Journal of Hydrology 397 (2011) 93–104

Contents lists available at ScienceDirect

Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol

A simple thermal mapping method for seasonal spatial patterns

of groundwater–surface water interaction
Christian Anibas a,⇑, Kerst Buis b, Ronny Verhoeven c, Patrick Meire b, Okke Batelaan a,d
a
Department of Hydrology and Hydraulic Engineering, Vrije Universiteit Brussel (VUB), 1050 Brussels, Pleinlaan 2, Belgium
b
Department of Biology, Ecosystem Management Research Group (ECOBE), University of Antwerp, 2610 Antwerp, Universiteitsplein 1, Belgium
c
Civil Engineering Department, Hydraulics Laboratory, University of Ghent, 9000 Ghent, Sint-Pieternieuwstraat 41, Belgium
d
Department of Earth and Environmental Sciences, K.U. Leuven, 3001 Leuven, Celestijnenlaan 200e, Belgium

a r t i c l e i n f o s u m m a r y

Article history: A simple thermal mapping method for simulating seasonal and spatial patterns of groundwater–surface
Received 2 October 2009 water interaction is developed and tested for a segment of the Aa River, Belgium. Spatially distributed
Received in revised form 14 November 2010 temperature profiles in the hyporheic zone of the river are measured in winter and summer seasons of
Accepted 25 November 2010
three consecutive years. Inverse modeling of the one-dimensional heat transport equation is applied to
This manuscript was handled by P. Baveye,
estimate vertical advective fluxes using the numerical STRIVE model and an analytical model. Results
Editor-in-Chief, with the assistance of D.P. of the study show that seasonal flux estimates for summer and winter can be derived with a minimum
Ahlfeld, Associate Editor data input and simulation effort. The estimated fluxes are analyzed via non-parametric statistical tests,
while spatial interpolation techniques are used to generate maps of distributed flux exchange. The esti-
Keywords: mated seepage is compared with volumetric flux obtained from piezometer measurements and output of
Heat transport modeling a groundwater model. The thermal method shows higher discharge rates in winter and that the relative
Temperature contribution of exfiltration to the river discharge is higher in summer. A higher flux and a more hetero-
Hyporheic zone geneous flow pattern are observed in the upper reach of the river compared to the lower reach. This spa-
Groundwater–surface water interaction tial difference shows the importance of the local geomorphology and to a lesser extent the hydrogeologic
FEMME setting on hyporheic flux exchange in the river. A significantly higher flux is noted on the banks than in
Belgium
the center of the river, which is driven by the relatively high hydraulic conductivity of the river banks. It is
concluded that bank flow in groundwater–surface water interaction deserves more attention. The main
channel of the Aa River alone accounts for about 15% of the total river discharge at its outlet. As the devel-
oped thermal method is cost-effective, simple and fast, it is recommended for use in identifying zones of
interest in initial stages of field investigations of groundwater–surface water interaction.
Ó 2010 Elsevier B.V. All rights reserved.

1. Introduction Directive (European Commission, 2000, 2006) mandate the protec-

An important process in ecohydrology is groundwater–surface
water interaction at the interface of lake or riverbeds (Stanford,
1998; Woessner, 2000). The accurate quantification of exchange
of water and energy between groundwater and surface water sys-
tems remains a challenging task due to the problems of heteroge-
neity and scale (Becker et al., 2004; Kalbus et al., 2008).
The saturated zone beneath and beside streams, rivers, lakes
and wetlands where groundwater and surface water actively mix
and exchange is called hyporheic zone (Brunke and Gonser,
1997; Boulton et al., 1998; Hayashi and Rosenberry, 2002; Smith,
2005). The processes in the hyporheic zone are variable in space
and time on relatively small scales (McClain et al., 2003). National
and international regulations like the European Water Framework
⇑ Corresponding author. Tel.: +32 26495877; fax: +32 26293022.
E-mail address: canibas@vub.ac.be (C. Anibas).
tion of the links between groundwater and surface water systems.
To do this, reliable and transferable conceptual models (NRC, 2004;
Smith, 2005; Schmidt et al., 2008) are required to assess mass
fluxes and related biochemical processes across groundwater–sur-
face water interfaces.
Different methods have been discussed and used to quantify the
exchange flux between groundwater and surface water systems
(Kalbus et al., 2006). Seepage meters are the only means by which
groundwater–surface water exchange is directly measured (Lee,
1977; Rosenberry, 2008). Although seepage meters are technically
simple, there are uncertainties in the measured flux due to opera-
tional problems in the field (Shaw and Prepas, 1990; Murdoch and
Kelly, 2003). Indirect methods estimate the groundwater–surface
water interaction by deriving the fluxes across the interface of
the hyporheic zone from other parameters. Groundwater head
measurements from piezometers and boreholes in riverbeds or
near streams (Cey et al., 1998; Baxter et al., 2003) can be used to

