Professional Documents
Culture Documents
TROPICAL CLIMATOLOGY,
METEOROLOGY AND HYDROLOGY
tJ
RMI KAOW·ARSOM
2004
Second International Conference on
TROPICAL CLIMATOLOGY,
METEOROLOGY AND HYDROLOGY
tJ
RMI KAOW·ARSOM
2004
ROYAL METEOROLOGICAL INSTITUTE RoYAL AcADEMY
OF BELGruM OF ÜVERSEAS SCIENCES
av. Circulaire / Ringlaan 3 rue Defacqzstraat 1/3
B- 1 1 80 BRUSSELS (Belgium) B-1000 BRUSSELS (Belgium)
Tel. : 02.373.05.08 Tel. : 02.538.02. 1 1
Fax : 02.373.05.28 Fax : 02.539.23.53
E-mail : rmi_info@oma.be E-mail : kaowarsom@skynet.be
Web : www.meteo.be/IRM-KMI/ Web : http//www.kaowarsom.be
ISBN 90-75652-36-4
D/2004/0 149/5
CONTENTS
Tropical Meteorology
,
F. LuPo M. CoRLAZZOLI & E. F. LAMBIN, Monitoring Natural Disasters and
"Hot Spots" of Land-cover Change with SPOT4 Vegetation to
Assess Region at Risk........................ . . """"" ".." 1 23
E. C. KlPKORIR, S. M. GACHUIRI, J. MuKABANA & D. RAEs, Evaluation of the
Onset of the Growing Season for Various Climatic Zones in Kenya
by means of a Soil Water Balance Method for Different Soil Types. 1 37
c. B . S. TEH, L. P. SIMMONDS & T. R. WHEELER, Modelling the Partitioning
of Solar Radiation Capture and Evapotranspiration in Intercropping
Systems ............................... ...... ........... ...... ........ 151
Historica) Climatology
Tropical Hydrology
Tropical Climatology
Foreword
December 2001 at the Royal Museums for Art and History, Pare du
The Academie Opening Ceremony took place on Wednesday 1 2th
Opening Speech
by
Jean-Michel SwALENS *
Excellency,
Permanent Secretary,
Chairman,
Ladies and Gentlemen,
Chairman,
Ladies and Gentlemen,
by
Rafique AHMED *
in a. major monsoon region of the world. Moreover, the results will be useful for
agricultural and regional planning purposes, and for the management of floods
and water resources of the country.
1. Introduction
Bangladesh is located at the entry and exit point of the Bay of Bengal
Branch of the South Asian Monsoon circulation (SPATE & LEARMONTH
1 984, AHMEn 1 998), thereby making it one of the most important mon
soon regions of the world. The summer monsoon season in Bangladesh
prevails from early June to mid-October, when it is dominated by south
westerly winds from the Bay of Bengal (AHMED & KARMAKAR 1 993).
The summer monsoon season in Bangladesh has a profound effect on
the overall economy of the country - especially on its agriculture,
hydrology and drainage system (JoHNSON 1 982). Several floods occur
every year, causing extensive damage to crops, livestock, life and proper
ties (JOHNSON 1 982, AHMED 1 998). Sometimes floods may reach calarnity
proportions, such as those of 1 987 and 1 988 (B RAMMER 1 990), requiring
large-scale international relief operations.
Using 35 years' ( 1 964 through 1 998) data, geographic distribution of
the duration of the summer monsoon and its variability, summer monsoon
rainfall amount and its variability, average number of raio days and its
variability in Bangladesh are presented and discussed in this paper. In
addition, the time series analyses of the duration of summer monsoon,
number of rain days, and summer monsoon rainfall at three selected sta
tions are presented and discussed in this paper.
3. Methodology
88' �·
26'
26
24'
'
24
22'
22
& 8'
Fig. 1 . - Physical setting of Bangladesh and the location of 19 stations used in this
study. Shaded areas are in excess of 250 m in elevation. Some areas are in excess of
1 ,000 m in elevation (Meghalaya Plateau and Assam Hills in India).
4. Data
Daily rainfall (mm), and daily wind direction (azirnuth), for each day
of May 1 through October 3 1 , for each year of the 35-year period from
1 964 through 1 998 were collected for 1 9 stations (fig. 1 ) from the
Bangladesh Meteorological Department, Dhaka.
(a) (b)
Fig. 2.- (a) Mean arrival dates, and (b) withdrawal dates of the summer monsoon in
Bangladesh (after AJ.!MEo & l<ARMAKAR 1 993).
of the country (fig. 2a). The summer monsoon begins to withdraw in the
northwestern part of the country, the mean date of which is September 30,
and the withdrawal from the country through its southeastern corner is
completed by October 17 (fig. 2b).
Geographic distribution of the average duration of the summer mon
soon in Bangladesh, and its standard deviation are shown in figure 3 .
Average duration of the summer monsoon varies from 1 1 0 days in the
west-central part of the country to over 1 35 days in the southeastern part
(fig. 3a), and its standard deviation ranges from 1 0 days in the northwest
to 1 4 days in the west-central region and in the southeastern corner of the
country (fig. 3b).
Average number of rain days and the standard deviation of rain days
during the summer monsoon in Bangladesh are shown in figure 4.
Average number of rain days during this season varies from 60 days in
the extreme west-central region to 1 00 days in the northeast and south
east (fig. 4a). Standard deviation of the number of rain days varies from
1 0 days in the extreme west-central part to 1 2 days in the northeast and
southeast (fig. 4b).
135 12
14
(a) (b)
Fig. 3.-(a) Average duration (days), and (b) standard deviation of the duration of
summer monsoon season (days) in Bangladesh.
(a) ( b)
Fig. 4.
- (a) Average number of rain days during the summer monsoon in Bangladesh
and (b) its standard deviation (days).
SUMMER MONSOON RAINFALL 21
(a) (b)
Fig. 5. - (a) Average rainfall (mm), and (b) the coeffi.cient of variation of summer
monsoon rainfall (%) in Bangladesh.
Tlmt8orlnol111tDurallonofS-MoMoan
Sllllon: �L Perlod: 1ff4.19N
m11,1 111111111
� 1 1 � s 5 i i 1 1 ! 1 1 1 1 ! 1 1
-
Fig. 6. - Time series of the duration of summer monsoon at three selected stations in
Bangladesh ( 1 964-1998).
Fig. 7. - Time series of the number of rain days during the summer monsoon season
at three selected stations in Bangladesh (1964- 1998).
Tlmo-otS..mmorManooon Ralnbll
Stlllon: fbl•hlN. Portod: 19"-1"9.
�-.,,---Q
s1atiOn: SylloL ...olod: 1-19N.
Fig. 8. - Time series of the summer monsoon rainfall at three selected stations in
Bangladesh (1964- 1998).
REFERENCES
by
1. Introduction
t
4"S.-
CONGO
6°S.--------:-
'
s·s.--------�-------------
sons of ENSO years, preliminary data suggest that the air temperature is
short, strong wind events can occur during this season. For the dry sea
higher and the wind weaker. This seems to cause variability in catches of
several species of pelagic fishes, affecting the economy and the food
stock of the neighbouring populations (PLISNIER 1 997, PL1sNIER et al.
2000). Furthermore, over recent years, the lake hydrodynamics has been
seen to exhibit variability related to climate change (PLISNIER 1 997,
2000).
Understanding the lake hydrodynamics and its variability is important
for the management of its resources, as well as understanding limnologic
al conditions in the framework of paleoclimatic studies such as the ongo
ing CLIMLAKE project (DESCY et al. 2002). In this respect, numerical
modelling is an invaluable tool. Our objective is to build a three-dimen-
THE WIND-INDUCED OSCILLATIONS OF THE THERMOCLINE 29
sional model of the lake hydrodynamics and ecology. The first stage of
this undertaking is concerned with the study of one of the most striking
features of the lake hydrodynamics, i.e. the tilting of thermocline occur
1 999, PLISNIER & CoENEN 200 1 ) . These phenomena, which are believed to
be induced by the wind forcing, are studied herein by means of a reduced
gravity model.
In the next section, the equations and the surface forcing of the
reduced-gravity model are established. Then, this model is simplified to
a one-dimensional, linear model, of which an analytica! solution is
derived and analysed. Finally, numerical results of the complete reduced
gravity model are obtained, and compared cursorily with field data and
one-dimensional, analytica! solutions.
The thermocline is present all year round over a large fraction of the
lake. This is the main reason why a two-layer model is believed to be a
relevant tool for representing the motions of the thermocline in an ideal
ized manner. In such a model, the prognostic variables related to each
"2" are associated with the top and the bottom layer, respectively, h;, u;
layer are assumed to be vertically homogeneous. If the subscripts " l " and
and V; (i = 1 ,2) denote, for the i-th layer, the unperturbed depth, the com
ponent of the horizontal velocity along the x-axis and the component of
d
: i +V • (H1u1) = 0 , (1)
d(
�t U)
i i +V • (H1U1U1) +fez x (H1U1) = - gH1VT] + D1 + _!_,
Pi
(2)
where t is time ; U; = u;e, + v;ey is the velocity vector in the i-th layer ; e,
and ey are the horizontal unit vectors associated with the x and y coor
dinate axes, while e, = e, x ey is the vertical unit vector, pointing upward ;
f is the Coriolis factor - which is relatively small and negative in the
domain of interest - and g is the gravitational acceleration ("' 9.8 m s-2).
In the governing equations above, the vector is the wind stress, which
is evaluated by means of the SMITH & BANKE ( 1 975) parameterization,
where p. ("' 1 kg m3) is the air density and v. is the wind velocity,
where Ax and Ar are the horizontal eddy viscosities, which are taken to be
much smaller than its "length". If the constant p; (i = 1 ,2) represents the
different in the x- and y-directions because the "width" of the lake is
water density in the i-th layer, the relative density difference E is defined
to be
p,
.
E = P2 -
(7)
P2
The density is computed from the UNESCO ( 1 9 8 1 ) equation of state
of the water, in which the salinity is set to zero while the pressure is as
sumed to be equal to one atmosphere - as density variations are mainly
due to temperature variations in Lake Tanganyika. Given the range of the
available in situ temperature profiles, E is likely to be smaller than 1 0-3•
The displacement of the lake surface is assumed to be much smaller
than that of the pycnocline. Therefore, by virtue of continuity equations
( 1 ) and (3), the layer heights and transports satisfy dH,/ dt "' - dH2fdt "'
- V • (H,u,) "' V • (Hiu2). Thus, the order of magnitude of the top layer
transport, Hiu,, is equal to that of the bottom layer, H2U2 , implying that
the order of magnitude of the left-hand side of momentum equation (2) is
equivalent to that of the left-hand side of (4). As the thickness of the
hypolimnion ranges from about 1 00 m to over 1 ,000 m while the depth
of the epilimnion rarely exceeds 50 to 1 00 m, the ratio H, / H2 is general
ly much smaller than unity. Therefore, the contributions to the pressure
force prevailing in the hypolimnion are likely to be the only dominant
terms in (4), i.e. -gH2V(ri - E l;) "' 0. By virtue of continuity equations ( 1 )
and (3), the lake-averaged values of ri and 1; must be constants, which are
THE WIND-INDUCED OSCILLATIONS OF THE THERMOCLINE 31
thermocline
P2
Fig. 2.
Bathymetry (in
metres) of Lake Fig. 3. The main parameters and variables of a two-layer
-
Tanganyika. model of the lake, which upon assuming that h, << h, gives rise to
a reduced-gravity model.
(8)
�
neglecting TJ relative to . dropping the subscripts " 1 ", it is readily seen
Substituting (8) into the pressure force in the right-hand side of (2),
êJH
at + V • (Hu) = 0, (9)
a�Hu)
t
+ • (Hun) +Je, u = - EgHV� + + _.!._
v x D
p
' ( 1 0)
32 J. NAITHANI, E. DELEERSNIJDER, P.-D. PLISNIER & S. LEGRAND
ty(t)
: duration et dry season
wind stress:
-- 1d Tw
time: t
t=O t - Id t-T t - T + 1d
( 1 2)
( 1 3)
the lake, with the thermocline being shallower in the south and deeper in
the north, i.e.
( 1 4)
V:n =
4t 4t
En = and (17)
EghLk; hLûJnkn '
34 J. NAITHANI, E. DELEERSNIJDER, P.-D. PLISNIER & S. LEGRAND
c 1 s)
j=O
00
v(t, y) = I [vH(t -jT, y)X(t -jn- vH(t -jT - L, y)x(t -jT - L) ] . ( 19)
j=O
-1
0 0 . 25 0.5 0 . 75 1
y/L
Fig. 5. The first three modes (n = 1 ,2,3) of the thermocline oscillations, as defined
-
Table 1
Qualitative illustration of the sensitivity of the amplitude and period of thermocline
oscillations to model parameters or forcing, in accordance with the linearized,
one-dimensional model solution
Wind velocity : v. î î
Amplitude Period
Stratification : E î J, J,
-
35
amplitude : Ei 20 ± I l metres
"
" " .
i i l l ,. ,. .
20 period :
2it
26 ± 4.5 days
1 5 ��
��
0 10 20 30 40 50 60
amplitude (metres}
Fig. 6. - Values of the amplitude E, and period 2 :rt / w, of the first - and only sig
nificant - mode of thermocline oscillations obtained by varying the wind velocity v., the
relative density difference E and the unperturbed depth of the thermocline h in the inter
vals 3 s v. s 7 (m s·'), 0.5 x 10-3 s E s 10-3 and 30 s h s 70 (m), which are believed to
represent the range of the admissible values of these parameters. The damping coefficient
µ is set to 2 year·', a value which has a minor impact on the amplitude and period of the
oscillations - but seems to be appropriate in view of the numerical results discussed
below. The mean and standard deviation of amplitude E, and period 2 :rt / w, are indicated.
36 J. NAITHANI, E. DELEERSNIJDER, P.-D. PLISNIER & S. LEGRAND
L\t - max
< { ïfl '
1 AxL\y } . (20)
[2gh(L\x2 + L\y2)] 1 12
40
20
g
ë
"
� 0
� � :� $ � " :
� l ... 1
Fig. 7. - Evolution during the second year of simulation of the downward displace
ment of the thermocline at the northem (solid curve) and southem (dashed curve) ends of
the lak:e, as evaluated by means of the complete, reduced-gravity model. The unperturbed
40
20
ë
"
lil
� 0
• 1
c.. •
B -20 '
:1 � .� , i �!�
• • ' ' ,, ., •
I I I t
J 1 , 1 1 1 f
� : � : \: �: �
-40
•�' ·:� i '!
'
1 20 240 360
Tme (days)
Fig. 8. - Evolution during the second year of calculation of the downward displace
ment of the thermocline at the northem (solid curve) and southem (dashed curve) ends of
the lake, as obtained from the analytical solution ( 1 8) of the simplified, linearized model.
The model parameters listed in the caption to figure 7 are equal to those selected to obtain
the present results. In addition, the friction parameter y is set to 4 year·• .
38 J. NAITHANI, E. DELEERSNIJDER, P.-D. PLISNIER & S. LEGRAND
2a ··-·-·-·-·-·-··
.i . . . . . . . . . . .i
1
· . . . . ·
•
-·-········· -··· ;· - · · · ···········i
.
•
.
•
. .
.. ..
z._ ___._____.__
_ ....;
__. ;..;;_ .._______.
__ ___.
..____-i
' 141116
Fig. 9. - Time series of the temperature measured in 1993- 1 994 at various depths in
the vicinity of Mpulungu. The numbers 1 to 1 1 correspond to the depths 1 , 5, 30, 50, 70,
90, 1 1 0, 1 50, 200, 250 and 300 m. These data are from FAO/FINNIDA .
5. Conclusions
ACKNOWLEDGEMENTS
The present work is carried out in the scope of the project "Climate Variability
as Recorded in Lake Tanganyika" (CLIMLAKE), which is funded by the Belgian
Federal Office for Scientific, Technical and Cultural Affairs under contract
EV/1 0/2D. Eric Deleersnijder is a Research Associate with the Belgian National
Fund for Scientific Research (FNRS). Figure 9 data are from the FAO/FINNIDA
THE WIND-INDUCED OSCILLATIONS OF THE THERMOCLINE 39
project LTR. The authors are indebted to A. Dierickx, P. Tréfois and D. Delvaux
of the Royal Museum for Centra! Africa for producing a digital bathymetry of
Lake Tanganyika.
REFERENCES
DESCY, J.-P., PLISNIER, P.-D., ANDRE, L., ALLEMAN, L., CmTAMWEBWA, D.,
(Ed.), Lake Tanganyika and its Life, Oxford University Press, pp. 49-75.
CocQuYT, C., DELEERSNUDER, E., KIMIRE1, I., NAITHANI, J., PHIRI, H.,
SINYENZA, D. & VYVERMAN, W. 2002. Climate variability as recorded in
Lake Tanganyik:a (CLIMLAKE). - Bulletin of the International Decade
for the East African Lakes, pp. 7-8.
NAITHANI, J., DELEERSNUDER, E. & PLISNIER, P.-D. 2002. Analysis of wind-induced
thermocline oscillations of Lake Tanganyika - Environmental Fluid
Mechanics, 3 : 23-39.
PLISNIER, P.-D. 1 997. Climate, Limnology and Fisheries Changes of Lake
Tanganyika. - FAO/FINNIDA Research for the Management of the
Fisheries on Lake Tanganyika. GCPIRAF/271/FIN-TD/73(En), 50 pp.
PLISNIER, P.-D. 2000. Recent climate and limnology changes in Lake Tanganyika.
- Verh. Internat. Verein. Limnol., 27 : 2670-2673 .
PL1sNIER, P.-D., CmTAMWEBWA, D., MwAPE, L . , Ts�ANGU, K . , LANGENBERG, V. &
CoENEN, E. 1 999. Limnological annual cycle inferred from physical
chemical fluctuations at three stations of Lake Tanganyika. -
UNESCO 198 1 . Tenth Report of the Joint Panel on Oceanographic Tables and
Standards. - UNESCO Technica! Papers in Marine Sciences, 36.
Second International Conference
"Tropical Climatology, Meteorology and Hydrology"
Proceedings edited by
G. Demarée, M. De Dapper, J. Alexandre
pp. 41-60 (2004)
by
1. Introduction
Over the East African Great Lakes, Victoria, Tanganyika and Nyassa
Malawi (fig. 1 ), the seasonal regime is primarily controlled by the migra
tion of the Intertropical Convergence Zone (ITCZ). This migration is
associated with intense convergence activity, which usually leads to abun
dant heavy rainfall. The ITCZ lies in the southern part of the Lake
Nyassa-Malawi during the boreal winter, and is located northwards of the
Lake Victoria during the summer. Therefore, the Great Lakes ' regions
experience their rainy seasons in boreal spring and autumn for the nor
thern Lakes and rather a single boreal winter rainy season for the southern
Lake Nyassa-Malawi. The spring rains are usually strong, while the
autumn rains are more variable. Rain generally falls in heavy thunder
storms and most of it occurs on the high escarpments surrounding the
lakes' ·basins. The entire region is under the prevailing influence of eas
terly circulation bringing moisture from the Indian Ocean.
In the intertropical region, the linkages between the ocean and the
atmosphere are known to be partly responsible for the interannual prec
ipitation variability. Atmospheric zonal circulation cells, similar to the
Walker types, have a predominant influence on the rainfall variability of
intertropical coastal regions (HASTENRATH 1 99 1 ) . In the East Africa
region, the interannual precipitation variability had been related to differ�
ent parameters associated to the intertropical ocean - atmospheric cir
culation (ÜGALLO 1 988 ; BELTRANDO & CADET 1 990 ; BELTRANDO &
CAMBERLIN 1 992 ; RICHARD 1 992, 1 994 ; CAMBERLIN 1 995 ; SELESHI &
DEMAREE 1 995 ; ROPELEWSKI & lIALPERT 1 996 ; CAMBERLIN 1 997). The
analyses of the interannual variability of the precipitation rate allow to
focus on the importance of the first rainy season (October-December) or
"first rains" period on the total annual rate. The rainfall anomalies in the
Great Lakes' region are subject to a marked and extensive interannual co
variability at the beginning of the rainy season (October-December,
RrcHARD et al. 1 998).
RELATIONSHIP BETWEEN THE ZONAL CIRCULATION 43
Fig. 1 .- Localization of the East African Great Lak:es, Victoria, Tanganyika and
Nyassa-Malawi.
The variability of this first rainy season has been significantly correla
ted with specific parameters of the zonal circulation associated to the
Indian and the Pacific Oceans. Some studies (OGALLO 1 988, N1cHOLSON
1 996, ROPELEWSKI & HALPERT 1 996) identify a relation or teleconnection
between the Southem Oscillation Index (SOi), which characterizes the
atmospheric circulation over the Pacific Ocean and the amount of prec
ipitation over the Grea� Lakes ' region. An excess of the precipitation in
East Equatorial Africa (the Northem and Central zone of the studied area)
appears preferentially in low SOi values (El Nifio Southem Oscillation
(ENSO) events) . This relation appears more diffuse over the southem
extremity of the Lake Nyassa-Malawi drainage catchment, and opposite
44 L. BERGONZINI & Y. RICHARD
over South Eastern Africa, where low SOI values are associated with a
relatively low precipitation rate (ROPELEWSKI & HALPERT 1 996, RICHARD et
al. 2000).
Some authors (BELTRANDO & CADET 1 990 ; RICHARD 1 992, 1 994) show
ed that this relation is not direct and systematic. They found that the
variability of the first rainy season precipitation rate over Bast Equatorial
Africa preferentially linked to the zonal circulation over the Indian
Ocean. Rains on the Bast African lakes tend to be deficient with enhan
ced low-level westerlies in the equatorial central Indian Ocean. This rela
tionship is particularly pronounced for the boreal autumn rainy season,
and can be captured by diverse indices (HASTENRATH 200 1 ).
It has been shown that the zonal circulations over the Indian and
Pacific Oceans are often, but not systematically, synchronous. A dipole
mode in the Indian Ocean with anomalously low (high) Sea Surface
Temperature (SST) off Sumatra and High (low) SST in the western
Indian Ocean, with accompanying wind and precipitation anomalies, is
shown to be independent of the El Nifio Southern Oscillation (ENSO)
(SAJI et al. 1 999). For example, the 1 997-98 anomalies - in spite of the
coincidence with strong ENSO events - may primarily be an expression
of the internal Indian Ocean dynamics (WEBSTER et al. 1 999) ; floods
observed during the 1 997 autumn over Bast Africa were comparable to
those noticed in 1 96 1 (KAPALA et al. 1 994), a year that does not corres
pond to an ENSO event.
The repercussions of the rainfall variability of October-December on
the hydrologie regimes over the region have not yet been investigated.
The main reason for this gap is the absence of continuous, available, flow
records. Several Bast African lakes appear to be key indicators of climatic
and environmental changes in the region. From previous studies, it has
been demonstrated that lake-level fluctuations, which present numerous
synchronisms, register the pulse of rainfall variability in the equatorial
tropics (SENE & PLINSTON 1 994, GROOVE 1 998, BERGONZINI 1 998,
NrcHOLSON 1 999). Indeed, fluctuations of the lakes level, the surface or
the volume mainly reflect changes in the precipitation and evaporation
integrated over the lakes and their catchment area. Therefore, their fluc
tuation constitutes a sensitive indicator of climatic trends from local to
global scales. Sediment records of this historica! evolution of lakes were
widely used in paleoclimate studies to evidence long-term trends of the
climate change during the previous thousand years (GASSE et al. 1 989,
FINNEY et al. 1 996). Several geophysical processes may affect the lake
level fluctuations, as tectonic processes or earthquakes, which can modify
RELATIONSHIP BETWEEN THE ZONAL CIRCULATJON 45
the morphology of the lakebed and such its capacity. However, for the last
century period, the main cause of the fluctuation of Bast African lakes'
levels has been related to the time variable changes in the amount of
water that fills the lakebeds. lt must be noticed here that human activity
at the outlets had induced some punctual but also often permanent
changes of the Bast African lake level. Nevertheless, if these changes had
influenced the absolute lake-level records they had just partially and not
really affected the lake-level variations times series. The water balance of
a given lake can be expressed in a simple way : where the changes in
levels or volumes are induced by the balance between the input and the
output. For a lake system, the inputs are the precipitation over the lake
surface and the runoff fluxes drained by various tributaries among the
lake's catchment basin. For an open lake system the outputs are the
discharge via the generally unique outlet and the evaporation from the
lake ' s surface. The variability of the runoff from the catchment is intrin
sically related to the variability of the precipitation over the upstream
drained area. Then, it clearly appears that the variability of the water
fluxes entering the lakes' budget is a function of the rainfall variability
integrated over both the lake and its catchment, assuming that the surface
conditions do not significantly evolve during the studied period. This can
become untrue when the main water input comes from a tributary whose
flow is regulated by humans. Thanks to their evolving capacity, lakes play
the role of natural regulators within their hydrographic basins by absor
bing part of the massive and rough water inflows. In spite of this proper
ty, variations of the outlet discharge are closely dependent on the lake
level, then related to the changes of input, and indirectly linked to the
rainfall variability. For intertropical lake surfaces, it is usually considered
that interannual variability of the evaporation rates is relatively small
compared to the variability of other components. However, the con
ditions, which are favourable to evaporation, are unfavourable to precip
itation and vice-versa. Therefore, when a lake balance is not affected by
a human supplementary input or output fluxes, the lake-level fluctuations
are mainly representative of changes in precipitation regimes and secon
darily of changes in evaporative conditions over the lakes' surfaces. The
collective records of the Bast African lakes can yield a spatially and tem
porally detailed picture of the region environmental history. lnstrumental
lake-level records, which are confined to the 20th century, are one of the
few available indicators of the hydrological variability, and can be used
to investigate the complexities between the water budget fluctuations and
the climate variability.
46 L. BERGONZINI & Y. RICHARD
Three East African lakes have been selected from the equator to around
1 5° of south latitude : Lakes Victoria, Tanganyika, and Nyassa-Malawi
(fig. 1 ). This choice relates to the predominance of direct rainfall over the
lake surfaces on the annually water budgets, respectively 86, 55 and 49 %
of the total input (BALEK 1 977, BERGONZINI 1 998), and also to the avai
lable long lake-level records. The other input component of these three
lakes is the drainage of the catchments which is linked to the annual rain
fall amount. For the selected lakes, the output is essentially the evapora
tion from the water surface, whereas the outflow - via the outlet -
RELATIONSHIP BETWEEN THE ZONAL CIRCULATION 47
-1
-2
-3
1 946 1 956 1 966 1 976 1 986 1 996
3
-1
-2
-3
1 946 1 956 1 966 1 976 1 986 1 996
-1
-2
-3
1 946 1 956 1 966 1 976 1 986 1 996
Malawi level records time series plots ; the recorded periods are : 1 950-97 (47 yr.), 1 946-
00 (54 yr.) and 1 946-89 (43 yr.) respectively. For visibility reasons the scales have been
removed each other.
RELATIONSHIP BETWEEN THE ZONAL CIRCULATION 49
6
• 0 •
5
-1
-2
-3 -+-.-�����
1 946 1 956 1 966 1 976 1 986 1 996
two consecutive October levels), time series plots, for respectively Lak:es Victoria in black
circles, Tanganyika in open circles, and Nyassa-Malawi in grey circles.
50 L. BERGONZINI & Y. RICHARD
-4 -+- SOl.Std
-o-- minus NIN03.Std
-3
-2
-1
3 +...
� .- T"-+-
� ....
� ... ��������...,.
� ... ���....
� ..- ��....
� ...- �
-2
-1
3 +---
� T"-+-
� ....
� ... ���....
� ...- ���...,.
� ... ��T"-+-
..�....- ��....
� ...- �
which limits our study to the past. Over the equatorial Indian Ocean, a
low-level zonal wind speed index from the COADS collection (ZWl
COADS) has been then calculated for 1 946- 1 992, October-December
quarter, in mis, between respectively 78° and 82° East and the equator
and 5° South. Significant relations between this variable have been alrea
dy demonstrated with rainfall amount over equatorial East Africa
(RICHARD 1 992, 1 994) as with Lake Tanganyika fluctuations (BERGONZINI
1 998). The second one also characterizes low levels of surface ( 1 ,000
52 L. BERGONZINI & Y. RICHARD
hPa) ZWI over the equatorial Indian Ocean, but has been derived from the
NCEP re-analyses (KALNAY et al. 1 996). Tuis ZWI-NCEP index has been
calculated between 78° and 82° East and between the equator and
6° South, for the same quarter (October-December). The re-analysed data
set allows to reconstruct the surface ZWI-NCEP, during boreal autumn,
along the 1 948-2000 period. The areas where the two indices, ZWI
COADS and ZWI-NCEP, have been defined, are close to the one defined
by HASTENRATH (200 1 ) and used to investigate East African climate varia
tions. Lastly, the two indexes have been standardized over their respect
ive reconstructed periods (fig. 4b) . As for the two Pacific indexes, the two
Indian indexes are well correlated at a 1 % level (r = 0.79 ; n = 44).
