Vous êtes sur la page 1sur 25

PUBLICATIONS

Reviews of Geophysics
REVIEW ARTICLE Karst water resources in a changing world: Review
10.1002/2013RG000443
of hydrological modeling approaches
Key Points: A. Hartmann1, N. Goldscheider2, T. Wagener1, J. Lange3, and M. Weiler3
We elaborate the importance of karst
water resources 1
Department of Civil Engineering, University of Bristol, Bristol, UK, 2Institute of Applied Geosciences, Karlsruhe Institute of
We provide a detailed overview of
karst modeling approach Technology, Karlsruhe, Germany, 3Chair of Hydrology, Freiburg University, Freiburg, Germany
We present new methods and direc-
tions for their improvement
Abstract Karst regions represent 712% of the Earths continental area, and about one quarter of the
global population is completely or partially dependent on drinking water from karst aquifers. Climate
simulations project a strong increase in temperature and a decrease of precipitation in many karst regions in
Correspondence to:
A. Hartmann, the world over the next decades. Despite this potentially bleak future, few studies specically quantify the
aj.hartmann@bristol.ac.uk impact of climate change on karst water resources. This review provides an introduction to karst, its
evolution, and its particular hydrological processes. We explore different conceptual models of karst systems
Citation: and how they can be translated into numerical models of varying complexity and therefore varying
Hartmann, A., N. Goldscheider, data requirements and depths of process representation. We discuss limitations of current karst
T. Wagener, J. Lange, and M. Weiler models and show that at the present state, we face a challenge in terms of data availability and
(2014), Karst water resources in a chan-
ging world: Review of hydrological information content of the available data. We conclude by providing new research directions to
modeling approaches, Rev. Geophys., 52, develop and evaluate better prediction models to address the most challenging problems of karst
218242, doi:10.1002/2013RG000443. water resources management, including opportunities for data collection and for karst model
applications at so far unprecedented scales.
Received 15 SEP 2013
Accepted 2 APR 2014
Accepted article online 8 APR 2014
Published online 7 AUG 2014
Corrected 11 MAY 2015
1. Introduction
This article was corrected on 11 MAY Karst regions cover 712% of the Earths continental area, and their aquifers are at least a partial
2015. See the end of the full text for
details. source of drinking water supply to almost a quarter of the worlds population [Ford and Williams,
2007]. Stress on groundwater resources has increased signicantly in recent decades [Wada et al.,
2010], (1) in terms of water quantity due to excessive irrigated agriculture [Aeschbach-Hertig and
Gleeson, 2012] and (2) in terms of quality due to pollution by fertilizers [Foley et al., 2011]. Projections
of 20 general circulation models using the A1B emission scenario for the years 20812090 [Christensen
et al., 2007] suggest that an increase of temperatures for North America and Europe, and a strong
decrease in precipitation in the more densely populated areas of North America and southern Europe
(Mediterranean), can be expected (Figure 1). In these regions, stress on karst water resources, in terms
of both quantity and quality, is likely to increase dramatically in the future. How will potential changes
in temperature and precipitation affect local or regional water availability in these karst regions? This
question is still very difcult to address given current tools and methods as we discuss in
detail below.
We need hydrologic models to understand the impact of climate change on water resources. Such models
transform scenarios of climate or land use change into their hydrological implications. Hydrologic models
require an adequate representation of karst specic processes, like the strong subsurface hydraulic
heterogeneity of karstied rocks [Bakalowicz, 2005], if they are meant to provide realistic water resources
simulations. Unfortunately, this heterogeneity is most often poorly characterized by available data and is
not reective of information required for modeling karst systems. Providing reliable simulations for the
sustainable protection and management of karst water resources is therefore still challenging, both
scientically and operationally.
In this review we provide an overview about the relevance of karst regions for human water supply and
discuss the importance of quantitative projections of future karst water availability. We (1) start with an
introduction to karst including its evolution and hydrogeology, (2) present an overview of karst exploration
methods with a particular focus on the data requirements of karst models, (3) discuss the present state of
hydrologic models for simulating karst water resources, and (4) discuss challenges and new directions in karst

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 218
Reviews of Geophysics 10.1002/2013RG000443

Figure 1. Location of carbonate rock outcrops in Europe (Williams and Ford [2006], modied) compared to expected mean change of temperature and precipitation
in (a and b) North America and (c and d) Europe from 19611990 to 20812090, derived from 20 general circulation models [Christensen et al., 2007].

modeling and exploration for better simulations in the future. Furthermore, we provide a glossary with the
most relevant expressions at the end of the review.