0022-1694/$ - see front matter Ó 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.jhydrol.2010.11.036
94 C. Anibas et al. / Journal of Hydrology 397 (2011) 93–104
derive vertical flow estimates. But the exact estimation of hypor-
heic-zone flux is hampered by difficulties in measuring in situ
hydraulic conductivity (Chen, 2000). Piezometers installed in river-
beds are also prone to scouring and damage by floods or debris.
Other indirect methods of estimating groundwater–surface water
exchange include differential discharge gauging (Becker et al., 2004),
numerical modeling (Cardenas and Zlotnik, 2003; Fleckenstein
et al., 2004; Saenger et al., 2005) and the use of conservative trac-
ers like dye, salt, chloride or stable isotopes (Carey and Quinton,
2005). Remote sensing, a relatively new technique in this field
(Becker, 2006), offers a spatially continuous qualitative estimation
of hyporheic-zone flux exchange (Loheide and Gorelick, 2006).
Heat can be used as a tracer to estimate groundwater–surface
water interaction (Stonestrom and Constantz, 2003; Anderson,
2005; Constantz, 2008; Constantz et al., 2008). The exchange pat-
terns can be inferred qualitatively from temperature variations at
the groundwater–surface water interface (Lowry et al., 2007) or
quantified by empirically relating mapped streambed tempera-
tures to volumetric fluxes (Conant, 2004). Hatch et al. (2006) and
Keery et al. (2007) presented methods to determine streambed
seepage rates using phase shifts and amplitude damping of ther-
mal time series data between pairs of sensors. Exchange rates have
also been successfully quantified by inverse modeling of tempera-
ture profiles (Arriaga and Leap, 2006; Schmidt et al., 2006; Anibas
et al., 2009).
Suzuki (1960) and Stallman (1965) presented analytical solu-
tions for using temperature as a natural tracer. Bredehoeft and
Papadopoulos (1965) introduced a graphical type curve solution
method based on temperature–depth profiles for the estimation
of vertical groundwater fluxes under steady-state thermal condi-
tions. As temperature distributions in the subsurface are transient
by nature, Lapham (1989) developed a numerical model that al-
lows quantification of groundwater–surface water interaction
using an annual temperature variation simplified by a sine func-
tion. Arriaga and Leap (2006) and Schmidt et al. (2006, 2007) pre-
sented applications of the estimation of groundwater–surface
water interaction based on steady state heat transport, while Ani-
bas et al. (2009) showed when the steady-state thermal assump-
tion can be used.
This paper presents a simple and fast method for quantifying
seasonally distributed groundwater–surface water exchange. The
proposed approach is a robust, adjustable and first-hand field
investigation tool for detailed monitoring of zones of interest.
The underlying hypothesis of the methodology is that steady-state
solutions for simultaneous vertical transport of heat and water in
riverbeds are able to detect groundwater–surface water interaction
with sufficient spatial and temporal resolution. Measured riverbed
temperature profiles from the Aa River in Belgium are used to test
the proposed method. Furthermore, the analytical solution method
of Arriaga and Leap (2006) is used in conjunction with the numer-
ical heat transport model STRIVE to statistically verify the pro-
posed method. The STRIVE model is based on the approach of
Lapham (1989) and the FEMME ecosystem modeling platform
(Soetaert et al., 2002).
2. Methodology
The temperature in the hyporheic zone varies seasonally and
diurnally as a consequence of heating and cooling of the land sur-
face. During summer months the groundwater temperature is gen-
erally cooler than the river temperature, whereas in winter it is
generally the opposite. The groundwater flow in the riverbed is as-
sumed to obey Darcy’s law consequently the natural temperature
distribution in the riverbed is influenced by the movement of
water. The geothermal gradient in the shallow subsurface is ne-
glected for our analysis hence the one dimensional, vertical, aniso-
thermal transport of heat through homogeneous, porous media is
described as (Suzuki, 1960; Stallman, 1965; Lapham, 1989).
@2T @T @T
k  v z cw qw ¼ cq ð1Þ
@z2 @z @t
where k is the thermal conductivity of the soil–water matrix in
J s1 m1 K1, T the temperature at point z at time t in the sediment
in K (°C), cw the specific heat capacity of the fluid in J kg1 K1, qw
the density of the fluid in kg m3, vz the vertical component of
the groundwater velocity in the sediment in m s1, c the specific
heat capacity of the rock-fluid matrix in J kg1 K1, and q the
wet-bulk density (density of the rock-fluid matrix) in kg m3. The
terms cwqw and cq represent the volumetric heat capacity of the
fluid and the rock-fluid matrix in J m3 K1 respectively. The first
term of the left hand side of Eq. (1) represents the conductive and
the second term the advective part of the heat transport. A positive
sign of the vertical groundwater velocity stands for water moving
from the surface into the hyporheic zone (e.g. groundwater re-
charge or losing stream reach) and a negative sign represents water
moving from the hyporheic zone into the river (e.g. groundwater
discharge or gaining stream reach). The groundwater velocity is ex-
pressed in mm d1, equivalent with a unit flux in L m2 d1.
In the case of a thermal steady state, the temperature distribu-
tion in the riverbed is supposed to be constant over time, which re-
duces the right hand side of Eq. (1) to 0:
@2T @T
k  v z cw qw ¼0 ð2Þ
@z2 @z
Notice that in Eq. (2), the thermal properties of the fluid–sedi-
ment matrix are only described by the thermal conductivity k.
For sediments as different as peat or gravel the thermal conductiv-
ity k varies less than an order of magnitude (Freeze and Cherry,
1979; Domenico and Schwartz, 1998); e.g. saturated sands usually
have k values between 1.4 and 2.2 J s1 m1 K1 (Lapham, 1989;
Schön, 1998; Stonestrom and Blasch, 2003). This small range of k
is a major advantage of the thermal method compared to the
hydraulic head method in which the hydraulic conductivity can
vary over orders of magnitude (Chen, 2000). Hence, in many cases,
the thermal conductivity can be parameterized by taking its value
from literature.
The thermal method is an indirect method, field data must be
processed with a model in order to derive quantitative estimates
of the flow velocity. We apply inverse modeling of the thermal
steady state by solving Eq. (1) numerically and Eq. (2) analytically.
Fig. 1 shows the concept of the steady state 1D vertical fluid and
heat transport for a saturated sediment column of the riverbed,
in which measured temperatures define the upper and lower
boundary conditions. The concept of the model does not allow
to determine lateral or longitudinal flow vectors (Fairley and
Nicholson, 2005; Hoehn and Cirpka, 2006). One or more measured
temperatures within the domain are needed in order to fit the sim-
ulated temperature distribution to the measured profile (Fig. 1a).
When groundwater flow occurs, the thermal profile shows a stron-
ger curvature with increasing flow velocity (Fig. 1b). For practical
reasons the presented methodology is limited to discharging flow
conditions if applied to shallow groundwater flow systems
(Schmidt et al., 2007). Mathematically Eqs. (1) and (2) allow esti-
mation of infiltration, a reversal of the flux under steady state
conditions however is producing a uniform temperature distribu-
tion with no significant thermal gradient. Hence it is not possible
to obtain a realistic fit of the model with the applied boundary
conditions (Fig. 1c).
To solve the thermal steady state problem a hydraulic
steady state has to be assumed, diurnal, seasonal and irregular
 C. Anibas et al. / Journal of Hydrology 397 (2011) 93–104
Fig. 1. Concept of vertical temperature distribution and simulation of vertical 1D
water and heat transport in a river or lake bed under thermal steady state
assumption of constant upper and lower boundary temperatures; (a) medium
upward groundwater flux; (b) high upward groundwater flux; (c) downward flux.
(meteorological) transient thermal influences in the measured data
should be insignificant, which is a limitation for very dynamic
river–aquifer systems. Seasonal thermal influences can be approx-
imated by a sine wave with a wavelength of an astronomical year;
close to the maximum (summer) and minimum (winter) of the
sine a steady state assumption is valid (Anibas et al., 2009). The
application of the thermal steady state assumption therefore is
limited to these periods. Diurnal temperature changes have a lim-
ited penetration depth in the sediment due to the high frequency
of its temperature signals. Schmidt et al. (2006) advises therefore
to exclude measurements close to the surface water–groundwater
interface from further analyses. Irregular transient influences,