Table 1
Sign and correlation coefficient (expressed in % ) between zonal circulation indexes,
defined for the quarter October-December, and annual Lakes Victoria, Tanganyika and
Nyassa-Malawi level variations calculated between two consecutive October
lakes' levels (in bold significant correlations, 1 and 2 asterisks indicate significance
at a 5 % and 1 % levels)
4 •
2 •
-1
•
VICTORIA
•• •
-2
-3
STANDARDISED ZWl/NCEP
-4 -3 -2 -1 0 2
0
(/)
z
i=
�
5
4
a:
<{ 0
3
> 0
_J 2
UJ
>
UJ
_J
UJ 0
�
<{ -1
_J
TANGANYIKA
_J -2
<{
::> -3
z
STANDARDISED ZWl/NCEP
z -4 -3 -2 -1 0 2
<{
5
4
2 •
• • "
• ••
• •
JI
• •
••
•
•
-1
0 • •
• • •
•
• ••
•
• •
J
NYASSA-MALAWI
•
-2
-3
STANDARDISEDminus Ni
-4 -3 -2 -1 0 2
03
or 5 % level, with the previous absolute annual lakes level. This specif
icity is clearly related to the outlet capacities and to the lake water body
properties. Por the same circulation indexes, or for the same climatic
conditions, lake-level variations will be function of the previous level sta
tus. If the levels are low/high the level variations induced by equivalent
climatic conditions will be relatively higher/lower because of the outlet
capacity control on the outflow (higher outlet capacity for higher levels).
This shows the complexity of the relation between climate and lake
levels. lndeed lake-level records are integrative and result on both annual
inputs and earlier level, underlining the complexity and the richness of
this type of hydrological signals.
The two Lakes Victoria and Tanganyika present almost the same coef
ficient in their respective correlations, whereas disconnections appear
with Lake Nyassa-Malawi. Disconnection, as well as opposition between
Nyassa-Malawi and Tanganyika, Victoria proxies had been already not
iced from sedimentological analysis and paleoenvironmental reconstruc
tions. During the last glacial maximum, Lake Nyassa-Malawi level was
similar to its present one (FINNEY et al. 1 996), whereas Lake Victoria and
Tanganyika levels were lower than the present ones (JoHNSON 1 996,
GAssE et al. 1 990). These similarities and differences must be regarded as
the expression of existing climatic coherence over Victoria and
Tanganyika basins whereas different dependencies occur in the Lake
Nyassa-Malawi southem region. These space-time pattems of the inter
annual lake-level variability, as the spatial correlation coherence, are in
good agreement with regional climate variability studies (ROPELEWSKI &
HALPERT 1 996, 0GALLO 1 998). In concordance with the rainfall
(RoPELEWSKI & HALPERT 1 996, RICHARD et al. 2000), Lake Nyassa-Malawi
interannual fluctuations have been positively correlated to Pacific time
series, whereas insignificant negative correlations are obseived between
northem lakes and ENSO events. During ENSO years, relative low prec
ipitations occur over southeastem Africa and thus induce decreasing or
relative low-level rising. Even if the precipitation amount over Equatorial
Eastem Africa has been already negatively correlated to SOi (RoPELEWSKI
& HALPERT 1 996, 0GALLO 1 998, BELTRANDO & CADET 1 990), no significant
correlation has been registered with Lake Victoria and Tanganyika fluc
tuations. Nevertheless, the sign of the insignificant observed correlations
between SOi and northem lake levels are negative. In addition, this
exploration shows that overall negative significant correlations between
the Victoria and Tanganyika level variations have been only obtained by
using the equatorial Indian Ocean wind indexes. Low westerly or
56 L. BERGONZINI & Y. RICHARD
5. Conclusion
REFERENCES
40.
Biogeography, 22 : 945-952.
by
1. lntroduction
theory by RmNn ( 1 995). However, the search for fossil landslides and
Middle Holocene times could add geomorphological evidence for the
measured. They are inserted in the wall of a Young-pit (YoUNG 1 960), the
66 J. MOEYERSONS, J. NYSSEN, J. DECKERS, H. MITIKU & J. POESEN
wall being oriented in the direction of the local slope (MoEYERSONS 1 989,
NYSSEN 200 1 ) .
The absolute movements of rock fragments and boulders, embedded in
the creeping soil mass of flow 4 in Ethiopia have been measured by
means of a laser theodolite total station, using fixed points from in situ
bedrock outcrops.
- Scar
_ Creep or flow
LEGEND
� Incised valley
toe ·
llfT Biconvexity
---
-
••"" Flat bottom valley
5 km
Fig. 1 . - The study area near Butare, Rwanda. The morphology of the hills reflects
creep and landsliding in the past.
68 J. MOEYERSONS, J. NYSSEN, J. DECKERS, H. MITIKU & J. POESEN
Most of the creep and earthflow lobes end in the flat-bottomed marshy
valleys. A creep lobe descending from Rwaza Hill (fig. 2) has been dug
out along the Kadahokwa River and has already been described else
where (MoEYERSONS 200 1 ). This section (fig. 2B) shows that the creep
lobe sits between two stratigraphical units. The unit, older than the lobe,
is an erosional river terrace, containing clast-supported gravels and peb
bles, varying in diameter from 1 to 1 0 cm, and still showing the original
bed and imbrication structures. This type of gravier sous berge is very
frequent in Rwanda and is generally attributed to erosional and braided
"high energy" river systems from LGM time (PEYROT 1 983). After dep
osition of the lobe, accumulation of peat and fine silt layers took place in
the Kadahokwa river. On the basis of 14C-datations a few hundred metres
downstream (RocHE 1 996) in this Late-Holocene terrace, it can be
concluded that the creep flow arrived in the valley before ± 3,000 BP.
This geologically recent date is not surprising because so many earth
movement bodies are still well recognizable in the field.
A detailed soil-mechanical and stratigraphical study of the creep lobe
at Rwaza Hill shows that it is still not far from the verge of failure
(MOEYERSONS 1 988, 1 989b, 1 990), but it is evident that its stabilization
before ± 3,000 BP is due to reduced seepage pressure and hence a
lowering of the level of subsurface seepage storm flow on the hill slopes.
During the period of active creep and other mass movements, the red
earths of the hurnic ferrallitic soil were characterized by higher water
contents than today, with saturation over a considerable part of their
thickness (MOEYERSONS 200 1 ). Important hydraulic slope erosion can be
ruled out during that time because the developing convex landforms
would have been quickly incised and dissected. Pollen analysis (RocHE
1 988) confirms that the Central Plateau in Southem Rwanda saw since
the end of the arid LGM, about 1 5,000 years ago, the development of
grass savannah vegetation, becorning increasingly forested during
Holocene times before 5,000 BP.
�cm c
Fig. 2. - A : frontal view of the Rwaza creep lobe (localization, see fig. 1), about 1 0 m
thick at its toe ; B : about 100 m to the right of A, the stratigraphical context of the toe of
the creep lobe is visible ; C : the creep movement between 1977 and 1 984, as measured
along a transverse section over Rwaza Hili, a few tens of metres to the left of A.
forest, interrupted by two stable periods. The base of the upper silt
corresponds with intense use of wood for iron melting and associated
forest clearing (VAN GRUNDERBEEK, DouTRELEPONT & RocHE 1 984). Up to
now, archaeological investigations do not inform about the nature of the
two former phases of forest recession. However, the disappearance of
forest is known in the region to reduce infiltration and to increase the
runoff coefficient as to decimate source discharge (Géomines-Somirwa
1 9 8 1 ). Even if the first two phases of forest recession have to be ascribed
to human activities, they should have contributed to the löwering of water
tab les in two ways : first of all by the reduction of the feeding of the water
table as explained by the Géomines-Somirwa study. However, the Late
Holocene deforestation was also the onset in large parts of Africa for gul
lying (MoEYERSONS 2000). Hence gulling leads to incision into the water
table and to the definitive lowering of the latter if the gullies are not
threatened to fill up.
Besides draining watçr tables, gulling aften modifies the slope con
figuration, which can lead to the undercutting of slopes, one of the main
reasons for landsliding (THOMAS 1 996). It is a general phenomenon in the
Butare area and also on the Lualaba-Nile divide that rivers today tend to
incise and undercut the valley borders. This way, the central part of the
creep lobe at Rwaza Hill, mentioned above, is actively undercut by the
Kadohokwa River. Measurements show that creep above this incision
amounted to 48 cm from 1 977 to 1 984 (fig. 2C). First-time movements,
mostly of the soil slip type (CHORLEY et al. 1984), have been clearly
demonstrated to be anthropogenic in origin (MoEYERSONS 200 l b).
The study area seats on the western shoulder of the Danakil rift depres
sion and covers about 200 km2 centred around the village of Hagere
Selam ( 1 3 °40' N ; 39° 10' E), some 40 km to the west of the town of
Makelle. The local geology consists of a subhorizontal series of altema
ting hard and soft Antalo limestone banks, ± 400 m thick, overlain by
Amba Aradam sandstone (HuTCHINSON & ENGELS 1 970). This succession
represents the Cainozoic transgression-regression cycle. Two series of
Tertiary lava flows, separated by silicified lacustrine deposits (MERLA et
al. 1 979), cover this succession. The dissection of the landscape, in
response to 2,500 m uplift during Tertiary and Quatemary times, resulted
PRESENT-DAY REMOBILIZATION 71
in stepped and tabular landforms. The highest levels on the upper basalt
series occur in the area at about, 2,750 m a.s.l. Other structural levels cor
respond to the top of the lower basalt series, the Amba Aradam sandsto
ne and hard layers within the Antalo limestone. The valleys in the study
area are about 500 m deep. Rainfall in the study area varies largely bet
ween 600 and 1 ,000 mm/year (NYSSEN et al. 2004) and falls for > 80 %
between June and September.
Numerous ancient landslide lobes occur in the area (fig. 3). Seventeen
individual slides could be distinguished. In addition, flights ( 1 ) of basalt
colluvium and silicified lacustrine deposits cover the Amba Aradam sand
stone cliff over a distance of about 6 km, west of May Rahaya. Also lobes
(2), (3), (9), ( 1 1 ) and ( 1 5) have their origin in the silicified lacustrine
deposits between the two basalt series. The diamicton of solifluction
lobes (4), (5), (6) and (7) contains some pieces of silicified lacustrine
deposits, but their bulk comprises materials derived from Vertisols, espec
ially present at the top of the lower basalt series. More statistica! infor
mation about the slides in the study area can be found elsewhere (NYSSEN
et al. 2004).
lt is difficult to put a precise age on most of the mass wastings, but they
are still morphologically recognizable and, therefore, geologically spea
king, rather young. Some morphological and stratigraphical elements
have been used to estimate the maximum age of flow ( 1 2). This sol
_
ifluction lobe arrived in the Tsigaba river at the moment when the level
of this river was forced to rise to its actual elevation because of the dev
elopment of a tufa dam about 1 km downstream. The approximate age of
the dam in question bas been defined by the Uffh method on a stalactite.
The obtained age amounts to 1 3 .7 (+0.6/-0.5) ky (sample # Mons 6604 in
NYSSEN 200 1 ) . On the other hand, some elements allow to estimate the
minimum age of Lobe ( 1 5) of May Rahaya (fig. 4). This lobe dams the
valley. Upstream of this dam, to the left of the flow, the climatic signif
icance of Late-Pleistocene and Early- to Middle-Holocene mass
movements and their present-day remobilization in Rwanda and Ethiopia
black earths of the Vertisol type have been deposited afterwards. These
earths can be interpreted as the product of the erosion phase, resulting
from deforestation. At Tsigaba, equivalent deposits have been dated near
their base by the 14C-method as 3,090 ± 30 BP (GrN-25665, NYSSEN
200 1 ) .
72 J. MOEYERSONS, J. NYSSEN, J. DECKERS, H. MITIKU & J. POESEN
-
Fig. 3. Topographical map of the region around Hagere Selam (Makelle, Ethiopia).
·
1 7 important ancient landslides have been defined. One square equals one km2•
Fig. 4. - Flow maya Rahaya ( 1 5), indicated on fig. 3, dams the valley. Black Late
'
Holocene deposits have been accumulated behind the dam.
Table 1
Geotechnical characteristics used in equation ( 1 )
for infinite slope analysis (SELBY 1 993) bas been used to assess the stab
ility of the steep portion of the May Ntebteb flow. In this equation, the
symbols, not mentioned in table 1 , are :
18
dry volume weight (kN/m') 1 1 .77
apparcnl cohcsion al fidd capacily
(kN/m') 6. 1 &
Fig. 5. - A : ground-view of the May Ntebteb flow (4), indicated on fig. 3. The steep
est part of the flow is on the verge of failure. Compression "lines" are developed down
slope ; B : aerial photograph ; C : the part visible in A. The white lines indicate the gen
era! morphology of the flow and its feeder. The black lines are deep gully incisions. The
toe is crossed by a gully incision and the detached part, indicated as B, does not anymore
fully participate to restrain the steeper part A.
Taking into account that flow (4) is about 30 m thick in the middle part
of its steep section (24-25°), equation ( 1 ) indicates a very precarious
equilibrium with m = 0 but with the water content w at field capacity.
This is a situation occurring at the start of the rainy season when springs
at the foot of the flow do not yet flow. Later in the rainy season, seepage
forces should also be taken into account and equation ( 1 ) indicates that in
such conditions the failure plane should be at a depth of much less than
30 m. Obviously, the steep part of the old landslide is held in equilibrium
by the much less inclined toe-part of the tongue.
PRESENT-DAY REMOBILIZATION 75
Several reasons for the slight remobilization of flow (4) can be for
warded :
- lt appears from figure 5B that gully incisions are very active at both
sides of the flow. They are installed on the lithological boundary be
tween the flow - mainly black clays - and the slope itself - Antalo
limestone. In the middle part of the steep section, these gully incisions
are about 20 m deep and it is clear that they partially release the lat
eral anchoring of the flow.
- Figure 5B shows also that the toe of flow (4) is crossed by a gully
from the feeder from the right. Tuis gully disconnects an important
part of the toe, which retains the steep part of the tongue, and reduces
its confining efficiency. Tuis is a typical case of toe unloading by gul
•
lying.
- lt is argued elsewhere (NYSSEN et al. 2004) that the redevelopment
of the vegetation on the May Ntebteb slide, after it was given the
status of exclosure, increases considerably the infiltration capacity of
the soil. This might cause the rise of the water table and an increase of
the hydraulic gradient and hence of the seepage forces towards the
gully incisions.
severely affected by local but heavy tectonic activity with mass move
ments as a result. One example is the situation in Bukavu Town near the
Rwandese-Congolese frontier, at the Southem side of Lake Kivu. Since
its construction, this town has been in constant degradation. A recent
study (TREFOIS 2002) shows a spatial relationship between the localiza
tion of slow creeping and/or sometimes surging macro landslides of the
translational type (DIKAU et al. 1 996) and a double active fault step,
traversing the town in a N-S direction.
Secondly, it appears from the Bukavu and other examples that neotec
tonic movements, even in the East-African rift, seem to be rather local in
nature. It would be an enormous coincidence that in the three regions, as
far apart as from l 0°S to l 3°N, seismic activity should be so closely
contemporaneous.
In the absence of convincing evidence that mass wasting would be a
proxy for seismicity, arguments in favour of a climatic significance can
be forwarded. It appears indeed that the occurrence of extensive mass
wasting in Rwanda, Ethiopia and Malawi coincides with the climatic
amelioration after the LGM. In East Africa and Ethiopia, this ameliora
tion has been documented by dozens of researchers, extracting their data
from glacial moraines, lake-level vàriations, fluvial and lacustrine dep
osits along the Nile, pollen analysis, soils and past geomorphic processes.
An extensive overview can be found in NYSSEN (200 1 ) . Also closer to the
Butare area, research has shown that the climate became wet at ± 1 3 ,000-
1 2,000 BP (MoEYERSONS 200 1 a). As a matter of fact, the occurrence of
mass movements gives additional information about the type of humid
environment, showing that water tables were higher than today, which
implies a better availability of water for vegetation growth.
ACKNOWLEDGEMENTS
REFERENCES
L 'anthropologie, 81 ( 1 ) : 27-47.
et sur leurs enseignements sur Ie paléo-environnement quaternaire des
Hautes Terres de l' Afrique Centrale. -
PRESENT-DAY REMOBILIZATION 79
by
Jozef Hus *
1. Introduction
aries were used as key reference ages to tie Ö'80 deep-sea records to in
K-Ar method (MANKINEN & DALRYMPLE 1 979). Recorded reversal bound
(HELLER & Lru 1984). Based on field reversals recorded in the sequence,
they showed that loess started to accumulate on the Loess Plateau at least
2.5 Ma ago. The onset of eolian loess accumulation on the Loess Plateau
has been extended backwards in time, when recently much older loess
like sediments were found in Lingtai, indicating that the cooling in the
Northem hernisphere may have started at least 7 Ma ago (DING et al.
1 999).
Amongst the best palaeoclimate archives on the continent are indeed
loess-palaeosol deposits. In Northem China the altemation of loess units
A PROXY OF THE PALAEOENVIRONMENT AND PALAEOCLIMATE 85
mate, on a soil property (such as MS) all the remaining factors have to be
(t) : S = f (cl, o, r, p, t). In order to identify any single factor, such as cli
kept constant. The Loess Plateau in China lends itself very well as a test
region to look for a relation between MS and climate. The loess parent
material is quite uniform and originates from the deserts north of the
Plateau. Topography (p) varies little and vegetation (v) co-varies with cli
mate (cl). Hence, S can be reduced in this case to a function of time and
climate, where time represents the duration of weathering. In pedology
two chronofunctions are currently used : linear growth and approach to a
steady state. We have to know how the pedogenic magnetic susceptibility
(Kped) builds up in time, or otherwise stated, to know the rate of change
of Kped and the duration of weathering or residence time in the weather
ing zone. Some investigators think that soils develop very quickly and
hence also Kped, others on the contrary claim that the residence time in
the weathering zone may be very long.
Previous investigations suggested a relation between Kped of the
palaeosols and the present-day precipitation based on a linear correlation
between bulk MS of soil unit S5 and present-day mean annual precipita
tion on the Loess Plateau (Lru et al. 1 99 1 ). Soil S5 was chosen as it is a
marker horizon, easy recognizable and displaying the strongest soil for
mation. This may be criticized as S5 is in reality a soil complex with three
distinct soils separated by loess and because the present-day precipitation
and palaeoprecipitation when S5 was formed are unrelated. Hence, it is
important to look for modem equivalents. On the Loess Plateau of
Northem China the modem soils have MS values comparable with
palaeosols. Apparently, there seems to be a relation between the MS of
the modem soils and precipitation and temperature. Annual average pre
cipitation (MAP) varies over the Loess Plateau from more than 600 mm
to less than 100 mm and average annual temperature (MAT) from
6 to 1 3 °C (PORTER et al. 200 1 ). There is a sixfold increase in MS, from
30 1 0-s to 200 1 0-s, from the southem margin of the Mu Us desert to the
northem margin of the Qinling Shan mountains.
There are many pitfalls and difficulties to face : it is hard to find places
where the modem soil <lid not undergo anthropogenic influences.
Localities with a long record of climatic parameters are limited and for
mean annual evaporation either scarce or absent. Moreover, there should
be no confusion between Holocene soils and the modem soil.
Nevertheless, results obtained by HAN et al. ( 1 996), by MAHER &
THOMPSON ( 1 999) and also by PORTER et al. (200 1 ) are encouraging. HAN
et al. ( 1 996) found that MS increases with both the MAP and MAT in the
90 J. HUS
5. Conclusion
REFERENCES
HAN, J., Lu, H., Wu, N. & Guo, Z. 1996. The magnetic susceptibility of modem
soils in China and its use for paleoclimate reconstructions. - Studia geo
physica et geodaetica, 3 : 262-275.
HAN, J., Hus, J. J., PAEPE, R., VANDENBERGHE, R. & Lru, T. S . 199 1 . The Rock
Magnetic Properties of the Malan and Lishi Forrnations in the Loess
Plateau of China. - In : Lru, T. S. (Ed.), Loess, Environment and Global
Change, Beijing, China, Science Press, pp. 30-47.
HEINRICH, H. 1 988. Origin and consequences of cyclic ice rafting in the Northeast
Atlantic Ocean during the past 1 30,000 years. - Quaternary Research,
29 : 143-152.
HELLER, F. & EvANs, M. E. 1995. Loess Magnetism - Reviews of Geophysics,
33 : 2 1 1 -240.
HELLER, F. & Lru, T. 1984. Magnetism of Chinese loess deposits. - The
Geophysical Journal of the Royal Astronomical Society, 77 : 1 25- 1 4 1 .
HEsLOP, D . , LANGEREIS, C. G . & DEKKERS, M . J . 2000. A new astronornical time
scale for the loess deposits of Northem China. - Earth and Planetary
Science Letters, 184 : 125-1 39.
HoUNsLOw, M. W. & MAHER, B . A. 1999. Source of the climate signa! recorded
by magnetic susceptibility variations in Indian Ocean sediments. -
Journal of Geophysical Research, 104 (B3) : 5047-505 1 .
HovAN, S. A . , REA, D. K . , PrsIAs, N. G. & SHACKLETON, N. 1989. A direct link
between the China loess and marine Ö'80 records : aeolian flux to the north
Pacific. - Nature, 340 : 296-298.
Hus, J. J. & BATER, B. 1 993. Magnetic hysteresis parameters of bulk samples and
particle-size fractions of the loess/palaeosol sequence in Centra! China. -
Geologica Carpathica, 44 (5) : 325-333.
Hus, J. J. & GEERAERTS, R. 1 986. Palaeomagnetic and rock magnetic investiga
tion of Late Pleistocene loess deposits in Belgium. - Physics of the Earth
and Planetary Interiors, 44 : 21 -40.
Hus, J. J. & HAN, J. 199 1 . The magnetic susceptibility profile of the loess
palaeosol sequence in North Centra! China - a proxy measure of climate.
- Acta Geologica Taiwanica, 24 : 57-76.
Hus, J. J. & HAN, J. M. 1992. The contribution of loess magnetism in China to
the retrieval of past global changes : some problems. - Physics of the
Earth and Planetary Interiors, 7 : 154- 168.
JENNY, H. 194 1 . Factors of Soil Forrnation. - New York, McGraw-Hill, 28 1 pp.
KuKLA, G. J. 1987. Loess stratigraphy in Centra! China. - Quaternary Science
Reviews, 6 : 1 9 1-219.
KUKLA, G., HELLER, F., Lru, x. M., Ku, T. c. Lru, T. s & AN, Z. S. 1988.
Pleistocene climates in China dated by magnetic susceptibility. -
Geology, 16 : 8 1 1-8 14.
KuKLA, G., AN, Z. S., MELIECE, L., GAVIN, G. & XIAo, J. L. 1 990. Magnetic sus
ceptibility record of Chinese Loess. - Transactions of the Royal Society
of Edinburgh : Earth Sciences, 81 : 263-288.
A PROXY OF THE PALAEOENVIRONMENT AND PALAEOCLIMATE 93
Lm, X " Lm, T" SHAw, J " HELLER, F" Xu, T. & YuAN, B . 1 99 1 . Paleomagnetic and
Paleoclimatic Studies of Chinese Loess. - In : Lm, T. S. (Ed.), Loess,
Environment and Global Change. Beijing, China, Science Press, pp. 61-8 1 .
LUND, s . P" ACTON, G " CLEMENT, B " HASTEDT, M " ÜKADA, M " Wrr..LIAMS, T. &
the ODP Leg 1 72 Scientific Party 1 998. Geomagnetic field excursions
occurred often during the last million years. - EOS, Transactions
American Geophysical Union, 79 : 178- 1 79.
MAHER, B. A. & TuoMPSON, R. 199 1 . Mineral magnetic record of the Chinese
loess and paleosols. - Geology, 19 : 3-6.
MAHER, B. A. & THOMPSON, R. 1 992. Paleoclimatic Significance of the Mineral
Magnetic Record of the Chinese Loess and Paleosols. - Quaternary
Research, 37 : 1 55- 1 70.
MAHER, B. A. & THOMPSON, R. 1 995. Paleorainfall Reconstructions from
scale for the interval 0-5 Myr BP. - Journal of Geophysical Research, 84 :
MANKINEN, E. A. & DALRYMPLE, G. B . 1 979. Revised geomagnetic polarity time
6 1 5-626.
NOWACZYK, N. R" FREDERICHS, T. W" EISENHAUER, A. & GARD, G. 1 994.
Magnetostratigraphic data from late Quaternary sediments from the
Yermak Plateau, Arctic Ocean : evidence for four geomagnetic polarity
events within the last 1 70 Ka of the Brunhes Chron. - Geophysical
Journal International, 117 : 453-47 1 .
PORTER, S . C " BALLET, B " Wu, X . & AN, Z . 200 1 . Dependence of Near-Surface
Magnetic Susceptibility on Dust Accumulation Rate and Precipitation on
the Chinese Loess Plateau. - Quaternary Research, 55 : 27 1 -283.
ROBINSON, S . G. 1986. The Late Pleistocene palaeoclimatic record of North
and soils from the Kirsa valley, Lappland, Sweden, and their relevance to
palaeoenvironmental reconstruction. - Terra Nova, 5 : 258-270.
THOMPSON, R. & ÜLDFIELD, E. 1 986. Environmental Magnetism. - London,
Allen and Umwin, 227 pp.
THOUVENY, N. J., DE BEAULIEU, J., BüNIFAY, E., CREER, K. M., GULOT, J.,
IcoLE, M . , JoHNSON, S . , JouzEL, J., REILLE, M., WrLLIAMS, T. &
WILIAMSON, D. 1994. Climate variations in Europe over the past 1 40 kyr.
deduced from rock magnetism. - Nature, 371 : 503-506.
THOUVENY, N., MORENO, E., DELANGHE, D., CANDON, L., LANCELOT, Y. &
SHACKLETON, N. J. 2000. Rock magnetic detection of distal ice-rafted
debris : clue for the identification of Heinrich layers on the Portuguese
margin. - Earth and Planetary Science Letters, 180 : 6 1 -75.
VANDENBERGHE, R. E., Hus, J. J. & DE GRAVE, E. 1 998. Evidence from Mössbauer
spectroscopy of neo-formation of magnetite/maghemite in the soils of
loess/paleosol sequences in China. - Hyperfine Interactions, 1 17 : 359-
369.
WoLLIN, G" ERICSON, D. B., RYAN, W. B. F. & FosTER, J. H. 1 97 1 . Magnetism of
the earth and climatic changes. - Earth and Planetary Science Letters,
12 : 1 75- 1 83 .
Zttou, L . P., ÜLDFIELD, F., WINTLE, A. G., RoBINSON, S. G. & WANG, J. T. 1 990.
Partly pedogenic origin of magnetic variations in Chinese loess. - Nature,
346 : 737-739.
Second International Conference
"Tropical Climatology, Meteorology and Hydrology "
Proceedings edited by
G. Demarée, M. De Dapper, J. Alexandre
pp. 95- 1 19 (2004)
by
developed for the Ethiopian Highlands : Ekvo1 36.65 ( 1 - (0.6 / I)) (R2 0.99,
The following relation between rain intensity (I) and kinetic energy (Ek) was
= =
=
n 1 8), (Ekvo in J m2 mm1 , I in mm h·1 ).
1
Due to the occurrency of large drop sizes, this relation yields, within the inten
sity range [0.6 - 84 mm h·1 ] , larger values for Ek compared to elsewhere in the
world. It is recommended to use this new relationship for calculating Ek of rain
in the Ethiopian Highlands, as well as for the computation of (R)USLE's rain
erosivity factor.
1. Introduction
Dso = aP (1)
is expected between the median volume drop diameter (Dso) and rain
intensity (1), with a and b constant in a given region (HUDSON 1 97 1 ). This
type of relationship is however questioned for high intensities, since rain
drops have a maximum size (HUDSON 1 97 1).
The calculation of kinetic energy of a raindrop
(2)
involves transforrnation of average drop diameter into mass (m), assu
rning that raindrops are spherical, and an assessment of the terminal fall
velocity (v) of raindrops of different sizes, as experimentally obtained by
LAws ( 1 94 1 ). Calculations of kinetic energy generally show that there is
an increase up to an intensity of about 75 mm h·1, above which Ek remains
constant (WISCHMEIER & SMITH 1 958, HUDSON 1 97 1 , JAYAWARDENA &
REZAUR 2000, SALLES et al. 2002). Measurements of kinetic energy and
98 J. NYSSEN et al.
raindrop sizes are in most cases not readily available ; hence the devel
opment of empirica! relationships between rain intensity and kinetic ener
gy, which is generally considered volume-specific, and expressed in
J m2 mm' . Time-specific kinetic energy
gives generally better correlations with rain intensity (SALLES et al. 2002).
In this section, we will particularly study (a) the spatial variation of
rain in the study area, and how this is influenced by elevation (range :
2,200-2,750 m. a.s.l.), slope aspect and geographical position ; (b) the
temporal variations of rain ; and (c) rain erosivity and the different fac
tors deterrnining it, such as rain intensity and rain drop size.
r� ·� -�
ellacklish Adi (5) 1u"tr 1 - 2 .c....u>
- u::.
lllDJD 3 r-._ 4 •
Fig. 1 . - Elevation map of the study area with position of the rain gauges (white :
2,000-2, 1 00 m a.s.!. ; black : 2,700-2,800 m a.s.l). Location map : ( 1 ) river ; (2) Rift
Valley escarpment ; (3) boundary ; (4) study area.
area, on a total area of approximately 40 krn2 (fig. 1). The rain gauges
Twelve rain gauges were installed at various locations in the study
were constructed from simple materials : a cylinder made from two metal
tins was fixed in cement, on top of a 1 - 1 .5 m high tower. A strong plastic
battle, with a funnel fixed at its top, was inserted in the tins (fig. 2).
Orifices of the funnels were horizontal.
The rain gauges were more or less evenly spread over the study area.
They were located near houses for security reasons, which sometimes
resulted in obstacles (especially eucalyptus trees) in their vicinity. Por
most rain gauges, these obstacles were at the recommended distance of
twice their height ; the distance is always larger than the height of the tree
itself, as recommended by some meteorological services. Sometimes rain
gauges were placed on rooftops, to avoid such obstacles. Due to the na
ture of the terrain, the rain gauge sites are generally inclined (fig. 3),
which is also not recommended, but which allowed introducing slope
aspect in the analysis. Rain gauges were read daily by secondary school
SPATIAL VARIABILITY OF RAIN 101
-
Fig. 2 . Cross-section o f rain gauge 7. Note the two metal tins fixed i n cement
(arrows), which hold a plastic bottle with funnel (diameter = 1 2.7 cm).
