HYDROLOGICAL PROCESSES

Hydrol. Process. 29, 965–978 (2015)

Published online 7 May 2014 in Wiley Online Library
(wileyonlinelibrary.com) DOI: 10.1002/hyp.10209

Groundwater temperature evolution in the subsurface urban

heat island of Cologne, Germany
Ke Zhu,1,3 Peter Bayer,2* Peter Grathwohl1 and Philipp Blum3
1
Center for Applied Geosciences (ZAG), University of Tübingen, Hölderlinstr. 12, 72074, Tübingen, Germany
2
Department of Earth Sciences, ETH Zurich, Sonneggstrasse 5, 8092, Zurich, Switzerland
3
Institute for Applied Geosciences (AGW), Karlsruhe Institute of Technology (KIT), Kaiserstraße 12, 76131, Karlsruhe, Germany

Abstract:
Long-term heating of shallow urban aquifers is observed worldwide. Our measurements in the city of Cologne, Germany
revealed that the groundwater temperatures found in the city centre are more than 5 K higher than the undisturbed background.
To explore the role of groundwater flow for the development of subsurface urban heat islands, a numerical flow and heat
transport model is set up, which describes the hydraulic conditions of Cologne and simulates the transient evolution of thermal
anomalies in the urban ground. A main focus is on the influence of horizontal groundwater flow, groundwater recharge and
trends in local ground warming. To examine heat transport in groundwater, a scenario consisting of a local hot spot with a length
of 1 km of long-term ground heating was set up in the centre of the city. Groundwater temperature-depth profiles at upstream,
central and downstream locations of this hot spot are inspected. The simulation results indicate that the main thermal transport
mechanisms are long-term vertical conductive heat input, horizontal advection and transverse dispersion. Groundwater recharge
rates in the city are low (<100 mm a 1) and thus do not significantly contribute to heat transport into the urban aquifer. With
groundwater flow, local vertical temperature profiles become very complex and are hard to interpret, if local flow conditions and
heat sources are not thoroughly known. Copyright © 2014 John Wiley & Sons, Ltd.

KEY WORDS subsurface urban heat island (UHI); simulation; groundwater temperature-depth (GWTD) profile; dispersion;
Cologne
Received 1 October 2013; Accepted 26 March 2014

INTRODUCTION tunnels, as well as from buried electrical power lines, district

heating networks, sewage water canals and leakages (Balke,
In the last century, a strong warming trend in urban
1977; Pollack et al., 1998; Menberg et al., 2013b).
subsurface has been revealed in many large cities such as
Temperature fluctuations at the surface or in shallow
Tokyo, Bangkok and Berlin (Taniguchi et al., 2007;
depth penetrate further into the ground. They can be seen in
Menberg et al., 2013a). Field studies in North America
deeper borehole temperature logs as deviation from the
(Wang et al., 1994; Ferguson and Woodbury, 2007), Europe
unperturbed, natural temperature profiles (Huang et al.,
(Balke, 1977; Bodri and Cermak, 1997; Perrier et al., 2005;
2000; Taniguchi et al., 2007). Numerous studies in borehole
Menberg et al. 2013a) and Asia (Taniguchi et al., 2007;
climatology examined vertical temperature logs that are not
Huang et al., 2009; Yamano et al., 2009) have indicated a
suspected of being affected by non-climatic factors. Most of
rise of the regional urban underground temperature by 2–5 K.
them identified a warming trend and increasing downward
The factors that raise the atmospheric temperature in cities
heat flux during the last century (e.g. Pollack et al., 1998;
and lead to the evolution of atmospheric urban heat islands
Huang et al., 2000; Perrier et al., 2005). Only a few focused
(UHIs) are also responsible for ground warming. These
on perturbed profiles (e.g. Taniguchi et al., 1999;
factors are, for example, solar heating of massive urban
Gunawardhana and Kazama, 2011). Urban ground surface
structures, such as buildings, and sealed surfaces, such as
temperature history was interpreted, for example, by
roads. UHIs in the subsurface are augmented by further non-
Yamano et al. (2009), who logged more than 100 boreholes
climatic perturbations, such as in-ground heat losses from
of generally 100–250 m depth in the Asian megacities
buildings, landfills, underground parking lots and subway
Bangkok, Jakarta, Taipei and Seoul. It was discovered that
ground surface temperatures substantially increased during
the last century, much more than could be inferred by
*Correspondence to: Peter Bayer, ETH Zurich, Department of Earth
Sciences, Sonneggstrasse 5, 8092 Zurich, Switzerland analytical inversion from logs in rural surroundings.
E-mail: bayer@erdw.ethz.ch Repeated measurements in some boreholes were carried

Copyright © 2014 John Wiley & Sons, Ltd.

966 K. ZHU ET AL.

out by Hamamoto et al. (2008) to inspect the stability of the evolution of a subsurface UHI, and especially to explore the
temperature logs. Instability at greater depths (without role of regional horizontal groundwater flowing at much
seasonal influence) was considered to be an indicator in higher groundwater flow velocities than studied by
variability of groundwater flow velocity. Perturbation of the Ferguson and Woodbury (2004). We focus on the city of
logs by horizontal and vertical advective heat flow was Cologne, Germany where repeated measurements in the
assumed to be the reason that unrealistic surface cooling was shallow groundwater revealed elevated temperatures by
inferred for the time before 1900, because their forward more than 5 K (Zhu et al., 2010; Menberg et al., 2013a).
model assumes vertical conduction only. Little is known about the long-term effects that lead to the
Plenty of previous studies to analyse borehole logs and extensive subsurface UHI reaching a depth of more than
subsurface heat transport are based on the assumption that 100 m. The currently available information is insufficient for
conduction is the dominant factor (e.g. Pollack et al., 1998; a detailed numerical study including all natural and
Beltrami et al. 2006). Although, the assumption of one- anthropogenic urban heat sources. Instead, here, a realistic
dimensional vertical heat conduction might not be suitable scenario is studied by considering potential long-term trends
in mixed, conduction–advection controlled systems (Bense of increased ground temperatures, which are implemented
and Beltrami, 2007), which are characteristic in the presence into a site-specific, simplified numerical model.
of groundwater. For instance, downward groundwater flow
might change temperature profiles into a shape that looks
similar to the one caused by surface warming, and the initial METHODOLOGY
and boundary conditions are not always linear (Kurylyk and
MacQuarrie, 2013). Accordingly, for subsurface UHIs, the In the following, first, the case study of Cologne is
applicability of analytical reconstruction methods introduced, and the data survey and measurements in the
established in borehole climatology is restricted. The early study area are described. Then, a three-dimensional (3D)
works by Suzuki (1960) and Stallman (1963) provide groundwater flow model is developed to capture the regional
solutions for solving one-dimensional conduction– hydrogeological conditions. On the basis of this ground-
advection equations, which can be used to estimate the water flow model, a vertical cross-section was chosen for a
vertical groundwater flow rate, but require a given surface two-dimensional (2D) flow and heat transport model. We
temperature distribution. Even though both vertical and/or examine different scenarios with variable assumptions for
horizontal advection components and surface temperature elevated ground temperatures, recharge rates and advective–
changes can be included in analytical solutions (Domenico dispersive effects from horizontal groundwater flow. By
and Palciauskas, 1973; Lu and Ge, 1996; Kurylyk and comparison of simulated to measured groundwater tem-
MacQuarrie, 2013), limitations remain, for instance, typical perature-depth (GWTD) profiles, the main factors that
layering of aquifers and thermal dispersion are neglected. influence subsurface UHI heat transport are elucidated.
Furthermore, for many cities, deep boreholes are typically
rather scarce. This limits monitoring of continuous
temperature profiles down to the undisturbed geothermal DATA COLLECTION
gradient. Alternatively, a numerical modelling framework
may be adopted (Ferguson and Woodbury 2004; Epting Data survey
et al. 2013). Although a numerical model can be more This case study is conducted in the city of Cologne, where
flexible and accurate, this comes at the expense of an the climatic, such as past air temperatures (Figure 1),
enormous data requirement. Ideally, the selected numerical geological and hydrogeological information was collected.
approach finds a compromise between data requirements The city lies on the Rhine River. It is one of the largest
and accuracy, which is intended in our study. Ferguson and cities in Germany, with a population of around one million.
Woodbury (2004), for example, set up generic 2D numerical We focus on the main city part, which is located on the
models to simulate the effect of horizontal groundwater flow western side of the Rhine. The study area is depicted in
on the long-term development of a heat anomaly beneath Figure 2, which also illustrates the subsequent model
single and multiple neighbouring buildings. Their study is implementation.
oriented at the conditions representative for the city of The mean annual air temperature back in 1900, when the
Winnipeg, with a minor groundwater flow velocity of city in Cologne started growing fast, is about 8.7 °C
4.0 × 10 7 m s 1 (0.03 m day 1) in the simulated carbonate (Figure 1). Because no earlier air temperature measurements
aquifer layer. Groundwater flow is shown to have only a are available for Cologne, we added data from 1900 to 1945
minimal effect on the evolution of bulb-shaped thermal recorded in Aachen (DWD, 2010), which is only 70 km
anomalies evolving in the modelled vertical cross sections. away. Linear regression for the whole period since 1900
The main objective of our work is to understand the reveals a mean linear increase of around 0.03 °C per year.
interplay of conductive and advective forces during the The average regional annual precipitation from 1961 to

