en hautes eaux, toutefois nous remarquons
que les réponses impulsionnelles obtenues par transformation s'inscrivent en presque
totalité entre les temps d'arrivée et de fin de passage des traceurs définies pour les
réponses de février. Les maxima des réponses impulsionnelles tenant compte des divers
taux de restitutions sont assez proches. Il est donc possible d'admettre que le système
"physique" (Thémines-Théminettes-Besaces) défini par l'espace occupé par la masse de
traceur au cours de son transfert,varie peu entre les périodes de basses et de hautes eaux.

REFERENCES
Beaudoing, G., Astruc, J. G., Baradat, J. M., Bouziges, M., Charentus, T., Coustou,
J. C , Getto, D., Mouyou, P., Ricard, J., Sauty, J. P., Tarrisse, A. & Vicente, A.,
1989, Traçages et protection des captages dans le karst: détermination des paramètres
de transfert et prévision de la propagation des pollutions dans le réseau karstique de
l'Ouysse Causse de Gramat (Lot, France). Hydrogéologie, n° 4, 279-292.
Duffy, C. J. & Gelhar, L. W., 1985, A frequency domain approch to water quality
modeling in groundwater: Theory. Wat. Resour. Res., 21 (8), 1175-1184.
Dzikowski, M., Carlier, E., Crampon, N. & Marsily, G., de, 1991, Relations entre
réponses impulsionnelles et conditions hydrodynamiques des systèmes dans le cadre
de traçages artificiels: Théorie et application sur colonne de laboratoire.(à paraître)
/. Hydrol., 125, 129-148.
Lepiller, M. & Mondain, P. H., 1986, Les traçages artificiels en hydrologie karstique:
Mise en oeuvre et interprétation. Hydrogéologie, n° 1, 33-52.
Mangin, A., 1975, Contribution à l'étude hydrodynamique des aquifères karstiques.
Thèse de doctorat en sciences naturelles de l'université de Dijon.
Margarita, R., Guizerix, J., Corompt, P., Gaillard, B., Calmels, P., Mangin, A. &
Bakalowicz, M., 1984, Réflexions sur la théorie des traceurs: Applications en
hydrologie isotopique. IAEA, Vienne, 270/84, 653-678.
Molinari, J., 1976, Interactions avec le milieu et développements récents dans l'emploi
des traceurs artificiels. La Houille Blanche, n° 3/4, 197-276.
Shirley, J. & Dreiss, 1989, Regional scale transport in a karst aquifer 2. Linear systems
and time moment analysis. Wat. Resour. Res., 25 (1), 126-134.
Smart, C. C , 1988, Quantitative tracing of the Maligne karst system, Alberta, Canada.
J. Hydrol., 98, 185-204.
Stanton, W. I. & Smart, P. L., 1981, Repeated dye traces of underground streams in the
Mendip Hills: Somerset. Proc. Univ. Bristol Spelaeol. Soc., 16 (1), 47-58.
Villermaux, J. & Van Swaaij, W. P. M., 1969, Modèle représentatif de la distribution
des temps de séjour dans un réacteur semi-infini à dispersion axiale avec zones
stagnantes: Application à l'écoulement ruisselant dans les colonnes d'anneaux
Raschig. Chem Engng Sci. GB, 24, 1097.
Walliser, B., 1977, Systèmes et Modèles: Introduction Critique à l'Analyse des Systèmes.
Ed.du Seuil.
404 M. Dzikowski et al.

Zuber, A., 1986, On the interpretation of tracer data in variable flow systems. /
Hydrol, 86, 45-57.
Bydrogeological Processes in Karst Terranes (Proceedings of the Antalya Symposium and Field Seminar, October 1990). . „
IAHS Publ. no. 207, 1993. 3 3V 405

NEW RESULTS ON CRETACEOUS LIMESTONE KARST IN WESTPHALIA,

GERMANY

ECKEHARD P. LOEHNERT
University of Munster, Department of Geology and Palaeontology, D-4400 Miinster, Germany

ABSTRACT

The open zone of the deep karst system which is developed in Upper Cretaceous limestones
and marls in eastern Westphalia, was subjected to tracer tests in combination with
recharge/discharge and hydrochemical measurements. The results are interpreted with
respect to the separation of short-term and long-term discharge components of springs.