caused by meteorological events, cannot be corrected and will
result in noise in the vertical heat distribution. Violations of the
thermal steady state assumption cannot be observed from the
Fig. 2. Location of the Aa River study site.
95
goodness of fit between the simulated versus measured tempera-
ture profile (Anibas et al., 2009). Hence, it is important that the
processed profiles selected for analysis respect the steady state
assumption.
Two models based on the differential Eqs. (1) and (2) have been
used in this paper, called ‘STRIVE’ and ‘Arriaga and Leap (2006)’
respectively.
2.1. Arriaga and Leap (2006) model
We applied the analytical type curve method of Bredehoeft and
Papadopulos (1965) as described by Arriaga and Leap (2006) using
Microsoft Excel with the Solver add-in. This constitutes a simple
and fast tool to derive quantitative estimates of the vertical
groundwater–surface water exchange.
2.2. STRIVE model
A vertical heat transport model was implemented in STRIVE
(STReam RIVer Ecosystem), which is a modular package of FOR-
TRAN subroutines for the dynamic ecosystem modeling platform
FEMME (Soetaert et al., 2002). STRIVE simulates different physical
and biogeochemical transport and exchange processes including
vertical 1D heat and water transport (Eq. (1)). The heat transport
model obtains a fit by minimizing the differences between mea-
sured and simulated temperature distributions according to inter-
nal user defined integration and fitting routines (Soetaert et al.,
2002) by adapting the flow velocity (vz).
For this study the STRIVE model is descritizised as a vertical,
one-dimensional saturated and homogeneous sediment column
of 5 m length and is composed of 100 layers. To reduce descritiza-
tion errors, the spacing of the model layers follows a sinusoidal dis-
tribution, resulting in layer thicknesses of 0.001 m at the upper and
lower boundary, while the thickness of the layers is maximum
about 0.08 m in the center of the model domain.
The simulation results from STRIVE and Arriaga and Leap (2006)
allowed the use of statistical tests for the comparison of spatial,
temporal and conceptual differences of the models, as well as for
comparison of solution procedures of STRIVE and Arriaga and Leap
(2006).
96 C. Anibas et al. / Journal of Hydrology 397 (2011) 93–104
Fig. 3. Digital elevation model showing the examined section of the Aa River with the location of 26 measurement points and three piezometer nests. The river flow is from
weir 3–4. The dotted line indicates the boundary between the occurrence of the Formation of Kasterlee in the upstream part of the river section and the Formation of Diest in
the downstream part (DOV, 2010).
3. Field site and measurements
We collected temperature versus depth profiles along a 1425 m
long stretch of the lower part of the Aa River, Flanders, Belgium
(Fig. 2). The river has a total length of 36.7 km and a drainage area
of 235 km2; its catchment comprises 15% of the Nete basin. The
measurement site is located around 5.5 km upstream of the conflu-
ence of the Aa with the Kleine Nete River. At the site the Aa flows in
WSW direction and has an average water depth of around 1.1 m, a
width of 14 m, an average discharge of about 1.9 m3 s1 and a mea-
sured slope of 0.48‰. The Aa is a typical Flemish lowland river,
which was stretched and canalized in the 1960s. This section of
the Aa contains three meanders and is upstream and downstream
limited by weirs, respectively numbered 3 and 4 (Fig. 3). Ground-
water tables in the area are shallow, about 1 m below the land sur-
face. The cross-section of the river is rectangular with a flat river
bottom. An exception is the area around points 10 and 11
(Fig. 3), where erosion and sedimentation processes established
pronounced riverbed topography. Although strongly determined
by the weirs, the discharge regime is influenced by the growth of
macrophytes in summer (Bal and Meire, 2009). Discharges and
flow velocities therefore are higher in winter. The combined effects
lead to relatively constant water levels throughout the year, dis-
rupted only by occasional peak discharges.
The riverbed is composed of fine sand containing a varying frac-
tion of organic matter, especially at the inner banks of meanders.
The sediments are usually compacted along the banks; while in
the middle of the river the sediments are loose, showing some pool
and riffle structures. The local geology is determined by tertiary
Formations of Diest and Kasterlee; the former occurs in the down-
stream part while it is overlain by the latter in the upstream part
(Figs. 3 and 4) (DOV, 2010). The Formation of Kasterlee consists
of fine sands and fractions of clay. The Formation of Diest is de-
scribed as a heterogeneous sand layer. In regional groundwater
models the hydraulic conductivity of the Kasterlee and Diest For-
mation was parameterized as 12.5 and 9.8 m d1 respectively
(Woldeamlak, 2007). Both formations together with the underlying
Berchem Formation constitute one aquifer system. It is underneath
concealed by the Boom Aquitard (Fig. 4; Marechal and Laga, 1988).
Temperature can be measured rapidly and easily as sensors are
technically simple, cheap, widely available and can be handled by
unskilled personnel. We performed ‘roaming surveys’ (Keery and
Binley, 2007) of temperature profile measurements, using a 2.0 m
long self-made instrument, the ‘Temperature stick’ (‘T-stick’)
(Anibas et al., 2009). Fig. 5 shows the measurement setup and
the instrument, consisting of a metal tube, a T-shaped handle at
the top and a pointed tip on the bottom. Close to the bottom end
a slotted hole holds the temperature sensor, a thermistor (Davis
Instruments Model 7817; Hayward, CA, USA). The sensor is con-
nected to an electric multimeter (ABB Metrawatt); the value of
the electric resistance of the sensor is converted to a temperature
value via a calibration function. The position of the temperature
sensor in the riverbed is determined by reading scaled marks with
respect to the surface water level.
Five summer and six winter measurement campaigns, each
consisting of two consecutive days have been conducted at the
 C. Anibas et al. / Journal of Hydrology 397 (2011) 93–104
Fig. 4. Longitudinal cross-section along the Aa River between weir 3 (right) and
weir 4 (left) showing the geology (DOV, 2010).
Fig. 5. Scheme for measuring temperature profiles in the riverbed of the Aa with
the T-stick instrument.
Aa River field site between August 2004 and February 2007. During
each campaign 32 temperature profiles were measured at 26 spa-
tially distributed points (Fig. 3), six of the points were investigated
on both days. We assumed, based on Anibas et al. (2009) that the
time between 10 July and 10 September and between 10 January
and 10 March, referred to as ‘summer’ and ‘winter’ respectively,
are most favorable for a thermal steady state investigation. In total
around 250 temperatures profiles fulfilling the requirement for a
thermal steady state assumption were used for analysis.
Fourteen points along the centerline of the river were measured
during the first day of a campaign. Point 1 is located around 30 m
from weir 4, point 14 around 40 m from weir 3, resulting in a dis-
tance of about 100 m between each point (Fig. 3). However, mea-
surement point 12 is not used for most of the analysis, as it was
not possible to reach the center of the river. The spatial position
of the measurement points was established with a Garmin etrex
GPS receiver resulting in a spatial error of about 1.5 m during the
different campaigns. As indicated in Fig. 5, the temperature profiles
97
consist of measurements at five depths: 0.0, 0.10, 0.25, 0.50 m and
a maximum depth. We measured the temperature profile starting
from the interface between river and riverbed sediment indicated
as 0.0 m depth up to a maximum depth generally between 0.40–
1.20 m (average 0.84 m). Deeper measurements were either pre-
vented by a compacted sandy layer or by the limited length of
the instrument. The measurements started usually between 9
and 11 a.m. and ended between 3 and 6 p.m.; around 15 min were
required to measure a temperature profile with the T-stick
instrument.
During the second day of a campaign measurements were taken
at the left bank (L), center (C) and right bank (R) of six transects at
the locations 1, 4, 7, 11, 13 and 14 (Fig. 3). The points L and R are
located a quarter of the width of the river away from respectively
the left and right bank. All measurement campaigns were per-
formed in a similar pattern.
Furthermore, at location 4, 7 and 13 piezometer nests were in-
stalled in the riverbed. Data loggers recorded groundwater levels at
three different depths (0.1–0.15 m, 0.5–0.6 m and 0.9–1.2 m) with
hourly frequency. Slug tests were performed to determine hydrau-
lic conductivity values.
For the simulations presented in this work, the thermal conduc-