Copyright © 2014 John Wiley & Sons, Ltd. Hydrol. Process. 29, 965–978 (2015)
 GROUNDWATER TEMPERATURE EVOLUTION IN THE SUBSURFACE URBAN HEAT ISLAND 967

14
15 wells were repeatedly recorded in the city. The
13 measurements indicate that the water table is relatively
12
stable in most of the wells distant from the river, but in the
Air temperature (°C)

northeastern part near the river, the water levels vary by

11 more than 1 m (data not shown). Despite this local and
10
temporal variability, however, the derived main groundwa-
ter flow direction is always from west towards east with a
9 calculated hydraulic gradient ranging between 5 × 10 4 and
8
Cologne
7 × 10 4, and no river infiltrating conditions are observed in
Aachen
the study area.
7
1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010
In the present work, we focus on the groundwater
Year (A.D.) temperature data as reported by Zhu et al. (2010). GWTD
profiles at 1 m intervals were recorded using SEBA
Figure 1. Mean annual air temperature recorded from 1900 to 2010 by a
weather station in Stammheim in the northeastern part of Cologne, and by KLL-T logging equipment. Well diameters range from 2
a weather station in Aachen (DWD, 2010) to 5 in (0.05–0.127 m), and the wells reach depths
between 20 and 44 m. The positions of these wells are
spatially distributed and associated with different types of
1990 was around 774 mm (DWD, 2006). Groundwater land use, including built environment, green spaces and
recharge in Germany is commonly only about 16% of the agricultural area. From the inspected 46 wells, ten wells
precipitation (BGR, 2008) and is mainly concentrated in the (Figure 2) with characteristic GWTD profiles (Figure 3)
winter period, when vegetation is least active. for undisturbed agricultural area (wells 1, 2), green spaces
Cologne is underlain by Quaternary terrace deposits that (3, 4) and built environment in the city (5–10) are chosen
host a shallow aquifer with a thickness of around 20 m as representatives to examine the conditions for different
(Klostermann, 1992), and the major components of the land-use types.
urban aquifer are sands and gravels (Losen, 1984). The In agricultural areas, the measured GWTD profiles
underlying aquitard consists of clays, silts, lignite and soft typically are smooth and not very variable. The measured
coals mixed in with thin sandy layers reaching to a depth of temperature is around 1 °C below the current mean annual
about 200 m (Balke, 1973; Hilden, 1988). The horizontal air temperatures in the city area (Figure 1). The closer the
groundwater flow velocity in the aquifer is around 1 m GWTD measurements are located to the city centre,
day 1 (Balke, 1973). The groundwater flows towards the generally, the higher the observed temperatures are, and
northeast and discharges into the Rhine River. the profiles tend to be less uniform. This reflects the growing
number and superpositioning of anthropogenic heat sources,
Field measurements such as buildings and paved ground (Ferguson and
In three measurement campaigns in October 2009, Woodbury 2004, Menberg et al., 2013a, b). Built-up areas
September 2012 and December 2012, groundwater levels commonly are associated with highest temperatures,
were measured. In 2009, data from 72 wells in the city, reaching more than 16 °C in the city centre. Lateral heat
suburban and surrounding rural area were collected (Zhu transport from these sources apparently generates slightly
et al., 2010). Out of these, there are 46 wells located within higher temperatures beneath the city green spaces than
the study area (Figure 2). In 2012, the piezometer levels of found in remote rural regions.

Figure 2. Three-dimensional conceptual model of the case study of Cologne including hydraulic and thermal boundary conditions, observation wells and
a simplified vertical hydrogeological profile (CHB, constant head boundary; R, groundwater recharge; Tini, initial temperature; Tu, urban temperature; qn,
natural geothermal heat flux). Cross-section A–B indicates the studied two-dimensional flow and heat transport model

Copyright © 2014 John Wiley & Sons, Ltd. Hydrol. Process. 29, 965–978 (2015)
968 K. ZHU ET AL.

Temperature (°C)
perpendicular to the reported groundwater head contours,
0
10 11 12 13 14 15 16 17
and hence no flow boundaries are assigned (Figure 2). The
western boundary is chosen by the mean groundwater head
5
isoline of 42 m above sea level in the contour map, which is
far beyond the built urban area. The Rhine River is
represented in the model as an eastern head-dependent
10 Wells Wells boundary. The water table, i.e. the upper boundary of the
7&8 9 & 10 alluvial aquifer, is simulated as a free-surface boundary.
Depth below groundwater level (m)

15 Groundwater recharge, which in urban space is not generally

well known because of surface sealing, was subsequently
Wells
20 3&4 determined by model calibration.
To obtain a representative model setting, first, ranges of
Wells uncertain hydraulic model parameters are investigated,
25 1&2
which are listed in Table I. These ranges guide automatic
Well 6 calibration of the model described in the results section.
30 Further details on data sources and calibration procedures
are given in the Supporting Information.
35