INTRODUCTION
There are two main karst terranes in the Federal State of Northrhine-Westphalia, namely
the Devonian massive reef limestones in the Rhenish Schiefergebirge and the Upper
Cretaceous limestones and marls. This paper deals with the latter which outcrop at the
southeastern margin of the Miinster basin in what is called the " High Plateau of Paderborn"
(Fig. 1). This area is famous for its early study by Hans Stille, the renowned German
structural geologist, nearly a century ago. In 1970 the Paderborn area was studied from the
karst-hydrological point of view by Baskan. Stille (1903), Bolsenkôtter (1967) and Baskan
(1970) did a number of tracing tests mainly applying dyes. They obtained flow velocities
in the range of 100-450 m h' 1 . Similar tests were combined by our group with
hydrochemical and recharge/discharge measurements in July of the years 1982, 1985 and
1989 when the karst reservoir was expected to be most depleted (Loehnert, 1988).

Results of tracer experiments and simultaneous flow volume and water quality
measurements need to be interpreted taking into consideration the varying climatic
conditions. The neglect of the latter is one of the major disadvantages of previous tests.
Our objective was to delineate eventually a conceptional model of flow conditions and
dissolved solid transport in the karst system. Recently the basin of Miinster was subjected
to flow system analysis by Michel & Struckmeier (1985). According to this study the
Paderborn plateau is part of a large regional system, comprising mainly the basal
Cretaceous aquifer unit (i.e. Cenomanian, Turonian and Lower Coniacian). At the same
time the Paderborn plateau is a marginal subsystem characterized by high flux and
recharge from both infiltrating precipitation and water-losing streams resulting in
considerable spring discharges, the major being the Pader springs in the city of
Paderborn with an average of 5.4 m3 s"1.

The author confirmed, by evaluating runoff data, the classification of the catchment
region of the Pader springs into the group of open deep karst (Loehnert, 1990). Recharge
rates were calculated between <410 mm year"1 and 571 mm year"1, based on the
imprecisely known size of the catchment area. These rates are roughly double the ones
over the total Miinster basin which is mainly covered by Quaternary deposits.

RESULTS
Tracer tests
We selected equal injection points as used by Stille and subsequent researchers. These
localities are seen in Fig. 1, and Fig. 2 presents a hydrogeological cross section along
major flow paths already identified by Stille (1903). All injection points are located in
406 Eckehard P. Loehnert

——«C Perennial Streams

-—*Ct Seasonal Streams

H S ^ Underground Flow Systems
4^0$^ Recharge Area of Pader
Springs
O Ponora »# Injections
P" Points
Upper Cretaceous
KSS&SI Lower Cretaceous 4 km
and older
Quaternary to the NW

Fig. 1 - Map of the study area (modified from Stille, 1903).

UPPER CRETACEOUS ' tOWÊR CRETACEOU5 CARBONIFEROUS MUSCHILKAIK

WATER TABLE UNDERGROUND HATER FLOW

2 LATE S U M M £ * * - - \ _ _ . LAIE SUMMER

JiFlR AOUITAM)

Fig. 2 - Hydrogeological cross section (modified from Koch & Michel,

1972).
 New results on Cretaceous limestone karst in Westphalia, Germany 407

the northeastern half of the catchment area where streams tend to lose their waters in
ponors which eventually dry up completely.
The period during which there is flow in these streams varies due to the rainfall in the
Egge mountain. Volumes of stream discharges are compared with discharge rates of those
Pader springs where tracer transit was proved. Tracers employed were dyes (uranine),
coloured lycopodium spores and, only in July 1989, fluorescent polystyrene
microspheres. The latter technique of applying drifting particles with a diameter between
0.5 and 90 pm was introduced by W. Kàss and H. Hôtzl (poster at the IAH 21st Congress
of I AH, Guilin, China, October 1988). Particles could be identified in spring waters but
no transit curves could be drawn. It is thought that contamination of samples occurred
while measuring uranine fluorescence. Spores yielded, in 1982 and 1985, good qualitative
results but are omitted in the following. Uranine transit curves were best in 1989 due to
dense sampling.
Selected examples are shown in Fig. 3 and calculated flow velocities are listed in Table
1. Tracers reappeared in six of a total of 10 sampled springs, all except one located in
the eastern Pader spring region. Consequently, this paper will focus on this region
comprising the groups of the Maspernpader (including Haxthauser Hof), Dielenpader,
Spiilpader and Rothebornpader. The latter three are made up of pools containing water
from numerous separate springs.

Hydrochemical data

Some selected parameters of traced streamwaters in the recharge area and spring waters
were measured on the spot, among others: electrical conductivity, temperature, pH,
hardness and chloride. In 1989, after completion of the tracing experiments, some
graduate students (M. Linke and others) continued to follow up hydrochemical and
discharge variations. Some preliminary results are included in this contribution.

10 6-

MASPERNPADER-*.

DIELENPADER

103-

10 2-
ROTHEBORNPADER

10a.m.
' "A" ' * I ' " ' "T t i i i j i i i i i i i i I i i i i i i i i i i I
Tl 12 13 14 15 16 17 19 20 JULY 1989
Injection : 2 k g U r a n i n e , E l l e r b a c h E of Dahl
3kg Uranine, Sauor N of Lichtenau

Fig. 3 - Five selected uranine transit curves of eastern Pader springs.