tivity is assumed to be 1.8 J s1 m1 K1; taking into account the
organic content of the uppermost hyporheic sediments. The other
model parameters are given in Table 1. At a depth below 4 m the
groundwater shows an almost constant temperature the year
round, about 1–2 °C above the mean annual surface temperature
(Anderson, 2005). The average temperature of about 650 tempera-
ture profiles (the complete set collected at the Aa River site) was
12.2 °C. This value corresponds well to the average temperature
measured in a piezometer on the southern edge of the flood plain.
Hence, the value is used as a constant for the lower boundary at
5 m depth, whereas the upper boundary of the model domain is
defined at 0.1 m depth and uses the value of the respective temper-
ature profile (Schmidt et al., 2006).
4. Results and discussion
The temperature–depth variation in the river is indicated in
Fig. 6, which shows the average temperatures of 13 profiles along
the center line of the river for 11 measurement campaigns. The
maximum temperature recorded at the interface (0.0 m depth)
was 23.6 °C on 3 July 2006 and the absolute minimum was 3.2 °C
on 26 January 2005. At the maximum depth (0.84 m) the temper-
ature varied between 17.2 °C on 25 August 2005 and 7.6 °C on 9
February 2006. The temperature differences between the interface
and the deepest measurement point vary between 1.6 and 7.3 °C,
with an average of 2.9 °C for all measurement campaigns. Hence,
this temperature difference was found both in winter and summer
seasons.
The relative large dataset of about 850 simulations performed
with the Arriaga and Leap (2006) and with STRIVE allowed statis-
Table 1
Physical properties and boundary conditions used for the steady state STRIVE model
and the modeling procedure of Arriaga and Leap (2006).
Parameter Value Unit
Thermal conductivity, k 1.8 J s1 m1 K1
Density of water, qw 1000 kg m3
Specific heat capacity of water, cw 4180 J kg1 K1
Upper boundary condition, T0a Variable °C
Lower boundary condition, TLb 12.2 °C
a
Located at the interface between surface water–riverbed or 0.1 m below the
interface.
b
Located 5.0 m below the interface between surface water–riverbed.
98 C. Anibas et al. / Journal of Hydrology 397 (2011) 93–104
Fig. 6. Measured vertical profiles of riverbed temperatures averaged for 13
longitudinal points and presented for 11 measurement campaigns, showing
different temperature gradients between summer and winter seasons.
tical analysis of the results (level of significance p = 0.05). First, we
tested if the distribution of the results is normal (Kolmogorov–
Smirnov and Shapiro–Wilk test; N = 202). Since the thermal steady
state analysis is limited to negative fluxes both tests indicated that
the simulated exchange fluxes are not normally distributed, hence
non-parametric statistical tests were applied for further analysis.
Since six points were measured on both days of the two-day
campaigns, we examined with a Wilcox on matched pairs test
(N = 42) and the Sign test (N = 38) if a significant difference exists
between the estimated fluxes for the center locations. The tests
show no significant differences and hence all thermal profiles are
treated as a single data set.
By comparing results of the STRIVE model simulations with
Arriaga and Leap (2006) (Kolmogorov–Smirnov two sample test;
N = 109) no significant differences were found. This verifies STRIVE
with the approach of Arriaga and Leap (2006), while it also shows
that both approaches are applicable for shallow thermal profiles
measured in a riverbed.
The measurement setup described in Fig. 5 has an estimated
uncertainty for the depth determination of ±0.01 m for a single
measurement resulting in a possible maximum cumulative error
for the measured domain of 0.10 m. A Kruskal–Wallis ANOVA test
by ranks and median (N = 192) with Arriaga and Leap (2006) for
domains of ±0.10 m showed no significant difference in resulting
fluxes. This proves that the T-stick is sufficient accurate for collect-
ing temperature–depth profiles.
For steady state boundary conditions the error between the
modeled and measured T-profile is dependent on the thickness
of the modeled domain (Schmidt et al., 2006). This was confirmed
by a Kolmogorov–Smirnov two sample test (N = 108), which com-
pared simulations using a model domain equal to the depth of the
measured profile (Fig. 5) with a model domain equal to 5.0 m.
Hence, the lower boundary should be defined at a depth where
the temperature is not changing between the different simulated
profiles, thus where a constant groundwater temperature can be
assumed.
An important question is the influence of diurnal transient ther-
mal signals on the applicability of the steady state assumption. The
Aa River shows diurnal temperature changes of about 2–3 °C in
winter and 2–5 °C in summer, which are lower than the variations
measured in the small stream in Germany examined by Schmidt
et al. (2006). By applying a Wilcox on Matched Pairs (N = 202)
and the Sign test (N = 195) using different upper boundaries,
including or excluding the uppermost measurement point at the
interface (0.0 m depth; Fig. 1) a significant diurnal influence was
detected in the uppermost 0.1 m of the sediment.
4.1. Spatial patterns
Figs. 7a and 8a show the temperature distribution in the river-
bed interpolated along a vertical, longitudinal cross-section. The
bar graphs of Figs. 7b and 8b suggest that the examined river reach
can be divided into two zones; a zone where higher fluxes are ob-
served in the upstream part of the river (points 9, 10 and 13), and a
zone of lower fluxes in the downstream part (points 5–8). This dif-
ference is confirmed by a Kolmogorov–Smirnov two sample
(N = 127) and a Mann–Whitney U test (N = 127) tests and coincides
with the change in geology as indicated in Figs. 3 and 4. At the cen-
ter line the variability of the fluxes between the measurement
campaigns is relatively similar along the whole river stretch. The
winter data however shows a relative higher variation compared
to the summer.
Measurements at six river cross-sections were performed; the
median fluxes for each measurement point for (a) winter and (b)
summer seasons are plotted in Fig. 9 versus their position along
the river. Characteristic is that the average fluxes of 79 and
53 mm d1 and their ranges at the left bank and right bank
respectively are higher than at the center (32 mm d1). A Krus-
kal–Wallis multiple comparison p values (2-tailed) test (N = 145)
showed significant differences between the center and the banks.
As the composition of the riverbed sediments is relatively uniform,
the difference in flux can be explained by different hydraulic gradi-
ents and heterogeneity of the underlying geologic layers (Conant,
2004; Kalbus et al., 2009), in this case the Formations of Diest
and Kasterlee. As indicated in Fig. 4, the upstream section of the
river is dominated by the higher conductive Kasterlee Formation
causing a more local flow system compared to the downstream
part. There, the Quaternary is underlain directly by the thick Diest
Formation, creating deeper, more regional discharges.
The analysis of the transect measurements across the river indi-
cate that the bank areas of the river not only contribute relatively
more groundwater discharge compared to the center of the river,
but that also the flow mechanism at the banks is different. The
temporal variability of the fluxes at the banks is higher, indicating
shallow transport processes. Fig. 9 shows that the left bank gener-
ally shows higher fluxes than the right bank. In summer higher
fluxes occur at the left bank in the upstream part whereas in the
downstream part higher fluxes are observed at the right bank.
The differences in flux between the left and the right banks as well
as the differences between upstream and downstream can be ex-
plained by hydromorphological patterns (Woessner, 2000). Due
to topography (Fig. 3) higher gradients are apparent upstream.
Moreover, a meander forces the groundwater flow to converge to
the left bank, while the flow pattern towards the right bank is
diverging. In the downstream part, the valley is widening and the
groundwater head gradients are lower leading to smaller differ-
ences between right and left bank.
In the upper reach however, points 11 and 14 show dissimilar,
relatively low fluxes compared to the neighboring points. At point
11, this can be explained by the uneven bathymetry of the river
cross-section due to the meander at this location. The right and
center measurements are located in the shallow inner bend of
the meander, while the measurement at the left bank is located
in the deeper part and shows higher fluxes (Fig. 9).
Piezometer nests (Fig. 3) were used to verify the two heat trans-
port models. The observed head differences indicate an upward
gradient during most time of the year. The horizontal hydraulic
 C. Anibas et al. / Journal of Hydrology 397 (2011) 93–104 99