Well 5 agricultural area

40 FLOW AND HEAT TRANSPORT SIMULATION
green spaces
built-up area We chose steady state to approximate the long-term mean of
45
built-up area recurrent and periodic conditions in the flow regime, while
(city center) keeping the modelling effort on a reasonable level.
50 However, urban groundwater temperatures are expected to
Figure 3. Measured groundwater temperature-depth profiles in Cologne. follow long-term trends, and thus transient conditions are
From the 46 profiles, those from wells 1–10 are selected as representatives favoured in the subsequent coupled 2D flow and heat
for specific urban environments and examined further in Figure 11
transport model.
The values of the thermal properties are chosen on
the basis of previous studies in the Cologne area
REGIONAL 3D GROUNDWATER FLOW MODEL
(Balke, 1977) and further literature sources (Gelhar
A 3D numerical groundwater flow model that delineates et al., 1992; VDI-4640/1, 2000; Maier and Grathwohl,
the major hydraulic conditions of the shallow aquifer 2006) (Table I). To eliminate the vertical influence of
system was set up with FEFLOW (Version 6.0, Diersch, seasonal temperature fluctuation, the groundwater level,
2009), a finite element groundwater flow and transport which is around 10–12 m below the surface, is taken as
simulator. The modelled area comprises 120.14 km2 the top boundary for the heat transport simulation. The
(Figure 2). It covers the main city area on the western simulation starts in the year 1900 with an initial
side of the Rhine and part of the surrounding suburban temperature at the top boundary of Tini = 8.7 °C according
areas. The numerical model consists of 40 layers, is about to measured mean air temperature in the region (Figure 1).
200 m thick and has a total of around 260 000 nodes. The The upward geothermal heat flux value in Cologne region
upper 20 layers, each with a thickness of 1 m, represent of qN = 59 mW m 2 (Balke, 1977) is expressed by a
the main aquifer followed by 20 layers with variable constant flux boundary at the bottom of the model
thicknesses from 2 to 10 m corresponding to the aquitard (Figure 2). Unspecified heat boundaries are assigned to
in the succeeding texts. The aquifer is unconfined, and the sides, with the river at the east having the same initial
steady-state hydraulic conditions are simulated. The high temperature as in the atmosphere (8.7 °C). The initial
resolution of the aquifer is chosen to characterize small- undisturbed temperature distribution is obtained by
scale vertical temperature variabilities, with steepest running the model for a burn-in phase of 108 days.
thermal gradients expected in the upper part of the Within this period, the model reaches quasi-steady-state
model. To limit the computational effort, the vertical grid conditions, with a maximum temperature change of less
size in the aquitard increases with depth. than 10 5 K within 100 years.
Boundary conditions of the model are obtained from The developed model describes undisturbed thermal
the groundwater level contour map and river head conditions. By selecting such settings for 1900, we
provided by the local water association (Erftverband, presume that more than one century ago, main potential
1995). The boundaries at the north and south are oriented anthropogenic heat sources were absent. This is a crucial

Copyright © 2014 John Wiley & Sons, Ltd. Hydrol. Process. 29, 965–978 (2015)
 GROUNDWATER TEMPERATURE EVOLUTION IN THE SUBSURFACE URBAN HEAT ISLAND 969

Table I. Reported value ranges, selected and calibrated model parameter values for the 3D groundwater flow and 2D heat transport
model

Model Calibrated
Parameter Value range input (this study) References

Hydraulic conductivity of aquifer (m s 1) 8.0 × 10 4–7.0 × 10 3

3.0 × 10 3
Voigt and Kilian (2007)
Hydraulic conductivity of aquitard (m s 1) 1.0 × 10 8–1.0 × 10 6
1.6 × 10 7
Freeze and Cherry (1979)
Groundwater recharge (mm a 1) 39 – 237 67.2 Erftverband (1995)
DWD (2006)
Porosity of aquifer ( ) 0.15–0.25 0.2 Balke (1973)
Porosity of aquitard ( ) 0.14–0.57 0.3 McWhorter and
Sunada (1977)
Thermal conductivity of porous media in 1.7–5.0 2.1* VDI-4640/1 (2000)
aquifer (W m-1 K-1)
Thermal conductivity of aquitard (W m 1 K 1) 1.1–5.1 1.9* VDI-4640/1 (2000)
Water thermal conductivity (W m 1 K 1) 0.6 VDI-4640/1 (2000)
Volumetric heat capacity of minerals 2.1–2.4 2.2 VDI-4640/1 (2000)
(MJ m 3 K 1)
Water heat capacity (MJ m 3 K 1) 4.2 VDI-4640/1 (2000)
Longitudinal dispersivity (m) 5–20 10 Gelhar et al. (1992)
Transverse dispersivity (m) 0.05–1.0 0.03 Maier and Grathwohl (2006)
1 1
*Please note that FEFLOW input values for thermal conductivity of aquifer and aquitard, 2.1 and 1.9 W m K , refer to the arithmetic mean of solids
1 1
and water saturated pore space. The corresponding geometric mean values are therefore 2.3 and 2.2 W m K , respectively.

assumption, and Ferguson and Woodbury (2004) already as the urban reference case. Although this linear increase is
emphasized in their study on heat loss of buildings that the most straightforward option, we might also expect much
time of construction plays an important role for the more dynamic evolution of thermal inputs in the ground. As
development of the heat anomaly. Although, it is hard to alternatives, step functions are defined, which, for example,
pinpoint the exact starting time of subsurface warming in could reflect the instantaneous effect of heat loss from
Cologne, a large amount of heat flowing into the additional new buildings. Comparison between the simu-
underground since the beginning of last century can be lated results for these different trends will facilitate to judge
expected. However, little is known about the temporal their sensitivities.
temperature trends. Hence, various scenarios of increasing Detailed simulations of spatial development of poten-
temperatures at the top are investigated, including back- tial heat sources over time as a consequence of city
ground temperature rise (linear increase as air temperature in growth are beyond the scope of this study. Instead, an
Figure 1), linear and step increase from 1900, 1955 and 1975 idealized hot spot of fixed size is initially defined, which
(Figure 4). The total simulation time is set to 110 years, from represents the city centre. It encompasses a circular area
1900 to 2010. Among the studied scenarios, linear increase with 1 km diameter, where the effects of the different
from 1900 (8.7 °C) to 2010 (15 °C, measured in the transient top boundaries are assessed. The hot spot can be
groundwater in the city centre of Cologne) is considered interpreted as a cutout of the city, an urban district with
elevated temperature, which is separately examined. This
means, potential lateral interaction, such as different
15 adjacent heat sources, or large scale expansion of the area
Top boundary temperature (°C)

Step 1955 Step 1975

14 of increased ground heat flux, is not currently accounted
for in the model. This aspect is discussed consequently
13
with the modelling results.
12 At the hot spot, we are primarily interested in the
75

55
19

19
11 ar evolution of the induced thermal anomaly in the vertical
r
ea

e
Lin
n

and groundwater flow direction (along A–B, Figures 2

Li

0
190 Background
10 ear
Lin
and 5), and thus specifically inspect the cross-section
9
through the centre (Figure 2). To simplify the model and
8
1900 1910 1920 1930 1940 1950 1960 1970 1980 1990 2000 2010
shorten simulation time, for this cross-section A–B, an
Year (A.D.) equivalent 2D vertical steady-state flow and transient heat
transport model is constructed and validated with the 3D
Figure 4. Scenarios of transient temperature boundary conditions (from
1900 to 2010) applied to control groundwater temperature at hot spot of model (see Supporting Information). Analogous to the 3D
the model model, the hydraulic boundary on the western side (A) of

Copyright © 2014 John Wiley & Sons, Ltd. Hydrol. Process. 29, 965–978 (2015)
970 K. ZHU ET AL.