Fig. 3. - A typical location for a rain gauge : on a fence near a house (rain gauge 5).
until the stains started overlapping. During the transfer period from insi
de the house to the measuring place, and back to the house again, the
cards were covered by a tray. Time was registered on the card and stains
were immediately outlined by pencil by a team of 3-5 people working
inside.
Underlying principle for this method is that a drop falling on a unif
ormly absorbing surface creates a stain with a diameter proportional to
the drop diameter. POESEN ( 1 983) established a calibration curve for stains
0.6-6 mm on blotting paper of 280 g m-2 :
0st = 2.44 * 0dr1.34 (r2 = 0.99 ; n = 9) (4)
where 0st = diameter of stain on blotting paper (mm) and 0dr = diameter
of water drop (mm).
This calibration equation was tested with the blotting paper we used in
the experiment. Measurement of the stain diameters was carried out with
a precision of 0.25 mm . In case of elongated stains, these were consider
ed to be the result of a coalescence of two stains, and measured as such.
In total, 65, 1 00 raindrops were outlined and measured (VANDENREYKEN
200 1 ) .
The tipping bucket rain gauge was functioning at some metres ' distan
ce from the place where the drop-size measurements were done, which
allowed to link rain intensity to each measurement of raindrop sizes. For
rain with low intensity, the time span between two successive tips was
used to calculate the intensity. Measurements at the beginning and at the
end of a shower, when the bucket in the gauge <lid not tip, as well as
SPATIAL VARIABILITY OF RAlN 103
i:
Table 1
Rain gauge characteristics and average annual rain (n = 4)
Rain gauge Location Years of X coordinate Y coordinate Elevation Aspect' Average Range
number observation (UTM) (UTM) (m a.s.l.) annual rain
(mm)
1 Hechi (Yohannes) 4 5223 1 3 1 507376 2275 228 743 689-782
2 Hechi (Mebrahtu) 4 521603 1508279 2330 1 94 787 692-890 �
3 Gabla Emni 4 520479 1507234 2400 26 712 660-794
4 Dingilet 4 5 19723 1 506467 2530 103 794 734-838
�
en
en
5 Haddish Adi 4 5 19050 1 505530 2730 42 776 665-8 8 1
�
6 Harena (Abraha) 4 520052 1 50794 1 2390 74 775 706-853
�
7 Harena (Argaka) 4 520434 1 508661 2350 173 791 697-9 14
8 Zenako 4 520289 1 509434 2540 150 779 678-937
!?--
9 Adi Kalkwal 4 52 1 7 1 9 1 508945 2470 1 66 789 695-965
10 Khunale 1 5 1 9648 1 5 1 0092 2560 340 654
11 Adi Worho 1 521 1 89 1 5 1 1450 2460 350 683
12 Miheno 1 522197 1 5 10 1 88 2520 39 846
Table 2
Regression equations for the different explanatory factors of average yearly rain
P ( 1 998-2001 ) (n = 9)
Factor Equation r2
Il
0
0
I 1600
i5i
� l:IDD tf1
0
Il
D
800
400
Elevation (m a.s.!.)
Fig. 4. - Annual rain depth vs. elevation for Ethiopia (r' = 0. 1 2) . Squares after KRAUER
1988. Circles correspond to measurements in this study.
8.SO �������
825
•
• •
700
0
s E N o
675
-180 - 1 35
650
-90 -45 0 45 90 1 35
and more downslope, from where they move upslope towards the NW,
stations (partly) in rain shadow receiving less precipitation. The path of
such intense showers can possibly be explained by the fact that this large
valley, in the SE of the study area (fig. 1), is a preferred flowpath for air
masses during the rainy season. Another explanation might be the pres
ence of limestone in the low-lying areas ; its higher albedo would create
convective movements more rapidly than in the upper areas on dark
basalt-derived materials. We found however no correlation between year
ly rain and distance to limestone area.
Correlations between geographical coordinates and average annual
rain are very weak, probábly due to the small dimensions of the study
area (tab. 2). The most relevant variables were combined in a multiple
regression model (fig. 6), integrating the models for the single variables :
P 0.06 z + 43. 1 sin(a+ 142.8) + 596 (n = 9 ; r2 = 0.76 ; P < 0.01 ) (6)
=
Application of this model to the DTM (fig. 7) shows how, at the same
elevation, differences in aspect would result in differences in rain depth
SPATIAL VARIABILITY OF RAIN 107
of 50 to 100 mm. Within a slope with the same aspect, average yearly
precipitation increases with elevation. This observation might be impor
tant for geomorphic processes (and for agriculture), since elevated land,
oriented south to east, receives most rain and would be more subject to
erosion and landslide risk.
Lqend·
D
D
< 700 mm
700 - 725 mm
-
725 - 750 mm
-
750 - 775 mm
-
775 - BOO mm
> BOO mm
liii iiiil
_....
lHOAIO
Fig. 7. -
Estimation of the yearly rain distribution in the study area, based on a non
linear multiple regression, using elevation and aspect of 9 rain stations. Eq. (6) was
applied to the DTM (NYSSEN 1995) and a mean filter (5 x 5) was used.
Table 3
Rain events by their moment of initiation
300
Yearly average:
790 mm
250
,-..., 200
s
s 150
�
'-"
100
50
0 L_
\� �� # t'Ç,<;;- �� \V.t;;>'Q \$> � c.:;ä 0 � <Y
<!}- 0..::.. f{;v
vó,;
should be stressed that the average rain measured in January is the result
of exceptional rains in January 1 999 only.
The precipitation coefficient, i.e. the average monthly precipitation
divided by one twelfth of the average yearly precipitation allows to char
acterize the different months (GAMACHU 1 977). October to May are dry
months (coeff. < 0.6), June and September have coefficients ( 1 .0 - 1 .9).
A very high concentration of rain is observed in July and August
(coeff. > 3 .0). Remarkably, there are no months with intermediate pre
cipitation coefficients (2.0 - 3.0).
succession of dry years between the late 1970s and late 1980s produced
the driest decade of the previous century in the Ethiopian Highlands,
there is no evidence for a long-term trend or change in the region's annual
rain regime (CoNWAY 2000).
Dogda Tcrri>ien
Maybar
5%
o 0-18 rrm1l
0 18-30 rrm1l
o J0-42 rrm1l
Q42-54 nmh
• 54-66 nmh
• 66-208 mmlh
b.
these thresholds caused > 50 % of total soil loss on SCRP plots (KRAUER
1 988).
Table 4
Table 5
Median volume drop diameter (D,o) for each rain intensity
• one tip of the tipping bucket rain gauge (resolution : 1 min.) corresponds to an inten
sity of 1 2 mm h·1• Time spans > 1 min. between two successive tips were only taken into
account if there was continuous rain ; intensity was calculated by dividing 1 2 mm h-1 by
the number of minutes between two successive tips.
5.0
4. 5
-.------.
4.0
•
s 3.5
•
g 3.0
2. 5 y = 0. 541.n(x) + 1 . 8
cS r = 0.89 18 p < 0.001
2.0 n=
1.5
1.0
0 10 20 30 40 50 70 80 90 100
-1--"""'T"--r-"""'T"-.....-"""'T"-r--r---t
60
16 r '
14
1 1
,....._ 1 \
1
�
12
"-"'
]0
10
------ 0.57 mm/h
- 1 . 7 1 mm/h
> """" 4 mm/h
·s
ç::: 8 -
6
24 mm/h
'""'
�
------ 84 mm/h
'iij 4
�
-
2
<I)
0
0 1 2 3 4 5 6 7 8
important. October-April are nearly dry ; July and August take the largest
share in yearly rain.
Rain intensity is lower than expected : 88 % falls with an intensity
< 30 mm h-1 • High intensities have a short duration ; maximum rain depth
over 1 h is nearly the same as the one over 24 h.
Por all rain intensities, the median volume drop diameters (Dso) are
larger than those found in other regions. Large rain erosivity is due to
large drop sizes, rather than to high rain intensities.
A relation between rain intensity and kinetic energy was devel
oped ( 1 2). Due to large drop sizes, it yields, within the range [0.6-84 mm
h·1], larger values for Ek than elsewhere in the world. lt is suggested that
this relationship ( 1 2) be used for calculating Ek in the Ethiopian
Highlands, and for the computation of the (R)USLE's rain erosivity
factor.
ACKNOWLEDGEMENTS
This study is part of a research programme funded by the Fund for Scientific
Research - Flanders, Belgium. Financial support by the Flernish Interuniversity
Council (VLIR, Belgium) is acknowledged. Thanks go to Berhanu Gebremedhin
Abay for assistance to all the fieldwork. Numerous farmers, the local
Agricultural Office, REST (Relief Society of Tigray) branch and the authorities
of the concemed villages and district facilitated the research. Many thanks go to
all people who faced the difficult kremti rainy seasons with us, and especially the
secondary school students who daily read the rain gauges as well as our fellow
rain drop outliners.
REFERENCES
ABEBE, Y. & APPARAO, G. 1989. Annual rainwater potential and its variability in
drought years over Ethiopia. - In : Conference on climate and water
(Helsinki, Finland, September 1 989), Vol. l . Suomen Akaternian
[Helsinki], pp. 2 1 9-235 .
AHRENs, C. D. 2000. Meteorology today. - USA, Brooks/Cole, 528 pp.
AINA, P. 0., LAL, R. & TAYLOR, G. S. 1 977. Soil and erop management in relation
to soil erosion in the rainforest of western Nigeria. - In : FosTER, G. R.
(ed.), Soil erosion : prediction and control. Ankeny, Soil Cons. Soc. Am.,
pp. 75-82.
ARKIN, Y., BEYTH, M., Dow, D. B . , LEVITTE, D., TEMESGEN, H. & TSEGAYE, H.
1 97 1 . Geological map of Mekele sheet area ND 37- 1 1 , Tigre province, 1 :
250.000. - Addis Ababa, Imperia! Ethiopian Govemement, Ministry of
Mines, Geological survey.
SPATIAL VARIABILITY OF RAIN 1 17
HunsoN, N. 197 1 . Soil conservation. - Great Britain, Billing & Sons Ltd,
320 pp.
HuLME, M. 1992. Rainfall changes in Africa : 193 1 - 1 960 to 1 96 1 - 1 990. -
International Journal of Climatology, 12 : 685-699.
1 069- 1 082.
MERLA, G., ABBATE, E., AzzAROLI, A., BRUNI, P., CANUTI, P., FAZZUOLI, M.,
SAGRI, M. & TACCONI, P. 1979. A geological map of Ethiopia and Somalia
( 1 973) 1 :2.000.000 and comment. - Italy, University of Florence.
MEssERLI, B. & RoGNON, P. 1 980. The Saharan and East African uplands during
the Quaternary. - In : WILLIAMS, M. & FAURE, H. (eds.), The Sahara and
the Nile. Quaternary environments and prehistorie occupation in northern
Africa, Rotterdam, Balkema, pp. 87- 1 32.
Mm-IR, P. 1 963. The geology of Ethiopia. - University College of Addis Ababa
Press, 270 pp.
MooRE, T. R. 1979. Rainfall erosivity in East Africa. - Geografiska Annaler,
61A : 147-156.
NYSSEN, J. 1995. An approach of soil erosion in the Tigray Highlands (Ethiopia) ;
case study in Dega Tembien. - Unpub. MSc thesis, Faculty of Sciences,
University of Liège, 325 pp.
NYSSEN, J. 200 1 . Erosion processes and soil conservation in a tropical mciuntain
catchment under threat of anthropogenic desertification - a case study
SPATIAL VARIABILITY OF RAIN 1 19
Catena, 24 : 275-287.
OBI, M. E. & SALAKO, F. K. 1 995. Rainfall parameters influencing erosivity in
southeastem Nigeria. -
27 1 .
SCRP 2000. Area of Maybar, Wello, Ethiopia : Long-term Monitoring of the
Agricultural Environment 1 98 1 - 1 994. Soil Erosion and Conservation
Database. Soil Conservation Research Project, Beme and Addis Ababa,
8 1 pp.
STRAHLER, A. H. & STRAHLER, A. N. 1 992. Modem physical geography. - New
York, John Wiley & Sons, 638 pp.
YILMA, S. & DEMAREE, G . 1 995. Rainfall variability i n the Ethiopian and Eritrean
Biogeography, 22 : 945-952.
highlands and its links with the Southem Oscillation Index. - Journal of
AGRICULTURE, FOOD
AND
WEATHER PREDICTION
Second International Conference
"Tropical Climatology, Meteorology and Hydrology "
Proceedings edited by
G. Demarée, M. De Dapper, J. Alexandre
pp. 123- 135 (2004)
by
Europe and the whole of Africa over a time period from May 1998 to April 2000,
and over Honduras and Nicaragua during the hurricane Mitch pass-over in
December 1998. We generated to continental scale maps of change in surface
attributes over the growing seasons of every biome. Finally, a validation of the
change patches identified allows us to interpret the type of change and its
environmental significance. For each important change, information on the
process of change and its environmental significance is presented.
1. Introduction
All recent scientific evidence clearly points to the fact that the impact
of global change on land surface attributes will not be uniformly distrib
uted geographically. Assessing the regions "at risk" of rapid land-cover
changes and/or natural disasters is therefore a priority for global change
research and for policies aimed at mitigating the impact of these changes.
The objectives of this project are : (1) to use SPOT VEGETATION data
to monitor over large regions the impact on ecosystems of natural disas
ters such as droughts, fires, floods and vegetation diseases, as well as
land-cover change "hot spots" ; (2) to validate and interpret SPOT VEG
ETATION-based maps of natural disasters and extreme land-cover
changes with collateral data on natural disasters and "hot spots" of land
cover change ; (3) to integrate this validated product in the current efforts
of the global change scientific community, sponsored by the International
Geosphere-Biosphere Programme (IGBP) and International Human
Dimensions Programme on Global Environrnental Change (IHDP) ; (4)
to assess regions "at risk" of rapid environmental change in order to focus
research on the most vulnerable areas and support the design of appro
priate mitigation policies.
2. Methodology
3 .4. REsULTS
The results are :
- A digital map representing the impact of the natural disasters and
rapid land-cover changes that occurred during the growing seasons in
the different agro-ecological regions of Africa and Europe between
May 1 998 and April 2000 ;
- A detailed validation of this map, with collateral data on natural di�as
ters (droughts, floods), rainfall, and information on the process of
change and their environmental significance.
In Africa, the regions were defined from the Ecosystem Map of White.
The period corresponding to the growing season for each region was
defined from rainfall and spectra! vegetation index profiles. In Europe,
the limits of the agro-ecological regions and their growing season are
MONITORING NATURAL DISASTERS 1 27
.....
"
,
"
high change
., . "
Fig. 1 . - Map of land-cover changes in Europe and Africa according to the growing
seasons of different agro-ecological regions from May 1998 to April 2000 . The change
intensity is measured by the change vector magnitude. The clouds are represented in
white (mask) and the red circles delimitate the area where a detailed validation has been
done ([W8]).
1 28 F. LUPO, M. CORLAZZOLI & E. F. LAMBIN
A large area bas been affected (fig. 2, black circle) with 1 17,000 acres
of farmland destroyed and with 290,000 people homeless (followed by
Over the past twö years, four lakes have been created from Nasser's
excess water, bringing new lakes to this part of the Sahara for the first
time in 6,000 years. Figure 4 shows NDVI images from SPOT Vegetation
compared with Landsat7 scene from August 23, 2000.
Fig. 2. - Monthly composites of NDVI images, Ghana, Lake Volta, in October 1 998
and 1 999 from SPOT VEGETATION.
MONITORING NATURAL DISASTERS 1 29
20·:0 .�
1
-·•
1!KJ •• ·•
Fig. 4. - SPOT Vegetation Monthly .composites in August 1 998, 1 999 and 2000 and
a Landsat7 scene in August 2000. West of lake Nasser, Southem Egypt.
In 1 997- 1 998 the global climate system was perturbed by the largest El
Nifio/Southem Oscillation (ENSO) event observed this century
(SHERIDAN et al. 1 998). During the period 25-3 1 October 1 998, the
Hurricane Mitch stroke Centra! America and heavy rainfalls caused
extensive flooding and triggered numerous destructive landslides.
Impacts of floods and landslides in land cover over the pacific coasts
of Honduras and Nicaragua were detected using a SPOT Vegetation sub
scene of 200 x 200 pixels covering a region in Centra! America across El
Salvador, Honduras and Nicaragua. The rainy season normally occurs
from the end of May to mid-November. In this particular context, we
1 30 F. LUPO, M. CORLAZZOLI & E. F. LAMBIN
- Ocean
- very low change
-
- low change
-
-
- s
••
high change
D
D very high Change
O clouds
_
Fig. 5. -
Location of the two hot spots of interest inside the red circles, the Choluteca
River and Choluteca city in Honduras and the San Cristóbal volcanic complex in
Nicaragua. The red crosshatched part represents the Hurricane Mitch pathway.
chose to look only over a time period from August to December respec
tively of 1 998 and 1999. Over these short periods, one expects to detect
mostly the influence of interannual climatic variability and natural
hazards.
The result is a digital NDVI-based change vector magnitude map
representing most of the change patterns biome seasonality and dynam
ics, but also the impact of natural disasters associated with the Hurricane
Mitch event and rapid land-cover changes. We focused more particularly
towards two test sites : the first along the Choluteca river in Honduras and
the second around the Casita volcano in Nicaragua (fig. 5).
MONITORING NATURAL DISASTERS 131
Figs. 6a and 6b. - SPOT VGT pseudo-coloured NDVl monthly composite of October
(a) and November (b) 1 998. The Choluteca River is well detected in yellow-green inside
the blue ellipse (b).
132 F. LUPO, M. CORLAZZOLI & E. F. LAMBIN
Figs. 7a, 7b and 7c. - Maps showing : (a) the location of the San Cristóbal complex
and Casita volcano within the Cordillera de Marabios, (b) the extent of the 1 998 debris
avalanche and (c) the "hot spot" of land-cover change location from the SPOT Vegetation
NDVI-based change vector magnitude map.
Choluteca river peak. Floodwaters remained high for several days cau
sing substantial lateral scour of channel margins and aggravation of sed
iment in the Rio Choluteca, as well as in the surrounding floodplain. In
Choluteca there were 260 registered deaths, 400 persons missing and
approximately 2,500 homes destroyed, primarily along the left bank of
the River.
Although Hurricane Mitch did not enter Nicaragua at all, its large
scale cyclone circulation, together with a strong low-level southwesterly
MONITORING NATURAL DISASTERS 133
flow, produced the highest accumulated precipitation rates over the north
westem part of Nicaragua. Casita volcano, one of the five principal
volcanic edifices of the Marrabios Range, was the site of a catastrophic
landslide and lahar in 1 998 .
.On October 30 more than 700 mm of precipitation had been recorded.
A part of the southwestem flank, some 250 m SW and 100 m below the
summit, composed of hydrothermally altered and highly fractured rock,
broke off. A rockslide of about 200,000 m3 ( 1 10 x 60 x 30 m) descend
ed the slope, and was joined by a second flank section, located just on the
western edge of the primary source. On the way it destroyed the towns of
El Porvenir and Rolando Rodriguez (figs. 7a, 7b and 7c). The first news
reports ([W9], [Wl O], SHERIDAN et al. 1 998) from the disaster area
described a large mudflow that had destroyed the towns and killed
2,000 people.
Tuis new land surface product represents the impact on surface attrib
utes of the major natural disasters, which are mainly caused by rainfall
anomalies. Most of the validation information is related to severe floods,
heavy rains, cyclones or important difference in rainfall distribution
between two growing seasons. High change magnitudes detected over
known deforestation "hot spots" are too close to the clouds to be valida
ted with confidence. The validation highlights the good performance of
the method. There is a need for improvements in the pre-processing of
SPOT VEGETATION data as the incomplete cloud mask and sensor noi
ses forced us to work with monthly composites, even though droughts
and floods would be better detected on decadal composites.
ACKNOWLEDGEMENTS
REFERENCES
244.
LAMBIN, E. F. & STRAHLER, A. 1 994. Remotely-sensed indicators of land-cover
15 ( 1 0) : 2099-21 19.
change for multitemporal change-vector analysis. - Int. J. Remote Sens.,
BoRAK, J. S., LAMBIN, E. F. & STRAHLER, A. 2000. Use of tempora! metrics for
1 859.
SHERIDAN, M. F. , SIEBE, C., BüNNARD, C., STRAUCH, W., NAVARRO, M.,
CRuz CALERO, J. & BuITRAGO, N. 1 998. Report on the October 30, 1 998
avalanche and breakout flow of Casita Volcano, Nicaragua, triggered by
Hurricane Mitch. - Smithsonian Institution's Global Volcanism
Program's World-Wide Activity Notices (November).
Web sources
by
* Institute for Land and Water Management, K.U. Leuven University (Belgium).
** Kenya Meteorological Department, Nairobi (Kenya).
1 38 E. C. KIPKORIR, S. M. GACHUIRI, J. MUKABANA & D. RAES
another criteria to determine the onset dates of the growing season. The results
of the study have important applications for formulating guidelines for planting
dates on different soil types for different regions in Kenya.
1. Introduction
2. 1 . RAfNFALL IN KENYA
The rainfall climatology of Kenya, and Bast Africa in general, is gov
emed by rain-bearing systems associated with the passage of the
Intertropical Convergence Zone (ITCZ). The rainfall regime of Kenya is
dominated by two mainly dry seasons, and two rainy seasons associated
with the movement of the ITCZ. The most southerly position of the ITCZ
occurs in January when it is centred over Tanzania, resulting in the esta
blishment of the northeast trades (fig. 1). During December to February
EVALUATION OF THE ONSET OF THE GROWING SEASON 139
1 .0
0.0
- 3 .0 INDIAN
OCEAN
- 4 .0
the trades bring comparatively dry air of mainly continental origin, thus
bringing stable conditions and low rainfall. The ITCZ moves northwards
and is centred over the Kenyan Highlands in April. From March to June
the northeast flow is weakened and a low-pressure system over Lake
Victoria gives rise to convergent easterly flow bringing moist air from the
Indian Ocean. This produces the first rains of the year, known in Bast
Africa as the "long rains". The most northerly position of the ITCZ
occurs in July, over the Sudan. From June to September the southeast
trade winds bring maritime air from the Indian Ocean. Despite the
140 E. C. KIPKORIR, S. M. GACHUIRI, J. MUKABANA & D. RAES
maritime origin of the air, this is a dry season for large parts of the coun
try. From September through to November, the ITCZ moves southward,
resulting in easterly flow, bringing moisture in October and November.
This produces the second rains of the year. The second rainy season in
Kenya is known as the "short rains". This pattern describes a very sim
plified double wet season equatorial model for East Africa and Kenya
(DAVIES et al. 1985). However, in reality the local picture is more complex
because of the influence of the north-south trending mountain ranges, and
the Rift Valley.
2.3. DATA
Daily rainfall data were collected for the period 1 970- 1999 from the
six climatic stations (tab. 1). The actual record length per station after
accounting for missing data is shown in table 1 , column 5 . All the daily
rainfall data were obtained from the headquarters of the Kenyan
Meteorological Department, Nairobi, that coordinate and compile all the
weather information collected by the meteorological stations in the
Table 1
Characteristics of the six considered climatic stations
Station Longitude Latitude Altitude Record Jength P/ETo Agroclimatic
a.s.l (m) (years) (%) zone
Kisii 34° 47' -0° 40' 1 ,768 29 1 36 Humid
Embu 35° 48' -1° 00' 1 ,5 1 0 20 95 Humid
Mtwapa 39° 42' -3° 24' 25 27 85 Humid
Kabete 36° 44' -1° 15' 1 ,820 26 71 Sub-humid
Nakuru 36° 06' -0° 1 6' 1 ,901 27 69 Semi-humid
Narok 35° 50' -1° 08' 1 ,890 28 52 Transitional
P : mean annual rainfall (mm) ; ETo : mean annual reference evapotranspiration (mm).
EVALUATION OF THE ONSET OF THE GROWING SEASON 141
2.4. CRITERIA
Different criteria have been used in literature to determine the onset of
the growing period. Examples include those of FRERE et al. ( 1990),
STEWART ( 1990), RoBERTSON ( 1 970), BENOIT ( 1 977), ILESAMI ( 1 972). In this
The third is that of RoBERTSON ( 1 970), which is based on wet spells being
greater than dry spells (Cspells). A wet spell is when the root zone deple
tion is less than or equal to readily available soil water in the root zone.
A dry spell is when the root zone depletion is greater than readily avail
able soil water in the root zone. Readily available soil water is considered
equal to 50 % of the total available water in the root zone and the root
zone depth is 30 cm.
60
50
40
,•
10
Jan Feb Mar Apr May June July Aug Sep Oct Nov Dec
-
Fig. 2. Probability of a wet day for Narok (line), Nakuru (dotted line) and Mtwapa
(asterisk) stations.
accepted. Once the three search dates are specified, onset dates are deter
mined between these limits. The date on which a criterion is first met or
exceeded is the onset date. The described procedure confines the search
to only those periods with less risk of dry spells that may follow short wet
spells occurring outside the normal rainy season.
In order to obtain the operator-specified search dates, the probability of
a wet day was determined for each climatic data set by using a ten-day
moving average. In the analysis, a day was considered to be wet when
there was more than 3 mm of rainfall and dry when rainfall was 3 mm or
less. The probabilities of a wet day versus time for the considered agro
climatic zones are plotted in figure 2 and figure 3 . The normal, earliest
and latest specified search dates were extracted from these probability
graphs. For consistency in defining these dates, the normal date was taken
to be the day when the 25 % probability of a wet day is exceeded.
Exceptions were at Kisii (a very wet station) where the 30 % probability
was used and at Narok (a dry station) where a 20 % value was used. The
earliest date was taken to be a date just before the normal date and prefer
ably coinciding with the beginning or middle of the month whichever is
closer. The latest acceptable date was taken to be May 31 (normal end of
the long rains), except at Kisii where no search date was set.
EVALUATION OF THE ONSET OF THE GROWING SEASON 143
70
�
60
0
......
>- 50
ca
"C
-
Q) 40
3:
ca
- 30
0
�
:c
20
�
ca
.Q
10
c..
Jan Feb Mar Apr May June July Aug Sep Oct Nov Dec
Fig. 3. - Probability of a wet day for Kabete (line), Embu (dotted line) and Kisii (aster
isk) stations.
Table 2
Soil data for the three soil types used in the study
Soil type Field capacity Wilting point Total Available Water
(vol. %) (vol. %) (mm) water/0.3 m soil depth
Fine Sand 19.6 4.2 46.2
Silt Loam 33.8 9.2 73.8
Silt Clay 44.7 25.7 57.0
roots and soil water movement in the absence of a shallow water table.
The estimation of the amount of rainfall lost by surface runoff is based on
the curve number method developed by the US Soil Conservation Service
(USDA 1 964, RALLISON 1 980, STEENHUIS et al. 1995). The infiltration and
internal drainage is described by an exponential drainage function (RAEs
1 982) that takes into account the initial wetness and the drainage charac
teristics of the various soil layers. The drainage function mimics quite
realistically the infiltration and internal drainage as observed in the field
(RAEs 1 982, FEYEN 1 987, HEss 1 999, WrYo 1 999). With the help of the
dual erop coefficient procedure (ALLEN et al. 1 998) the soil evaporation
rate and erop transpiration rate of a well-watered soil is calculated. The
actual soil evaporation is derived from soil wetness and erop cover
(BELMANS et al. 1 983). The actual water uptake by plant roots is described
by means of a sink term (FEDDES et al. 1 978, HOOGLAND et al. 1 98 1 ,
BELMANS et al. 1 983) that takes into account root distribution and soil
water content in the soil profile.
The resulting onset dates per criteria were statistically analysed using
RAINBOW software package (RAES et al. 1996). The results are shown
in table 3. Only the long rains were analysed. Most of the data fitted log
normal distribution and the remainder fitted normal distribution. Results
indicate that criteria Caccum and Cstored give earlier onset dates on aver
age of about one week than criteria Cspells regardless of the climate. The
onset dates from criteria Caccum and Cstored are not significantly differ
ent, although those from Caccum were slightly earlier in most of the
years than Cstored except at Kisii (a very wet station).
month in Kisii, Embu, Mtwapa, Kabete and Nakuru. In Narok, á dry sta
Onset of the growing season ranges over the years from 2 weeks to one
tion, onset dates range about 3 months. The high variation means that
EVALUATION OF THE ONSET OF THE GROWING SEASON 145
Table 3
Early, mean and late onset dates derived from three criteria for six stations in Kenya
Station Onset date Caccum Cstored Cspells
in years
Fine Sand Silt Clay Silt Loam
out of 10
2 (early) 6 March 5 March 10 March 9 March 8 March
Kisii 5 (mean) 17 March 16 March 23 March 23 March 22 March
8 (late) 29 March 27 March 9 April 9 April 9 April
2 (early) 19 March 1 8 March 27 March 27 March 2 1 March
Embu 5 (mean) 29 March 29 March 7 April 6 April 5 April
8 (late) 9 April 10 April 19 April 1 7 April 20 April
2 (early) 8 April 8 April 1 1 April 1 3 April 12 April
Mtwapa 5 (mean) 17 April 17 April 24 April 24 April 24 April
8 (late) 26 April 27 April 8 May 6 May 7 May
2 (early) 1 7 March 16 March 1 April . 25 March 26 March
Kabete 5 (mean) 30 March 30 March 13 April 9 April 1 1 April
8 (late) 1 3 April 14 April 25 April 25 April 28 April
2 (early) 6 April 6 April 1 1 April 7 April l O April
Nakuru 5 (mean) 19 April 20 April 28 April 22 April 23 April
8 (late) 3 May 7 May 16 May 9 May 8 May
2 (early) 2 1 Nov. 2 Dec. 17 Dec. 16 Dec. 19 Dec.