Fig. 7. Temperature distribution and vertical groundwater discharge in winter. (a) Two dimensional, longitudinal cross-section (vertical axis exaggerated) along the river
centerline indicating the measured temperature distribution in the riverbed; crosses show measurement locations. (b) Bar graph indicating the measurement points and the
estimated vertical groundwater flux averaged for all winter seasons as simulated with STRIVE.

Fig. 8. Temperature distribution and vertical groundwater discharge in summer. (a) Two dimensional, longitudinal cross-section (vertical axis exaggerated) along the river
centerline indicating the measured temperature distribution in the riverbed; crosses show measurement locations. (b) Bar graph indicating the measurement points and the
estimated vertical groundwater flux averaged for all summer seasons as simulated with STRIVE.

conductivity, Kh, of the streambed was calculated from rising and
falling head slug tests performed in piezometer nest 13. The aver-
age estimated Kh of the riverbed was 2.7 105 m s1. Anisotropy ra-
tios of horizontal over vertical hydraulic conductivity, Kh/Kv, of 3–
10 were tested to calculate vertical groundwater fluxes. A ratio of
nine resulted in a lowest absolute difference of 22.4 mm d1 be-
tween the estimated fluxes based on the hydraulic head gradients
and the heat transport simulation. With a comparable approach
Schmidt et al. (2006) obtained a similar difference. The exchange
fluxes based on the measured hydraulic heads follow the spatial
trend observed from the heat transport investigation, indicating
highest discharge at location 13, followed by a decreasing trend to-
wards location 4. The temporal trend also corresponds with esti-
mated fluxes from the heat transport; significantly higher
discharges are observed in winter than in summer. We conclude
that the flux estimated with thermal and hydraulic head methods
show similar temporal and spatial patterns and acceptable abso-
lute differences.
4.2. Seasonal differences
Seasonal and diurnal transient influences can be taken into ac-
count prior to the measurement or simulation by measuring at cer-
tain times in the year or by excluding the area close to the interface
100 C. Anibas et al. / Journal of Hydrology 397 (2011) 93–104

Fig. 9. Bar graph of the vertical groundwater fluxes of measurement points along the Aa River (1 is downstream; 14 is upstream) as derived from the STRIVE model; (a) shows
the median of the winter results; green, blue and red bars indicate the measurements at the left bank, centerline and right bank respectively; (b) presents the results for the
summer season. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

from the simulation. However, this is usually not possible with
irregular, hence meteorological events such as phases of relative
cold or warm weather. The effect of the irregular influences on
the profiles becomes clearly visible when temperature profiles of
different winter seasons are compared. Two groups, profiles with
a steep gradient (e.g. 26 January 2005 or 03 July 2006) and with
a relative low gradient (e.g. 14 January 2007 or 08 September
2006) can be distinguished in Fig. 6 by their different shapes and
gradients. When these profiles are fitted with the steady state heat
transport model higher estimated fluxes are obtained for the steep
profiles as confirmed by a Mann–Whitney U test (N = 127) and a
Kolmogorow–Smirnow test (N = 127).
The strong thermal variations in deeper layers as shown in Fig. 6
is a phenomenon not explained by flux changes unless a significant
infiltration is assumed. Infiltration is however not supported by
piezometer measurements. If we assume that irregular meteoro-
logical influences are responsible for the different temperature dis-
tributions within the seasons, our analysis shows an error of the
flux estimation of around ±50%. This is exemplified by the median
results of the summer campaigns, which vary between 21 and
38 mm d1, and the results for winter between 23 and
49 mm d1.
Since a thermal steady state is approached both in summer and
winter (Anibas et al., 2009), we hypothesize that seasonal differ-
ences in groundwater–surface water interaction can be evaluated
with the presented methodology. Kolmogorov–Smirnov two sam-
ple (N = 135) and Mann–Whitney U tests (N = 135) confirm that
the Aa River has a different hydrological flow regime in winter
and summer. The median flux of all winter profiles along the center
of the river is 36 mm d1 while the summer flux is 27 mm d1.
A transient thermal simulation with STRIVE over several
months in 2006 (Anibas et al., 2009) results in an average flux of
24 mm d1 at measurement point 7, showing that the magnitude
of flux can be reproduced well with the steady state simulation
(Figs. 7 and 8).
4.3. Net flows
Spatially distributed plots of the groundwater–surface water
interaction along the examined river reach were produced by
interpolating the median flux values of all 26 points on the river
surface area of 20,400 m2. For the interpolation a multilog radial
bases function (Surfer 8.04, Golden Software, 2003) was used with
R2 = 1800, anisotropy ratio 2.5 and an angle of 20°, representing the
morphology and orientation of the river course (Fig. 10). GIS tech-
niques were used to determine the net fluxes between groundwa-
ter and surface water body (Table 2).
The spatial interpolation (Fig. 10) resulted in a median flux of
44 and 55 mm d1 for summer and winter respectively (Table 2,
No. 1), i.e. a seasonal difference of 23%. This is equivalent to net
groundwater discharges of 10.5 L s1 in summer and 12.9 L s1 in
winter (Table 2, No. 2). These groundwater discharges are respec-
tively 0.7% and 0.4% of the average summer and winter discharge
at weir 4. Hence, the groundwater discharge has a relatively higher
contribution during summer than in winter. Extrapolating these
discharges over the full length of the Aa River (about 37 km) and
comparing with long-term average total river discharge we esti-
mate a groundwater contribution to the main channel of about
15%. Baseflow separation for the Aa River resulted in a baseflow
percentage of 75% (Batelaan and De Smedt, 2007). Hence, it is clear
that although groundwater discharge through the riverbed is sig-
nificant (15%), the major component (60%) of baseflow is drainage
from tributaries and ditches in the Aa River catchment, which is
obvious from presence of the dense network (Fig. 3).
Beside minimum and maximum fluxes (Table 2, Nos. 3–4), Ta-
ble 2 (Nos. 5 and 8) also shows discharge values of the 90th and
10th percentiles of the interpolated fluxes. The 90th percentile is
not changing significantly between the seasons (Table 2, No. 5),
whereas the 10th percentile shows an increase of 52% from sum-
mer to winter (Table 2, No. 8). The maximum fluxes in the river
thus do not change significantly over the seasons (Table 2, No. 3),
 C. Anibas et al. / Journal of Hydrology 397 (2011) 93–104 101

Fig. 10. Spatial interpolation of the median groundwater flux along the Aa River of (a) six winter and (b) five summer campaigns 2004–2007. The width of the river is five
times exaggerated.

Table 2 Table 3
Summary of estimates of flux, net discharge and discharge area as derived from Advantages and disadvantages of the steady-state thermal method for estimating
spatially interpolated groundwater–surface water interaction along the Aa River for groundwater–surface water interaction.
summer and winter.
Advantages
Parameter Unit Summer Winter Cost-effective
Simple field measurement
Median net flux mm d1 44 55
Allows spatially distributed data collection
Net dischargea L s1 10.5 12.9
Small range of parameter values
Maximum flux mm d1 75 77
Minimal modeling effort
Minimum flux mm d1 22 36
Short simulation time
90th percentile flux mm d1 73 75
Eludes limitations of other methods
Discharge area 90th percentile %b 9.6 13.4
Net discharge 90th percentile %c 16.2 18.6 Disadvantages
10th percentile flux mm d1 25 38 Indirect method
Discharge area 10th percentile %b 7.9 22.6 Point estimate
Net discharge 10th percentile %c 4.8 15.5 Purely vertical discharge estimation
Only exfiltrating conditions
a
Surface area of the river stretch 20,400 m2. Limited temporal resolution
b
Percentage of discharge area of total river surface area of 20,400 m2 with dis- Measurement only useful in summer and winter
charge higher than 90th (or lower than 10th) percentile. Significant but unknown error band
c
Percentage of the net discharge.