2. Introduction to Karst
Raindrops collect atmospheric carbon dioxide (CO2) when they form in the atmosphere. Vegetation and
microbial processes in the soil further increase the CO2 concentration in the water after the rain has fallen on
the land surface and inltrated in the soil. Soil moisture will further percolate and, if the underlying bedrock is
composed of carbonate rock, its CO2 will dissolve the bedrock material. In this way, the landform of karst
develops over tens of thousands of years, creating a landscape with specic surface and subsurface features.
Karst landforms, such as karren, dolines, swallow holes, dry valleys, and poljes, indicate the presence of
surface karstication processes (Figures 2a2c). Even if no supercial karst features are present, subsurface
dissolution can create hierarchically organized networks of open fractures, karst conduits, and caves in the
subsurface that often drain to large karst springs (Figures 2d and 2e).
The term karst also refers to the particular hydrologic behavior of karst regions. Due to the dissolution
processes, enclosed depressions at the surface (swallow holes, dolines) channel water to dissolution-
enhanced fractures (karst conduits). Entire streams sink into karst conduits and emerge again as large springs
in well-developed karst systems.

3. Specic Characteristics of Karst


3.1. Karst Evolution
Karst typically develops from carbonate rocks, such as limestone (consisting of the mineral calcite, CaCO3)
and dolomite rock or dolostone (consisting of the mineral dolomite, CaMg(CO3)2). Karst can also develop
from gypsum or halite. However, contributions of gypsum and halite karst systems are unimportant for water

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 219
Reviews of Geophysics 10.1002/2013RG000443

Figure 2. Typical surface and subsurface karst features: (a) karren eld in the North of Mlaga, Spain (photo by Andreas Hartmann), (b) karren eld with doline in the
Swiss Jura Mountains, Switzerland (photo by Nico Goldscheider), (c) cross section through the epikarst zone and a funnel-shaped doline observed in a quarry in the
Swabian Alb, Germany (photo by Nico Goldscheider), (d) cave in the Ardche region, France (photo by Remy Wenger, SISKA), and (e) spring of the Loue River, France
(photo by Nico Goldscheider).

resources management, and we will therefore only focus on carbonate rock in the subsequent discussion.
Karst landscapes and aquifers result from intense water-rock interaction over long time periods. The solubility
of carbonate minerals in pure water is very low, but the presence of CO2 strongly increases this solubility. The
dissolution of carbonate rock, here represented by calcite, is described by the following chemical equilibrium:

CaCO3 H2 O CO2 Ca2 2HCO2


3 (1)

The products of this reaction are dissolved calcium (Ca2+) and bicarbonate (HCO3). Carbonate rock
dissolution depends on lithological factors, such as chemical and mineralogical purity of the rock (see
Goldscheider and Drew [2007] for more details), and physicochemical factors, such as temperature and CO2
partial pressure [e.g., Buhmann and Dreybrodt, 1985]. The CO2 contained in the water originates from the
atmosphere (approximately 400 ppm) [Mauna Loa Observatory, 2013] and, to an even larger extent, from
biological processes in the soil, such as respiration of plant roots and decomposition of buried plant material.
Soil CO2 partial pressures often range between 5,000 and 50,000 ppm [Liu et al., 2007].
The formation of karst aquifers and caves, a process also referred to as karstication, is controlled by
dissolution kinetics. Water rich in CO2 enters a narrow fracture and rapidly dissolves calcite in the rst few
meters until it reaches 75% calcite saturation. Beyond this point, dissolution rates drop to very low levels
[Berner and Morse, 1974; Dreybrodt, 1990; Plummer and Wigley, 1976]. This means that water does not reach
full saturation with respect to calcite quickly but remains slightly undersaturated when it enters fractures and
causes initial karstication at very slow rates. The karstication process shows a positive feedback when
dissolution causes wider fractures, allowing for higher ow and therefore increased calcite dissolution, which
increases fracture size and so forth (see Figure 3). For the same reason, karstication is a selective process.
Fractures that are initially only slightly wider than others have higher initial ow and calcite dissolution rates
and thus grow faster than the narrower fractures. As a result, karstication transforms fractured carbonate
rock into a karst aquifer that includes a hierarchically organized network of hydraulically connected open
fractures, conduits, and caves, sometimes drained by only one major spring (Figure 2) [Worthington and
Ford, 2009].
At the land surface, dissolution together with other hydrologic and geomorphologic processes give rise to
characteristic karst landforms (Figure 2) [Ford and Williams, 2007; Goeppert et al., 2011]. These landforms
depend on climatic conditions and can be used as archives for past environmental changes [De Waele et al.,
2009]. For example, karren that form under soil cover tend to be rounded, while karren that form on exposed