Resulting travel times see Table 1.
408 Eckehard P. Loehnert

Table 1. Field velocities (m/h).

July 1982 July 1985 July 1989
Maspernpader val = 387 val = 145
va2 = 178 va2 = 331 va2 = 100-121
Haxthauserhof val = 195 val = 371 val = 139
va2 = 172 va2 = 303 va2 = 97-115
Dielenpader val = 404 val = 132
va2 = 164-168 va2 = 342 va2 = 87
Spiilpader val = 407 val = 114
va2 = 356 va2 = 86
Rothebornpader val = 365 val = 131
va2 = 305 va2 = 88
Eastern region (val = 195) val = 365-407 val = 114-145
va2 = 164-178 va2 = 303-356 va2 = 86-121
val = maximum flow velocity (first tracer arrival)
va2 = dominant flow velocity (peak transit of tracer)

Table 2. Discharge of Pader springs (1/s).

17-31 July 1982 3-5 July 1985 10-14 July 1989
Maspernpader 279 386-404 273-360
Haxthauserhof 37 62-66 21-27
Dielen- and 924-1115 1389-1611 742-1211
Spiilpader
Rothebornpader 460-539 660-676 385-426
Eastern region 1742-1824 2530-2722 1423-2024
Western region 1633-1700 1915-2048 1447-1687
Total 3443-3457 4578-4637 2870-3611

In Table 2 figures of differing springs discharges are listed. Figures 4 and 5 present
variation of total hardness (Ca 2+ + Mg 2+ ) split up into carbonate/permanent and
calcium/magnesium hardness respectively. Simultaneously sampled losing streams are
also shown in the diagrams. In July 1982 and 1989 streamwaters occupied a distinct
position while in July 1985 streams approach or overlap spring waters.

Furthermore, an increase in hardness of spring waters is established according to the

following succession: Maspernpader/Haxthauserhof, Dielenpader, Spiilpader and
Rothebornpader. Calcium is limited to max. 7 meq l"1 and magnesium to max. 2.2
meq l*1, both alkaline-earth metals reaching highest values in the Rothebornpader.
Further interpretation is prohibited due to random sampling in different years. However,
hardness variation of the Rothebornpader is somehow peculiar in July 1989 when waters
tend to be free of non-carbonate hardness.

In Fig. 6 total hardness is plotted against conductivity for two time periods with
distinctly differing discharges as measured by M. Linke: July-September 1989 and
December 1989-February 1990. Discharge increased by a factor of 1.7-2.0, and the
objective was to identify hydrochemical changes depending on discharge differences.
 New results on Cretaceous limestone karst in Westphalia, Germany 409
"*> moq/l 10 meq A

JULY 1985

5 —•Ga2**Mg2* lOmaq/l

Maapernpader, Haxthauserhof
Dtelen - a Spfllpacter
Rothebornpader
Eilerbach
Sauer / Schmittwasser.Grundsteinhe.'rn
Sauer, Lichtenau

-"•Caî-tMos* 10meq/l
2+
Fig. 4 - Total hardness (Ca + Mg ) vs. carbonate hardness (HCO )
and permanent hardness respectively in a trilinear diagram.

Along with discharge duplication, the water temperature dropped by approximately 2°C
as already found by Stille. The average regression line in Fig. 6 obeys the equation:
Cond. [25°C] = 78 meq l_1(Ca + Mg) + 130. Rothebornpader spring plots are
commonly above the regression line while those of the others are below it.

DISCUSSION AND CONCLUSIONS

Spring discharges in July for the three years varied considerably, with total figures in
1985 close to the average and those in 1989 in the minimum range (Table 2). According
to Stille Pader spring discharges are dependent on rain amounts received 3-4 months
earlier. It can be concluded that rains of the preceding winter months were excessive
(1982 and 1989) or deficient (1985), and recharge to groundwater is assumed to vary
accordingly. It is obvious that the amount of winter precipitation contradicts spring
discharge of subsequent summer (July). Consequently, the short-term effect of days
resulting from losing streams is of major importance (Table 1).
The lost volume amounts to 300-450 1 s"1 in the Ellerbach and Sauer/Schmittwasser
(without including the Sauer sinkhole near Lichtenau). While streamwaters constitute an
overall of 16-18% of the eastern Pader springs, the portion may exceed one quarter when
spring discharge happens to be lowest, i.e. in July 1989.
410 Eckehard P. Loehnert
10m©q/! 10meq/l

oV \ JULY tSW

Maspernpader, Haxthauser hof

Dielen- & Spùlpader
Rothebornpader
Ei 1erbach
Sauer/Schmittwasser, Grundsteinh.
Sauer. Lichtenau

/ " \
5 —•Ca**»Mg*- 10meq/l
2-1 2+
Fig. 5 - Total hardness (Ca + Mg ) vs. calcium and magnesium
respectively in a trilinear diagram.