Figure 5. Temperature distribution after 110 years simulation (reference case with linear temperature increase at hot spot) for selected cross-section of 3D
model and of emulated vertical 2D model; ‘center’ indicates the city centre and ‘down’ is near river Rhine

the 2D model is a constant head of 42 m, and on the Table II. Flow balance of the 3D groundwater flow model of
eastern side (B), the depth of the river is 20 m, with a head Cologne
of 36.5 m. The 2D model has a length of 10 km and a Inflow Outflow
thickness of 200 m. Same spin-up simulation was applied, (103 m3 day 1) (103 m3 day 1)
and results show the same initial vertical temperature
distribution as the 3D model. To emulate the hot spot, Constant head boundary 81.33 —
different increasing temperature scenarios were assigned in the west
Groundwater recharge 19.65 —
at the top boundary of the 2D model, from 9 km to 10 km.
Sum of constant head 100.98 —
In the following presentation of the results, first, the boundary in the west
focus is set on the sensitivity of some potentially crucial and groundwater recharge
model parameters. The impact of recharge and thermal River boundary in the east — 101.00
dispersion is discussed under undisturbed condition and
with hot spots. In addition, the influences of different
temperature top boundaries and hot spot size are
examined. The model results are juxtaposed to measured vertical groundwater fluxes can cause temperature anom-
GWTD profiles to inspect the relevance of flow alies, and that in recharge areas, geothermal gradients are
conditions and local hot spots of temperature profiles smaller (e.g. Domenico and Palciauskas, 1973; Taniguchi
for the Cologne underground. et al., 2003; Ferguson et al., 2006). Furthermore,
groundwater recharge can accelerate shallow subsurface
warming (Kurylyk and MacQuarrie, 2013).
RESULTS AND DISCUSSION Because around 70% of the study area is highly
urbanized with sealed concrete surfaces and drainage
Influence of groundwater recharge on heat transport systems, it is reasonable to expect such a small local
The values of hydraulic conductivities (aquifer, effective recharge rate. Trial simulations with and without
3.0 × 10 3 m s 1; aquitard, 1.6 × 10 7 m s 1) and recharge recharge resulted in negligible differences in subsurface
rate (67.2 mm a 1) of the 3D flow model are adjusted temperatures (not shown here). Under steady-state condi-
within the given ranges by hydraulic head calibration (see tions and for constant surface temperature (i.e. initial
Supporting Information) with support of the parameter undisturbed temperature distribution), the simulated
estimation software PEST (Version 12, Doherty, 2010). GWTD profiles with recharge result in minor differences
The resulting water flow balance of the 3D groundwater (<1% relative discrepancy) compared with the ones with
flow model is provided in Table II. no recharge (data not shown). This is because the
The automatic calibration reveals that the inversion horizontal groundwater flow rate is around three orders
problem is not well posed, and a range of parameter value of magnitude higher than the recharge. A similar outcome
combinations produces equally satisfactorily predictions. was observed, for instance by Molina-Giraldo et al.
The best fit is obtained with 67.2 mm a 1 (Table I), (2011b), who could demonstrate the minor impact of
although several other close-optimal solutions exist. The groundwater recharge (using 300 mm a 1) for their heat
essential question here is how sensitive this model transport model of a river infiltrating in an aquifer.
specification and especially the recharge value is for the If we ignore the groundwater recharge at the top
simulated results. Several previous studies indicated that boundary in the model of the reference case, the

Copyright © 2014 John Wiley & Sons, Ltd. Hydrol. Process. 29, 965–978 (2015)
 GROUNDWATER TEMPERATURE EVOLUTION IN THE SUBSURFACE URBAN HEAT ISLAND 971

temperature beneath the centre of the hot spot only impact of the transverse thermal dispersion on vertical
decreases by an average of 0.03 K along the first 50 m temperature gradients. Thermal dispersion comprises heat
depth (results not visualized). Nevertheless, in densely conductivity and hydromechanical dispersion in groundwa-
populated urban areas with high air and ground temper- ter. Mechanical dispersion in porous media is caused by the
atures and higher groundwater recharges, more anthropo- movement of a heat carrying fluid, such as groundwater. The
genic heat might also penetrate into the ground and locally different flow pathways on the pore scale create differential
pronounced advective heat transport by recharge has to be advection, and the mixing of the pore-scale interstitial water
considered more carefully. Foulquier et al. (2009), for causes spreading of the temperature gradients. Macroscopic
example, observed that the thermal amplitude of ground- scale heterogeneity of a permeability field also contributes
water can be increased dramatically by stormwater additionally to dispersion, and increases uncertainty of the
infiltration at seasonal and event scales. spreading of heat plumes (e.g. Molina-Giraldo et al., 2011a).
In conclusion, groundwater recharge rates obtained by Finally, heat conduction in water and solids contributes to
fitting the numerical model are less than 100 mm a 1 in the overall dispersion in heat transport in groundwater
Cologne, and thus do not contribute much to shallow (e.g. Bons et al., 2013).
subsurface warming in this city. If we assume an increase In our high-velocity case, transverse dispersion can be
of surface temperatures by 4 K, the average heat transfer expected to influence the vertical heat flow and the GWTD
to the aquifer by recharge in a period of 100 years is profile. The value of the hydromechanical transverse
about 1.1 × 1011 kJ km 2, whereas, by heat conduction, it is dispersivity, however, is not exactly known and has not
more than fivefold (5.8 × 1011 kJ km 2). In contrast, local been examined for the Cologne area. In general, there are
anthropogenic heat sources, such as leakage of sewage or only few studies specifically dedicated to experimental or
water distribution networks, might play a more prominent model-based estimations of thermal transverse and longitu-
role (e.g. Menberg et al., 2013b). This is reported for cities in dinal dispersivities (e.g. Molina-Giraldo et al., 2011a; Rau
England. For example, the study by Lerner (1990) showed et al., 2012; Bons et al., 2013; Stauffer et al., 2013). Hence,
that urbanization reduces the groundwater recharge, but in our reference case, we initially approximate the field-scale
creates new and additional sources for recharge, such as hydrodynamic transverse dispersivity by Maier and
leaking from water mains and sewers. Grathwohl (2006) with αt = 0.03 m, which was derived for
a sand and gravel aquifer in south-west Germany. This field-
Influence of groundwater flow for undisturbed conditions scale derived (fitted) transverse dispersivity is therefore a
bulk parameter, which also accounts for possible effects of
Vertical and horizontal groundwater flow. At the study
heterogeneity or non-stationarity. The chosen value is at the
site, a shallow aquifer with substantial groundwater flow
lower range of reported field-scale transverse dispersivities
velocity (v) of around 1 m day 1 exists, and, in comparison
applied in field experiments and modelling studies ranging
with the study by Ferguson and Woodbury (2004),
between 0.005 and 10 m (Stauffer et al., 2013). Hence, even
horizontal advection is therefore expected to play an higher values might occur in more heterogeneous
important role for the evolution of the groundwater environments. In contrast, typically, laboratory-scale
temperatures. Lu and Ge (1996) demonstrated that when (local-scale) derived transverse dispersivities are much
the horizontal heat and fluid flow is greater than 30% of the smaller (αt < 1.8 × 10 4 m; e.g. Olsson and Grathwohl,
vertical one, it has a significant effect on the vertical 2007; Bauer et al., 2009). Furthermore, the transverse
temperature distribution. Ferguson et al. (2006) concluded dispersivity may depend slightly on the flow velocity of the
that when the downward Darcy flux is smaller than studied system (e.g. Chiogna et al. 2010). The latter
2.0 × 10 8 m s 1 (0.63 m a 1), it does not significantly however is not known for the studied aquifer; thus here,
change the temperature profile under steady-state thermal only a range of transverse dispersivities is examined in more
conditions. In our study area, which is in a river valley, the detail.
dominant flow direction in the urban aquifers is Groundwater temperature-depth profiles simulated for
horizontal, with significant flow rates of about 1 m day 1 undisturbed conditions are compared for different transverse
(365 m a 1); whereas the vertical groundwater flux due to dispersivities (αt = 0, 0.03, 0.1 and 1 m). With increase of
recharge is only around 1.9 × 10 4 m day 1 (0.07 m a 1). dispersivity, the simulated vertical temperature gradient
Simulation results show that consistent with the criterion along the aquifer becomes smaller in the aquifer because of
set by Ferguson et al. (2006), it only has a minor influence the increase in vertical mixing (Figure 6). This is a similar
on the temperature profile with a temperature change effect as that from vertical groundwater flow, such as
<0.03 K. described, for example, by Taniguchi et al. (2003). Thus,
macroscale mechanical dispersion reveals to be an important
Influence of transverse thermal dispersion. A mecha- process for the heat transport in vertical direction. The
nism, which has not been studied in this context, is the hydromechanical dispersion coefficients (Dt = v × αt) and