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 220
Reviews of Geophysics 10.1002/2013RG000443

Figure 3. Schematic description of the karstication process and its inuence on the hydrodynamic behavior of spring discharge.

limestone tend to be sharped rimmed. Rounded karren on exposed rock surfaces therefore indicate recent
soil erosion [Goldscheider, 2012; Liu et al., 2007].

3.2. Karst Hydrology


There are three types of porosities in karst systems: (1) micropores that develop during the genesis of the
carbonate rock, (2) small ssures and fractures that develop due to tectonic processes, and (3) large fractures
and conduits that develop due to karstication (section 3.1). The rst two porosities are usually referred to as
the matrix, while the latter are called (karst) conduits. These three types of porosities result in a strong
heterogeneity of water movement at the surface and in the subsurface [Bakalowicz, 2005]. Figure 4 provides a
schematic illustration of the functioning of karst systems based on a general conceptual model for karstic
aquifers [White, 2003]. Internal runoff and diffuse inltration represent autogenic recharge, while external
runoff and sinking streams provide allogenic recharge from the surrounding areas [Goldscheider and Drew,
2007]. The uppermost layer of the carbonate rock is referred to as the epikarst. It develops close to the
topographic surface through rapid dissolution (see section 3.1) [Williams, 2008]. In the epikarst, storage and
further concentration of the downward ow can occur, which is then routed downward to the karst conduits
[Aquilina et al., 2006; Williams, 1983]. Soils that develop on carbonate rock are usually high in clay content
[Ford and Williams, 2007] and therefore exhibit low inltration capacities when analyzed at small spatial
scales. However, larger-scale karst systems usually have high inltration capacities, which is due to lateral ow
toward larger fractures and swallow holes on the surface and in the epikarst [Jeannin and Grasso, 1997]. While
diffusively inltrated water percolates slowly through the matrix, ow in the karst conduits is often fast and
turbulent [White, 2002]. These water pressure gradients can produce exchange ow between the matrix and

Figure 4. Conceptual model of a karst system including all characteristic karst processes; dark green and red dashed lines
represent the soil/epikarst and the groundwater subsystems (see section 5.2).

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 221
Reviews of Geophysics 10.1002/2013RG000443

the karst conduits [White, 2003]. Additionally, varying recharge conditions can result in water losses to
neighboring systems through conduits (piracy routes) [Jukic and Denic-Jukic, 2009] or through moving
groundwater divides [Le Moine et al., 2007]. Depending on the degree of karstication, discharge takes place
at one or several main springs contributing to rivers or subterraneous lakes or the ocean [Fleury et al., 2007a].
Overow springs are activated when the conduit carrying capacity is exceeded [Worthington, 1991]. Overall,
the hydrological behavior of karst systems shows a duality in its process and storage dynamics [Kiraly, 1998]:
(1) Duality of inltration and recharge processes: diffusive, slow inltration and recharge into the matrix,
concentrated, rapid inltration, and recharge into the conduits; (2) duality of the subsurface ow eld:
low ow velocity in the matrix and fast ow velocity in the karst conduits; and (3) duality of discharge
conditions: low and continuous discharge during dry periods when the system is dominated by ow through
the matrix and high discharge with high temporal variability during rainfall events when ow through the
conduits is dominant.

4. Investigation Techniques for Karst Systems


Sustainable karst water resources management requires an adequate knowledge of the functioning of karst
systems. Understanding the characteristics of a karst system will allow us to assess its potential for drinking
water supply and other uses [Sheffer et al., 2010], its sensitivity to dry spells and pumping [Hao et al., 2012],
and its vulnerability to contamination [Pronk et al., 2009]. Monitoring and understanding the spatial and
temporal variability of ow and storage behavior will further enable the development and application of
appropriate prediction models for water management under potential future conditions. Furthermore, eld
data are necessary to tune the model parameters to the conditions of the specic karst system under study.
Karst systems require specic investigation methods that capture their duality due to their particular
dynamics, i.e., both slow/diffuse and fast/concentrated ow and storage dynamics [Goldscheider and Drew,
2007]. In the following paragraphs we discuss those methods and methodological adaptations that are
particularly suitable and relevant for analyzing karst systems.