whereas the minimum fluxes alter between winter and summer
(Table 2, No. 4). Consequently, there are seasonal differences in
the discharge areas, as follows from the 90th and 10th percentiles,
which occupy respectively 10% and 8% of the river surface area in
summer, while in winter these areas comprise respectively 13%
and 23% (Table 2, Nos. 6 and 9). Thus the increased total ground-
water discharge (Table 2, No. 2) in winter is mostly caused by an
increase of low net discharges (Table 2, No. 10) and its correspond-
ing discharge area (Table 2, No. 9). The areas indicating high net
discharges show however a gentle increase (Table 2, No. 7).
The spatially interpolated fluxes can be used to verify ground-
water models or vice versa. Woldeamlak et al. (2007) and Dams
et al. (2008) developed a groundwater model for the catchment
in which the Aa River is situated. WetSpass (Batelaan and De
Smedt, 2001, 2007) was used to simulate groundwater recharge
to condition this steady state MODFLOW model (Harbaugh and
McDonald, 1996) with a 50 m grid cell resolution. Drain and river
leakages were extracted for the Aa River section from this model,
resulting in a spatial average of the leakage flux of 15.3 mm d1.
This value is significantly lower as estimated with the heat trans-
port model (Table 2, No. 1). The spatial distribution of the fluxes
however is similar, showing high fluxes upstream and low fluxes
downstream. The lateral extent of the groundwater model is signif-
icantly larger and the horizontal and vertical resolution of the
model is much lower than the spatial scales of the roaming T-stick
surveys. It is therefore likely that the differences in estimated
fluxes can be attributed to the differences in resolution and the
steady state character of the groundwater model.
102 C. Anibas et al. / Journal of Hydrology 397 (2011) 93–104
5. Conclusions
The presented methodology indirectly detects the movement of
water in the subsurface by using heat as a tracer. The thermal steady
state investigation results in an acceptable flux approximation.
Moreover, measurements of thermal profiles are easy to perform
(e.g. with the T-stick), very robust, cost-effective and have proven
their reliability in different applications. A major advantage is cer-
tainly the fact that spatial differences can be detected by performing
‘roaming surveys’, collecting data from multiple locations. Such a
spatial distribution cannot be achieved with temperature measure-
ments in boreholes or piezometers, or with most other field methods.
Furthermore, practical problems of temperature measurements in
piezometers can be overcome with a T-stick, including thermal
instabilities of the water column or problems related to heat trans-
port along and across the piezometer pipe. Opposite to the high sen-
sitivity of hydraulic gradient measurements, the thermal method
shows a high robustness regarding vertical misplacement of the tem-
perature sensor. This robustness is explained by the fact that the
maximum measurement error is small with respect to the whole
model domain. The measurement errors have the largest impact
when they occur close to the upper boundary. The cumulative mea-
surement error close to the interface however is still small in case a T-
stick like instrument is used. Despite the uncertainties of the thermal
steady state methodology it can be used as a ‘stand alone’ technique
to estimate the groundwater–surface water interaction at first hand.
If lower uncertainty is required, we recommend thermal modeling in
transient mode and/or in combination with other methods. The used
models have small computational requirements, while their input
parameters have a limited range, which can be assumed from litera-
ture. Table 3 briefly summarizes the advantages and disadvantages
of the presented methodology.
The steady state approach of Arriaga and Leap (2006) proved to
be very efficient as the data preparation, model setup and calibra-
tion was less time consuming than for the STRIVE model. However,
STRIVE can also be used for transient simulation (Anibas et al.,
2009) if continuous data sets are available. Beside the discussed
use, temperature is an important ecological parameter and moder-
ates biological and chemical processes in groundwater and surface
water systems; hence the measured temperatures can be used as
input parameter for various other models examining biogeochem-
ical transport, reaction and retention processes. The thermal steady
state model performed best with a domain of several meters (e.g.
5.0 m) thickness with a constant groundwater temperature at the
lower boundary. The fluxes are integrated over the model domain,
which can explain the difference between the estimated fluxes and
results of other models or data from piezometer nests, as they have
different horizontal, vertical and temporal scales. The presented
models however show similar spatial and temporal trends.
Since a thermal and hydrological steady state has to be assumed
the applicability of the methodology is limited to summer and win-
ter and to locations where no sudden changes of the hydrological
conditions are expected. Diurnal thermal transient influences have
an effect at the upper model boundary, regardless the size of the
river system investigated. We confirm Schmidt et al. (2006) find-
ings that the temperatures of the uppermost section of the river-
bed should not be used for a steady-state thermal analysis. This
leads to a loss of measured information and limits the model do-
main that can be investigated with an instrument like the T-stick.
To maximize the trade-off between temperature information and
diurnal transient influences we propose to use a temperature mea-
sured between 0.10 and 0.20 m below the interface as upper model
boundary.
Corrections for diurnal and seasonal thermal influences can be
considered before the analysis. This is however not the case for
the effects of irregular meteorological events. Measuring at times
of strong changes in weather conditions (e.g. at a drop in temper-
ature with an incoming rain front) is therefore not recommended.
Preferably, the temperature measurements should be performed
during stable weather conditions of three or more days. These in-
creases on one hand the accuracy of the method, but also reduces
its flexibility. Temperature variations between entire winter and
summer seasons of different years (Fig. 6) cannot be avoided what-
soever. We regard these influences on the simulated flux results as
acceptable with respect to the limitations of the thermal method as
well as in comparison to uncertainties of other methods used for
estimation of groundwater–surface water exchange.
For relatively shallow rivers of up to several kilometer length,
we suggest a hierarchical setup of measurement campaigns, where
an initial survey using the thermal method can detect places of
special interest (McClain et al., 2003). Consequently, as presented
in this article, the survey can be extended by successive investiga-
tions, allowing detection of seasonal differences or spatial varia-
tions. Results of the initial survey are guiding continued research
efforts and help in establishing measurement locations for gauges,
piezometers, seepage meters, etc. This hierarchical setup assures
capturing the significant features of groundwater–surface water
interaction and increases the efficiency of the employed
instrumentation.
From the spatial interpolation for the River Aa, and supported
by statistical analysis it is concluded that in winter the absolute
flux of groundwater–surface water exchange is higher than in
summer. This difference is attributed to the in general higher
groundwater head gradients in winter. But, as the surface water
discharge changes significantly with the seasons, the relative con-
tribution of exfiltration to the river discharge (i.e. the baseflow) is
higher during the summer season. The bank areas contribute rela-
tively more to the exchange than the center part of the river
stretch. They also show stronger variability of fluxes. This indicates
the importance of the banks in groundwater–surface water inter-
action. Local shallow groundwater flow systems and likely higher
hydraulic conductivities of the bank zones are thought to be
responsible for these phenomena.
The upper reach of the examined stretch of the Aa River showed
in general a higher flux and more heterogeneous flow patterns
than the lower reach. Different mechanisms can be detected, the
hydrogeology of the subsurface of the riverbed (Fig. 4), shallow
groundwater flow mechanisms at the banks, deep flow at the cen-
ter of the river channel, the morphology of the river with respect to
the surrounding alluvium, the topography (Fig. 3) and the bathym-
etry of the riverbed. It is concluded that these difference between
the upper and the lower part are the expression of a combination
of causes of which topography and hydrogeology are dominant.
Spatial interpolation techniques can be used to calculate net
discharge fluxes from the riverbed towards the river, on a scale
were other methods like differential gauging of surface water dis-
charge are below their detection limit. The estimated fluxes indi-
cate that the riverbed of the main channel of the whole Aa River
account for about 15% of the total river discharge at its outlet. It
is concluded that the remaining about 60% of the baseflow of the
Aa is produced all over the catchment by diffuse discharge in small
tributaries, ditches, drains, swamps, etc. The spatial interpolation
indicates that the significant increase in net flux between summer
and winter is due to an increase in low fluxes as well as an enlarge-
ment of the zones with these relatively low fluxes. On the other
hand regions with high fluxes increased moderately in extent
and flux rates.
Eventually, the flux results of the thermal method are useful as
input for calibration or validation of distributed hydrological or
groundwater flow models. For reducing model uncertainty these
fluxes are complementary to the traditional head calibration data.
The presented methodology for estimation of fluxes by thermal
 C. Anibas et al. / Journal of Hydrology 397 (2011) 93–104 103