Copyright © 2014 John Wiley & Sons, Ltd. Hydrol. Process. 29, 965–978 (2015)
972 K. ZHU ET AL.

Temperature (°C)
8 9 10 11 12 13 14 15 16 17 dominant and complete mixing of the aquifer happens
0
α
(Figure 6).
t = 1at
With m= 1 m

20 α t = 0.1
With at =m0.1 m Influence of groundwater flow with local heat sources
40
α t = 0.03
With at = m
0.03 m
Different temperature top boundaries. As illustrated in
no
No flow
groundwater flow Figure 6, the effect of horizontal groundwater flow under
Depth below groundwater level (m)

60 steady-state thermal conditions on ground temperature

evolution depends on the degree of transverse dispersion,
80 and the influence is more evident when dispersivities are
larger than 0.1 m. Figure 7 shows simulated GWTD
100 profiles at different locations (upstream, centre and
downstream) of the hot spot, given transient conditions
120
with different top boundary temperature increasing
scenarios (Figure 4). The dispersivity is not varied here
and set to the default value of αt = 0.03 m. The simulated
140

profiles are comparable in the upper part, down to the

160
aquifer basis at 20 m, as well as in a depth of more than
180
100 m below the water table. Even if slightly increased
temperatures are simulated at greater depth than 100 m,
200 all curves converge to a linear trend representing
undisturbed conditions. Because of horizontal groundwa-
Figure 6. Simulated initial undisturbed groundwater temperature-depth
profiles (without hot spot) under conduction-dominated condition (solid
ter flow, all the simulated profiles at upstream are almost
line) and flow conditions with different transverse dispersivity (αt) values; the same (Figure 7a), and at centre and downstream, the
1
groundwater recharge is 67.2 mm a in each scenario differences become larger.
As expected, temperature increasing scenarios with
earlier onset of the hot spot generate deeper penetration,
thermal diffusivities (κ = λ/Cv, λ and Cv are thermal and for the different linear trends, the scenario starting
conductivity and volumetric heat capacity of the aquifer) from 1900 (‘Linear 1900’ in Figure 7c) generates a more
can be used to quantitatively compare the influence of pronounced anomaly than the scenario initiated in 1955
dispersion and conduction. For a horizontal flow velocity of (‘Linear 1955’ in Figure 7c). Because of thermal diffusion
1 m day 1, if αt equals to 0.1 m, the calculated value of Dt and dispersion, step functions assigned to the top
(1.2 × 10 6 m2 s 1) is already larger than κ (8.8 × 10 7 m2 boundary also induce smooth profiles. Temperatures are
s 1). When αt ≥ 1 m, dispersion becomes even more higher for stepwise than linear increase starting from the

a) up Temperature (°C) b) center Temperature (°C) c) down Temperature (°C)

10 11 12 13 14 15 16 17 10 11 12 13 14 15 16 17 10 11 12 13 14 15 16 17
0 0 0

20 20 20
Linear 1900
Depth below groundwater level (m)

40
Depth below groundwater level (m)

Depth below groundwater level (m)

40 Linear 1955 40
Linear 1975
60 Step 1955 60 60
Step 1975
80 80 80

100 100 100

120 120 120

140 140 140

160 160 160

180 180 180

200 200 200

Figure 7. Simulated groundwater temperature-depth profiles at different locations (upstream, centre and downstream) of hot spot assuming no
groundwater flow and horizontal flow with different transverse dispersivity (αt) values

Copyright © 2014 John Wiley & Sons, Ltd. Hydrol. Process. 29, 965–978 (2015)
 GROUNDWATER TEMPERATURE EVOLUTION IN THE SUBSURFACE URBAN HEAT ISLAND 973

same point in time (e.g. the scenario ‘Step 1955’ shows

higher temperatures than ‘Linear 1955’ in Figures 7b and
7c), and this can be easily explained by the higher average
temperature at the top boundary for this scenario and thus
a higher amount of heat introduced to the subsurface. In
all profiles, because of horizontal groundwater flow, the
influence of different increasing temperature top bound-
aries is only marginal in the shallow parts (<20 m depth).

Transverse dispersion effects at different locations. The

simulated profiles show that the footprint of elevated
urban heat flux can be tracked to a maximum of about
100 m, which is similar to the deviation depth observed in
other cities, such as Winnipeg with a depth of ~130 m
(Ferguson and Woodbury, 2004) and Tokyo with a depth
of 140 m (Taniguchi et al., 2007). The different trends of
past temperature increase obviously do not have substan-
tial impact on the shallow part of temperature profiles,
given that the initial (Tu = 8.7 °C) and final (Tu = 15 °C)
conditions are known. Thus, in the following, we stick to
the linear increase from 1900 (urban reference case), and
concentrate on the role of transverse dispersion. This is
investigated by visualizing the thermal conditions beneath
the hot spot, and by comparing the temperature profiles
simulated at the central as well as upstream and
downstream fringe position (Figures 8 and 9).
Figure 8 shows that horizontal groundwater flow in the
shallow aquifer deviates from the thermal anomaly,
evolving beneath the hot spot, in the downstream direction.
This is also reflected in the vertical temperature profiles in
Figure 9. In the upstream position (Figure 9a), horizontal
flow therefore prevents the evolution of deeper temperature
alterations. In comparison, for conduction-dominated con-
ditions, simulated by switching off groundwater flow in the
model, the thermal anomaly is not deviated, and vertical
thermal diffusion is most effective. Still, despite the high
flow velocity of 1 m day 1, horizontal advection does not
act as a thermal barrier. In downstream direction,
conduction-induced profiles become similar to those
conditions with small to moderate values of macroscale Figure 8. Simulated temperature distribution after 110 years (urban
transverse dispersivity, which is αt ≤ 0.1 m (Figure 9c). reference case, with 1 km hot spot) of vertical 2D models with different
transverse dispersivity (αt) values (b–d) and no horizontal groundwater
Figures 8 and 9 illustrate that most pronounced flow (a). The groundwater flows from left to right with a velocity of about
effects occur for conditions with high transverse dispersion 1
1 m day , and a cold plume develops below the hot spot. Temperature
(αt = 1 m). This scenario could represent substantial profiles at upstream, centre and downstream position of the hotspot of
each scenario are compared and shown in Figure 9
macrodispersive effects as a consequence of aquifer
heterogeneity. For simulating the aquifer of Cologne, we
assume αt = 0.03 m, which, however, is highly uncertain and systems using larger thermal transverse dispersivities
can not be validated here. In fact, the formation hosting the (αt > 1 m) are rare (Stauffer et al., 2013). For example,
aquifer consists of terrace deposits with interbedding gravels Smith and Chapman (1983) applied a thermal transverse
and sands. These structures and heterogeneities are not dispersivity of 10 m in their 2D flow and transport model.
currently resolved in our simplified model, which assumes However, they considered a sedimentary basin on kilometre
homogeneous aquifer and aquitard layers. Hence, in the scale (40 km wide and 5 km deep) and not an urban scale.
field, macrodispersion may play an even more prominent Because the temperature profiles at center, up or
role than simulated by the presented models. Although, such downstream locations are distinct, the excess heat stored