Narok 5 (mean) 1 Jan. 14 Jan. 6 Feb" 2 Feb. 12 Feb.
8 (late) 1 8 Feb. 4 March 5 April 29 March 7 April
mean onset dates have little meaning and that the risk farmers wish to
take should be considered when planning agronomie activities.
The onset dates per criteria were evaluated by computing the satisfac
tion of the erop water requirement for the successive 30 days from plant
ing by means of the BUDGET soil water balance model. The magnitude
and frequency of the occurrence of water stress was computed and used
to evaluate the performance of the criteria. The results are shown in
table 4. Analysis of the results yields that :
- Applying one or another criteria to determine the onset of the growing
season does not guarantee that the erop will not experience water
stress during the sensitive initial growth stage. At the humid agrocli
matic zones (Kisii, Embu, and Mtwapa), analysis of the results shows
that the erop is likely to survive the initial growth stage without great
damage. In the sub-humid (Kabete) and semi-humid (Nakuru) zones
the risks of erop failure or damage is already more likely to occur. At
the dry Narok station, the crops failed completely. As such the criteria
are not valid for dry regions.
146 E. C. KIPKORIR, S. M. GACHUIRI, J. MUKABANA & D. RAES
Table 4
Frequency (in percentage) of obtaining no water stress
during initial growth stage (30 days)
Station Fine Sand Silt Loam Silt Clay
Kisii 83 76 76 97 1 00 100 90 90 93
Embu 81 76 81 86 86 81 86 81 86
Mtwapa 74 82 70 93 96 89 89 96 82
Kabete 58 63 72 79 75 92 67 71 79
Nakuru 50 61 50 72 86 75 50 65 70
Narok 18 18 21 29 32 44 18 18 30
85 mm for Kabete. The results are shown in table 5 . Results indicate that
the value of 75 mm was quite correctly selected. Lowering the value in
humid regions (Kisii) or increasing the value for sub-humid regions
(Kabete) did not yield more reliable results.
In order to examine if there exists a shift in onset date with time, cor
relation analysis was performed at an a = 0.05 on the determined onset
dates in each station per criteria. The results are shown in table 6. The
results indicate that there is a noticeable decrease (early) in onset dates
with time at Kabete station, suggesting a climate change in the region
around Kabete. This observed trend could be very important for agricul
ture in the region. Any major shift in the onset of the growing season
would affect the timing of planting and the availability of water supply
during sensitive stages of erop growth.
EVALUATION OF THE ONSET OF THE GROWING SEASON 147
Table 5
Frequency (in percentage) of obtaining no water stress during the initial growth stage
(30 days) as detennined with Caccum criteriafor different values of accumulated
amounts of rain
Accumulated amount of rain
Station 65 mm 75 mm 85 mm
Sand Loam Clay Sand Loam Clay Sand Loam Clay
- - -
Kisii 79 86 86 83 97 90
- - -
Kabete 58 79 67 63 79 71
Table 6
Coefficient of correlation of onset dates with time for each criterion in each station
Station Caccum Cstored Cspells
Fine Sand Silt Loam Silt Clay
Kisii -0. 1 3 0.03 0.04 -0.01 -0.06
Embu -0.03 -0.04 0. 1 2 -0.20 0.01
Mtwapa -0.06 -0.05 -0. 1 5 -0.05 -0.04
Kabete -0.52 -0.52 -0.27 -0.34 -0.40
Nak:uru 0.00 0.00 0.06 0.20 0.28
Narok 0.04 -0.01 0.02 0.3 1 0. 1 1
4. Conclusions
The onset dates of the growing season for six selected stations in the
major climatic zones of Kenya were determined on the basis of three cri
teria. There was high variation in onset dates for the stations studied.
Therefore, the risk that the farmers wish to take should be considered
when planning agronomie activities.
Applying one or another criteria to determine the onset of the growing
season does not guarantee that the erop will not experience water stress.
Analysis of the results indicates that when a erop is planted after the onset
date of the growing season determined by one of the three criteria, the
erop is likely to survive the sensitive initial growth stage without great
damage in the humid regions. In the sub- and semi-humid regions, erop
damage is more likely to occur, but yield depressions might still be
acceptable. In the dry regions the criteria have little meaning, since the
erop has a high percentage of complete failure. Occurrence of water
stress is only slightly influenced by the water storage capacity of the soil
1 48 E. C. KIPKORIR, S. M. GACHUIRI, J. MUKABANA & D. RAES
type, indicating that all the three criteria can be used for various soil
types.
There was no significant difference in water stress arnong the three cri
teria studied, despite Caccum and Cstored starting on average one week
earlier than Cspells. Because of the ease of its computation and its
reliable results, preference is given to Caccum, a criterion that is based on
the accumulated rainfall of FRERE et al. ( 1 990). Sensitivity analysis of the
criteria indicated that the "75 mm" value was quite correctly selected.
There was a noticeable decrease in onset dates with time only at
Kabete station. Therefore there is need for continuous monitoring of the
growing season parameters, especially the onset in the regions under rain
fed agriculture.
REFERENCES
ALLEN, R., PEREIRA, L. S., RA.Es, D. & SMITH, M. 1998. Crop evapotranspiration
- Guidelines for computing erop water requirements. - Rome, ltaly, FAO
Irrigation and Drainage Paper N° 56, 300 pp.
RoBERTSON, G. W. 1970. Rainfall and soil water variability with reference to land
use planning. - Tech. Series no. 1 , WMO/UNDP and Dept. of
Meteorology, Univ. Philippines.
SoMBROEK, W. G., BRAUN, H.M.H. & VAN DER Pouw, B .J.A. 1982. Exploratory
soil map and Agroclimatic zone map of Kenya. - Nairobi, Kenya soil
Survey.
STEENHUIS, T. S., WINCHELL, M., ROSSING, J., ZOLLWEG, J. A. & WALTER, M. F.
by
1. Introduction
Factors affecting the plant-radiation regime are the amount and quality
of incident radiation, canopy architecture, and the optical properties of
leaves and soil (SINOQUET & CALDWELL 1 995). The simplest way to model
how irradiance varies within canopies is to assume that irradiance varies
only vertically ; that is, in one dimension. However, this assumption is
only valid in homogenous canopies, achieved when canopies are closed
and randomly mixed. Consequently, some workers (SrNOQUET &
BONHOMME 1992, DE CASTRO & FETCHER 1 998) have extended the model
ling of radiation attenuation in two or three dimensions in attempts to
produce more representative and accurate models.
Compared to radiation modelling, evapotranspiration modelling is
usually more difficult because it involves a simultaneous mathematica!
analysis of above and below ground spaces (TouRNEBIZE et al. 1 996).
Evaporation from the soil or transpiration from the plants is often estim
ated using the Penman-Monteith (PM) equation (MoNTEITH 1 965), but the
PM equation can only be used to estimate either soil evaporation or plant
transpiration, but not both simultaneously. A recent, important extension
of the PM equation is the Shuttleworth-Wallace (SW) equation
(SHUTTLEWORTH & WALLACE 1985) because it specifies explicitly the ener
gy exchanges at the soil and canopy, thereby making it possible to distin
guish the fraction of water transpired and that evaporated from soil
(FARAHANr & AHV1A 1996). The SW equation is still relatively new and not
extensively used (FARAHANI & AHUJA 1 996), but field tests of this equa
tion have been promising (LAFLEUR & RousE 1990, WALLACE et al. 1 990,
FARAHANI & BAUSCH 1995).
Thus, the main objectives of this study were : 1) to model the part
itioning of captured radiation and evapotranspiration in a maize-sun
flower intercropping system in partial- and fully-grown canopies, and 2)
to deterrnine the degree of complexity that is required for radiation mod
elling.
2. Development of Models
CD; = _n______
k; L; • �o; (1)
I [kj Lj . �oJ
j= l
where k; and L; are the extinction coefficient and leaf area index o f erop
species i, respectively ; n is the total number of crops in the system ; and
s; is the leaf scattering coefficient of radiation for crops species i
(ToURNEBIZE & SINOQUET 1 995) which is 0. 1 4 and 0. 1 0 for sunflower and
maize, respectively (MAsoNI et al. 1994).
where gL(rL) is the leaf normal distribution function which expresses the
probability that a leaf normal is around direction rL, where direction rL is
described by a very narrow normal inclination range (Ji + ael and normal
azimuth range �L + d�L (LEMEUR 1973a, b) ; cos rlr is the eosine angle
1 54 C. B. S. TEH, L. P. SIMMONDS & T. R. WHEELER
between the leaf normal direction rL and sun direction r, and is calculated
by
COS rL r = COS 8 COS 8L + Sin 8 Sin 8L COS (!{') - !i'>L)
and g(rL> / 2Jt 1 cos rL rl dQL is the projection of the part of the leaf area with
normals within the solid angle dQL around the sun direction r.
Consequently, G(r) is regarded as the average projection per unit foliage
area in the sun direction r, where the sun direction is described by inclin
ation 8 and azimuth !{'). The function g(rL) or g(8L, !i'>L) is simplified by
· assuming that the leaf normal inclination 8L is independent of the leaf
normal azimuth !i'>L, so then g(8L, !i'>L) = g(8L) g(!i'>L), where g(8L) is the
•
distribution function for leaf normal inclination, and g(!i'>L) is the function
describing leaf normal azimuth distribution (LEMEUR 1 973a).
G(8, !?>) is corrected to account for two situations : 1) radiation
scattering by leaves, and 2) when leaves are not randomly distributed but
clumped as found in row crops :
G (8, !?>) = G (8, !?>) . �O; . Q (8)
A
where G(8, !?>) is the corrected G-function ; and Q(8) is the dumping fac-
tor (TouRNEBIZE & SINOQUET 1 995 ; CAMPBELL & NoRMAN 1 998, KusTAS &
NoRMAN 1999a, b). Q(8) is determined empirically by
Lc =fc-L
where fc is the fractional canopy cover which is the fraction per unit
ground area occupied by canopy cover and can be approximated by ta
king the ratio of canopy width to row spacing (CAMPBELL & NoRMAN
1 998). Extinction coefficient k is related to the G-function by
k=
G (8, !7>)
cos 8
INTERCROPPING SYSTEMS 1 55
h=
( n ê ce, f/J)j )
j=l cos e Lj
( 1 - p) Io,dr exp - I
where Io.dr is the amount of direct radiation above canopy ; n is the total
number of crops ; and p is the mean canopy reflection coefficient calcul
ated by
( 1 - �a) x
p=
( 1 + �a)
2
( 1 + 1 .6 cos8)
-----
(GOUDRIAAN 1977, 1988). Diffuse radiation within mixed canopies is
calculated by
2:rc"12 [ n ê ce, f/J)j Lj] cos8 sine d8 df/J
j=l
Idf = ( 1 - p) J J B (8, f/J) exp - I
0 0 cos e
and integrated numerically using the 5-point Gaussian method
(GOUDRIAAN 1 988). The B(8, f/J) is the brightness function or the amount
of diffuse radiation component coming from sky direction 8 inclination
and f/J azimuth (CHARLES-EDWARDS et al. 1986).
The total amount of direct radiation Ad, and diffuse radiation Ad! inter
cepted by all crops are calculated by
Adr = fo,dr - fdr
Ad! = lo,dJ - Id!
The amount of radiation captuied by erop species i is then calculated
by
Adr,c, ; = Adr • W;
(2)
Adf,c,i = Adf • W;
where Adr.c.• and AdJ.c,i are the amount of direct and diffuse radiation cap
tured by erop species i, respectively ; and w, is determined from Eq. ( 1).
the soil surface to the canopy top is divided into N, horizontal layers of
thickness E" and N, vertical sections of thickness E,. The horizontal cell
thickness E, needs not be equal to the vertical cell thickness E,.
Fig. 1 . - The canopy space is divided into a network of cells in the 2D model.
The probability (Pk) of total radiation intercepted within the k-th cell
visited by a single bearn is calculated by
[ ( � GAr) PJJk
1 - exp - • • sk • �aj)]
where Wk; is determined from Eq. (1). The height and width of cells were
fixed at 0.2 m in this study (E, = E, = 0.2 m).
As shown in Eq. (3), three properties must be determined for each cell
in the grid network to determine the irradiance within the canopies : the
horizontal and vertical distributions of 1 ) G-function G(r), 2) leaf area
density P1> and 3) distance of beam travel s. Unlike the I D model, the 2D
model used a simpler form of the G-function :
I L/dj
j= I
where Lkij is the area of the j-th leaf of erop species i in the k-th cell ; and
N is the total number of leaves in k-th cell from erop species i
(THANISAWANYANGKURA et al. 1 997). Information about the leaf area
density in each cell are obtained from the plant profile method as
described by STEWART & DWYER ( 1 993). Calculations for the distance of
beam travel are based on simple geometry as described by GuZEN &
GOUDRIAAN ( 1 989) and SINOQUET & BONHOMME ( 1 992).
A beam with the same inclination and azimuth can enter any point on
the cell, thus the number of beams "pushed" into the cell was pre-deter
mined by several trial runs to obtain the minimum number of beams that
can be used without sacrificing accuracy if more beams were used. This
study used two beams to be pushed into a cell for a given sun azimuth and
inclination. In several runs of radiation simulation, the error of using two
beams instead of ten beams did not exceed 5 %.
Each computed se is then substituted into Eq. (3) to determine Pk so that
the mean probability of intercepted radiation for a given beam direction
can finally be calculated (Pk). Direct radiation intercepted by all crops
within cell k is calculated as
Ad<k = (1 - p) • fo.dr • Pk
where Io,dr is the amount of direct radiation above canopy ; and p is mean
erop reflection coefficient. Diffuse radiation intercepted by all crops
within cell k is calculated as
2.4. EvAPOTRANSPIRATION
The Shuttleworth-Wallace (SW) ( 1 985) equation was extended to
include the transpiration from two or more crops and evaporation from
the soil (WALLACE 1997). The energy budget of the system is described in
a series of equations, which are the sum of the various latent heat, sen
sible heat and radiation fluxes (fig. 2).
� �
mean
canopy t
Hc.1 Hc.2
flow Hs
+- -
� �
(T5,es)
Rs
soil Il "
saturated
layer ewrrs>
�s
G s
the system, fust erop, second erop and soil, respectively ; Rn and Rsn are net
radiation fluxes into the system and to the soil, respectively ; G is heat conduc
tion into the soil ; Tr, Tf, 1 , Tf,2, To and Ts are temperatures for the reference
height, fust erop, second erop, mean canopy flow (canopy source height) and
soil, respectively ; er and eo are vapour pressure at the reference height and mean
canopy flow, respectively ; ew(T) is saturated vapour pressure at temperature T ;
raa is aerodynarnic resistance between the mean canopy flow and reference
height ; rc, ls and rc,2s are bulk stomata! resistance for the fust erop and second
erop, respectively ; rc, la and rc,2a are bulk boundary layer resistance of the
canopy for the fust erop and second erop, respectively ; rsa is aerodynarnic resis
tance between the soil and mean canopy flow ; and rss is soil surface resistance.
.,
With some algebraic manipulations, it can be shown that the total
=
latent heat flux of the system with n crops is given by
n
ÀE = � Cc. PMc, i
;
+ C,PMs
-
i l
where
cC,I'
-
-
( . 1/R +
l
-+
Î. l /KJ
j= l,J'' l
J
Rc,i R.
Cs = { l + RsRa / K (R + R.)}-1
-
Rs = (� + Y) r,: + Y r;
Mc.2 + (pcpDo)lr!,·2
� + y ( 1 + rsc,2 / r:·2
ra
Do = D+ -0
- {M - (� + y)ÀE}
pep
r:: (resistance between mean canopy flow and reference height) ; r;·; (bulk
The SW model required several resistance components to be known :
4. 1 . RAf>IATION CAPTURE
Por the l D model, there was a close clustering of points along the 1 : 1
ratio line, and there was no trend of estimation error (fig. 3). The mean
error (simulated minus measured) was nearly zero (0.01 ) and 95 % of
these errors were limited to a narrow range (- 0.09 to 0. 1 1). The more
complex 2D model, however, tended to underestimate when the fraction
al radiation interception was around 0.80 - 0.90 (fig. 3). Mean error was
- 0.04 which indicated an overall tendency to underestimate, but com
pared to the l D model, 95 % of the prediction errors from the 2D model
were limited to a narrower range (- 0. 1 3 to 0.06).
Fig. 3.- Comparisons between simulated and measured fraction of total incident
radiation intercepted, where simulation was by the (a) lD model, and (b) 2D model.
!
Cl> 0.80
�
.Sl
.s
u
c
0
0 .60
�
9 11 13 15 17 19
hour
Fig. 4. -
Diumal fraction of total incident radiation intercepted in the 1 998
experiment on DAS 78.
Table 1
Simulation error of transpiration (mm h·')
Model Intercrop Mean Error Mean - 2SD Mean + 2SD
Component (simulated-measured)
I D model Maize 0.00 - 0.06 0.05
Sunflower 0.00 - 0.08 O.ü7
2D model Maize - 0.01 - O.ü7 0.05
Sunflower - 0.01 - 0.09 O.ü7
SD = standard deviation.
164 C. B. S. TEH, L. P. SIMMONDS & T. R. WHEELER
.s .s
E E
"
N 0.12 0. 1 2
'öj -0 -0
E Q)
1il êii
:; 0.06 :; 0.06
E E
ïn "(ij
0.00 0.00
0.00 0.06 0.12 0. 1 8 0.24 0.00 0.06 0.12 0.18 0.24
(c) (d)
0.60 0.60
1 :1 1 :1
-::-- 0 -::--
.i= .i=
.s .s
.... E 0.40 E 0.40
" Oo cP
0
'§"
�
0
-0 -0
Q)
"' êii 0.20 1il 0.20
:; :;
E E
"(ij '(ij
o_oo 0.00
0.00 0.20 0.40 0.60 0.00 0.20 0.40 0.60
measured (mm h ·1) measured (mm h -1)
Fig. 5. - Comparisons between simulated and measured transpiration for the intercrop
components in the 1998 and 1999 experiments.
ly broader error range than the l D model, and the mean error by the 2D
model was - 0.01 compared to 0.00 by the l D model. This revealed that
the 2D model was slightly less accurate than the l D model. Nevertheless,
both models tended to underestimate transpiration slightly for intercrop
maize and intercrop sunflower when measured transpiration exceeded
0. 15 and 0.40 mm h-1, respectively (fig. 5). And this underestirnation was
slightly larger for the 2D model than the l D model. Prediction by the 2D
model was less accurate than the l D model because, as shown in fig. 3,
the 2D model predicted radiation interception in the intercrop slightly
less accurately than the l D model. It followed that because the 2D model
tended to underestimate radiation when measured intercepted radiation
exceeded 0.80 ; this would also lead to an underestimation of trans-
INTERCROPPING SYSTEMS 1 65
(a) (b)
0.25 � measured
.s .s
0.20
E -e- 20 E
c: c:
0. 1 5
0 0
� �
"ë.. "ë..
g g
"' "'
c: c:
9 11 13 15 17 19 9 11 13 15 17 19
0.25 0. 1 8
J:: .i::
.s .s
0.20
E E
0.12
0. 1 5
c: c:
0 0
� 0. 1 0 �
"ë.. ·g.
g g
"' 0. 06
c: c:
0.05
7 9 11 13 15 17 19 0 6 12 18 24
hour hour
(e) (f)
.s:: .s::
.s .s
E E
c: c:
0 0
� �
"ë.. "ë..
g g
"' "'
c: c:
9 12 15 18 21 24 0 6 12 18 24
hour hour
Fig. 6. Comparisons between simulated and measured diumal transpiration for the
-
intercrop maize on : (a) DAS 7 1 , 1998 ; (b) DAS 78, 1998 ; (c) DAS 95, 1998 ; (d) DAS
44, 1999 ; (e) DAS 48, 1999 ; and (f) DAS 53, 1999.
1 66 C. B. S. TEH, L. P. SIMMONDS & T. R. WHEELER
(a) (b)
0.60
0.60
.c. -+- measured .c.
.s .s
E --e- 1 0 E
0.40
c -20 c
0 0 0.40
� �
·a. ·a.
g g
(/) (/)
c c
0.20
0.20
0.00 0 00
(c) 9 11 13 15 17 (d) 9 11 13 15 17 19
0.60
<=-- 0. 1 6
.c. .l::
.s .s
E E
0.40 0. 1 2
c c
0 0
� � 0.08
·a. ·a.
g g
(/) (/)
c 0.20 c
0 . 04
0 . 00 0.00
7 9 11 13 15 17 19 0 6 12 18 24
hour hour
(e) (f)
0. 1 6 0. 1 6
.l:: 0. 1 2 .l:: 0. 1 2
.s .s
E E
c 0.08 c 0.08
0 0
� �
ëï 0.04 ·a. 0.04
g g
(/) (/)
c c
0.00 0.00
9 12 15 18 21 24 0 6 12 18 24
hour hou r
Fig. 7. Comparisons between simulated and measured diurnal transpiration for the
-
intercrop sunflower on : (a) DAS 7 1 , 1 998 ; (b) DAS 78, 1 998 ; (c) DAS 95, 1 998 ; (d)
DAS 44, 1 999 ; (e) DAS 48, 1999 ; and (f) DAS 53, 1999.
INTERCROPPING SYSTEMS 167
(a) 2.5
� 2.0
.s
E
1.5
c
0 -e- 1 0
�0 1.0
--e-- 2D
0.
� 0.5
70 80 90 1 00
DAS
(b)
2.0
�
-+- measured
1 .5
'O
.s
E
c 1 .0 -e- 1 0
0
g_
� --e-- 2D
0.5
�
40 45 50 55 60 65
DAS
Fig. 8.
- Comparisons between simulated and measured daily soil evaporation for the
(a) 1998 experiment, and (b) 1999 experiment.
0.69 mm day·1). This showed that though the I D model was, overall, more
accurate than the 2D model, 95 % of the estimation errors by the former
model occurred in a slightly wider range than by the latter model.
5. Conclusion
1 3-27.
CHARLES-EDWARDS, D. A., DoLEY, D. & TuoRPE, G. M. 1 986. Modelling plant
growth and development. - New South Wales, Academie Press.
els for maize - bare soil to closed canopy. - Trans. ASAE, 38 : 1 049-1059.
FARAHANI, H. J. & BAUSCH, W. C. 1995. Performance of evapotranspiration mod
20.
GouDRIAAN, J. 1 977. Crop micrometeorology : a simulation study. -
Wageningen, Simulation monograph, Pudoc.
Bio. Plantarum., 19 :
KucERA, J., CERMAK, J. & PENKA, M. 1977. lmproved thermal method of contin
-
ua! recording the transpiration flow rate dynarnics.
413-420.
KusTAS, W. P., CHOUDHURY, B. J., KUNKEL, K. E. & GAY, L. W. 1989. Estimate of
KusTAS, W. P. & NoRMAN, J. M. 1 999b. Reply to comments about the basic equa
Meteorol. , 94 : 275-278.
tions of dual-source vegetation-atmosphere transfer models. - Agric. For.
135-153.
LEMEUR, R. 1973a. Effects of spatial leaf distribution on penetration and inter
ception of direct radiation. - In : SLATYEAR, R.O. (Ed.), Plants response to
clirnatic factors. Proc. Uppsala Symp., 1 970 (Ecology and conservation, 5),
UNESCO, pp. 349-356.
LEMEUR, R. 1973b. A method for simulating the direct solar radiation regime in
MASONI, A., ERcou, L., MARIITTTI , M. & BARBERI, P. 1 994. Changes in spectra!
Plantarum. , 91 : 334-338.
properties of ageing and senescing maize and sunflower leaves. - Physiol.
Ross, J. 198 1 . The radiation regime and architecture o f plant stands. - The
Hague, W. Junk.
Ross , J. & NrLSON, T. 1 965. Propuskanye pryamoi radiatsii solntsa
855.
272.
by
1. Introduction
* Advisor, Cerro del Cubilete, No. 302, C.P 04200, México D.F. (México).
176 M. FRANCISCA NARANJO
There was a time when the current level of the Mexican basin was
located at the bottom of the sea. This area was inhabited by fish, ammon
ites and micro-fauna of the Cretaceous era. Nevertheless, the current sur
face was developed during the Cenozoic era. Throughout this era, while
the floor level was emerging by rising higher and higher, there was a
repeated sequence of eruptions that completely buried the old seabed.
The Mexican basin was exactly in the middle of the formation known as
the Trans-Mexican Volcanic Belt (TVB) (Memoria de las obras... 1 975).
Currently the tectonic plate of the TVB continues in an emerging pro
cess : every day we get closer to the stars.
We can see that there are common traits in all the TVB, and these same
natural phenomena were the ones that synthesized a sole vision of the
cosmos in the pre-Hispanic period. We will examine the special case of
the Mexican basin. It is known that the current basin was an open basin
where a series of free open flows, during different periods, were in
contact with the south and north water flow. Here and there the emergen
ce of new volcanoes could enclose water in the same areas, but in gen
eral, it was an open basin.
The north mountain range emerged in the late Miocene. At the end of
the Pleistocene era and at the beginning of the Holocene the most relevant
geologie happening for the basin took place : the emergence of the huge
wall of the south that encloses the basin as a damned door. During the
same period also the construction of the great volcanoes took place.
Now let us talk about the lake system. Before it closed, the drainage of
the basin was from north to south, therefore it is natural that its deepest
part would be in the south during the period of 32,000 to 22,000 years .
ago. There is another period of very superficial waters from 22,000 to
1 8,000 years ago and a low level, but not as much as in the above
mentioned period from 1 8,000 years ago. We also have to consider a
CHANGE IN THE HYDROLOGICAL MANAGEMENT 177
personalities were mixed with theirs. They believed that mountains were
containers full of water and that during rain season they liberated that
water, while during dry seasons they kept it. They believed that the inter
ior of the earth was humid, that cavems communicated with an aquatic
paradise inhabited by an aquatic god. They believed in a world that at first
sight would appear completely mythical, but after examining the hydro
logie performance the basin presents, the millenary observations of a
natural world included the complex religious and mythical structures.
In fact, the mountains of this basin are full of water. They keep inside,
as a lava pitcher, the atmospheric humidity and make it emerge, all year
long, through its pores. It is true, the inside of the earth is humid, almost
aquatic, the cavems present this humidity and sometimes they present an
underground world, broad and connected, magnified by legends, myth
and a cosmovision of another world ; an underground heaven populated
by gods, nurturing and destiny (HEYDEN 1 983, EsPINOSA 1 960).
Things in Spain were different. The geographic location and in partic
ular the politica! position from the last third of the 1 5th century, created
the proper climate for science, technologies and art to receive very differ
ent influences. The early times of the Spanish Renaissance were divided
between two tendencies : one influenced by the Dutch countries
(Flandes) and the other with Italian influence : modem and Renaissance.
The first influence stood for a Catholic world, for realism and everyday
reality. The second influence cultivated beauty, form and the ideal (TovAR
Y DE TERESA 1 979).
Other European spiritual currents had a decisive influence on the
Spanish culture. The rejection to accept the modem Italian influence with
a modem and scientific vision of the world was explained by the persis
tency of religion, the scholastic and the satisfactory attitude towards the
glorious past of the Middle Ages that concluded with the definite
cortquest. This was different to what happened in Italy, which in the past
was seen as a nightmare, especially by liberal thinkers who were awaken
ing and returning to the people the faith in themselves by incorporating
and reconciliating the Christian religion with the thinking of the classic
ancient time (TERRERO 1 972).
The people lost in the medieval religious thought consoled themselves
with the idea of being the centre of creation. One day Copernicus
demonstrated that the Sun did not rotate around the Earth, but that the
Earth rotated around the Sun. At that moment, mankind lost its sense of
location and was converted into a Universe vagabond. From this moment,
humans started to doubt everything : the values that were believed to be
CHANGE IN THE HYDROLOGICAL MANAGEMENT 179
permanent, the moral values that were thought to be statie. The above
mentioned are just some examples. All these changes made eastem men,
as it was pointed out by Hausere, fall into a crisis of which, up to this
time, they have not been able to recuperate and that is reproduced in
modem times (TOVAR y DE TERESA 1 979).
In 1 492 the conquest ended and a force of national pride that could not
be defined until decades after, emerged. This is the reason why Spaniards
in the late 1 5th century showed, on the one hand, the thinking and
conception of the world at the end of the Medieval Era, and on the other
hand, the conceptions of an Arabic (Mudéjar) world. Bebind these pers
pectives, the Renaissance conception had infiltrated and was finally
adopted.
In the lak:e system there was not only the City of Mexico-Tenochtitlan
but other cities as well. In the lower regions of the system, as it was the
eastem part, there was Lak:e Texcoco, which was salty, while the fresh
water lak:es were located in the southem region. The part of the lak:e
where Mexico City was located was salty. Lak:e Zumpango, in the north
east region, had fresh water, whereas Lak:e Xaltocan, also in the northeast
region, was salty (GmsoN 1 97 1 ).
When reading a hydrologie map, the first impression of the fluvial
sources is a very closed net of water threads that fulfil the river plain in
large amounts. In general, they are stationary flows or affluents of a lar
ger flow.
In the aforementioned thinking schema, where the world is conceived
according to cycles and close to the spherical shape, the handling of the
basin was structured. lts function started in the hills, where earthen dams
controlled the water speed. These structures were also used to irrigate the
terraces, to raise fish, to stop erosion, to facilitate water infiltrat1on, but,
above all, they were scattering regulators of speed and energy. Tuis func
tion was very important due to the fact that the most intensive farming
was next to the lak:e (NARANJO 200 1 ).