Spring water temperatures were also closely investigated and found in the July range
reported by Stille. Absolute readings increase from Maspernpader/Haxthauserhof (9.8-
10.5°C)- Dielen/Spiilpader (9.8-11.8°C) to Rothebornpader (10.4-12.0°C). In January/
February 1990, with the duplication of flow rate, the temperature dropped by 2°C on
average. This is the annual fluctuation of springs classified by Stille (1903) as either
always clear or turbid for a time but with considerable temperature fluctuations of up to
4°C.

In the eastern spring region only one exception to this group was identified, namely the
"Pfalzquelle" discharging into the Rothebornpader.

The temperature fluctuation is only 0.9°C and its chemical composition is unique
reaching highest figures of hardness (with one third magnesium) and sulphate but
relatively low chloride and nitrate contents (M. Linke, unpublished). We assume this
spring water to represent baseflow (or "karst groundwater" in the sense of some authors).

On the other hand, only one of the many Rotheborn spring sources showed visible dye
during our tracer test in 1982, and this source water had lowest hardness and
temperature. Hence long-term discharge ("karst groundwater") and short-term discharge
"karst water") occur in differing portions in Pader spring flows and may be separated by
using hydrochemical parameters. So far eastern Pader springs and the Rothebornpader
in particular received highest portions of long-term storage in July 1982 and possibly
lowest in July 1985 when losing streamwater composition after heavy rain overlapped
 New results on Cretaceous limestone karst in Westphalia, Germany 411

"i ' 1 • 1 « 1 -;—p—,.

4 5 6 7mBql 'ca + Mg
Fig. 6 - Plot of electrical conductivity (25°C) vs. total hardness (Ca 2+ +
Mg 2+ ) for two periods.

those of springs (Figs 4 and 5). A quantification of the complex hydraulic mixing process
will be subject to future research.

In this context reference is made to the publication by Geyh & Michel (1987) who
suggested, based on environmental isotopes, a mixing of three components of a thousand
years to months residence time, of which two are supposed to occur in the springs. The
shortest component, proved by tracer tests, has however not been taken into
consideration. The isotopic identification of this component would require dense sampling
of precipitation and streams in the recharge area as well as spring waters.

REFERENCES
Ba.skan, M. E., 1970, Hydrogeologische Verhâltnisse am Sûdostrand des Munsterischen
Kreidebeckens und im Eggegebirge unter besonderer Beriicksichtigung der
Karsthydrologie. Fortschr. Geol. Rheinl. u. West/., 17, 537-576, Krefeld.
Bolsenkôtter, H., 1967, Fàrbe- und Impfversuche im Einzugsgebiet der Paderquellen.
Woher kommt das Wasser der Paderquellen. Dechenia, 118, 212-215, Bonn.
Geyh, M. A. & Michel, G., 1987, Grundwasserbewirtschaftung und Isotope. Z. Dt.
Geol. Ges., 138, 261-271, Hannover.
Koch, M. & Michel, G., 1972, Hydrogeol. Karte Kreis Paderborn 1:50 000. Erl.,
Krefeld.
Loehnert, E. P., 1988, Routine tracing experiments in the Upper Cretaceous Karst
system of Westphalia. In: Karst Hydrogeology and Karst Environment Protection
(Proc. 21st Congress of IAH, Guilin, China, October 1988), vol. 2, p. 937. IAHS
Publ. no. 176.
Loehnert, E. P., 1990, Beitrag zur Geohydrologie des Karst-Aquifersystems von
412 Eckehard P. Loehnert

Paderborn (Nordrhein-Westfalen) (in German, English abstract) N. Jb. Geol.

PalHont. Abh., 181, 519-530, Stuttgart.
Michel, G. & Struckmeier, W., 1985, The Cretaceous basin of Munster - a general
groundwater system in response to multiple impacts (water supply, spas, deep
mining). In: Hydrogeology in the Service of Man (Proc. 18th IAH Congress,
Cambridge, September 1985), vol. 2, 150-159. IAHS Publ. no. 154.
Stille, H., 1903, Geologisch-hydrologische Verhâltnisse im Ursprungsgebiet der
Paderquellenzu Paderborn. Abh. Kgl. preuss. Geol. L.A. u. Bergak., N. F., H.28,
Berlin (Reprint: Geol. Jb. ReiheC, H. 14, Hannover 1976).