Copyright © 2014 John Wiley & Sons, Ltd. Hydrol. Process. 29, 965–978 (2015)
974 K. ZHU ET AL.

a) up Temperature (°C) b) center Temperature (°C) c) down Temperature (°C)

10 11 12 13 14 15 16 17 10 11 12 13 14 15 16 17 10 11 12 13 14 15 16 17
0 0 0

Depth below groundwater level (m) 20 20 20

Depth below groundwater level (m)

Depth below groundwater level (m)
40 40 40

60 60 60

80 80 80

100 100 100

120 120 120

140 140 140

αt = 1 m
160 αt = 0.1 m 160 160

180 αt = 0.03 m 180 180

no flow
200 200 200

Figure 9. Simulated groundwater temperature-depth profiles with different transverse dispersivity (αt) values at different locations (upstream, centre and
downstream) beneath the hot spot (Figure 8). The dotted blue lines indicate the lower boundary of the aquifer. Additionally, the profile derived for no
horizontal groundwater flow in the aquifer is shown

at such locations is also different. By heat content Temperature (°C)

calculation for the reference case (Zhu et al. 2010), the 10 11 12 13 14 15 16 17
0
total extra heat content at upstream, center and down-
stream are Qup = 3.62 × 1010 kJ km 2, Qcenter = 1.08 × 1011 up
20 center
kJ km 2 and Qdown = 1.53 × 1011 kJ km 2, respectively.
down
Size of hot spot. The hot spot scenario is based on the 40

extension of the 15 °C temperature area on the groundwater

Depth below groundwater level (m)

temperature contour map (Zhu et al. 2010), and a size of 60

1 km is just an approximation of local conditions in the city
of Cologne. In fact, the distribution of heat sources in the city 80
is very heterogeneous, and thus, the superimposing heat
sources could cover a larger area. If, for example, we extend 100
the hot spot to 3 km, the simulated temperature profile in the
upstream is the same as for the 1 km case, but the profiles at
120
the centre and downstream show higher temperatures for
3 km (Figure 10). However, for the case studied here, the
140
difference between the 3 km centre and downstream profiles
is only minor, which indicates that the size of the hot spot
has a limited influence until 1.5 km from the upstream, and 160
after this distance, the impact of cold groundwater plume
coming from upstream almost vanishes. 180
1 km hot spot
3 km hot spot
Comparison between simulated and measured 200
temperature-depth profiles
Figure 10. Comparison of simulated groundwater temperature-depth
During the 2009 field campaign, GWTD profiles were profiles at different locations (upstream, centre and downstream) with 1
recorded at 46 wells in the study area. Because of spatial and 3 km hot spot, respectively
variation in land-use type, differences in heating history, the
local presence of different heat sources and hydrogeological The numerical model is used to reproduce these profiles
influences, temperature profiles in the city show pronounced to identify governing factors and to study, in general,
spatial variability. Here, ten wells with characteristic the consistency of our modelling approach. Simulated
temperature profiles covering all different land-use types profiles are compared separately for the measured profiles
and the measured temperature ranges are chosen (Figure 2). at given land use types. Again, a hot spot of 1 km extension

Copyright © 2014 John Wiley & Sons, Ltd. Hydrol. Process. 29, 965–978 (2015)
 GROUNDWATER TEMPERATURE EVOLUTION IN THE SUBSURFACE URBAN HEAT ISLAND 975

is implemented. In the four simulated temperature variants Undisturbed conditions are found in measurements in the
(10.8 °C, 13.5 °C, 15 °C and 17 °C), a standard linear trend surrounding rural area of Cologne: Wells 1 and 2 are located
since 1900 is assumed with the final temperature at the in agricultural land, and the vertical measured GWTD
hot spot oriented at average values characteristic for profiles are very similar with a nearly constant temperature
different land-use types. These are 10.8 °C in 2010 for of 10.8 °C. The measured profiles at these wells are
undisturbed conditions, 13.5 °C and 15 °C for built reproduced by the model, which shows that the model is
envorinment and 17 °C as observed in the highly urbanized accurately calibrated to these undisturbed conditions. The
city centre. measured temperature in urban green spaces is about 1 °C

Figure 11. Simulated and measured groundwater temperature-depth profiles from different land-use types in Cologne (Figure 3)

Copyright © 2014 John Wiley & Sons, Ltd. Hydrol. Process. 29, 965–978 (2015)
976 K. ZHU ET AL.

higher than the simulated undisturbed temperature. When perturbs GWTD profiles. The influence of this mechanism
moving towards the city centre, measured groundwater substantially depends on the effective transverse dispersivity
temperatures increase and the temperature-depth profiles are and is more pronounced in more heterogeneous media,
perturbed, which is revealed to be specific for each well. especially with local heat sources, than under undisturbed
When simulating the hot spot with different temperatures in conditions. In the Cologne case, the chosen values of
2010, the profiles that are found in upstream, centre or transverse dispersivities ranging between 0.03 m and 1 m are
downstream position may fit in the uppermost, shallow already more influential than vertical heat transport by
profile, however, not in the deeper parts of the wells. Here, groundwater recharge for both undisturbed conditions and
typically measured temperatures remain high and the hot spot scenarios. When αt is larger than 1 m, instead of
profiles appear less inclined than the simulated ones conduction, dispersion becomes the dominant process of
(Figure 11b–d). On the basis of the previous analysis on vertical heat transport in the aquifer. Furthermore, our study
crucial model configurations, we can identify several shows that, under undisturbed conditions, the transverse
potential reasons for this discrepancy. dispersion could lead to a concave upward temperature
One principal reason is that the generic hot spot scenario distribution (Figure 9b and 9c), which could also be caused
is only an approximation of the site-specific and local by downward groundwater flow. In urban regimes,
conditions of the city of Cologne. The true heterogeneity exclusion of transverse dispersion in aquifers with high
of spatially and temporally variable features can hardly horizontal groundwater flow velocity, when analysing
be resolved with our simplified scenarios. Comparison with GWTD profiles, thus may result in erroneous estimates of
measured profiles confirms that for a city, a much more surface warming rate or groundwater recharge.
extensive hot spot or, on fine resolution, multiple over- When horizontal advection is the dominant heat transport
lapping hot spots existing in the entire city area can be process, a subsurface UHI might be moved downstream.
expected. Downstream of an area with elevated heat flux, Consequently, the GWTD profiles in urban aquifers are
groundwater temperature profiles tend to be more vertical as strongly influenced by the relative position and distance
a consequence of transverse mixing. This also means that in (upstream or downstream) to the anthropogenic heat sources,
this case, transverse dispersion may play an important or which were also demonstrated by Ferguson and Woodbury
even dominant role. Alternatively, merely assuming more (2004). It is shown by numerical modelling that different
intense macrodispersion than specified in our reference increasing temperature trends and the size of the heat source
model would lead to more realistic temperature profiles. (simulated as hot spot) also play a role on temperature
This could be an indication that the role of subsurface evolution, especially at centre and downstream location, if
heterogeneity is underestimated. Finally, an apparent reason distances are less than 1.5 km. Horizontal groundwater flow
for higher temperatures in the deeper wells is higher vertical causes a cold plume below the city and heat discharge into
heat fluxes in the past than estimated by the linear trend, the river. With an outflow rate of 1.0 × 105 m3 day 1 and a
which might date back before the year 1900, i.e. time zero in temperature gradient of 4 K, the amount of heat that is
the current model. discharged to the river per year is about 6.1 × 1011 kJ.
The comparison of measured and simulated GWTD
profiles also indicates that even in idealized scenarios,
temperature profiles are rather complicated. They are hard to
CONCLUSIONS
interpret and thus it is difficult to capture the driving
During the past hundred years, because of urbanization, a subsurface heat transport mechanisms with streamlined
large amount of anthropogenic heat has entered the scenarios. The heat flow in urban subsurface depends on
subsurface of cities. The anthropogenic heat discharge many local and site-specific parameters, and a more detailed
elevated the temperature of a local urban aquifer beneath the resolution of underground geological structures and the
city of Cologne, Germany, by up to more than 5 K. On the temporally variable heat urban sources would be desirable.
basis of simulation results from site-specific modelling, we
can conclude that, for urban conditions such as in Cologne,
average groundwater recharge rates are very low, and
ACKNOWLEDGEMENTS
recharge does not play a significant role in urban
groundwater temperature evolution. Because the dominant We would like to thank two anonymous reviewers for their
flow direction in the Cologne urban aquifers is horizontal, fruitful comments. We also would like to acknowledge the
with significant flow rates, the influence of horizontal assistance of Steffen Rehner during the field campaign in
groundwater flow on subsurface temperature evolution has Cologne, and Valentin Wagner for technical support with
to be addressed. the numerical simulation. Furthermore, we would like to
It is shown that field-scale transverse dispersion causes thank the Erftverband and the RheinEnergie, particularly
additional vertical mixing and vertical heat flux, which Stefan Simon and Stefan Schiffmann for providing data, as