modeling proves to be a simple and reliable method, improves the European Commission, 2000. Directive 2000/60/EC of the European Parliament and
of the Council establishing a framework for Community action in the field of
understanding of groundwater–surface water interaction.
water policy. Official Journal of the European Community L327, 1–72.
Fairley, J.P., Nicholson, K.N., 2005. Imaging lateral groundwater flow in the shallow
subsurface using stochastic temperature fields. Journal of Hydrology 321, 276–
Acknowledgements
285. doi:10.1016/j.jhydrol.2005.08.017.
Fleckenstein, J., Anderson, M., Fogg, G., Mount, J., 2004. Managing surface water–
Financial support from the Research Foundation-Flanders groundwater to restore fall flows in the Consumnes River. Journal of Water
(FWO) for the work on the Aa River as part of the project ‘A funda- Resources Planning and Management – ASCE 130 (4), 301–310. doi:10.1016/
(ASCE)0733-9496(2004)130:4(301).
mental study on exchange processes in river ecosystems’ Freeze, R.A., Cherry, J.A., 1979. GroundWater. Prentice-Hall Inc., Englewood Cliffs,
(G.0306.04) is greatly appreciated. Further support was received NJ.
from the Belgian Science Policy Office in the framework of the MA- Harbaugh, A.W., McDonald, M.G., 1996. User’s documentation for MODFLOW-96, an
update to the US Geological Survey modular finite-difference ground-water
MUD (SR/00/105) and FRAC-WECO (SD/TE/02) projects. flow model. US Geological Survey Open-file Report 56, 96–485.
We would also like to thank the numerous assistants for their Hatch, C.E., Fisher, A.T., Revenaugh, J.S., Constantz, J., Ruehl, C., 2006. Quantifying
invaluable help in the field. surface water–groundwater interactions using time series analysis of
streambed thermal records: method development. Water Resources Research
42 (10), 1–5. doi:10.1029/2005WR004787.
References Hayashi, M., Rosenberry, D.O., 2002. Effects of groundwater exchange on the
hydrology and ecology of surface water. Ground Water 40 (3), 309–316.
Hoehn, E., Cirpka, O.A., 2006. Assessing residence times of hyporheic ground water
Anderson, M.P., 2005. Heat as a groundwater tracer. Ground Water 43 (6), 951–968.
in two alluvial flood plains of the Southern Alps using water temperature and
Anibas, C., Fleckenstein, J., Volze, N., Buis, K., Verhoeven, R., Meire, P., Batelaan, O.,
tracers. Hydrology and Earth System Sciences 10, 553–563.
2009. Transient or steady-state? Using vertical temperature profiles to quantify
Kalbus, E., Reinstorf, F., Schirmer, M., 2006. Measuring methods for groundwater–
groundwater–surface water exchange. Hydrological Processes 23, 2165–2177.
surface water interactions: a review. Hydrology and Earth System Sciences 10,
doi:10.1002/hyp.7289.
873–887.
Arriaga, M.A., Leap, D.I., 2006. Using solver to determine vertical groundwater
Kalbus, E., Schmidt, C., Reinstorf, F., Krieg, R., Schirmer, M., 2008. How streambed
velocities by temperature variations, Purdue University, Indiana, USA.
temperatures can contribute to the determination of aquifer heterogeneity.
Hydrogeology Journal 14 (1–2), 253–263. doi:10.1007/s10040-004-0381-x.
Grundwasser 13 (2), 91–100. doi:10.1007/s00767-008-0066-9.
Bal, K., Meire, P., 2009. The influence of macrophyte cutting on the hydraulic
Kalbus, E., Schmidt, C., Molson, J.W., Reinstorf, F., Schirmer, M., 2009. Influence of
resistance of Lowland Rivers. Journal of Aquatic Plant Management 47, 65–68.
aquifer and streambed heterogeneity on the distribution of groundwater
Batelaan, O., De Smedt, F., 2001. WetSpass: a flexible, GIS based, distributed
discharge. Hydrology and Earth System Sciences 13 (1), 69–77.
recharge methodology for regional groundwater modeling. In: Gehrels, H.,
Keery, J., Binley, A., 2007. Temperature Measurements for Determining
Peters, J., Hoehn, E., Jensen, K., Leibundgut, C., Griffioen, J., Webb, B.,
Groundwater–Surface Water Fluxes. Environment Agency-science Report,
Zaadnoordijk, W.J. (Eds.) Impact of Human Activity on Groundwater
SC030155/SR9, Bristol, United Kingdom.
Dynamics. IAHS Publ. 269, IAHS, Wallingford, UK, pp. 11–17.
Keery, J., Binley, A., Crook, N., Smith, J.W.N., 2007. Temporal and spatial variability of
Batelaan, O., De Smedt, F., 2007. GIS-based recharge estimation by coupling
groundwater–surface water fluxes: development and application of an
surface–subsurface water balances. Journal of Hydrology 337 (3–4), 337–355.
analytical method using temperature time series. Journal of Hydrology 336
doi:10.1016/j.jhydrol.2007.02.001.
(1–2), 1–16. doi:10.1016/j.jhydrol.2006.12.003.
Baxter, C., Hauer, F.R., Woessner, W.W., 2003. Measuring groundwater–stream
Lapham, W.M., 1989. Use of Temperature Profiles Beneath Streams to Determine
water exchange: new techniques for installing minipiezometers and estimating
Rates of Vertical Ground-water Flow and Vertical Hydraulic Conductivity.
hydraulic conductivity. Transactions of the American Fisheries 132 (3), 493–
Water-Supply Paper 2337, US Geological Survey.
502.
Lee, D.R., 1977. Device for measuring seepage flux in lakes and estuaries. Limnology
Becker, M.W., 2006. Potential for satellite remote sensing of groundwater. Ground
and Oceanography 22 (1), 140–147.
Water 44 (2), 306–318. doi:10.1111/j.1745-6584.2005.00123.x.
Loheide, S.P., Gorelick, S.M., 2006. Quantifying stream–aquifer interactions through
Becker, M.W., Georgian, T., Ambrose, H., Siniscalchi, J., Fredrick, K., 2004. Estimating
the analysis of remotely sensed thermographic profiles and in situ temperature
flow and flux of groundwater discharge using water temperature and velocity.
histories. Environmental Science and Technology 40 (10), 3336–3341.
Journal of Hydrology 296 (1–4), 221–233. doi:10.1016/j.jhydrol.2004.03.025.
doi:10.1021/es0522074.
Boulton, A.J., Findlay, S., Marmonier, P., Stanley, E.H., Valett, H.M., 1998. The
Lowry, C.S., Walker, J.F., Hunt, R.J., Anderson, M.P., 2007. Identifying spatial
functional significance of the hyporheic zone in streams and rivers. Annual
variability of groundwater discharge in a wetland stream using a distributed
Review of Ecology and Systematics 29, 59–81.
temperature sensor. Water Resources Research 43 (10), W10408. doi:10.1029/
Bredehoeft, J.D., Papadopulos, I.S., 1965. Rates of vertical groundwater movement