There were doek causeways, aqueducts that crossed the lak:e. When
locating these areas on a map we can appreciate a pattem that irradiates
from the lak:e to the outer sides : to comrnunicate with all the population
that settled on the shore of the lak:e. The expansion was from the water to
the mountains, providing access to a broad spectrum of resources. Even
the intensive erop system was made by germinating the seeds on the
1 80 M. FRANCISCA NARANJO
6. Conclusion
REFERENCES
Cosro VILLEGAS, D. 1 985. Historia Genera! de México (T. 1). -:---- Colegio de
CORTES, H. 1 963. Cartas de Relación. - México, Editorial Porrua.
México A.C.
DIAZ DEL CASTILLO, B. 1963. Historia Verdadera de la Conquista de la Nueva
Espafia. - México, Editorial Porrua.
EsPINOSA PINEDA, G. 1996. El embrujo del lago. - U.N.A.M.
GrnsoN, C. 197 1 . Structure of the Aztec Empire, Handbook of Middle American
lndians (Vol. 1 0). - Austin, Texas, U.S.A, University of Texas Press.
HEYDEN, D. 1983. Mitologfa y simbolismo de la flora en el México prehispánico.
- México, Instituto de Investigaciones Antropológicas, U.N.A.M.
KuBLER, G. 1983. Arquitectura Mexicana del siglo XVI. - México, F.C.E.
LEvr, E. 1989. El Agua segun la Ciencia. - México D.F, CONACYT.
,
LLANAS FERNANDEZ R. 1994. Evolución de la Ingenieria Sanitaria y Ambiental en
México. - México, D.D.F.
LoPEZ AusTIN, A. 1989 (1" Ed, 1 973). Hombre-Dios. Religión y politica en el
mundo Nahuatl. - México, U.N.A.M., Instituto de Investigaciones históri
cas, Serie de Cultura Náhuatl, Monograffas, 15.
Nacional, México.
NARANJO PEREZ DE LEON, M. F. 1 989. Revista El Acueducto, México, l.M.T.A.
NARANJO PEREZ DE LEON, M. F. 200 1 . Le cycle hydrologique et les travaux d'irri
gation du Mexique préhispanique. - In : Origines et Histoire de
!' Hydrologie. Colloque International (Dijon, France, 9- 1 1 mai 200 1).
SAHAGUN, F. B . (de) 1985. Historia Genera! de las cosas de la Nueva Espafia. -
México, Edit. Porrua.
TERRERO, J. 1972. Historia Genera! de Espafia. - Barcelona, Espafia, Ed. Ramón
Sopena.
TovAR Y DE TERESA, G. 1 973. Pintura y escultura del Renacirniento en México. -
México, l.N.A.H.
Second International Conference
"Tropical Climatology, Meteorology and Hydrology"
Proceedings edited by
G. Demarée, M. De Dapper, J. Alexandre
pp. 187-198 (2004)
by
1 . Introduction
The climate change issue bas initiated an increasing need for empirica!
knowledge on natural climate variability. Climate description of the early
instrumental period is. hampered by the poor coverage outside Europe in
the early 1 9th and 1 8th centuries. Among the many blank spots is Japan,
where the official meteorological network started in the 1 870s.
8501 (Japan).
1 88 G. P. KONNEN et al.
The Dejima data were taken under responsibility of the Dutch govern
ment. Usually the Dejima medica! doctor in charge, or his assistant, took
the observations. The Dutch observations in Japan continued in the Dutch
Hospita! after the opening of the Japanese Empire. In the 1 870s, the of
ficial Japanese meteorological network was established. Nagasaki
Observatory was operational in 1 878 and overlaps by a half year with the
Russfan
Consulate Station readings
. --
Hakodate -
1 859-1 862 1 872
Fig. 1. - Pre- 1 900 availability of meteorological data in Japan. Black : official meteo
rological stations. Grey : data recovered in this study. The early Hakodate data were taken
at the Russian Consulate.
There is a 1 4-month gap between the two series ; the von Siebold series
bas a one-year gap (Nov. 1 825 - Oct. 1 826). The data of Cock B lomhoff
and von Siebold are in the von Siebold coUection (voN S IEBOLD, undated).
The observations were usually taken three times a day, referred to as
"morning", "midday" and "evening". Although there is no explicit men
tion of the corresponding observation hours, we could reconstruct them
from the two existing lists of two-hourly observations during full days in
the von Siebold documents as being 6:00, 1 2:00 and 22:00 LT. For Nov.
1 826 - Sept. 1 827, there are six observations a day. The close agreement
1 90 G. P. KONNEN et al.
between the mean diumal temperature course in these von Siebold data
and the modem Nagasaki data strengthens the credibility in the quality of
the von Siebold series.
1845-1863, Dejima Series : these are the official Dutch observations at
Dejima. The observations were taken by four observers in total ; the last
observer was the medical doctor J. L. C. Pompe van Meerdervoort ( 1 857-
1 862). The observations include temperature, pressure, humidity, rainfall,
wind direction and force (4-point scale), cloud cover and (before Oct.
1 848) cloud motion. The observation hours were basically 6:00, 9:00,
1 5 :00 and 22:00 LT. The 1 845- 1 856 data are published by STAMKART
( 1 85 1 ), and in the KNMI Yearbooks 1 855 and 1 856. The data 1 852/53
were reported missing in the Yearbook, and been lost for more than a cen
tury, hut were recently accidentally recovered from a forgotten KNMI
archive at Rotterdam Harbour. The 1 857/58 data are in PoMPE VAN
MEERDERVOORT ( 1 859a,b).
Dejima observations after 1 858 have not been located, although it is
known that Pompe van Meerdervoort continued observing till his depar
ture in 1 862 (GEERTS 1 875, General State Archive 1 857- 1 866). If Pompe
van Meerdervoort did carry his observations back to Holland, then they
were lost forever as he was shipwrecked (hut survived) on his return trip.
1871-1878, Nagasaki Hospita/ : the Nagasaki Dutch Hospital was
founded on 20 September 1 86 1 . The Hospita! was located 500 m south
east of Dejima (fig. 2). The reservoir of the barometer was at a height
37 m above sea level (GEERTS 1 875, KNMI Yearbook 1 875). The obser
vations were probably discontinued after Pompe van Meerdervoort's
departure in Nov. 1 862, but were resumed in Nov. 1 87 1 by the Hospital
pharmacist A. J. C. Geerts and continued from 1 874 by the . doctor
W. K. M. van Leeuwen van Duivenbode. The latter left Nagasaki in March
1 879 (Japan Daily Herald). Presumably he was the last Dutch doctor.
The daily data for Nov. 1 87 1 - Dec. 1 877 are in the KNMI Yearbooks
1 875- 1 877. The daily readings for 1 878 have not been published, hut
monthly averaged values of pressure and temperature for 1 878 are in the
KNMI Yearbook 1 878. The observation schedule is basically 7:00, 1 2:00,
1 8 :00 LT.
The Hospital pressure data exhibit a sudden downward jump of about
4.5 mb on New Years day 1 873 ; from 1 Nov. 1 874 onward (the first day
published in KNMI Yearbook 1 876) the pressure regained the old level.
The obvious explanation is that the observer or KNMI failed to apply the
height correction to the 1 873-74 readings.
JAPANESE INSTRUMENTAL METEOROLOGICAL OBSERVATION SERIES 191
Fig. 2 . - Location of the observation sites 1 8 1 8-2000 in Nagasaki. The coastline of the
early 19th century has also been indicated.
3. Data Homogenization
The 1 857-58 pressure data show after all corrections an erratic behav
iour, which we were unable to correct. We consider this behaviour as
indicative of a malfunctioning of Pompe's barometer during the entire
year and excluded his data from our summary graphs. figure 3 shows the
time series of annual temperature values 1 820- 1 999 ; figure 4 shows the
time series of annual pressure values 1 827- 1999.
19
18
�
:l
1;j 1 7
....
Cll
Q.
E
i!
16
15
1 820 1 840 1 860 1 880 1 900 1 92 0 1 940 1 960 1 980 2000
Fig. 3. Time series of Nagasaki annual mean temperatures 1 878- 1999 with its exten
-
sions back to 1 820. The black line is from Nagasaki Observatory ( 1 879-1999) ; the grey
line is from the data recovered in this study. The thick light-grey line represents a 20-year
smoother (CLEVELAND 1979).
1 94 G. P. KONNEN et al.
1 01 8
1 01 7
î 1 01 6
�
:::i
"'
� 1 01 5
a..
1 01 4
1 01 3
1 82 0 1 840 1 860 1 880 1 900 1 92 0 1 940 1 960 1 9 80 2000
Fig. 4. - Time series of Nagasaki annual mean pressures 1 878- 1998 with its exten
sions back to 1 827. The black line is from Nagasaki Observatory ( 1 879-1999) ; the grey
3 . 3 . HOMOGENEITY TESTS
ring in the period 1 935- 1 957. The SNHT, Buishand Range and Pettitt
tests point to homogeneity breaks in 1 957, 1 947, and 1 936, respectively.
None of these dates correspond to a relocation of Nagasaki Observatory.
lt remains to be investigated whether these breaks are real inhomo
geneities or whether they are related to the increase of the urbanization
effect in the Nagasaki record.
ACKNOWLEDGEMENTS
found a first indication of the existence of the 1 857 data. J.B. Wijngaard (KNMI)
1 96 G. P. KONNEN et al.
ran her homogeneity testing procedures on the data. T. Brandsma (KNMI) cal
culated the smoother over the data and M.H. Kaltofen (KNMI) digitized the
Dejima Documents.
The Bochum von Siebold collection is owned by the Ruhr University
Bochum, Germany, Department of East Asian Studies. P. D. Jones is supported
by the US Department of Energy, Atmospheric and Climate Research Division,
under grant DE-FG02-86ER60397.
NOTE
[ 1 ] Dejima was a small artificial island of 190 m length and 75 m width in the
bay of Nagasaki, a short distance ( 1 2 m) offshore (fig. 2). lt was a trade fac
tory of the Dutch East India Company during the period 1 64 1 - 1 86 1 . During
that period, it was the sole gateway for trade and science between Japan and
the western world. lt is also referred to as Decima, Desima, Deshima,
Desjima, Dezima. We will refer to it as Dejima, which produces almost a
correct pronunciation if read by a native English speaker. The Dutch Dejima
staff consisted of only 8 men, augmented by a handful Javanese servants.
Dejima was visited each autumn by a Dutch trade fleet. The Japanese gov
ernment required that this fleet went back by 1 November at the latest.
REFERENCES
Climatol" 6 : 661-675.
ALEXANDERSSON, H. 1 986. A homogeneity test applied to precipitation data J. -
the late 1 8th and early 19th centuries. - Int. J. Climatol. , 22 : 1 1 19-1 1 42.
Madras (Chennai) mean sea level pressure using instrumental records from
J. Hydro!" 58 : 1 1 -27.
BrnsttAND, T. A. 1982. Some methods for testing the homogeneity of rainfall
records. -
J. Climate, 6 :
CHENOWETH, M. 1993. Nonstandard Thermometer Exposures at U.S. Cooperative
Weather Stations during the Late Nineteenth Century. -
1787- 1 797.
Climatic Records of Japan and the Far East Area 1954. Air pressure, Air
temperature, Amount of precipitation. - Central Meteorological
Observatory, Tokyo Japan (available from library KNMI).
DEMAREE, G. R. & MIKAMI, T. 2000. Some XVIl-th and XVIIl-th century Dutch
meteorological observations at Deshima, Nagasaki (Japan), seen as a com
plement to Japanese climatological historie documents. - In : Proceedings
of the International Conference on Climate Change and Variability (Tokyo,
JAPANESE INSTRUMENTAL METEOROLOGICAL OBSERVATION SERIES 197
J. Climate, 1 1 : 2325-2339.
sions of the Southern Oscillation Index using early lndonesian and Tahitian
meteorological readings. -
J. Climate, 16 : 1 1 8-
TsuKAHARA, T. 2003. Pre- 1 872 extension of the Japanese instrumental
meteorological observation series back to 1 8 19. -
131.
MIKAMI, T., ZAIKI, M., KoNNEN, G. P. & JoNES, P. D. 2000. Winter temperature
reconstruction at Dejima, Nagasaki based on historical meteorological doc
uments during the last 300 years. - In : Proceedings of the International
Conference on Climate Change and Variability (Tokyo, Metropolitan Univ
ersity, 13-17 Sept. 1999), International Geographical Union, pp. 103-106.
ONNEN, P. L. 1 844. Meteorologische waarnemingen te Buitenzorg op het eiland
Java. - Nieuwe verhandelingen der Eerste Klasse van het Koninklijk
ScHUEPP, M., B URNETT, F. M., RAo, K . N . & RouAUND, A. 1 964. Note on the stan
dardization of pressure reduction methods in the international network of
synoptic stations. - Geneva, Switzerland, WMO Technica! Note 6 1 ,
\XT1\,f()_l\Tr. 1 <; Ll 'T'P '7Ll W,...rlrl l\APtPf\rf\J ,... n ir<>l nrcrnni7:itinn 4() nn
1 98 G. P. KONNEN et al.
par
1. lntroduction
SEGALINE NIETO & CABRE 1988, RAMIREz et al. 1 995, RoNCHAIL & GronA
1 998).
Avant toute étude d' une longue série de données pluviométriques, il est
nécessaire de vérifier si celle-ci correspond à un ensemble homogène ; en
202 A. GIODA, J. RONCHAIL, Y. L' HOTE & B. POUYAUD
300
250
200 '� -
1 50 .,,,,,,,.
'
" /'.....
....i...
.. ...
�
1 00 ' I
-- /
50 r---... ,_ î/
....... "'
- __.,,.,.
-�-
0
--.... �
Q; Q; � � �
"§;
-
--,
.öi c: <::l e e
e e
"5
--,
c:
"" "'
� < � ·:;
--,
·:;
0
<
.c
E
.c .c
E
.c
E
"' IL Q)
J5
ë.
�
0
Q)
>
0
z
""
Q)
tl
0
ou ECRi est l' écart-type centré réduit de la pluie de l' année i ; Pi, la pluie
de l' année i ; Pm, la pluie moyenne sur N années ; E, l' écart-type des
pluies sur N années.
1 1
111 7 ,......_
1 890
1 892
,__-
"ll -1 '-
0 0 "Tl
1 894
'(Ï)'l>
1 896
� !il: �
o5 1.r
r
1 898 ._
� !!. i· �
�.......
1 900
z e: "' _ _
-- 1 1 1
1 902 ....� -...._
...,
;- �
<D .... .... �
...-- � �
"'O Dl 1 906
3 1 908
0- a ""
�� 1 1 1
1 9 1 0 - ..,. r--1---
(il
.... 1 9 1 2 - r- -.;
Q. ç
.'-
.... 1914 -- ---
�
!il 1 91 6
,r
0-
3
..-1
1918
"'.<!:: ;.- �
� 1 1 1
qci � 1 920 .,.
__....--
J � -
w 1 922 I> __... - �
� ,,,,
1 924
E. .....__. 1 1
1 926
111
-- - - -
1 928 ..... ....
';:s: - ;.;;:::o y
('Il
,(
�� t:
1 930 ::!:....
"'
"ll "ll S:: ) ....::::r
� -Tl 1 1 1
P>' 1 932
�-=i � <::.. l
t""' . . )> 1 934 -
P> � ....
� ....... �1 1 1 1
1 936
c Q- �
��
��
)> - .... 1 938 .... __.
§ ...... [ .......
1 940 11
Vl
{
!!.C> 1 942
D>-- !t �
� t--
1 944
� ..-
.,,
n D> �
7 �
0
1 1
1 946 - - -
s
� 1 948 17' �.
g
r-... "-
1 950 j \r- -- .
v ./ .,,.,.,.-
1 952
§ 1 954
...... r-... --
('Il, 1 956
('Il --- - '-
"'
-�
1 958
0
1 960
� _.. .,,,.
/ -- "'-.
i:l
1 962
- --
11 --
0.
1 1
1 964
�·
oo 1-- " --
('Il
1 966 ....
)
?' "ll en 1 968
�
1 972
' ........,,
.., en
�
s:: <D 1 974 i.r 1
-- 1..... --
)>>!. 1 976 ,
<:::
Dl 0-3
�
1 978
�r êi 1 980 -
;, �
:,...
n
1 982
��
/ -- - -
1 984
-- �-
v
1 986 �
�
2, 3 - � - -
::::: 0- 1 988
v �..... -
1 990
lot" < ,,. - -
1 992
1 994 -
.....
....,
..
1 996
1-- 1 998
c.. ' --
"'n""
observée, qu' il s' agisse de l' année complète ou des trois principales
saisons.
Les totaux annuels selon les années hydrologiques et les valeurs cor
respondantes du SOI sont illustrés sur un graphique à double échelle
d' ordonnées (fig. 4). On note une correspondance visuelle entre certains
des forts déficits de précipitations et les SOI les plus faibles ( 1 939-40 et
1 982-83, FRANcou & PIZARRO 1 985). Toutefois, cela n' est pas systéma
tique.
Le régime pluviométrique bolivien montrant une saison des pluies
unique, il est logique d' étudier les précipitations annuelles selon un
découpage préservant cette caractéristique. Par conséquent, nous consi
dérerons une année hydrologique comportant une saison des pluies suivie
par une saison sèche. Cette année commencera en septembre pour s' ache
ver en août de l' année suivante.
Les grandes oscillations du SOI - tout au moins les épisodes El Nifio
- sont généralement centrées, elles aussi, sur la fin d' année. Cette
approche en années hydrologiques «septembre-août» ne tronçonne pas la
période de référence des épisodes El Nifio. Elle est donc pertinente pour
relier ces derniers avec les précipitations. Pour les épisodes La Nifia, qui
suivent ou précèdent les épisodes El Nifio, ce découpage peut parfois ne
pas être aussi satisfaisant car il peut segmenter les effets d'un épisode La
Nifia parfois centré sur les mais de juin et juillet.
Une corrélation directe entre les indices de précipitations centrées
réduites et le SOI se révèle généralement décevante à toutes les échelles
de temps. Nous pouvons cependant dégager une tendance en classant les
précipitations annuelles centrées réduites selon les valeurs croissantes des
SOI correspondants (RoNCHAIL 1 995 , 1 998).
Cela a d' abord été fait à l' échelle annuelle ou les meilleurs résultats
sont obtenus avec les valeurs des années hydrologiques (fig. 5). Cette
figure montre qu' aux forts El Nifio (à gauche) correspondent le plus sou
vent des précipitations annuelles déficitaires. Cette tendance dépasse en
moyenne un demi écart-type. Elle est donc significative. L' examen des
situations La Nina (à droite) fournit un résultat original puisqu' elles sont
également associées à des précipitations souvent aussi déficitaires que
...... N (.o) � (JI O> ....... 00 <O
0 0 0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
1 890-91
.... -
1 892-93
1 894-95
'\
1 896-97
�
1 898-99 ,-
"
::!1
ë 1 9�1
-�
16' 1 902-03
11 1"-
-- -
1 904-05
�
.'"'>
1 906-07
f>- 1 908-09
i'6' �
1 9 1 0-1 1
i.--- i.---
["'
1 91 2-1 3
./
"
1 91 4- 1 5
0 --
1 91 6-1 7
§
......
1 91 8-1 9
,__
c:: 1 920-21
� t
, -
1 922-23 ... -�
"' 1 924-25
I»' C/)
�-..
1 926-27 - ......
t""' Q 1 928-29
I» C/) / -
�'
1 930-31
_,,.,.
),. ,
1 932-33
---
êl �
1 934-35
(/)
1 936-37
-
�
1 938-39
�
(") 1 940-41
-- �
:>< 1 942-43 --
1
.--
...�
8 1 944-45
......
-
..._
� 1 946-47 ....
(/) 1 948-49
0
" -
0
-<
....... 1 950-51
- ,......,
î
1 952-53
en 1 954-55
...... -
o,
Q �... i--
0 C/)
1 956-57 ...
�
�
�
"'
Er
1 958-59
:::>'"
�
�
'< 1 960-61
_,,...
'
1 962-63 ,-
1
g.
0 1 964-65
'\._ I '\.
<e. 1 966-67
.0 -...
c:: 1 968-69
0
"' 1 970-71
"'\
.g 1 .I
-
"' 1 972-73
L
1 974-75 -
s
�
1 976-77
0
" - ,.....
1 978-79 "
"" "
a "il
1 980-81
�
Il" ë 1 982-83
I» ëiï
-?
0 1 984-85
C::> ::J
"'
1 986-87
::J r--- �
c:
.!
1 988-89
!!. .....,_... �....
1 990-91
�,_.... �-
_,,..
1 992-93
_.,_
1 994-95 - -
1 996-97
1 998-99
.r --
' '
......
' _ N _ .,.. (JI O>
(.o) N
o (.o)
0 0 0 0 0 0 0
0 0 0 0 0 0 0 0 0 0
0 0 0
dans le cas des épisodes El Nino. Ce sont les situations moyennes ou nor
males - c' est-à-dite «ni El Nino ni La Nina» - qui sont associées à des
excès de précipitations, en moyenne tout au moins. En effet, pour ces
années-là, on note des excès ou des déficits de précipitations qui peuvent
atteindre un ou deux écarts-types. Ce dernier résultat de notre analyse
n'est pas propre à la station de La Paz et peut être obtenu pour !' ensemble
de l' Altiplano andin (VmLLE 1 999), y compris le péruvien, dans la zone
au nord-ouest de la Bolivie.
Sur cette figure 5, sont aussi reportées en vert clair les valeurs du SOI
de l' année précédente. La courbe de tendance est globalement croissante,
ce qui montre que les années à SOI négatif sont généralement précédées
par des années de même caractéristique. L' inverse est aussi vrai ; les
années à SOI positif sont elles-mêmes précédées par des années à SOI
positif. Cette constatation montre à l' évidence que la durée des événe
ments El Nino et La Nina est généralement supérieure à une année.
Un graphique comparable (non figuré ici) a été réalisé cette fois en
classant, en ordre croissant, les années hydrologiques selon leur hauteur
de précipitations centrées réduites. La courbe de tendance du SOI de
l' année examinée et celle de l' année précédente ne montrent rien de
significatif.
3,00 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1. 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
2 00 1 1 1 1 1 1 1 1 1 1 1 1 1 •
l • • �
1
1 1 00 \ ) I ,,, \ )i L� .-1
'
/
1I I - � I - 1 �\ 1
\} � A ' 0
1 \ / "\ \\
0,00 ll Il en
Aj I
{ � Il \ l
:v I
·
- 1 00 "'l"'H I Î � / \, \ -
1
if \J ' � r �
1
l�IJ I JI \ �
'� i/\���\· · \ �" 2:i
�" i \ v '
1
-2 00 î� � .
•
�
r::
en
-3,00 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1 1
.. ....
-4 , 00 ........ .. _.__
.. ....__._
.. ...._....__._ _. _._
.. _._...__... . ........
... _._
.. _.__....__._
.. ...._....__._ _. _._
.. _._...__... . ........
... _._
.. _.__....__._ .. -f
.. ...._...._... . �
M W N OO M m W � M m OO N O � � O W O M V � N N OO � � w � N M � � m � w o N OO � m v w m oo w v
� N m m m � � m � � � � � m 1 N 1 m q N � q � � � � � � m N � 1 m M m � � M � N q � m � � �
� � � � � � � � � � � � � � � é � � � � � o � � � g � � � � é � � � � � � � � � 8 � � � � �
m m m m m m m m m m m m m m m m m m m m m m m m m m m m oo m m m oo m m m m m m m m m m m m m
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �
-- SOi An-1 Hydro S-A - Plu ie annuelle -SOi Hydro S-A
1 - Poly. (Pluie annuelle) - Poly. (Pluie annuelle) - Poly. (SOi An-1 Hydro S-A) 1
Fig. 5. - Pluies (en valeurs centrées réduites) et SOi à La Paz - San Calixto (années hydrologiques : septembre à août), années classées par
SOi croissant.
N
0
\0
N
,_..
0
4,00
3,00
?'"
2,00 I� 9
0
i,. 7 �
1 ,00 J 1 � J & ' �{, :-
I ' -
'
I I IV lt... �
i..
��
0,00 ' Il ' IJ J I )) I � Il l\i "'
l'I ,
1
J
" \\ l.-Il 1 � '
!'\ 7 Î
·- ·� 1 v �
'\ 'I lf 1 1
' 1 1\
\ u
J j I
-1 ,00
.
v
,, \ r'I .•
.,
1
!
v
\' \ I \v l/
1
:-<
�
r:
1 ...- � \ -v\ \
-2,00 j
�
R<>
O:I
1/
-3,00
N � � � g m N � � m N M ID M � N � N @ O ID � � � m � m � � � � N M � m � N � ID � � � 2 � �
��
� m m � m � � � � m m � � � m m � � m � � � m m � m � � m m m � � m � m � m m m � � m m �
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � �
g
�
[ -so1 M-A - Pluie SOND - SOi S-D -- Poly. (SOi M-A) - Poly. (Pluie SOND) §
Fig. 6. - Pluies (valeurs centrées réduites) à La Paz - San Calixto. Quadrimestre de septembre à décembre, années classées par SOi SD crois
sant.
). _(,)
� 8 8 -8 8
� c..>
"' � 0 "" _.f>.
<!Cl 8 8 8 8 8
...J
,,.
1 983 ..
11
....""
1 905
' �
["' l - :...._
1 992
"'"""'-'" "'
Q. 0
0 "U (/) 1 987
l'
1 941 -
? ':< · - 1 978
"'
f(") :2
- ....
1 966
(/) -...:i �
c: 1 1 990
ëi) 0 - r::::t' -
�
1 994
l
s::
c... ..'....
o, 1 970
)>..
"
0
"' 1 981
..., -
(1),
a I
.... 1 973
- �
1 959
ii
-?
� 1 995 ....
--
I»' 1 991
�
" .7'
1 953 "'
� 1 9n
1 933
v �-
§
en 1 946
1t
--
1 988
(") 1 944
...... .......
� i'..
!=)'
..__
>:;• "U "U 1 986
o ... ....
- -c:
"'-
'< -·
1 957
• CD
tO 1 954
c:
0 s::
- c... ......._
I>'
- )>
(/) " 1 936
i 1 903 -
�
1 892 - - --
g-
(/) 1 955
1 "
0 1 951
-
l
1
....... 1 964 -
......
1 934
�· - ,.._
�
1 938
I»'
1 982
_...
-tf!...-
1 935
F: --.ia.
1 922
l(")
;r-.-..
1 975
-
11
"' 1 996 - -
�
1 923
�
�
"' -
"'
Q. 0
1 927
� _.
"U (/)
"' 1 943
'O
_i..-- i..-
1910
ei
':< -
-
\
en
1 898 ..... 1-... .._
"U c...
[. ):.
0 1 976
t---.
- _i..- i---
1 939 �
> �,_
-...___ .?"
c...
s::
1 925
ê.
)>..
"
1 ...... '
1 989
"'
� ,
.... 1 91 7
-
1 971
i,...""
107.d.
l lZ 3'1V11.LSilV NOI.LVTIIJSO
N
...-
N
4,00
3,00
l'I
2,00 j ?'
a
� -'"" 0
• � 0
1 ' 00 , "" � L
j 1 }>
•
/ I ' 1 :-
I'
0,00 l ' J�
u
� � h �. � N ' \) �
,. /. ' 1 ,, �v l\ z
1 ' � � i.... n
,_ ll7 J
i/ IJ v
"
-1 ,00 1� f f'. \! , I v
i.-' 1 �
-2,00 ,...
� .J �
:-<
r:
:i::
-3,00
?,<>
�
to
-4,00
� N O N • M � ID - � m ID m m N oo O • O M m � N ID O � OO M M O m N M N O N � oo - M • - oo m ID ID OO O
o oo • � m m ID • � N m � • � m - M • oo oo N m N ID O OO N M • � m • o � N ID • � � � � oo oo o � - M �
m m m m m m m m m m oo m m m m m m m m m m m m m o m m m m m m m m m m m m m m m m m m m m m m m
� � � � � � � � � � � � � � � � � � � � � � � � N � � � � � � � � � � � � � � � � � � � � � � � 8
�
--SOI J-A - Pluie MJJA - SOi M-A §
1 - Poly. (Pluie MJJA) -- Poly. (SOi J-A) - Poly. (Pluie MJJA)
1
Fig. 8. - Pluies (valeurs centrées réduites) à La Paz - San Calixto. Quadrimestre de mai à août, années classées par SOi MA croissant.
OSCILLATION AUSTRALE 213
dépendantes des valeurs du SOI, dans ses phases positives aussi bien que
négatives, que les pluies observées durant le reste de l' année.
Ces tendances ont été chiffrées en calculant, pour chaque quadri
mestre, les hauteurs de la pluie centrée réduite qui correspondent,
d' abord, aux 6 classes suivantes des valeurs de l' indice SOI : <- 1 ; - 1 à
- 0.5 ; - 0.5 à 0 ; 0 à + 0.5 ; + 0.5 à + 1 ; > + 1 . Ensuite, le découpage
adopté correspond aux quatre classes plus amples suivantes : <- 1 ; - 1 à 0 ;
O à + l ; > + 1 (tab. l).
Tableau 1
Moyennes des précipitations (en valeurs centrées réduites) par quadrimestre et pendant
l'année hydrologique (septembre-août) pour différentes classes de ! 'indice SOi
* NIN O, El Nifio fort ; Nifio, El Nifio moins prononcé. Idem pour NINA et Niîia.