Copyright © 2014 John Wiley & Sons, Ltd. Hydrol. Process. 29, 965–978 (2015)
 GROUNDWATER TEMPERATURE EVOLUTION IN THE SUBSURFACE URBAN HEAT ISLAND 977

well as assistance with the field measurements. This study Freeze RA, Cherry JA. 1979. Groundwater. Prentice-Hall: Englewood
Cliffs, NJ, USA.
was supported by the Federal Ministry for Education and Gelhar LW, Welty C., Rehfeldt KR. 1992. A critical review of data on
Research (BMBF) scholarship programme for International field-scale dispersion in aquifers. Water Resources Reseatch 28(7):
Postgraduate Studies in Water Technologies (IPSWaT), by 1955–1974. DOI: 10.1029/92wr00607.
Gunawardhana L, Kazama S. 2011. Climate change impacts on
the Swiss National Science Foundation (SNSF) under grant groundwater temperature change in the Sendai plain, Japan. Hydrolog-
number 200021L 144288, and the German Research ical Processes 25: 2665–2678. DOI: 10.1002/hyp.8008.
Foundation (DFG), under grant number BL 1015/4-1. Hamamoto H, Yamano M, Kamioka S, Nishijima J, Monyrath V,
Goto S, Taniguchi M. 2008. Estimation of the past ground surface
temperature change from borehole temperature data in the Bangkok
area. In From Headwaters to the Ocean, Taniguchi M (ed). Taylor and
REFERENCES Francis: Baton Rouge, FL, USA.
Hilden HD. 1988. Geolgie am Niederrhein. Geologisches Landesamt
Balke K-D. 1973. Geothermische und hydrogeologische Untersuchungen Nordrhein-Westfalen.
in der südlichen Niederrheinischen Bucht. Bundesanstalt für Huang SP., Pollack HN, Shen PY. 2000. Temperature trends ever the
Bodenforschung und den Geologischen Landesämtern der past five centuries reconstructed from borehole temperatures. Nature
Bundesrepublik Deutschland, Hannover. 403(6771): 756–758. DOI: 10.1038/35001556.
Balke K-D. 1977. Das Grundwasser als Energietraeger. Brennstoff- Huang SP, Taniguchi M., Yamano M, Wang CH. 2009. Detecting
Warme-Kraft 29: 191–194. urbanization effects on surface and subsurface thermal environment -- A
Bauer RD, Rolle M, Bauer S, Eberhardt C, Grathwohl P, Kolditz O, case study of Osaka. Science of the Total Environment 407(9): 3142–3152.
Meckenstock RU, Griebler C. 2009. Enhanced biodegradation by DOI:10.1016/j.scitotenv.2008.04.019.
hydraulic heterogeneities in petroleum hydrocarbon plumes. Journal of Klostermann J. 1992. Das Quartär der Niederrheinischen Bucht -
Contaminant Hydrology 105: 56–68. DOI: 10.1016/j. Ablagerungen der letzten Eiszeit am Niederrhein. 4. Auflage, Krefeld.
jconhyd.2008.11.004. Geologisches Landesamt Nordrhein-Westfalen.
Beltrami H, Bourlon E, Kellman L, Gonzalez-Rouco JF. 2006. Spatial Kurylyk BL, MacQuarrie KTB. 2013. A new analytical solution for
patterns of ground heat gain in the Northern Hemisphere. Geophysical assessing climate change impacts on subsurface temperature. Hydro-
Research Letters 33(6). DOI: 10.1029/2005gl025676. logical Processes. DOI: 10.1002/hyp.9861.
Bense V, Beltrami H. 2007. Impact of horizontal groundwater flow and Lerner DN. 1990. Groundwater recharge in urban areas. Atmospheric
localized deforestation on the development of shallow temperature Environment. Part B. Urban Atmosphere 24(1): 29–33. DOI: 10.1016/
anomalies. Journal of Geophysical Research-Earth Surface 112(F4). 0957-1272(90)90006-G.
DOI: 10.1029/2006jf000703. Losen H. 1984. Grundwasserstände und Grundwasserbeschaffenheit im
BGR. 2008. Water water cycle/water balance for Germany. Available südlichen Teil der linksrheinischen Kölner Scholle. Dissertation RWTH
from: http://www.bgr.bund.de/EN/Themen/Wasser/Bilder/ Aachen.
Was_wasser_startseite_wasserkreis_g_en.html. Federal Institute for Lu N, Ge S. 1996. Effect of horizontal heat and fluid flow on the vertical
Geosciences and Natural Resources. temperature distribution in a semiconfining layer. Water Resources
Bodri L, Cermak V. 1997. Climate changes of the last two millennia Research 32(5): 1449–1453. DOI: 10.1029/95wr03095.
inferred from borehole temperatures: results from the Czech Republic 2. Maier U, Grathwohl P. 2006. Numerical experiments and field results
Global and Planetary Change 14(3-4): 163–173. DOI: 10.1016/S0921- on the size of steady state plumes. Journal of Contaminant Hydrology
8181(96)00010-0. 85(1–2): 33–52. DOI: 10.1016/j.jconhyd.2005.12.012.
Bons PD, van Milligen BP, Blum P. 2013. A general unified expression McWhorter DB, Sunada DK. 1977. Ground-Water Hydrology and
for solute and heat dispersion in homogeneous media. Water Resources Hydraulics. Water Resour. Publications: Ft. Collins, CO, USA.
Research 49(10): 6166–6178. DOI:10.1002/wrcr.20488. Menberg K, Bayer P, Zosseder K, Rumohr S, Blum P. 2013a. Subsurface
Chiogna G, Eberhardt C, Cirpka OA, Grathwohl P, Rolle M. 2010. urban heat islands in German cities. Science of the Total Environment
Evidence of compound-dependent hydrodynamic and mechanical 442(0): 123–133. DOI: 10.1016/j.scitotenv.2012.10.043.
transverse dispersion by multitracer laboratory experiments. Environ- Menberg K, Blum P, Schaffitel A, Bayer P. 2013b. Long term evolution of
mental Science and Technology 44: 688–693. DOI: 10.1021/es9023964. anthropogenic heat fluxes into a subsurface urban heat island.
Doherty J. 2010. PEST Model-Independent Parameter Estimation User Environmental Science & Technology. DOI: 10.1021/es401546u.
Manual: 5th Edition. Watermark Numerical Computing: Corinda, Molina-Giraldo N, Bayer P, Blum P. 2011a. Evaluating the influence
Australia. of thermal dispersion on temperature plumes from geothermal systems
Domenico PA, Palciauskas VV. 1973. Theoretical analysis of forced using analytical solutions. International Journal of Thermal Sciences
convective heat transfer in regional ground-water flow. Geological Society 50(7): 1223–1231. DOI: 10.1016/j.ijthermalsci.2011.02.004.
of America Bulletin 84(12): 3803–3814. DOI: 10.1130/0016-7606. Molina-Giraldo N, Bayer P, Blum P, Cirpka OA. 2011b. Propagation of
DWD. 2006. Mittelwerte des Niederschlages, Deutscher Wetter Dienst. seasonal temperature signals into an aquifer upon bank infiltration. Ground
DWD. 2010. Monatliche Temperaturwerte, Deutscher Wetter Dienst. Water 49(4): 491–502. DOI: 10.1111/j.1745-6584.2010.00745.x.
Erftverband. 1995. Basisplan III zur Sicherstellung der Wasserversorgung Olsson A, Grathwohl P. 2007. Transverse dispersion of non reactive tracers in
im Bereich des ErftverbandsRep., Erftverband, Bergheim. porous media: a new nonlinear relationship to predict dispersion
Epting J, Händel F, Huggenberger P. 2013. Thermal management of coefficients. Journal of Contaminant Hydrology 92: 149–161.
an unconsolidated shallow urban groundwater body. Hydrology Perrier F, Le Mouel JL, Poirier JP, Shnirman MG. 2005. Long-term climate
and Earth System Sciences 17: 1851–1869. DOI: 10.5194/hess-17- change and surface versus underground temperature measurements in
1851-2013. Paris. International Journal of Climatology 25(12): 1619–1631. DOI:
Ferguson G, Beltrami H, Woodbury AD. 2006. Perturbation of ground 10.1002/joc.1211.
surface temperature reconstructions by groundwater flow? Geophysical Pollack HN, Huang SP, Shen PY. 1998. Climate change record in subsurface
Research Letters 33(13). DOI: 10.1029/2006gl026634. temperatures: a global perspective. Science 282(5387): 279–281. DOI:
Ferguson G, Woodbury AD. 2004. Subsurface heat flow in an urban 10.1126/science.282.5387.279.
environment. Journal of Geophysical Research-Solid Earth 109(B2). Rau GC, Andersen MS, Acworth RI. 2012. Experimental investigation of
DOI: 10.1029/2003jb002715. the thermal dispersivity term and its significance in the heat transport
Ferguson G, Woodbury AD. 2007. Urban heat island in the subsurface. equation for flow in sediments. Water Resources Research 48(3):
Geophysical Research Letters 34(23). DOI: 10.1029/2007gl032324. W03511. DOI: 10.1029/2011wr011038.
Foulquier A, Malard F, Barraud S, Gibert J. 2009. Thermal influence of Smith L, Chapman DS. 1983. On the thermal effects of groundwater flow:
urban groundwater recharge from stormwater infiltration basins. 1. Regional scale systems. Journal of Geophysical Research: Solid
Hydrological Processes 23(12): 1701–1713. DOI: 10.1002/hyp.7305. Earth (1978–2012), 88(B1): 593–608.