2007WR006145.
estimated form the Earth’s thermal profile. Water Resources Research 1 (2),
Marechal, R., Laga, P. (Eds.), 1988. Voorstel Lithostratigrafische Indeling van het
325–328.
Paleogeen. Nationale Commissie Voor Stratigrafie, Commissie Tertiair, p. 208.
Brunke, M., Gonser, T., 1997. The ecological significance of exchange processes
McClain, M.E., Boyer, E.W., Dent, C.L., Gergel, S.E., Grimm, N.B., Groffmann, P.M.,
between rivers and groundwater. Freshwater Biology 37 (1), 1–33.
Hart, S.C., Harvey, J.W., Johnston, C.A., Mayorga, E., McDowell, W.H., Pinay, G.,
Cardenas, M.B., Zlotnik, V.A., 2003. Three-dimensional model of modern channel
2003. Biogeochemical hot spots and hot moments at the interface of terrestrial
bend deposits. Water Resources Research 39 (6), 1141. doi:10.1029/
and aquatic ecosystems. Ecosystems 6 (4), 301–312.
2002WR001383.
Murdoch, L.C., Kelly, S.E., 2003. Factors affecting the performance of conventional
Carey, S.K., Quinton, W.L., 2005. Evaluating runoff generation during summer using
seepage meters. Water Resources Research 39 (6), 1163. doi:10.1029/
hydrometric, stable isotope and hydrochemical methods in a discontinuous
2002WR001347.
permafrost alpine catchment. Hydrological Processes 19 (1), 95–114.
National Research Council (NRC), 2004. Groundwater Fluxes Across Interfaces.
Cey, E.E., Rudolph, D.L., Parkin, G.W., Aravena, R., 1998. Quantifying groundwater
National Academic Press, Washington, DC, p. 85.
discharge to a small perennial stream in southern Ontario, Canada. Journal of
Rosenberry, D.O., 2008. A seepagemeter designed for the use in flowing water.
Hydrology 210, 21–37.
Journal of Hydrology 359 (1–2), 118–130. doi:10.1016/j.jhydrol.2008.06.029.
Chen, X.H., 2000. Measurement of streambed hydraulic conductivity and its
Saenger, N., Kitanidis, P.K., Street, R.L., 2005. A numerical study of surface–
anisotropy. Environmental Geology 39 (12), 1317–1324.
subsurface exchange processes at a riffle-pool pair in the Lahn River, Germany.
Commission, European., 2006. Directive 2006/118/EC of the European Parliament
Water Resources research 41, W12424. doi:10.1029/2004WR003875.
and of the Council on the protection of groundwater against pollution and
Schmidt, C., Bayer-Raich, M., Schirmer, M., 2006. Characterization of spatial
deterioration. Official Journal of the European Community L372, 19–31.
heterogeneity of groundwater–stream water interactions using multiple
Conant Jr., B., 2004. Delineating and quantifying ground water discharge zones
depth streambed temperature measurements at the reach scale. Hydrology
using streambed temperatures. Ground Water 42 (2), 243–257.
and Earth System Sciences 10, 849–859.
Constantz, J., 2008. Heat as a tracer to determine streambed water exchanges.
Schmidt, C., Conant Jr., B., Bayer-Raich, M., Schirmer, M., 2007. Evaluation and field-
Water Resources Research 44, W00D10. doi:10.1029/2008WR006996.
scale application of an analytical method to quantify groundwater discharge
Constantz, J., Niswonger, R.G., Steward, A.E., 2008. Analysis of temperature
using mapped streambed temperatures. Journal of Hydrology 347, 292–307.
gradients to determine stream exchanges with ground water. In: Rosenberry,
doi:10.1016/j.jhydrol.2007.08.022.
D.O., LaBaugh, J.W. (Eds.), Field Techniques for Estimating Water Fluxes
Schmidt, C., Kalbus, E., Krieg, R., Bayer-Raich, M., Leschik, S., Reinstorf, F.,
Between Surface Water and Ground Water. Techniques and Methods, US
Martienssen, M., Schirmer, M., 2008. Schadstoffmassenströme zwischen
Geological Survey Report (4-D2).
Grundwasser, Flussbettsedimenten und Oberflächenwasser am regional
Dams, J., Woldeamlak, S.T., Batelaan, O., 2008. Predicting land-use change and its
kontaminierten Standort Bitterfeld. Grundwasser 13, 133–146. doi:10.1007/
impact on the groundwater system of the Kleine Nete catchment, Belgium.
s00767-008-0076-7.
Hydrology and Earth System Sciences 12, 1369–1385.
Schön, J.H., 1998. Physical Properties of Rocks, Fundamentals and Principles of
Domenico, P.A., Schwartz, F.W., 1998. Physical and Chemical Hydrogeology, second
Petrophysics, second ed. Pergamon.
ed. John Wiley and Sons Inc., New York.
Shaw, R., Prepas, E., 1990. Groundwater–lake interactions: I. Accuracy of seepage
DOV, 2010. Databank Ondergrond Vlaanderen. Database Subsurface of Flanders.
meter estimates of lake seepage. Journal of Hydrology 119, 105–120.
<https://dov.vlaanderen.be/>.
104 C. Anibas et al. / Journal of Hydrology 397 (2011) 93–104
Smith, J.W.N., 2005. Groundwater–Surface Water Interactions in the Hyporheic
Zone. Environment Agency-science Report SC030155/SR1, Bristol, United
Kingdom.
Soetaert, K., De Clippele, V., Herman, P., 2002. FEMME, a flexible environment for
mathematically modeling the environment. Ecological Modelling 151 (2–3),
177–193.
Stallman, S., 1965. Steady one-dimensional fluid flow in the semi-infinite porous
medium with sinusoidal surface temperature. Journal of Geophysical Research
70 (12), 2821–2827.
Stanford, J.A., 1998. Rivers in the landscape: introduction to the special issue on
riparian and groundwater ecology. Freshwater Biology 40, 402–406.
Stonestrom, D.A., Blasch, K.W., 2003. Determining Temperature and Thermal
Properties for Heat-based Studies of Surface-water Ground-water
Interactions. USGS Circular 1260, Reston, Virginia, pp. 73–80 (Appendix A).
Stonestrom, D.A., Constantz, J., 2003. Heat as a Tool for Studying the Movement of
Groundwater near Streams. USGS Circular 1260, Reston, Virginia.
Suzuki, S., 1960. Percolation measurements based on heat flow through soil with
special reference to paddy fields. Journal of Geophysical Research 65 (9), 2883–
2885.
Woessner, W.W., 2000. Stream and fluvial plain ground water interactions:
rescaling hydrogeologic thought. Ground Water 38 (3), 423–429.
Woldeamlak, S., 2007. Spatio-temporal Impacts of Climate and Land-use Changes
on the Groundwater and Surface Water Resources of a Lowland Catchment. PhD
Thesis, Vrije Universiteit Brussel, Brussels, Belgium.
Woldeamlak, S., Batelaan, O., De Smedt, F., 2007. Effects of climate change on the
groundwater system in the Grote-Nete catchment, Belgium. Hydrogeology
Journal 15 (5), 891–901. doi:10.1007/s10040-006-0145-x.