<-1 NIN O
m
�
-1< <0
--
0< <+ l
>+ 1 1 NINA
* NINO, El Nifio fort ; Nifio, El Nifio moins prononcé. Idem pour NINA et Nifia.
N
......
VI
216 A . GIODA, J. RONCHAIL, Y. L' HOTE & B. POUYAUD
6. Conclusion
REMERCIEMENTS
NOTE
de lecture y sont grandes. Enfin, elle fut fermée dans les années 1950. Seule,
1882) mais la qualité des données est moindre ; les lacunes et irnprécisions
BIBLIOGRAPHIE
420.
FRANcou, B. & PIZARRO, L. 1985. El Nifio y la sequfa en los altos Andes centrales
(Pérou-Bolivie). - Bull. Inst. français Etudes Andines, 16 : 1 - 1 8 .
OSCILLATION AUSTRALE 217
KhronoStat version I .O. 1 998. Logiciel développé par l'Orstom UR 2-2 1 , l'Ecole
des Mines de Paris et l'UMR GBE de l' Université Montpellier II.
LUBES-NIEL, H., MASSON, J. M., PATUREL, J. E. & SERVAT, E. 1 998. Variabilité cli
matique et statistiques. Etude par sirnulation de la puissance et de la robus
RAMIREz, E., MENDOZA, J., SALAS, E. & RrnsTEIN, P. 1 995 . Régimen espacial y
-
RoNCHAIL, J . 1 995. L'aridité sur l ' Altiplano bolivien. Sécheresse, 6 : 45-5 1 .
RoNCHAIL, J. 1998. Variabilité pluviométrique . lors des phases extrêmes de
In : Serninario El
RoNCHAIL, J. & GrooA, A. 1998. San Calixto de La Paz : las lluvias y las fases de
la Oscilación Austral durante cerca de un siglo. -
by
Achim DREBS *
1. Introduction
Rautanen used this definition all through his observations. The number of
days fits well into the general behaviour of the parameter. The results are
presented in table 1 .
Table 1
Mean number of days, when the amount of precipitation exceeded the thresholds
� 0.0 mm, � 0. 1 mm, � 1 .0 mm, � 1 0.0 mm (Olukonda, 1905 - 1926,
without the years 1 9 1 6, 1 925)
Jan. Feb. Mar. Apr. May June July Aug. Sept. Oct. Nov. Dec.
2: 0.0 mm 1 8.3 1 5.4 1 3.4 6.9 1 .7 0.5 0.2 0.4 2.4 5.9 9.7 1 5.2
2: 0. l mm 10.5 1 0.4 8.8 3.7 1 .0 0. 1 0. 1 0.7 2.3 5.1 8.9
;::: 1 .0 mm 9.0 8.9 7.1 3.1 0.7 0. 1 0.3 1 .6 4.0 6.7
;::: 10.0 mm 3.1 4.0 2.6 0.7 0. 1 0. 1 0. 1 0.3 0.9 2.4
The Olukonda observations showed that some rainfall could have been
observed during the whole year ; the possibility to observe precipitation
amounts ;;:: 0.0 mm during wintertime (May to August), however, was
very small (tab. 1 ). The main rainy season occurred from November to
March with its peaks in January and February when more than three days
with precipitation (more than 9.9 mm) could have been observed. High
daily precipitation sums had different effects depending on the season. In
the early rainy season when the ground was definitely dry the run-off
could have been instantaneous and flooding occurred regularly. During
the growing season heavy rainfall could harm erop and livestock. The
orography form of the Olukonda area is like plateau and too much rain
fall in connection with high temperature had a great impact on the health
of the inhabitants in this area, too. The occurrences of malaria are con
nected to the combination of these two factors.
The next step was to calculate the mean values for the precipitation
sums for this period. After the first measuring period 1 884- 1 894 the sta
tion network was closed down and reopened in this area around
1 904/ 1 905. Therefore there were no stations to enlarge the data series.
The number of days with precipitation indicted that there was only one
rainy season situated in the southern summer hemisphere with its max
imum in January and February. To specify this result the mean monthly
precipitation sums for the period were calculated. The mean monthly
precipitation sums are shown in table 2. Table 3 shows the absolute high
est and lowest monthly precipitation sums for the period 1 905 to 1 926.
Table 4 shows the absolute highest daily amounts of precipitation.
222 A. DREES
Table 2
Mean monthly precipitation sums (mm)
(Olukonda, 1 905- 1926, without the years 1 9 1 6, 1925)
Jan. Feb. Mar. Apr. May June July Aug. Sept. Oct. Nov. Dec.
84.4 107.7 73.8 26.4 3.2 1.1 0.0 1 .9 10.3 29.8 66.8
Table 3
The highest and lowest monthly amount of precipitation (mm)
(Olukonda, 1 905- 1926, without the years 1 9 1 6, 1925)
Jan. Feb. Mar. Apr. May June July Aug. Sept. Oct. Nov. Dec.
28 1 .9 272.7 232.2 166.2 40.3 1 6.4 0.2 35.6 6 1 .3 80.0 223.2
0.8 32.6 6.9
Table 4
The highest daily amount of precipitation (mm)
(Olukonda, 1 905 - 1 926, without the years 1 9 16, 1 925)
Jan. Feb. Mar. Apr. May June July Aug. Sept. Oct. Nov. Dec.
83.5 1 80.5 6 1 .5 1 36.5 1 5.4 1 6.4 0.2 1 6.5 6 1 .2 59.0 1 14 . 1
Table 5
Rainy season (August-July) precipitation (mm) and its difference
to the mean rainy season precipitation (%)
(Olukonda, 1 905- 1926, without the years 1 9 1 6, 1 925)
Year mm % Year mm %
- -
1 905/1906 595.6 147.0 1 91 5/ 1 9 1 6
- -
1 906/1 907 503.9 1 24.4 1 9 1 6/ 1 9 1 7
1907/1908 343 . l 84.7 1 9 1 7/ 1 9 1 8 48 1 .5 1 1 8.8
1 908/1909 858.5 2 1 1 .9 1 9 1 8/ 1 9 1 9 148.3 36.6
1 909/ 1 9 1 0 533.7 1 3 1 .7 1 9 1 9/1920 303.0 74.8
1 9 1 0/ 1 91 1 1 82.3 45.0 1 920/192 1 7 10.0 175.2
191 111912 585 . 1 144.4 1 92 1 / 1 922 1 84.5 45.5
1 9 1 21 1 9 1 3 379.7 93.7 1 92211923 417. 1 1 02.9
1 9 1 3/ 1 9 14 390.8 96.4 1 923/1924 409.4 1 0 1 .0
- -
1 9 1 4/19 1 5 340.2 84.0 1 924/1 925
areas are marked in figure 1 . The right solid arrow indicates the date
when the 1 0-day moving average exceeds the 2 mm-threshold (December
9), this pointing the beginning of the vegetation period. The left solid
arrow shows the end of the vegetation period, when the 1 0-day moving
average feil below the 2 mm-threshold (March 1 9) . The duration of the
vegetation period is therefore 1 0 1 days on average. In this area of the
world, vegetation reacts immediately to small amounts of water, so the
definition of vegetation period is somewhat a statistica! one. The second
period marked with dotted arrows indicates a period with high amount of
daily precipitation (3.5 mm, 1 00 mm/month). During this period there is
a risk that heavy precipitation events harm the agricultural production.
This risk period started on average on January 3 1 and lasted until March
7 (36 days). The absolute maximum occurred on February 7.
5.0
4.5
4.0
3.5
3.0
E 2.5
E
2.0
1 .5
1 .0 1
0.5 1
" "
..., <(" ;;j_ en"' eng- ö ö z
0.0
� � "' .r.h
..." "... ..." u.."' u.."' ::;:" ::;:"
.0 .0
c. c.
<(
,., ... "" ...,"" 3
... ,., ..., 3
..., 3 C> C> Q_ >
0
>
0
z
u
"' 0�
0
�
7 ::;: � 7
..:. ;::;; .... cD "' r.h o
<} 9
"' "' - "' r.h ("')
d> "' ("') ""
- ..:.
"' o - "' "' .;,
"' cD "' d> "" ..:. r;:;
Table 6
Mean monthly precipitation sums (mm)
(Olukonda, 1 884-1 894 (calculated))
Jan. Feb. Mar. Apr. May June July Aug. Sept. Oct. Nov. Dec.
88.2 1 33.5 95.8 39.0 1 .2 0.0 3.1 14.0 45.5 73.4
Ottweiler pointed out that the northem part of South-West Africa was
less represented in his studies and more and denser observations were
needed to support his results. A quick look at Rautanen's diaries for this
· period indicates that Ottweiler was right. There is no single month with
complete observations. The number of precipitation observations made
varies between 40 and 90 %, for some other meteorological parameters
the number of observations was higher. Precipitation was first observed
by weighing a container made of glass. Rautanen wrote down the results
in grams. From 1 887 he used mm as the unit. Due to his duties Rautanen
left his missionary post quite often and the precipitation measurements
were probably summed over several days. For the monthly amount this is
not a problem, but crossing the end of a month made his results uncertain.
An uncertainty is the loss due to evaporation. This more-than-one-day
sum did not include a correction due to evaporation and should lead to
an underestimation of precipitation. In genera!, however, Ottweiler
EARLY WEATHER OBSERVATIONS IN OLUKONDA 225
5. Conclusions
1 60
1 40
1 20
1 00
-
�
m 80 - -
m �
60 - -
40 - - -
20
0
Jan
-
Feb
-
H
Mar
-
Nov Dec
• mean 1905-1926 • mean Ottweiler
800
700 //
600 //
E
E
500
400
� �á � �$$�
W##�ff/ff/a W�#ff/ff/M
w���
300
200
1 00 • -
1 1
�
0
IO r- 00 °'
s::: 00 0 N tri IO r-
-
0 0 0 0 s - s:! "'
- ::!; � � - � N N N
"'
N N N N
s � "'
°' °' °' °'
�' � @ - � � - .._
- @ � °' - °'
- °'
- � -
.._
°' °'
@
°'
- °'
- -
.._
°'
Fig. 3. - Rainy season precipitation (mm) : Olukonda, 1 905- 1926, without the years
1 9 1 6, 1 925, compared to Ottweiler rainy season precipitation mean.
ACKNOWLEDGEMENTS
REFERENCES
Original precipitation observation diaries 1905- 1 926, without the years 1 9 1 6 and
1925. Hamburg, Archives of the Finnish Missionary Society (in German)
and DWD.
by
1. Introduction
2. Hydrogeological Setting
I
}'
/'v
Geological fault
Nam La river
"_"_" Catch. boundary
� Doline/Depressio
� Sinkhole/Cave
,- ...
.I
Fig. 1 . - Geological sketch of the Cao Pha Pass area.
� "
"
s:: �"'"''t-- "=
�
:i:
:t
'$ '# - $ ,s t
F
::J � -lilt
•
' • •,. Hang Doi Resurgence
� �= =<l!! C=a=
::;.,_�r== o=P•
ha.�!���� =.==. ="••••• :::;> ••t====
Nam L a River "
..
.7'
•'
"
•'
,.
,.
•'
� " "$
•'
1 n-k -)( -
Cyx (k) = - I (
n i: 1
Yr - Y Xr+k - X ) (4)
n
l
a, = -JI
n i:1
(x, - x)2 _
(5)
23 6 V. THANH TAM, F. DE SMEDT & 0. BATELAAN
l n
ay = J -
n
� (y, - y)
t= I
_ 2
(6)
Note that the cross-correlation function is not symmetrical [rxy (k) =F ryx
(k)], i.e. if rxy (k) > 0 for k > 0, the input series x, influences the output
series y,, while if rxy (k) > 0 for k < 0, the output influences the input. The
delay time, defined as the time-lag between k = 0 and the occurrence of
the maximum rxy (k), gives an estimation of the peak impulse response
time of the system.
70
·•·••·•·· Sinkholes
--
60
Resurgence
'
'
50
1
i1
1 '
1
:1Il
20
10
o ������=--�-�--::::::::
:: � :: =:::�
: �:l:::�
::: '....:...
..__ .,_ __ ,_:
01/1 5/00 03/15/00 05/14/00 07/13100 09/1 1/00 1 1/10/00 01/09/01 03/10/01 05/09/01 07/08/01 09/06/01
Time (mm/dd/yy)
The data for the analysis are hourly streamflow hydrographs for the
period l st October 2000 till 8th October 200 1 , as presented in figure 3 .
Two automatic reading loggers recorded the data, one capturing the total
discharge upstream of the sinkholes and the other the discharge of the
resurgence. The data series were smoothed by a Tukey filter (JENKINS &
WATTS 1 968) to overcome bias in the deterrnination of the cross-correla
tion function. The result of the analysis is shown in figure 4.
The resulting cross-correlation function (fig. 4a) diminishes slowly
with increasing time-lag and reaches a zero value at a time-lag of 69 days,
implying that the duration of the impulse response of the system is quite
ESTIMATION OF UNDERGROUND RIVERS 237
- - --
,:;:i, ...... -
.... -.031 ·""'
�· .0224 .....
"
_,,, .1567 ·"''
..., ""'
· 0038
,
....
.
," \.\.
' '---
0.00
'-"""
-' �
"
0.05
...,�
,_
0.10
�����--'
0.15 0.20 0.25 0.30 0.35 0.-40 0.45 0.50
-
(a) (b)
Fig. 4. - Cross-correlation function (a) and cross-amplitude function (b) of the stream
flow of the Nam La sinkholes and the resurgence.
238 V. THANH TAM, F. DE SMEDT & 0. BATELAAN
In order to :find out how the high flooding level is related to the pan
ding in the Nong Lua dolines, the difference of the total inflow and out
flow, during the period 1 2 till 1 5 July 2000 and caused by a total rainfall
of 80 mm, is calculated as about 1 .4 1 06 m3• Tuis volume of surplus water
was filled in the dolines and the Bom Bay depression. A digital elevation
model (DEM), with a spatial resolution of 20 m, developed on basis of
the topographic map of scale 1 :50,000 and using the method described by
TuRcaITE et al. (200 1 ), was used as a baseline data for the filling process.
The water ftlling is an iterative process with the constraints : (i) the mini
mum water depth filled in the Bom Bay depression is 2 m (taking into
account the field observation that during high-stream discharge periods,
this depth of water is ponding in the depression but no groundwater rose
in the dolines) and its maximum elevation is 1 m lower than that of the
Ban Ai sinkhole (accounting for the estimated streamflow head loss
downstream to Bom Bay), and (ii) the maximum elevation of the water
column filled in the dolines is 2 m lower than that of the Bom Bay depres
sion (accounting for the estimated minimum head difference between the
water column in the depression and that in the dolines). These conditions
also warranted that the difference of the stream water level in the Ban Ai
sinkhole and the possible highest elevation of the water column filled in
the dolines is 3 m, an estimated minimum head loss for water flowing
through the underground conduit from the Ban Ai sinkhole to the Nong
Lua dolines. The GIS ARC/INFO 8.01 package was used to calculate the
water volume ponded in the dolines/depression given an initial trial water
column of 2. 1 m in the Bom Bay depression. Por each subsequent level
an increment of 0. 1 m of water column was used. The iteration process
was continued until a reasonable match with the surplus water volume
was obtained. The filling definitively showed that during the period of
interest the highest ponding water level is 1 3 m above the bottom of the
Nong Lua dolines, which is approximately equal to the flooding traces
left on the rock walls in the doline. The estimated maximum water pon
ding volume of approx. 1 .0 1 07 m3 using the àbove-mentioned upper
filling constraints thus shows how important these dolines are in flood
mitigation of the study area.
Entrance
En trance
.; -
tm
tm
(a) (b)
Fig. 5. -
(a) Vertical profile (upper) and horizontal profile (lower) of Queen Cave ; the
two straight lines represent geological faults. (b) Vertical profile of Nong Lua Cave ; the
arrows show the direction of groundwater flow (the figures are adapted with permission
of SPEKUL, Speleoclub of the K.U.-Leuven).
ESTIMATION OF UNDERGROUND RIVERS 24 1
conveyers (TAM et al. 200 1 ) . In this study four cave profiles are analysed
to deduce the possible underground flowpaths from Ban Ai to Nong Lua
and from Bom Bay to Hang Doi.
A relatively comprehensive study on the cave structure and develop
ment in the study area revealed that most of the caves developed in differ
ent geological stages, resulting in the multi-step cave profile as seen
today (DINH 2001). This is illustrated in the typical cave development pro
file of the Queen cave (fig. 5a) ; the two fault systems, one in NW-SE and
the other in SW-NE direction, drive the cave development and hence the
groundwater follows the same directions. The stepping-down (or shaft)
segments of the cave resulted from the crossing of the two fault systems.
The change of cave development from the NW-SE to SW-NE direction is
due to the fact that the SW-NE fault system occurred after the NW-SE
system. Such a typical cave development pattem explains why the begin
ning and end of a cavem conduit can link surface watercourses located at
largely different levels, as in the study area.
Because of limited cave expedition data, the connection between the
Bom Bay sinkholes and the Hang Doi resurgence, as well as between Ban
Ai and Nong Lua, cannot be completely visualized (fig. 6). However, if
/'v
__ Oeological ti u lt
Nam L a ri ve r
_ _ _
Catch ment bou ndary
Cave development
-- and OW flow direction
e Sinkhole/oa...,
Fig. 6. - Plan of cave development and groundwater flow direction in Nam La study
area.
242 V. THANH TAM, F. DE SMEDT & 0. BATELAAN
the first part of the cave develops in the direction of the resurgence, then
it is regarded as an indication for a possible existence of a connection. In
case of the Nong Lua cave (fig. 5b ), an inflow conduit is located at 463 m
a.s.l, 79 m below the ground surface, and the main outflow conduit is at
435 m a.s.l, 107 m below the ground surface. The strearnflow disappear
ing in the Ban Ai sinkhole is at 572 m a.s.l, and likely steps down, along
an about 2.5 km path, to the Nong Lua inflow conduit. Consequently, it
drops down along the cave shaft, at the bottom of the doline, to the main
outflow conduit, which is thought to lead to the main resurgence at Hang
Doi. The conveying capacity of the Nong Lua main outflow conduit is
likely lirnited to 20-25 m3/s, which means that whenever the inflow
exceeds this capacity, the groundwater rises in the shaft and emerges
above the cave entrance, resulting in the flooding of the Nong Lua
dolines.
Finally, a schematization of the underground river system is shown in
figure 2, with the following assessment :
- The Nam La streamflow follows three major underground rivers : (i)
from Ban Ai to Nong La and further to Hang Doi ; (ii) from Cao Pha
to Hang Doi ; and (iii) from Bom Bay to Hang Doi ; all emerge in
Hang Doi resurgence.
- The functioning of these rivers is dependent on the discharge regime.
If the Nam La strearnflow is less than 25 m3/s the Ban Ai - Nong Lua
- Hang Doi underground river is the major drainage path, while the
Cao Pha - Hang Doi and the Bom Bay - Hang Doi are in that case of
minor importance. However, if the strearnflow exceeds 25 m3/s, then
the latter flow paths become more important.
- The Nong Lua dolines play an important role in temporally storing the
Nam La strearnflow during extensive rainy periods. It is therefore
recommended that the dolines should be under protection.
Two tracer experiments were carried out by the end of a rainy season
to verify the above-described flow paths. The first tracer test was execu
ted with 2 kg of Sulforhodarnine B (Cz1H29N2Na01S2) and injected in the
Ban Ai sinkhole ; the second tracer test was executed half an hour later
with 2 kg of Uranine (C20H10Na2Üs), injected in the Cao Pha sinkhole. For
both tracer tests sampling was done during four days at the Hang Doi
ESTIMATION OF UNDERGROUND RIVERS 243
7. Conclusions
ACKNOWLEDGEMENTS
This work has been carried out within the project A32 10 "Rural development
in the mountain karst area of NW Vietnam by sustainable water and land man
agement and social learning : its conditions and facilitation (VIBEKAP)" funded
by the Flemish University Council (VLIR). The authors are grateful to all
VIBEKAP's participants for their contributions. Special thanks are paid to the
former project coordinator, Mr Koen Van Keer, for his warm encouragement and
comments on this study.
244 V. THANH TAM, F. DE SMEDT & 0. BATELAAN
REFERENCES
LAROCQUE, M., MANGIN, A., RAzACK, M. & BANTON, 0. 1 998. Contribution of cor
89.
4) : 225-242.
YEvrnvrcH, V. 1 972. Stochastic Processes in Hydrology. - Fort Collins, CO,
Water Resources Publications, 302 pp.
Second International Conference
"Tropical Climatology, Meteorology and Hydrology"
Proceedings edited by
G. Demarée, M. De Dapper, J. Alexandre
pp. 245-25 1 (2004)
by
Gaston R. DEMAREE *
the Gumbel (or GEV l ) probability distribution. Based on this evidence, the IDF
curves are then built using three different methods.
the late Franz Bultot ( 1924- 1 995). It was also F. Bultot who published the
out special studies in statistica! climatology. This office was headed by
The ordinary rain gauge used for daily readings is of the "Miniagri"
type that was quite common in the Belgian Congo and in Ruanda-Urundi
of the recipient measures 400 cm2 and its rim is at 65 cm above ground
[Rwanda, Burundi] . This model has been used since 1 9 1 1 . The opening
receiving area is the same as for the ordinary rain gauge. The rim of the
uced by Richard in Paris. M. Frère has improved the instrument.· The
recipient is at 1 .50 m above the ground level. Tlie rainfall chart is chan
ged at 08 h 00 TML. However, in the case of heavy precipitation, the
chart may be changed between 08 h 00 and 09 h 1 5 at the latest.
Consequently, the daily time span corresponding to the recording rain
gauge extends between 08 h 1 5 , the time of the provoked siphoning on
the day of its placement, until approximately 08 h 1 5 the next day.
Therefore, the possible differences between daily precipitation amounts
of the non-recording rain gauge and from the recording rain gauge should
not be forgotten. The time step used in the manual processing of the prec
ipitation charts is 1 5 minutes. Therefore, the precipitation amounts prod
uced by the recording rain gauge need to be seen as clock-time data
(CRABBE 1 97 1 ) .
However, all recording rain gauges of the network of the
Meteorological Service, except one, were of the Hellniann-Fuess type.
248 G. R. DEMAREE
The receiving area measures 200 cm2• The precipitation charts were pro
cessed with a 1 5-minutes' time step.
Monthly and annual means and its extremes of precipitation (in mm) at
Yangambi-Km 5 ( 1 93 1- 1 980)
January February March April May June July August September October November December Year
January February March April May June July August September October November December Year
�
0
z
Mean 9 9 14 15 15 15 13 14 16 19 18 13 1 69
Max. 23 16 20 23 22 20 21 23 24 28 25 24 236
Min. 3 4 7 6 9 8 8 6 8 11 9 3 1 16
><
1
8
Table 3
(/)
Largest and smallest monthly and annual extremes of the daily precipitation amounts (in mm) at Yangambi-Km 5 ( 193 1 - 1 980)
�
January February March April May June July August September October November December Year
Largest 82 1 12 89 1 12 146 81 1 70 1 26 1 23 1 34 98 1 19 1 70
Smallest 3 11 12 22 16 8 8 16 17 14 25 8 3
N
�
'°
250 G. R. DEMAREE
The data used in this study are twofold : (a) the monthly and annual
maximum of precipitation depths expressed in mm in 1 5 , 30, 45, 60 and
in mm from the ordinary rain gauge. These data were available for the
long-term reference period 1 950- 1983. The data have been provided by
Mr Marcel CRABBE and were produced in manuscript form by Mr Bernard
Totiwe T'Essabe (TOTIWE T'EssABE 1 985) from the "Institut National
pour l'Etude et la Recherche agronomiques, Réseau d'Ecoclimatologie",
Yangambi, to whom both our sincere gratitude is expressed here.
The reference period used in this study is quite exceptional for the
tropical region of Centra! Africa. Therefore, the establishment of the IDF
relationship for precipitation in a station of the Centra! Region of Congo
is of great importance from both the theoretica! and practical points of
view.
REFERENCES
77- 1 00.
BuLTar, F. 1 950. Cartes des régions climatiques du Congo Belge établies d' après
les critères de Köppen. - Bruxelles, Bureau Climatologique, Publications
de l' Institut National pour l' Etude Agronornique du Congo beige (INEAC),
Communication n° 2, 15 pp. + 1 planche.
BuLTar, F. 1 956. Etude statistique des pluies intenses en un point et sur une aire
au Congo Belge et au Ruanda-Urundi. - Bruxelles, Bureau Climatolo
gique, Publications de l' Institut National pour l'Etude Agronornique du
Congo beige, Communication n° 1 1 , 90 pp.
INTENSITY-DURATION-FREQUENCY CURVES 25 1
162 pp.
RUWET, A" SENGELE, N" AGANA, P. & TOTIWE, T. 1985. Paramètres moyens et
extrêmes principaux du climat des stations du réseau INERA, T. 1 . -
République du Zaïre, Institut national pour · l'Etude et la Recherche
Agronomiques (INERA), Section de Yangambi, 1 99 pp.
TOTIWE T'EssABE, B. 1 985. Intensité maximale journalière des pluies en mm -
Cote udométrique maximale journalière en mm enregistrée à Yangambi de
1 950 à 1 983. Régime de l'élément. - Yangambi, Congo, Institut National
pour l'Etude et la Recherche agronomiques, Réseau d' Ecoclimatologie.
VANDENPLAS, A. 1 943. La pluie au Congo Belge. - Mémoires, Institut Royal
Météorologique, Vol. XVI, 1 30 pp. + cartes.
Second International Conference
"Tropical Climatology, Meteorology and Hydrology "
Proceedings edited by
G. Demarée, M. De Dapper, J. Alexandre
pp. 253-265 (2004)
by
.
l = --
a ( T) . (2)
b (d)
a (T, p l )
i= (3)
b (d, p2)
where p 1 and p2 are two vectors of parameters. The estimation of the IDF
curves results in the estimation of those parameters. Figure 1 is an
example of a set of IDF curves for the station of Uccle/Ukke l where the
vectors p 1 and p2 each have two dimensions.
Figure 1 shows that the IDF relation forms a group of parallel decea
sing curves. The intensity decreases with the duration and increases with
the return period. In figure 1 , the denorninator of the relation (3) bas been
chosen equal to
b (d) = (d + 8)� (4a)
where 8 and rJ are two parameters to be estimated. Tuis function b(d) is
purely empirica! and will be considered again in this paper. On the other
hand, variants of the relation such as
b (d) = (d � + 8) (4b)
have been used in DEMAREE ( 1 985) and others.
Return period(T)
(frequency)
l Intensity of
precipitatfon
T o 200
T - SO
T • 10
duration (min.)
° 1 2 9 4
10 10 10 10 10
Fig. 1 . - Example of a set of IDF curves for precipitation at the station of Uccle/Ukke l
(Belgium).
INTENSITY-DURATION-FREQUENCY CURVES 255
a ( T) = ( �)
Fy-1 1 _ (6)
(7)
where Xo and a are respectively the parameters of location and scale (to
be estimated). In that case :
a ( T) = a { : ( ( �)) }·
- 1n - ln 1
.
- (8)
a( D = À 1/J +
[- 1n (1 - _!_)r- 1
T ( 1 0)
x
Thus, in this latter case, there are three parameters (À , 1/J and x) to
estimate. An introduction tb extreme value theory and its applications is
found in BEIRLANT et al. ( 1 996).
Let us fix a duration d. We have then a sample of n data points (in our
case, we have 33 annual maximum precipitation depths for different dura
tions at the station Yangambi-Km 5, Congo - see Part I of this paper)
coming from the random variable /(d) : Xi. Xz, . . . , x". These data are sup
posed to be independent. We can test this assumption by computing the
autocorrelation function for each duration. Figure 2 represents the graph
of the autocorrelation function corresponding to the duration of one hour.
It is observed that the values of the autocorrelation function are small in
absolute value (close to zero) . This means that the series of the annual
0.8
0.6
0 .4
0.2 *
• •
" •
•
"
--02
•
•
--0.4 �--�--�---'----'---'
0 10 12
Different probability distributions can fit our data, assuming that the
other durations with the same conclusions.
( 1 1)
where again x0 and a are respectively the location and scale parameters.
To estimate these parameters, the Kimball estimators are used, which are
an approximation of the Maximum Likelihood (ML) estimators (KlMBALL
1 956). Table 1 contains the numerical results of the Kimball type estim
ates x03 and a3 for the Gumbel distribution function of the annual max
imum precipitation depths (expressed in 0 1 mm), corresponding to diffe
rent durations at the station of Yangambi, Congo.
Table 1
Numerical results of the parameter estimates for a Gumbel distribution function corres
ponding to the different durations at the station of Yangambi, Congo
15 4 1 .0 248.2
30 64.9 369. 1
45 89.2 454.9
60 1 12.7 5 1 1 .4
1 20 1 39.9 594.4
The Kolmogorov-Smirnov and the x2 (or Chi-square) tests are two non
parametric tests that are well known and whose objective is to test if a
given sample comes from a distribution fixed in advance. The assumption
that the annual maximum values of rainfall amounts fallen in Yangambi
follow a Gumbel distribution must be checked. To test this assumption
the Kolmogorov-Smirnov and the x2 tests werercarried out for each dura
tion. The results of these tests are given n:n able 2. lt is noted that the test
258 B . MOHYMONT & G. R. DEMAREE
statistic never exceeds the lirniting 95 % value and this for each of the two
tests. Consequently, the assumption made is not rejected and the assump
tion according to which the data follow a Gumbel distribution for each
aggregation time is accepted.
Table 2
Numerical results for the Kolmogorov-Smirnov test and the test for the annual maximum
values of the rainfall depth for the different durations at the station of Yangambi, Congo
1.