Copyright © 2014 John Wiley & Sons, Ltd. Hydrol. Process. 29, 965–978 (2015)
978 K. ZHU ET AL.

Stauffer F, Bayer P, Blum P, Molina-Giraldo N, Kinzelbach W. 2013. VDI-4640/1. 2000. Thermal use of the underground—fundamentals,
Thermal Use of Shallow Groundwater. CRC Press: Baton Rouge, FL, approvals, environmental aspects VDI (verein deutscher ingenieure).
USA; 287 pages. Voigt J, Kilian L. 2007. Grundwassermodell für das Rheinische
Stallman RW. 1963. Computation of ground-water velocity from Braunkohlerevier, in Modellbericht, RWE Power AG, Cologne.
temperature data. U.S. Geological Survey. Water Supply Paper 1544-H: Wang K, Lewis TJ, Belton DS, Shen PY. 1994. Differences in recent ground
35–46. surface warming in eastern and western Canada: evidence from borehole
Suzuki S. 1960. Percolation measurements based on heat flow through soil temperatures. Geophysical Research Letters 21(24): 2689–2692. DOI:
with special reference to paddy fields. Journal of Geophysical Research 10.1029/94GL02670.
65(9): 2883–2885. DOI: 10.1029/JZ065i009p02883. Yamano M, Goto S, Miyakoshi A, Hamamoto H, Lubis RF, Monyrath V,
Taniguchi M, Shimada J, Tanaka T, Kayane I, Sakura Y, Shimano Y, Taniguchi M. 2009. Reconstruction of the thermal environment evolution in
Dapaah-Siakwan S, Kawashima S. 1999. Disturbances of temperature- urban areas from underground temperature distribution. Science of The Total
depth profiles due to surface climate change and subsurface water flow: Environment 407(9): 3120–3128. DOI: 10.1016/j.scitotenv.2008.11.019.
1. An effect of linear increase in surface temperature caused by global Zhu K, Blum P, Ferguson G, Balke KD, Bayer P. 2010. The geothermal
warming and urbanization in the Tokyo metropolitan area, Japan. Water potential of urban heat islands. Environmental Research Letter 5(4):
Resources Research 35(5): 1507–1517. DOI: 10.1029/1999WR900009. 044002. DOI: 10.1088/1748-9326/5/4/044002.
Taniguchi M, Shimada J, Uemura T. 2003. Transient effects of surface
temperature and groundwater flow on subsurface temperature in
Kumamoto Plain, Japan. Physics and Chemistry of the Earth 28(9-11):
477–486. DOI: 10.1016/s1474-7065(03)00067-6. SUPPORTING INFORMATION
Taniguchi M, Uemura T, Jago-on K. 2007. Combined effects of
urbanization and global warming on subsurface temperature in four
Asian cities. Vadose Zone Journal 6(3): 591–596. DOI: 10.2136/ Additional supporting information may be found in the
vzj2006.0094. online version of this article at the publisher’s web-site.

Copyright © 2014 John Wiley & Sons, Ltd. Hydrol. Process. 29, 965–978 (2015)