- Gumbel fitting for each given duration. Tuis step has been done in sec
tion 2 and the results, the estimated parameters, are shown in table
- Estimation of the quantiles for el!ch given duration. Suppose that the
data follow a distribution F. The quantile x*T having a return period T
is defined by the value x* that verifies the expression
F(x*) = 1 _ _!_ ( 1 2)
T
( 1 3)
l" * T,d
X *T
= d ( 1 4)
expression ( 1 5). To perform the regression one rninirnizes the obj ective
function given by the expression ( 16). Relative errors are preferred ins
tead of absolute errors in order to give an "equal" weight to each error
associated with each experimental quantile.
IDF rurves oblained by the lhree stages model lor the station of Y�ambi
10' �---��---���--���
10'
f
"
�
1:
"'
� 10'
10· ������-����
,� ,� 1� 1�
OURATICN IN MINUTES
Fig. 3. - IDF curves for precipitation at Yangarnbi, Congo, obtained by the classica!
three-step method.
m = jI,= I nj
k
kKW =
12 (- --
I, n · r· - m +
k
1 )2
( 1 7)
m(m + 1 ) j=I ' ' 2
where the ranks rj1 are assigned to the valuesyjl (with respect to the global
sample). The average rank of the nj values of the j-th group is denoted by
rj. For a suffi.ciently large number of groups and a sufficiently number of
elements in each group, the sampling distribution of kKW is well approx
imated by the x2 distribution with a number of freedom equal to (k - 1 ).
If the computed value of kKW is equal or larger than the critica! value of
the X2 distribution for a previously determined significance level and a given
number of freedom, then the null hypothesis Ho of equal distributions
may be rejected. However, it appears from the relation yj, = ij1b(dj) that the
ranks rj1 will depend on the numerical values of the parameters 8 and 71 of
the function b (d). The result is that the estimation problem is reduced to
rninirnizing the test statistic kKW as a function of the parameters 8 and 71.
A derivative free numerical search technique is most appropriate for
solving the problem. The rninirnizing technique can be speeded up when
only results in the region of higher intensities are needed by using only a
part of the data values of each group instead of the complete series. In this
application only the highest half of the intensity values for each duration
were used.
Using the optirnized values 83 and 713, all values of yj, = ij1b(dJ form a
unique sample. Selecting a probability distribution (in our case the
x0
Gumbel distribution function) and an appropriate estimation technique
(in our case the Kimball technique), the parameters 8 and of the func
x�,
tion a (T) are then estimated.
Finally, the estimates i�d(a3, 83, 713) for the extreme intensities are
given by the expression :
· 3 (a3, xo,
z T,d
"l n -
o 3 , 11 3) -
a (T, a3, x�) ( 1 8)
b (d, ()3, 713)
262 B. MOHYMONT & G. R. DEMAREE
The IDF curves following this two-stage methodology for the rainfall
data of the station of Yangambi, Congo, are given in figure 4. The estim
ated parameter values are given in table 4.
lDF curves obtained by the two slages model lor the slation ol Yangambi
10' �--����--���-�
10'
10'
r - aio
T - 50
Ε 1 0
î·2
10· ���-�����-��-�
10' 10'
DURATICN IN MtNUlES
Fig. 4. - IDF curves for precipitation at Yangambi, Congo, obtained by the two-step
method.
- n1 - 0. l 2
T - ( 1 9)
jl l - 0.44
is assigned. So, for each data value we have a triplet of numbers (i11, �1,
d} On the other hand, for a specific form of a (T), the modelled intensity
is given by
(20)
(21 )
In our case, the specific form of a (T) is again given by the equation (8).
The overall mean squared relative error is given by
e2 = -1
:S -1k
! e/
k j= I nj 1= 1
n
(22)
Table 4 gives the estimated parameter values for the three methods and
the mean squared relative error between the estimated values of the IDF
IDF curves otxa.ined by the one stage model lor the Sation ol Yangambi
10· �--���----��-�
10'
"
�
�
�
�
�
en
10'
OURAT/00 IN MINUTES
Fig. 5. - IDF curves for precipitation at Yangarnbi, Congo, obtained by the one-step
method.
264 B. MOHYMONT & G. R. DEMAREE
curves and the empirica! quantiles. Although the three methods are based
on totally different philosophies, they provide very similar results. In all
three cases the mean relative error between the function values of the IDF
curves and the empirica! quantiles is of the order of 3 to 5 %. The one
step technique has the smallest mean squared relative error while the esti
mated parameter values for the three techniques are numerically very
similar. In fact, the difference in the graphs corresponding to the three
techniques is hardly discemable. Note that the methods in one or in two
stages do not try to minimize the relative error between the curves and the
empirica! quantiles ; they have in fact totally different objective func
tions.
Table 4
Method a3 xî. 83 7/ 3 e
When compared to the IDF curves for precipitation for the station of
Uccle/Ukkel, Belgium (DEMAREE 1 985), it is noted that the slope of the
Yangambi IDF curves is larger than the Brussels ones, which is caused by
the short duration of high intensity rainfall. Furthermore, the Yangambi
4. Conclusion
The application of two different statistical tests has shown that the
annual maximum values of precipitation in Yangambi, Congo, follow a
Gumbel distribution function for the durations considered. Based upon
this preliminary result, IDF relationships covering the range of durations
from 1 5 minutes until 1 ,440 minutes ( 1 day) and using. only four model
parameters were assessed by three different methods. In all three
methods, the separation of the functions a (D and b (cl) is maintained and
a (D is taken as being the inverse function of the Gurrtbel distribution
function. Although the underlying philosophies are different, particularly
INTENSITY-DURATION-FREQUENCY CURVES 265
ACKNOWLEDGEMENTS
The authors are particularly indebted to Mr Marcel Crabbé who provided the
manuscript with the extreme rainfall data in Yangambi, Congo. His excellent
Furthermore, Mr Crabbé was willing to read an early draft of part 1 of this paper
memory was most useful when writing the historica! part of the text.
REFERENCES
by
Yilma SELESID *
located at the farthest north part of the Eastem Africa Rift Valley.
The data condition is that there are neither meteorological nor hydrological
recorded data in the Melkei river catchment of 562 km2• The river is ephemeral
with flashy flood. Daily rainfall data outside the catchment had to be used. The
topography of the catchment is changing in an irregular manner from 1 ,000 m at
the head water to about 50 m at the outlet of the catchment. The catchment
consists of rock escarpments and hills with no vegetation cover.
Design flood is estimated by two different methods : the Soil Conservation
Service (SCS) method, and two Regional Equations. The estimated design flood
of 50 years' return period with the SCS method is 692 m3/s and with two region
al flood estimation equations, 5 1 9 m3/s and 4 1 0 m3/s.
The design flood estimated by the SCS method 692 m3/s is selected. Reasons
for its selection are discussed. Finally the selected design flood depth is cross
checked with the historical flood mark at the site.
1. Introduction
The purpose.of this report is to estimate a design peak flood of the river
Melkei at the bridge site on the Serdo Afrera road. The bridge is located
at km 1 29 + 500 from Serdo. The project area is found in the northem part
of the East African Rift Valley in the Afar Regional Government (fig. 1).
The topography o f the catchment i s changing i n a n irregular manner
from 1 ,000 m at the headwater to about 50 m at the outlet of the catch
ment, at the bridge site. The catchment area of river Melkei is approx
imately 562 km2 up to the bridge site. The catchment consists of rock
escarpments and hills covered with no vegetation. The river is ephemeral.
The flood is characterized as flashy. The river at the bridge site is very
wide (about 1 50 m) with gravel and sand bed. Apparently, the bridge site
is located at the middle of alluvial fan. The left bank of the river at the
bridge site is well defined while the right bank is not well formed (fig. 2).
1 . 1 . CLIMATE
An arid climate dominates the catchment. Annual rainfall varies from
500 mm at the headwater of the catchment to nearly 1 00 mm at the brid
ge site. According to the climate classification of Koppen System, the
study area is located in BWH (hot arid climate) where the landscape is
dominated from harren to sparse vegetation. Strong wind, high temper
ature, low relative humidity and little clouds usually characterize it. Mean
maximum and minimum temperatures are 35° C and 22° C respectively.
2. Hydrometeorological Data
3. Methodology
In section 2 it has been stated that the river Melkei has no measured
river flow data. In such a case, for medium-size catchments the use of the
DESIGN FLOOD ESTIMATION UNDER INADEQUATE DATA 269
Cros s i ng 1 2 9 + 500
�
...
0
0
t. t.
• •
12" 4S' N
2
A = 562 Km
L � 50 Km
12° :50' N
Scale 1 : 2 50 , 000
_._ _ _, Catchment area water-shed
� S tream
---�- Contour l i ne s
Fig. 2. - Upstream view of the Melkei catchment seen from the bridge site at the
crossing km 1 29 + 500 on the road from Serdo to Afrera. In this ardi catchment, exposed
rocks of recent volcanic formation cover hills and escarpments. Top photo : right bank of
the river at the crossing which is relatively well formed ; bottom photo : left bank with no
defined boundary. Transportation of gravels and boulders of different sizes from the
uplands have formed alluvial fans at the foot of the hills. Photo taken during a field visit
made on May 1 , 1 999.
DESIGN FLOOD ESTIMATION UNDER INADEQUATE DATA 27 1
(P - 0.2s) 2
Q=
P + 0.8s
s = 254 ( l OO - 1)
CN
---- -
0.208 AQ
qp =
0.5 D + 0.6tc
'E 20
.• . .
� 10 ..... :�
... . " . .,_ .. ". ..·>: -·
..L..li::l-
: ..L-L
� -L-,...L..,- .= .""'1...,...
. ;.u._
.Ji: -"'-'L-.L-l..
- ....�
0
�;..
... ---�....... ""-1
;;;- 60 -
J! 50 -
!i 40
c
� 30
·-
�
20
ns
E 10 -
"
E
:::J
0
')(
[b'O
ns
�
,,_C!>
Fig. 3. - (A) Mean monthly rainfall (mm) at Dubti ; (B) Annual daily maxirilum rain
fall at Dubti over the period 1988- 1 997.
qp = ( 1 + 25S) a N
where qp = estimated flood peak of return period T (m3/s) ; S = the aver
age river slope (rn/m) ; A = catchment area (km2) ; a = 35.70 and b =
0.364 for T = 50 years.
DESIGN FLOOD ESTIMATION UNDER INADEQUATE DATA 273
qp = 1 .426 ( 1 + 0.25S) A0 1 Xr
where C}p = estimated flood peak of return period T (m3/s) ; S = the aver
age river slope (m/m) ; A = catchment area (km2) ; XT = frequency factor
corresponding to a return period T and given by the following formula :
Xr = u + (alk) ( 1 .0 - (-lnFY)
where u, a, and k are location, scale, and shape parameters. For the study
catchment we used the parameter values estimated for Tekeze catchment.
Thus u = 0.7708, a = 0.344 1 , and k = - 0.08274. F = 0.98 for T = 50 years.
4. 1 . DESIGN DATA
Data used to estimate a peak design flood of the river Melkei at the
bridge site km 1 29 + 500 on the Serdo-Afrera road are given below.
during the time of concentration t,, = 6.3 hrs. This procedure provides
of maximum daily rainfall value 77 mm bas been used as if it rained
Table 1
Estimated design peak floods at the crossing 1 29 + 500 on the road from Serdo to
Afrera based on the SCS method for T = 50 years and T = 100 years
5. Conclusion
Table 2
Summary of estimated peak design floods by three methods at the crossing 1 29 + 500
on the road from Serdo to Afrera
IM 410 By regionalization
276 Y. SELESHI
Table 2 shows that SCS method gives a higher flood peak value than
the other two AMPM and IM methods which are based on regional coef
ficients and catchment area. This is explained partly by the fact that for
this arid catchment without any vegetation cover, the flood is character
ized by a high peak as compared to catchments, which have moderate
vegetation cover. It is known that the regression methods of AMPM and
IM have been developed for the catchrnent with moderate vegetation
cover in wet seasons and this results in attenuated flood peak. Thus, the
design flood peak value of 692 m3/s having 50 years' return period is
acceptable for safe design of the bridge at the crossing 1 29 + 500 on the
road from Serdo to Afrera. The depth of this flood at the crossing with a
bottom width of about 1 50 m (fig. 4) and a manning roughness of 0.03
(for sandy gravel bed and banks) and an average river bed slope of 0.01
is about 1 .3 metres. This depth of flow apparently corresponds to the
recent flood marks left on the right bank of the river Melkei at the cross
ing.
Table 3
Altemative I : bridge design parameters at the crossing km 1 29 + 500 on Serdo-Afrera
road, taking clear bridge span 30 m
Table 4
REFERENCES
by
Maurice MucHINDA *
1. Introduction
may involve management of both supply of, and demand for, water, e.g.
tion) and water supplies. Responses are mainly from local authorities and
2. Data
Annual rainfall data were collected from 29 stations (fig. 1) and an
alysed on the basis of the three main agro-ecological regions as follows :
Region I : Sesheke, Livingstone, Katondwe ; Region II : Senanga,
Petauke, Mt. Makulu, Lusaka International Airport (Lusk.I.A), Lundazi,
-9.
-10.
KASAMA
•
"" - MPIKA
MWINILUNGAI •
-12.
. � � too SOLWEZ
e FIRONDA
-18. 0<>-1--�-�-.,....:=::'���-�-�-�----�-�--1-
22.00 23.00 24.00 25.00 26.00 27.00 28.00 29.00 30.00 31.00 32.00 33.00
LONGITUDE
N.B. MT.MAKULU (·15.31S,28.14E) mo LUSAKA l.A. (·15.185,28.27E) ."not """"" on the map 1
3. Methodology
sities (tab. 1), and these are the intensities used in this study.
The variation of the Rainfall Index across the three main agro-ecol
ogical regions was examined, with a view to determining regional differ
ences in drought incidence over the three regions.
DROUGHT INCIDENCE IN ZAMBIA 285
Table 1
Table 2 presents the main rainfall statistics across the three agro-ecol
ogical regions, and figures 2, 3 and 4 present maps of the mean annual
rainfall, standard deviation and coefficient of variation. Figures 5, 6, 7
and 8 show graphs of Rainfall Indices averaged over Zambia, and over
Regions III, II and 1. A notable feature of the Mean Annual Rainfall map
in figure 2 is the absence of the 1 ,400 mm isohyet (naw replaced by the
1 ,300 mm isohyet) in comparison with similar maps (not shown here) for
previous climatological periods, i.e. 1 96 1 -90, 1 95 1 -80, 1 94 1 -70, etc. This
is an indication of a decline in the rainfall regime. Indeed, an update of
ten-year annual rainfall means for selected Zambian stations (MucHINDA
1 992) (tab. 5) shows that the 1 99 1-2000 ten-year period recorded the
lowest annual rainfall.
The mean annual rainfall countrywide is around 903 mm, with stan
dard deviation of 1 94 mm and coefficient of variation of about 22.6 %.
The average rainfall of 903 mm encompasses mean regional variations
from 1 , 1 80 mm in agro-ecological Region III (R III), to 669 mm in agro
ecological Region I (R I). The stations that exhibit high values of standard
deviation (fig. 3) are Ndola in Region III (SD 237 mm), Mumbwa in
Region II (SD 233 mm) and Choma in Region II (SD 23 1 mm). The cor
responding CVs (fig. 4) are 20.0 %, 26.4 % and 29.2 %, respectively.
Other regional differences are discussed below.
4. 1 . REGION III
Although Kasama and Solwezi have similar annual mean rainfall, the
coefficient of variation indicates that the rainfall is more variable at
Solwezi (CV 1 7.0 %) than at Kasama (CV 13 .3 %). These differences in
the rainfall regime may be crucial, especially from the agronomie stand
point of determining the suitability of erop cultivars in different localities.
286 M. MUCHINDA
Table 2
Main rainfall regime statistics across the agro-ecological Regions III , II and I
-9.
-10.
-11.
-12.
w
0
�
:5
-13.
-14.
-15.
-
16.
-17.
Rainfall variability is probably the largest single factor bebind the variab
ility of erop yields and livestock productivity.
The stations with the highest CV are Ndola and Kasempa (CV 20.0 % ),
whilst the lowest CV is at Kawambwa (CV 1 1 .7 %).
4.2. REGION II
Notwithstanding the similarity in mean rainfall totals between
Mumbwa and Lukulu (88 1 mm and 883 mm, respectively), the rainfall is
slightly less variable at Lukulu (CV 24.6 % ) than at Mumbwa (CV
26.4 %).
Still in Region II, the highest variability in rainfall is exhibited by
Choma (CV 29.2 %) and the lowest is at Mt. Makulu (CV 20.7 %). Such
high variability in rainfall can have far-reaching implications on water
availability for such diverse sectors as agriculture, irrigation and energy.
288 M. MUCHINDA
-9.
- 1 1.
-12.
ê
0
-13.
:5
-14.
22.00 23.00 24.00 25.00 26.00 27.00 28.00 29.00 30.00 31.00 32.00 33.00
LONGITUDE
4.3. REGION I
Region I is partieularly known for its high rainfall variability, with
Sesheke exhibiting higher variability (CV 29.5 %) than the other two sta
tions, i.e. Livingstone and Katondwe, whose mean rainfall totals are sim
ilar, albeit the rainfall climate at Katondwe is more variable (CV 26.7 %).
In sueh areas where there is high rainfall variability, there is need to adapt
eropping strategies to the natural climatie eonditions by developing
appropriate tillage praetiees that eonserve soil moisture. Drought-tolerant
erop varieties, as well as erop diversifieation (including development of
organized markets for these erops), should also be vigorously promoted
in this region.
DROUGHT INCIDENCE IN ZAMBIA 289
-9.
-10.
-11.
·12.
w
Cl
§
-13.
-14.
·15.
-16.
-17.
·18.
22.00 24.00 26.00 28.00 30.00 32.00
LONGmJDE
2 .,...-----.
>< 1 .5
w
c
�
:j
�
o.5
0 ;--.-..rr-.-......
..- ..,...
.. ,-,..
., ........
., ..-...
. ._.,_.,....,_,,.,,...
.. __.. .......-
..., .-.....i .
i•ZAMBIAI.
�
z
-0.5
-1
-1 .5 �----'
71 73 75 77 79 81 83 85 87 89 91 93 95 97 99
TIME {YEARS)
�
w
0.5
..J
..J
� -0 . 5
z
� -1
71 7 3 7 5 77 79 81 83 8 5 8 7 89 9 1 93 95 97 99
TIME {YEARS)
1 .5
><
w
c
z 0.5
-' 0
-'
l'.t -0.5
z
� -1
- 1 .5
-2
71 73 75 77 79 81 83 85 87 89 91 93 95 97 99
TIME (YEARS)
January 1 995), it was the southern half that was mainly affected. Extreme
drought afflicted mainly Regions II and ID during the 1 972/73, 1 99 1/92,
2.5
2
><
w 1 .5
c
z 1
...J 0.5
...J
0
�
z -0 . 5
� -1
-1 .5
-2
71 73 75 77 79 81 83 85 87 89 91 93 95 97 99
TIME (YEARS)
Table 3
Table 4
Table S
5. Conclusion
Although Zambia has recorded many drought episodes during the per
iod 1 97017 1-1999/2000, these have mainly been mild to moderate, with
a few falling in the severe or extreme drought category. The mean annual
rainfall countrywide is around 903 mm and is considerably more than
what some of Zambia's neighbouring countries (e.g. Zimbabwe) receive.
Zambia should therefore be capable of producing regular food surpluses
than is the case at the moment.
However, in spite of the relatively high annual rainfall, some parts of
Zambia (especially agro-ecological regions I and II) experience high
rainfall variability (CV 2 1 -30 %). In these areas, adoption of tillage prac
tices and cropping strategies that conserve soil moisture should help
rninimize erop production shortfalls arising from drought.
The Rainfall Index provides a rapid and handy approach to drought
analysis aimed at exploiting the climate potential. It is a useful tool for
agronornists (e.g. the index can be correlated with erop yields), plant
breeders, land evaluation specialists, etc. who need to determine the suit
ability of particular localities for agricultural purposes. Water resources
planners, who usually rely on quantitative indices such as the RI to ascer
tain whether or not a drought exists, would also find it useful.
294 M. MUCHINDA
REFERENCES
MASCARENHAS, 0 . 1 992. Policy Issues Relating to Maize and Food Security i n the
SADC Region. - In : First Climate Change Conference on Climate
Change (Windhoek, March 2-6),
McKEE, T. B., DoESKEN, N. J. & KLEIST, J. 1993. The relationship of drought fre
quency and duration to time scales. - In : Eighth Conference on Applied
Climatology (January, 1 7-22, Anaheim, California), pp. 1 79- 1 84 (pre
prints).
McKEE, T. B., DoESKEN, N. J. & KLEIST, J. 1 994. Drought Monitoring with mul
tiple tirnescales. - In : Proceedings of the Ninth Conference on Applied
Clirnatology, American Meteorological Society, Boston, pp. 233-236.
MOYO, S. (Ed.) 1 99 1 . Zimbabwe's Environmental Dilemma : Balancing
par
Mars-CLES. -
Régimes pluviométriques ; Congo ; Angola.
REsUME. - Une classification des régimes pluviométriques a été faite en uti
lisant la méthode des centroïdes. Elle a porté sur les données pluviométriques de
95 stations météorologiques distribuées sur l'ensemble du territoire étudié. Pour
caractériser objectivement les tendances de pluviosité dans chaque site, quatre
critères ont été définis au préalable. Ces critères sont relatifs à la hauteur annuel
le des précipitations, à la succession des fortes et des faibles pluviosités, définies
à partir de la moyenne annuelle et, enfin, au contraste pluviométrique.
Les résultats obtenus ont montré que Ie territoire Congo-Angola est partagé
en quatre principaux régimes pluviométriques, très nuancés, et <lont la distribu
tion spatiale est essentiellement zonale.
Il s'agit des régimes équatorial, subéquatorial, tropical et aride ou semi-aride.
Les doubles maxima et minima pluviométriques équatoriaux se réduisent en
unique maximum et minimum vers les régions tropicales. Dans les régions sub
équatoriales, les variations des régimes pluviométriques sont déterminées par les
facteurs liés à la circulation atrnosphérique, principalement à la position du FIT
(Front InterTropical), par Ie mouvement apparent du soleil ainsi que par l'in
fluence du courant marin froid de Benguela. Les régions montagneuses ne sont
pas soumises à l' influence de l' altitude quant à leur distribution annuelle de plu
viosité. La classification des régimes pluviométriques coïncide approximative
ment avec celle des climats, ce qui traduit Ie röle déterminant de la pluviosité
dans la classification des climats tropicaux.
1. Introduction
2. Méthodes
L' utilisation pratique de ces critères peut être illustrée à travers les sta
tions météorologiques de Basankusu et Lubumbashi Luano, représentant
respectivement les régions humide et sèche (tabx 1 et 2).
CLASSIFICATION DES REGIMES PLUVIOMETRIQUES 297
Tableau 1
Mois Janvier Février Mars Avril Mai Juin Juillet Août Sept. Oct. Nov. Déc.
PR 69 128 178 1 79 173 1 67 147 204 211 203 213 1 23
Tableau 2
Mois Janvier Février Mars Avril Mai Juin Juillet Août Sept. Oct. Nov. Déc.
PR 241 272 208 56 2 1 2 0 2 34 1 62 254
Z · = X;1 - µ1
IJ
a1
(2) Le calcul du coefficient de corrélation de Pearson suivi de l' établis
sement d'une matrice de corrélation :
� (x - x)(Y; - Y)
R = -;=;;;;==;==========
V2.i (xi - x)2 2. (Y; - Y)2 •
j
298 J. C. MOLIBA BANKANZA
Ces quatre critères ont été transformés suivant la méthode des cen
troïdes, de manière à obtenir quatre nouvelles variables non corrélées, qui
permettent, sans difficulté, le regroupement des différents sites. Le résul
tat de ce regroupement se traduit par un dendrogramme repris au point
suivant.
3. Résultats
��: ·Aö1�1u'f;dfäl�
Mains h.un1kle a court minima d'h.iver
Tre11iuo.1!1 1 "'mini ma d 'h.iver
i é a tres coun
PR.: ltoJn1m-2000 mn1
S:l; G=l; K:=O.l0-0.37
s�6�ó!o<f;�'o. 14-0.20
A luviosite nioyenne et long niinima d'hiver
g.
:i filu���te g�en�= f!J��s long minima d'h.iver
Mf.:>l l�nm>; �=7; 1 G=l; K:=0.050.0.015
REGIME lROPICAL
REGIME EQUATORlAL
�10�� K.::0.54.0.64
--P,.�-0.66
Fig. 1 . - Dendrogramme .
300 J. C. MOLIBA BANKANZA
Régime équatorial
Régime subéquatorial
Régime tropical
Régime aride
• Stations météorologiques
Namibie
1 5" 20·
- Les régimes caractérisant les régions très humides, c ' est-à-dire dont la
hauteur annuelle des pluies ne descend pas en dessous de 1 500 mm :
- Le régime à faible oscillation entre maxima et plusieurs minima
secondaires : il caractérise les sites à l ' instar de Basankusu et de
Kisangani.
- Le régime à deux maxima et deux minima distincts ou équatorial
comme à Basoko, Bumba, ...
- Le régime à deux maxima et deux minima distincts, à long mini
mum pluviométrique. C ' est le cas d'Eala, Rumangabo, Loeka.
- Le régime à minima forts et à faible contraste pluviométrique à
Kikwit, Kiyaka Plateau, ...
- Le régirne très humide à très court minimum d' hiver : il est marqué
par une très forte pluviosité ( 1 600 - 2 000 mm) et dont le minimum
d' hiver ne dépasse pas 3 mois. Il caractérise les stations de Lukolela
et Isiro.
- Le régime humide à court minimum d' hiver : c'est le régime caracté
ristique des stations qui connaissent moins de quatre mois de séche
resse et <lont la hauteur annuelle des pluies est supérieure à 1 500 mm.
C' est le cas de Mulungu, Lusambo, Yalingimba.
- Le régime moins humide à court minimum d' hiver : celui-ci concerne
les sites de Kin Ndjili et Ilebo qui connaissent moins de quatre mois
de sécheresse et dont la hauteur annuelle des pluies se situe autour de
1 500 mm.
4. Conclusion
REMERCIEMENTS
Nous remercions sincèrement Ie «RNDr Petr BOSHEK» qui nous a aidé dans
l'utilisation des méthodes statistiques de classification.
BIBLIOGRAPHIE
ANNEXE
Diagrammes pluviométriques
300
��l1.1.l.l.1. '
1 z 3 4 5 6
Mois
7
'
8
•
9
1.l.1.
10 11 12
ld.1. . . .
1 2 3 4 5 6 7
Mois
8 9
.- . •.
10 1 1 1 2
by
Elena NIEPLOVA *
days and the seasorial number of days with different precipitation intensity in the
are no trends in seasonal precipitation amounts, the total number of precipitation
1962-2000 period.
1. Introduction
tropics, conditions have become drier (IPCC 2001 ). Some of the regional
studies show that regional temperature and precipitation variation
influenced by circulation and local orography often do not match those of
the globe or hernisphere (METAXAS et al. 1 99 1 , TURK.ES et al. 1 995, MBAYE
1 996, NASRALLAH & BALLING 1 996, BUFFONI et al. 1 998, MAUGERI &
NANNI 1 998, NANNI et al. 1 998, AKINREMI et al. 1 999, BRUNETII et al.
2000). Most of the detailed studies have been conducted in North
America, Europe and Australia. There are only a few studies dealing with
temperature and precipitation changes in the Middle East region.
NASRALLAH & BALLING analysed temperature trends over the Middle East
region during the 1 950- 1 990 period ( 1 993a), as well as temperature and
precipitation trends over the Arabian Gulf region during the 1 89 1 - 1 990
period ( 1 993b) and over the Arabian Peninsula region for the same per
iod ( 1 996). Analysing temperature data from Jones data set (JoNEs et al.
1 986) selected for grid points within the respective region, they found
statistically significant positive trends over both the Arabian Peninsula
(+ 0.63 °C/100 y) and the Arab Gulf region (+ 0.65 °/100 y). With regard
to the Arabian Peninsula region, a significant warming was observed only
during the 1 9 1 1 - 1 935 period, whereas in the Arab Gulf region most of the
warming occurred during the last forty years of the period. Analysis over
the Middle East region showed statistically significant warming during
the 1 950- 1 990 period (+ 0.07 °C/ l 0 y). Precipitation records for both the
Arabian Peninsula and the Arab Gulf region showed statistically insignif
icant negative trends over the 1 950- 1 990 period. NASRALLAH et al. ( 1 990)
investigated as well the influence of rapid urbanization of Kuwait City on
temperature pattems, because the rapid population growth in urban areas
could alter the surface characteristics of cities, resulting in a heat-island
effect and a rise in recorded temperatures (NASRALLAH & BALLING 1 993a).
The results of NASRALLAH et al. ( 1 990) suggest that the Kuwait City tem
perature measurements are not significantly influenced by the heat-island
effect. The aim of our study is to analyse temperature and precipitation
trends in Kuwait during the period of instrumental measurements in the
country.
new climatological observatory was set up within the new Kuwait Airport
area (recently Kuwait International Airport, in next Kuwait Intl. Airport) .
From 1 962 onwards this observatory has provided hourly synoptic obser
vations. For our investigation we used :
- Mean daily temperature data of Kuwait Intl. Airport station from the
1 962-2000 period, calculated from the hourly readings (missing data
from July 3 1 , 1 990 to July 14,1991 due to Kuwait occupation).
- Maximum and minimum temperature observations from the climat
ological station of Kuwait Airport for daily means in the 1954- 1 96 1
period calculation. The daily means were calculated b y formula : T =
(Tmax+Tmin)/2 (Tmax is the daily maximum temperature, Tmin is
the daily minimum temperature). Maximum and minimum temperatu
re data from Kuwait City and Shuwaikh stations from the same period
were used for Kuwait Airport temperature data checking and comple
tion.
- Daily precipitation totals of Kuwait Intl. Airport station from the
1962-2000 period (missing data from July 3 1, 1 990 to July 1 4, 1 99 1
due to Kuwait occupation).
.
For temperature and precipitation series reconstruction and analysis,
next procedures were proceeded :
Table 1
Jan. Feb. March April May June July Aug. Sept. Oct. Nov. Dec.
0.998 0.955 0.961 0.982 0.968 0.941 0.903 0.882 0.943 0.950 0.94 1 0.970
2.033 1 .664 2.397 0.528 -0.452 2 . 1 1 1 0.956 3.249 0.963 2.583 3.079 1 .007
0.870 0.9 19 0.884 0.977 1 .032 0.963 0.990 0.924 0.997 0.935 0.865 0.957
Table 2
Jan. Feb. March April May June July Aug. Sept. Oct. Nov. Dec.
0.992 0.993 0.987 0.982 0.984 0.954 0.934 0.933 0.9 12 0.967 0.987 0.983
0.963 0.925 0.947 1 .026 1 .078 0.989 1 .0 1 0 0.923 0.83 1 0.955 0.934 0.967
0.395 1 .054 0.991 -0.662 -2.2 17 0.936 0. 1 27 3. 146 5.768 1 .049 1 . 1 85 0.3 1 4
TRENDS IN KUWAIT 311
3. Results
Table 3
Basic annual and seasonal temperature and precipitation characteristics of Kuwait Intl.
Airport station. Average values were calculated for the 1 96 1 - 1 990 reference period and
extremes were selected for the whole 1 962-2000 period
[" C]
1. -- Anomalies -- !"fovavi: 1 Year
-
Fig. 1 . Seasonal temperatures at Kuwait Intl. Airport expressed as anomalies from
the 1 96 1 - 1 990 averages (thin lines) and their 5-year moving averages (bold lines).
Mean annual temperature series exhibited some cool and warm per
iods. An apparent cool period appeared in the early 1 970s. The lowest
annual mean of the record was observed in 1 972 (24.5 °C ; - 1 .3 °C below
the 1 96 1 - 1 990 average ). Since this year we can observe gradual warming.
The warmest years of the record occurred in the 1 990s. The five warmest
312 E . NIEPLOVA
Table 4
Annual and seasonal temperature trends for Kuwait Inti. Airport station
(period 1 954-2000) ; b : Jinear regression coefficient, u : Mann-Kendall test value,
P(u) : associated probability
_rr
-2. 0
°'
\0
00
r-
......
00
\0
°'
-
°'
°'
°' °' °' °' °'
...... ...... ...... ...... ......
(° C] 1 -- Movavr -- Anomalies 1 Au
2.0
1.0
0.0
- 1 .0
-2.0
'<:t" r
V') V')
°' 0\
the 1 96 1 - 1 990 averages (thin lines) and their 5-year moving averages (bold lines).
Fig. 2. -
Seasonal temperatures at Kuwait Inti. Airport expressed as anomalies from
3 14 E. NlEPLOVA
Year Spring
6.0 6.0
. .
3.0
_ ... . . .... . .. . . -.
-
. ..._ " /
• �
3.0
t/
·-"·· "
·
-. -.
,t " . ...
- . " " . . . ...
.
1 -tri r-
0.0 0.0
u1
.J W"V v
1
1
-3.0 -3.0
· · · ····
· · · · · · · UI " u'i
-6.0 -6.0
1 950 1 960 1 970 1 980 1 992 1950 1960 1970 1980 1990
Summer Autumn
6.0 " ... . .. 6.0
- - .- - .. -
. ·. .. ..
3.0 / 3.0 .
..
-· .
�'
� � .
7\[7
0.0
�
0.0
1 1
'"'--""
ui
/\
-- I 1 u'i ·1
1
I
-- m
l
-3.0 ·
' -3.0
· · · · · · ·
U ·· · ····
-6.0 -6.0
1 950 1 960 1 970 1 980 1 992 1 950 1960 1 970 1 980 1990
Winter
6.0
- \.-.r""
..\- - .. .
..;.·�··
.., -.
, .. r
\Il v
0.0
-3.0 w1
u1i
! · · · · - ·
-6.0
1 950 1 960 1 970 1 980 1 990
Fig. 3 . .....:. Mann-Kendall test of trend. Progressive values of ui (continuous lines) and
u'i (dashed lines) for annual and seasonal temperature means at Kuwait Inti. Airport.
>=
the category 0 1 - 1 mm contributed only 4 % to the seasonal sum, while
days with precipitation 25 (less than 2 %) contributed 14 %. Neither
seasonal precipitation sums nor the total number of precipitation days
(No. 0 1 ) exhibited any significant trend during the 1 962-2000 period
(tab. 5, figs. 5 and 6). Figure 7 depicts seasonal precipitation sums ex
pressed in a percentage of the 1 96 1 - 1 990 seasonal average. Two periods
of higher sums of precipitation can be observed : in the 1 967- 1 982 per
iod and during the last decade. While the occurrence of larger amounts in
the 1 967- 1 982 period was accompanied with an increasing number of
precipitation days of all categories, the higher precipitation amounts
[%]
5 0 -.---- • Amounts D Number of events 1-------.
4 0 -+----1r-"1--����---1
3 0 -t---t
2 0 --
1 0 --
0
0, 1 - l mm l -5mm 5- l Omm 1 0-25mm >25mm
Fig. 4.- Contribution of the precipitation categories to the mean seasonal (rainy sea
son) precipitation sum and to the mean seasonal number of days with precipitation.
316 E . NIEPLOVA
Table 5
Mean number of days with precipitation of different categories and respective average
sums of precipitation for the 1 962-2000 period and for two relatively wet periods
( 1 967- 1 98 2 and 1 990-2000)
Table 6
Trends of the rainy season's precipitation sums and the number of days with precipitation
of different categories for Kuwait Intl. Airport station (period 1 954-2000) ;
b : linear regression coefficient, u : Mann-Kendall test value, P(u) : associated probability
+-4-+--1--1-+-+.=t=_=t=........ . ....
. �i-l-l
..
280 ..-.-,-..--.-,-.,.-; -- Movavr
- - · · - - · Average 1
- - - - - · · Average
90 -- Movavr
240
200
+-t-+1f-t.......:ll-+-t-++-t-+-9-:li....l-H 1
+-1-+-+-+-.r-+.-+-+-+-+-+-1-+-++.......+i
.
60
+-±-±1�--
160
120
'
30
,""
L- .....
I
I ��1 '\
•
-
- ---- -
-
0
N 'O 0 ..,. 00 N 'O ..,. 00
'O 'O r- r- r- 00 00 � °' °'
� � � � � � � � � �
10
(t
- No!0-25
15 10 �����.....-ri
-- Movavr -- Movavr
·· · · · Avcragc
- · · · · · · Avcragc
· ·
10 ,+-i�-+-+-t-+--+-t-+-+-+-i�-+-+-lr+-t-+1
- No25 (g
5 --
--Movavr
. . . . . .
�-
4 Avcragc
1
2
0
· f ��ftL�L.- . .
'°
....v.
. - ��1
00
°'
Fig. 6. - Seasonal number of days of different categories ((c)-(g)). The bold line is the
5-year running mean, the dashed line is the 1 96 1 - 1 990 average.
[%] D % of 1 96 1 - 1 990
�
average
�
- - -
-
- - - - -
"'
"'
- - - - - - -
- - "' ""
-
- - - - - -
"'
-- - .... - 1-
r
-
" ;::; ""'
-
Fig. 7. -
Precipitation sums of the rainy season expressed as a percentage of the 1 96 1 -
1 990 average sum.
80 - Precipitalio 2 2
60
40 -- Temperatur
20
0 o•c 0 ()"
-20 -20
-40 -1 -40 -1
.Q() -00
-80 -2 -80 -2
§ §
0 V"l O V"l O V"l O V"l O V"l 0 .,., 0 .,., 0 .,., 0 .,., 0 .,.,
.,., Vl \O \O r- f"""-- OO OO O\ O\ .,., .,., \Cl \Cl r- r- 00 00 0\ 0\
0\ 0\ 0\ 0\ 0\ 0\ 0\ 0\ 0\ 0\ (a) 0\ 0\ 0\ 0\ 0\ 0\ 0\ 0\ 0\ 0\
- -- - - - - - 0\ - N 0\ N (b)
Fig. 8. -
5-year running averages of annual (a) and rainy season's (b) mean temper
atures and precipitation sums expressed as anomalies from the 1 96 1 - 1 990 average.
4. Conclusion
REFERENCES
28 : 2277-2278.
and Minimum Temperatures in Kuwait City. - Atmospheric Environment,
1 965- 1 974.
JoNEs, P. D" NEw, M" PARKER, D. E" MARTIN, S. & RIGOR, 1. G. 1 999. Surface air
tions between 1 861 and 1 984. -
1 9 1 - 196.
MBAYE, C. 1 996. Etude des fluctuations de la température de l'air et des précipi
tations au Sénégal de 1 854- 1 990. - In : DEMAREE, G" ALEXANDRE, J. & DE
DAPPER, M. (eds.), Tropical Climatology, Meteorology and Hydrology in
Memoriam Franz Bultot ( 1 924- 1 995), Brussels, Royal Academy of
Overseas Sciences, 22-24 May, pp. 1 79- 1 90.
Int. J. Climatol" 11 :
METAXAS, D. A" BARTZOKAS, A. & VrsTAS, A. 1 99 1 . Temperature fluctuations in
the Mediterranean area during the last 1 20 years. -
897-908.
161.
Environment Conservation, 20 :
NASRALLAH, H. A. & BALLING, R . C. (Jr.) 1 993b. Analysis of Recent Climatic
Changes in the Arabian Gulf region. -
223-226.
245-452.
SNEYERS, R" VANDIEPENBEECK, M" VANLIERDE, R. & DEMAREE, G. 1 989. Climatic
Changes in Belgium as Appearing from the Homogenized Series of
Observations made in Brussels-Uccle ( 1 833- 1 988). - In : Symposium :
Climate Changes (22-26 June, 1 989, Bmo).
TURKES, M" SUMER, U. M" Kruc, G. 1 995. Variations and Trends in Annual
par
1. Introduction
1 2 +-------...---1 �------r
8 --------1 1------r
1 950 1955 1960 1965 1970 1 975 1980 1985 1990 1 995 2000
Fig. 1 . - Evolution des grandes catastrophes naturelles, par type, répertoriées dans Ie
monde de 1950 à 2001 (d'après Munich Rev. 2002).
2. Contexte
étendue géographique que par leur échelle de temps. D' après les respon
sables de la Veille Météorologique Mondiale, notre planète subit d' in
nombrables assauts au cours d'une année : de l' ordre de 1 00 000 orages,
1 0 000 inondations, des milliers de séismes, d' incendies de forêts, de
glissements de terrain, d' avalanches et de tomades, et des centaines
d' éruptions volcaniques, de cyclones tropicaux, d' épisodes de sécheresse
et d' infestations acridiennes. Seules les plus dramatiques de ces catas
trophes font les gros titres de la presse internationale, mais beaucoup
d' entre elles causent des pertes en vies humaines et des dégäts matériels.
Les grandes catastrophes naturelles [ 1 ] * de ces dix dernières années
( 1 990- 1 999) ont provoqué des pertes économiques d'un montant supé
rieur à 600 milliards de US$ et des pertes couvertes par les assurances de
l' ordre de 1 1 0 milliards de US$ (tab. 1 ). D' autres catastrophes moins
importantes ont fait doubler au moins le montant de ces pertes. Par rap
port aux années soixante, la charge des pertes économiques a été multi
pliée par buit, sans compter l' inflation, et celle des pertes couvertes par
les assurances, par 1 6. Avant 1 987, un seul sinistre avait coûté plus d'un
milliard de US$ au secteur des assurances. Depuis, ce montant a été
atteint pour 3 3 sinistres dont 1 5 se sont produits entre 1 998 et 2001
(Munich Rev. 2002).
Au cours des trente demières années, les catastrophes naturelles ont
coûté la vie à plus de trois millions de personnes et laissé plus d'un mil
liard d'individus malades ou sans abri, <lont 95 % dans les pays en voie
de développement. Ceci fait dire à plusieurs spécialistes que tant les
Tableau 1
3. Objectifs du cours
4. Structure de l'enseignement
Ces cours ont comme objectif de présenter et d' étudier les phénomènes
naturels susceptibles d' effets négatifs sur les activités humaines, de pré
voir leur risque d' occurrence et de proposer des moyens de lutte. Ces
phénomènes sont subdivisés en deux grandes orientations : celle relative
à l' agrométéorologie et à l'environnement (les climato-risques) et celle
relative aux risques hydrologiques, géodynarniques et en milieu cötier
(les géo-risques).
5. Bilan et conclusions
Des contacts réguliers sont maintenus avec chacun d' entre eux, retour
nés à leur poste au pays. En outre, l'un d' entre eux a entamé un DEA à
la Fondation Universitaire Luxembourgeoise (F.U.L.). Actuellement, la
deuxième promotion est constituée de seize étudiants, dont onze
L' ENSEIGNEMENT EN GESTION DES RISQUES NATURELS 329
NOTE
BIBLIOGRAPHIE
by
Pierre ÜZER *
1. Introduction
For the late 1 970s, there bas been a growing awareness of the impor
tance of <lust mobilization, transport, and deposition in the evolution of
landscapes in arid and semi-arid areas. Over the last decades, it bas been
shown that poor strategies of land management such as deforestation,
The analysis of the rainfall evolution during the 20th century until 1 998
underlines the extremely intense and long-term drought that has affected
all the Sahelian countries for the late 1 960s. The application of the PETIITI
( 1 979) statistical test on rainfall data of the last eighty years confirms the
non-stationarity of the rainfall pattems between 1 92 1 - 1 968 and 1 969-
650 �������--.1
600
550
500 ������fi.oc���boo����»
450
400
350
300
250
200 -..����....,,
1 92 1 193 1 1941 1 95 1 1961 1971 1981 1991
Years
Fig. 1. -
Yearly rainfall evolution within the Sahel from 1921 to 1998. Presentation of
the three periods used for the study of dust production.
ANALYSIS IN THE SAHEL 333
1 998 periods (fig. 1 ). Indeed, more than 1 00 mm separate these two per
iods, as the average yearly rainfall was 488 and 3 8 1 mm respectively.
The <lust production study presented in this article is based on the
1 95 1 - 1 997 period. These 47 years are divided into three shorter periods
of time :
Long before the onset of the drought, slow land degradation processes
and higher vulnerability of the Sahelian region to desertification resulted
from anthropogenic pressure on the environment. This pressure was
mainly driven by overgrazing and destruction of woody vegetation for
cropping extension purposes. This land degradation was noticed by
increasing soil surfaces suffering from wind erosion and new dunes dev
elopment (STEBBING 1 935 ; AUBREVILLE 1 949 ; TRICART 1 954, 1 959 ;
DRESCH & RoUGERIE 1 960 ; GROVE 1 960 ; PRESCOTI & WHITE 1 960).
Over the last three decades, together with severe drought conditions, a
number of anthropogenic factors link:ed to the explosion of human and
animal population have dramatically increased. In the Sahel, a 3 . 1 %
yearly increase of human population is currently observed ( 1 997- 1 998),
this represents about one more rnillion inhabitants per year. Due to the
increased sedentary population, pressure on cultivated land has led to a
shortening of the fallow period in the shifting cultivation cycle and to the
extension of cropping (multiplied by 3.3 in Niger since 1 96 1 , ÜZER 200 1 )
into the more precarious drier regions. Nomadic pastoralists were depri
ved of some of their best grazing lands as the cultivators moved in lea
ding to concentrated effects of overgrazing in many areas.
Destruction of woody vegetation has been hastened by the ever-increa
sing need for firewood and habitat to meet the demands of the larger
population (THOMAS & MIDDLETON 1 994, BENJAMINSEN 1 996, CHAMARD &
CoUREL 1 999). The destruction is especially noticeable around the rapid
ly growing urban centres, where the circle of deforested lands gets larger
every year. 95 % of the woody vegetation has disappeared around
Niamey since 1 972 (SPATH 1 997). A sirnilar evolution is noticed around
3 34 P. OZER
This study covers dust production frequency during the dry seasons,
October to April, for the 1 95 1 - 1 952 to 1 997- 1 998 period. The dry season
was selected because it is not influenced by any change in the rainfall pat
terns. Data on dust conditions were taken from the 3-hourly synoptic
observations archived at the meteorological offices of Senegal,
Mauritania, Mali and Niger. For this research, data from meteorological
observations made in 28 synoptic stations are used. The international
synoptic surface observation code (SYNOP code) (WMO 1 992) allowed
the identification of four classes of dust-related conditions :
- Dust being raised from the ground at the time of the observation
(07, 08) and reducing horizontal visibility to less than 5 km ;
- Dust storms at various degrees of intensity (09, 30, 3 1 , 32, 33, 34, 35,
36) reducing horizontal visibility to less than 1 km ;
- Dust suspended in the air but not being raised from the ground at the
time of observation (06, presumably remnants of earlier deflation
events), reducing horizontal visibility to less than 5 km ;
- Haze (05, presumably caused by dust) reducing horizontal visibility to
less than 5 km.
4. Results
Dusty days' frequency is plotted on maps (fig. 2) for the three referen
ce periods. The contrast between the "humid" ( 1 95 1 - 1 968) period and the
two following ones is evident. The analysis of average dusty days' fre
quency between the drought (fig. 2[B]) and the "humid" (fig. 2 [A]) per
iods shows a regional increase in all of the 28 stations. Later on, during
the "present" period (fig. 2 [C]), the trend is still on the rise. Some excep
tions can be observed in northern stations where dust frequency is broad
ly stable or slightly decreasing.
In short, between the "humid" ( 1 95 1 - 1 968) and the "present" ( 1 987-
1 997) periods, the dusty days' frequency has constantly increased up to a
336 P. OZER
20.00
[A]
15.00
10.00
-20.00 ·15.00 -10.00 -5.00 0.00 5.00 10.00
0
20.00
[B]
15.00
10.00
-20.00 ·15.00 -10.00 -5.00 0.00 5.00 10.00
20.00
[C]
15.00
10.00
·20.00 -15.00 -10.00 -5.00 0.00 5.00 10.00
-
Fig. 2. Distribution of the dusty days' frequency during the [A] "humid"( 1 95 1 -
1 968), [B] drought ( 1 969- 1986) and [C] "present" ( 1 987- 1 997) periods.
Fig. 3.- Evolution of the relative difference in dusty days' frequency between the
"present" ( 1987- 1 997) and the "humid" ( 195 1 - 1 968) periods. The legend shows the
evolution (%) between these two periods.
Dusty days
><
Cl)
" 1 .5 a
.E
.?:- 1
"'
E
0 0.5
c
"'
f/l 0
>-
"'
" -0.5
>-
-
f/l
:::J -1
c
-1 .5
� <.O <.O <.O <.O � <.O
l.Cl l.Cl <.O <.O ,.._ ,.._ CX) CX) Ol Ol
Ol Ol Ol Ol Ol Ol Ol Ol Ol Ol
Years
Fig. 4. - Sahelian dusty days anomaly index (LAMB 1 982) calcultated over the 1 95 1 -
1997 period.
6. Conclusion
REFERENCES
ANHUF, D., GRUNERT, J. & KoCH, E. 1 990. Veränderungen der realen bodenbe
Sécheresse, 10 : 1 1 - 1 8.
CHAMARD, P. C. & CoUREL, M. F. 1999. La forêt sahélienne menacée. -
de Climatologie, 12 : 342-3 5 1 .
observations synoptiques. - Publications de l 'Association Internationale
FoURQUET, N., CouREL, M. F., CHAMARD, P., LETERRIER, E., BERGES, J . C.,
BENIE, G., BOULAHYA, M., ANDIGUE, J., DJELLOULI, Y. & ÜZER, P. 200 1 . Third
International Workshop on Geomedical Systems. Paris, France (October
17- 1 8, 2001).
Aerobiologia, 17 : 3 1 -42.
Environmental variability and coccidioidomycosis (valley fever). -
Sécheresse, 4 : 95- 1 1 1 .
LE HouEROU, H. N. 1 993. Changements climatiques et désertisation. -
54-60.
MAINGUET, M. 1 99 1 . Desertification : Natural background and human mis
management. - Berlin, Springer-Verlag, 306 pp.
MAINGUET, M., DUMAY, F., ÜULD EL HACEN, M. L. & MAHFOUDH, A. 1 999.
Diagnostic par la télédétection d'un changement de rythme de la dyna
mique éolienne en Mauritanie : la désertification accélérée. - Article pré
senté aux VIII' Joumées scientifiques du Réseau Télédétection de l' AUF
(22-25 novembre 1 999).
McTAINsH, G. H. & PITBLADO, J. R. 1 987. Dust storms and related phenomena
de la désertification. - GEO-ECO-TROP, 24 : 1 -3 1 7.
ÜZER, P. 200 1 . Les lithométéores en région sahélienne : un indicateur climatique
PNUE/CCD 1995. Convention des Nations Unies sur la lutte contre la désertifi
cation dans les pays gravement touchés par la sécheresse et/ou la désertifi
cation, en particulier en Afrique, 71 pp.
Sécheresse, 2 : 1 0 1 - 109.
SALAMA, M., DECONINCK, J. N., LOTFY, M. F. & RlsER, J. 199 1 . L'ensablement de
Nouakchott : exemple de l' aéroport. -
Studien, 6 : 55-88.
and desertification in north and west Africa. Paderborner Geographische
TEGEN, 1. & FUNG, 1. 1995. Contribution to the atmospheric mineral aerosol load
22897-229 14.
TEGEN, 1., LAc1s , A. A. & FuNG, 1. 1996. The influence of mineral aerosols from
1 8707- 1 8726.
Sécheresse, 5 : 1 9 1 - 198.
Pergamon Press, 448 pp.
VALENTIN, C. 1 994. Sécheresse et érosion au Sahel. -
Catena, 37 : 23 1-256.
VALENTIN, C. & o' HERBES, J. M. 1999. Niger tiger bush as a natural harvesting
system. -
37 : 649-663.
WICKENS, G. E. 1 997. Has the Sahel a future ? - Journal ofArid Environments,
by
1. Introduction
There is evidence from the past that rainfall variations and droughts are
common events in the Sahel of West Africa. There has been information
on rainfall fluctuations during historica! times. N1cHOLSON ( 1 978, 1 998)
showed that conditions had been wetter than present from the 1 6th to the
1 8th century. During the 1 9th century, average rainfall was probably 20-
40 % higher than the current average. Since the early 20th century,
instrumental records have shown several severe short-term climatic
fluctuations. The wettest period occurred from the early 1 950s until the
mid- 1 960s. The most significant dry periods were recorded during the
early 1 9 1 0s, the early 1 940s, and the "big drought" which started in the
late 1 960s (SIRCOULON 1976, LAMB 1 982, DEMAREE 1 990, NDONG 1 995,
MOREL 1 998). This last drought was the most important rainfall deficit, in
length and in intensity, ever recorded worldwide during the 20th century
(HULME 1996, JONES & HULME 1 996, DAI et al. 1 998).
Since 1 999, wetter conditions have been reported in Senegal. Accord
ing to the results obtained by the application of the PErrrrr ( 1 979) statis
tica! test, one out of two synoptic stations has seen the end of the
drought (SENE & ÛZER 2002).
This last increase in rainfall was followed by floods in several urban
centres of the Sahel. These calarnities were particularly severe in the main
cities of Senegal such as Dakar, Saint-Louis and Kaolack in 1 999, 2000
and 200 1 . Such floods lasted for several weeks, sometimes up to three
months. Most of the time, affected areas have been abandoned and strick
en population has to be resettled in precarious conditions. Most analysts
consider these floods as a new phenomenon due to the exceptional heavy
daily rainfall. The aim of this article is to find out if these floods are the
result of extreme rainfall events as it was explained by decisions-makers
and try to highlight some factors, which may favour or generate such type
of disaster.
Rainfall analysis was based on daily rainfall records from ten synoptic
stations of Senegal (fig. 1 ) from 1921 to 2000. The selection of this long
period of analysis was done in order to include all recent climatic varia
tions. The analysis of the flood-rainfall relationship is only based on the
daily rainfall examination, without taking into account antecedent cum
ulated rain. The GUMBEL (1958) law was applied to daily rainfall data in
order to estimate the return period of the extreme rainfall events.
The results of the statistica! analysis are presented in figure 2 for the
stations of Dakar, Kaolack and Saint-Louis, which will be discussed
afterwards. For all the ten analysed stations, table 1 shows maximum
daily rainfall amount expected for some selected return periods. These
return periods are of 6, 1 0, 30 and 1 00 years. According to the inter-
EXCEPTIONAL RAINFALL EVENTS 347
LOUIS
DIOURBEL
•
�LACK
TAMBACOUNDA
•
-
Fig. 1 . Location of the synoptic stations in Senegal used in this study. Stations label
led by a square are the urban centres that were affected by severe floods over the last
years.
200
--
1 75
�
�
1 50
/.�
-- �·
A�
1 25
-<>- DAKAR
1 00 -- SAI NT LOUIS
75
r - KAOLACK
•
50
10 20 30 40 50 60 70 80 90 1 00
Fig. 2. -Return periods of maximum daily rainfall at the stations of Dakar, Saint
Louis and Kaolack.
Table 1
ÜZER (2002), it is likely that important daily rainfall may be more fre
quent than during the last decades (ÜZER & CRESSMAN 2000).
Heavy rainfall exceeding a 10-year return period previously recorded
in these three analysed stations have not been associated to damaging
floods in the selected urban centres. This shows that other causes besides
large rainfall amounts could be responsible for the floods observed these
last years.
EXCEPTIONAL RAINFALL EVENTS 349
4. Conclusion
Populations of the Sahel are used to the risk of drought. These last
years, urban populations have had to face another risk, which appears to
be new : floods. These two opposed natural risks require different solu
tions for their management, including new adaptations and new beha
viours. Drought in Senegal is maybe over. However, this probable return
to wetter conditions has led to floods unknown in the past. It appears that
rainfall amounts responsible for the beginning of each flood are not
uncommon. At most they are abnormal with a return period always below
10 years. These floods are therefore to be blamed on the new relationship
between people and their environment and the absence of urban planning
which are the major parameters controlling the development of this nat
ural risk. Wetter conditions are only revealing lots of bad management of
the actual land occupation.
The increase of urban population leads newcomers to build their
houses in areas considered repulsive before the onset of the drought.
There is therefore a strong necessity to take decisions concerning land
planning to avoid concentration of inhabitants in very vulnerable areas.
ACKNOWLEDGEMENTS
REFERENCES
1 39- 1 47.
GUMBEL, E. J. 1 958. Statistics of extremes. - New York, Columbia University
Press, 37 1 pp.
JoNES, P. D. & HuLME, M. 1 996. Calculating regional climatic time series for
107- 1 1 6.
ÜZER, P. & CRESSMAN, K. 2000. Return period analysis of the daily rainfall in
West Africa within the scope of the Météo-France daily rainfall database
import into SWARMS. - Rome, FAO, AGPP, DLIS, 7 pp.
Liège, 42 : 27-33.
nements pluvieux au Sénégal. - Bulletin de la Société Géographique de
by
1 . Introduction
2. Data
3. Method
For the description of the Pacific region the base system of points was
used, which most adequately describes the circulation fields in the equat
orial Pacific (fig. 1). The first point is situated in the region of Australian
summer monsoon ; the second and third one are localized in the wester
ly wind zone. The fourth and fifth are in the region of southeast trade
winds near the coast of Peru and Ecuador.
Fig. 1 . - The base system of points for the description of the Pacific region.
THE ASSESSMENT OF THE CONNECTION 355
o.os
0.04
o.oa
0.02
0,01
()
...#' "'->()
-0.01
".
-0.02 .
•. \
-0.03
-0,04
-0,05
Fig. 2. - Tempora! coefficients of the first SVD vector of the SST anomaly (solid line)
and zonal wind velocity (dotted line).
356 E. SOKOLIKHINA, N. SOKOLIKHINA & E. SEMENOV
0,02
,S>: ...�
q,�: er !"<:I ......�'' -.IP'
•
• • 1
?;
� .
�
.
r§>
-.'11
.S> ·� , A
-."i' f' ...�;
1 •
. : '
�
• 1
1
•
.
... .: . :
I
.
-0,02 ,
.
\•I:
. . .
1 • \j .
.
•
. .
'1..J'
.
1
.
1
1
-0,04 . .
: .·
.
'.i
..
-0,06 "
-0,08 _ _________________________,
....__
Fig. 3. - Tempora! coefficients of the second SVD vector of the SST anomaly (solid
line) and zona! wind velocity (dotted line).
4. Conclusions
60 60
Correlation betwe�n SST and tempora! coefficients
50 50
for 1 -st SVD of synoptical signal
40 40
30 30
� 20 20
10 10
�
:;; 0 1----_.,,_
c
0
-10 -10
__ �,__-----+------+-� 0
�0 -20 -20
-30
c..>
-40 -40
-30
� � �
'*'
30 30
.Q
1ii
8 20 20
ëi)
t::
10 10
0 0
-280 -240 -200 -160 -120 -80 -40 0
SST behind lags (days)
Fig. 4. - Lag-cross correlation function for the tempora! coefficients of the first (a) and
second (b) SVD vectors of SST anomalies and zonal wind velocity : t : tempora! lag ;
t < 0 : the SST anomalies are late from the wind ; i; > 0 : the SST anomalies go ahead.
ACKNODWLEDGEMENTS