4.1. Application of Articial and Natural Tracers


Articial tracers are substances added to moving water so that the actual travel ow paths and ow times
through hydrologic systems can be estimated. The ideal articial tracer is highly soluble, detectable at
extremely low concentrations (thus requiring only low injection quantities), behaves conservatively (i.e., no
degradation or adsorption), and is nontoxic to humans and the environment [Kss, 1998; Leibundgut et al.,
2009]. There are also natural tracers, i.e., natural substances added continuously and spatially to the water.
Widely used natural tracers include solutes of major ions or heavy isotopes of water [Leibundgut et al., 2009;
Mazor, 2004]. Oxygen-18, deuterium, and tritium can be regarded as ideal natural tracers since they are part
of the water molecule. Both articial and natural tracers are widely used to study hydrologic systems.
Articial tracers are particularly useful and applicable in karst aquifer systems where transit times are short
(thus limiting the required sampling and monitoring periods) and where the sole application of conventional
hydrogeological methods, such as piezometric maps and numerical groundwater models, often leads to
ambiguous and incomplete results [Worthington et al., 2012]. In karst hydrogeology, articial tracers can be
applied (1) to dene underground connections and to delineate spring catchment areas, (2) to estimate
geometric and hydraulic properties, such as conduit diameters and ow velocities, and (3) to identify the
origin or destination of contaminants, as well as to quantify contaminant transport processes [Goldscheider
et al., 2008].
Fluorescent dyes, such as uranine, sulforhodamine, eosin, or naphthionate, are the most commonly used
articial tracers, as they behave very close to ideal tracers. The detection limit of uranine, for example, is as
low as 2 g/m3 [Kss, 1998] so that long-distance tracer tests (order of tens of kilometers) are possible with
only small injection quantities (order of a few kilograms). It is possible to apply several tracers in so-called
multitracer tests since the abovementioned dyes have distinct uorescent properties allowing for their
analytical separation [Leibundgut et al., 2009].
Different methods are available to observe articial tracers at relevant monitoring sites, such as manual
or automatic sampling with subsequent laboratory analysis or the use of continuously recording eld
uorometers [Schnegg, 2002]. The primary result of a tracer test is a breakthrough curve (BTC) showing

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 222
Reviews of Geophysics 10.1002/2013RG000443

Figure 5. (left) Injection of tracer in the Blue Cave System, Blaubeuren, Swabian Jura, Germany (photo by Andreas Kuecha);
(right) example of a tracer breakthrough curve (BTC) and illustration of relevant parameters that can be directly obtained
from this curve.

concentration over time (Figure 5). BTCs allow direct determination of transit times and different ow velocities
(maximum, peak, and mean velocity). The recovered tracer mass (R) is obtained as follows:

R Q  cdt (2)
t0

where Q is discharge, c is concentration, and (Q c) denes the ux. Recovery can be expressed as an absolute
or a relative quantity. Under ideal experimental conditions (i.e., ideal tracer, injection into owing water,
and monitoring of the complete BTC), recoveries can be used to quantify underground ow rates and to
estimate the so-called system discharge. Based on such simple approaches, but with more data and
sophistication, it is possible to resolve complex conduit networks, determine conduit diameters, and to
estimate water volumes and conduit ow rates even at inaccessible locations [Goldscheider, 2009; Pronk et al.,
2005; Smart, 1988]. In a comparative study at Mammoth Cave karst aquifer, Worthington [2009] has shown
that the inclusion of tracer test results in numerical groundwater models can substantially improve the
validity of the model.
Natural tracers provide integrated information about karst systems without allowing for the specication of
input locations and times [Clark and Fritz, 1997; Mazor, 2004]. They can be used (1) to estimate the fractions
and mixing of water from different sources [Aquilina et al., 2006; Plummer et al., 1998], (2) to assess the
functioning of karst systems [Barber and Andreo, 2011; Mudarra and Andreo, 2011], and (3) to determine
origin and residence times of karst waters [Batiot et al., 2003; Long and Putnam, 2004]. Typical natural tracers
are major ions, trace elements, dissolved organic carbon DOC and natural uorescence, and water isotopes
[Hunkeler and Mudry, 2007; Leibundgut et al., 2009; Mazor, 2004]. While water isotopes are part of the water
molecule, and can therefore be regarded as the most ideal tracer, other natural tracers often show reactive
behavior in terms of exchange with soil and rock or in terms of decomposition. Simultaneous consideration of
natural tracers and spring discharge provides integrated information regarding key systems characteristics
(see section 4.2).
More advanced interpretation techniques make it possible to quantify dispersion, physical nonequilibrium
exchange between mobile (conduits) and immobile uid regions (e.g., matrix), as well as relevant
contaminant transport parameters [e.g., Field and Pinsky, 2000; Maloszewski et al., 2002]. These techniques
usually include the application of spatially lumped hydrologic models that conceptualize the transport
of the articial or natural tracer through the karst system considering convection, dispersion, mixing,
separate ow paths through different conduits, and exchange of tracer with the matrix [Maloszewski,
1994; Maloszewski et al., 1998]. That way, transit times (through articial tracers), residence times
(through natural tracers), mixing fractions, aquifer characteristics, and diffusion losses to the matrix
can be estimated.

4.2. Continuous Monitoring


Karst springs and karst aquifers typically show marked and rapid reactions to precipitation events in both
water quantity and quality variables. Therefore, monitoring at high temporal resolutions, ideally continuous
monitoring, is required to characterize the dynamic behavior and variability of karst systems. The combined

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 223
Reviews of Geophysics 10.1002/2013RG000443

interpretation of spring hydrographs and


chemographs is a key approach used in
karst hydrology (Figure 6) [Shuster
and White, 1971].
Groundwater models for alluvial aquifers
are often based on water table data from
observation wells (i.e., data at low temporal
resolution but dense spatial distribution).
Karst models, on the other hand, are
frequently calibrated and validated with time
series of discharge at karst springs (i.e., data
at high temporal but low spatial resolution)
[e.g., Fleury et al., 2007b]. However, these data
Figure 6. Example of combined reaction of discharge and chemophy- are only really useful for modeling purposes
sical variables; time period 1: when discharge Q begins to react on the if relevant input data series are available at
precipitation event P, water that was stored in the conduits is pressed the same time. This requires concomitant
out (rst peak of turbidity due to mobilization of sediments in the
monitoring of rainfall and sinking streams,
conduits), then solute-enriched water from the conduits or epikarst
arrives (small peak in specic electric conductivity SEC); time period 2: again in terms of quantity and its
event water arrives at the spring, bringing sediments and bacteria hydrochemical composition. Continuous
from the surface (peak of E. coli and second peak of turbidity), diluting monitoring usually involves the use of
the preevent water (decreases in SEC and temperature T), and washing probes and data loggers, often combined
out organic carbon from the soil (peak of total organic carbon TOC).
with data teletransmission, or autosamplers
for microbial and natural tracer analyses.
At karst springs, discharge variations by factors of 10 to 100 within hours or days are common.
Intermittent karst springs dry up during long dry periods but exhibit the largest discharges which can
be tens of m3/s following intense precipitation [Kresic and Stevanovic, 2009]. Water tables in caves and
karst aquifers can vary by several tens of meters or more [Bonacci and Roje-Bonacci, 2000]. Pressure
probes are commonly used for the continuous observation of water tables. They can also be used to
monitor spring discharge if a stage-discharge curve has been established through independent
ow measurements.
Physical properties of water, such as temperature or specic electrical conductivity (SEC), can also easily be
monitored and are therefore often used as natural tracers to obtain information on karst aquifer behavior.
Rainwater generally has a different temperature and a lower SEC than groundwater. The arrival of freshly
inltrated rainwater at a karst spring can thus be recognized by changing water temperature and decreasing
SEC [Ford and Williams, 2007] (Figure 6). However, the inverse effect has also been observed, caused by
dynamically changing catchment boundaries and groundwater ow directions in the aquifer system [Ravbar
et al., 2011].
Hydrochemical data series at high resolution can be obtained by means of ion-specic probes, in situ wet
chemistry systems, or autosampling with subsequent laboratory analyses. Chemical parameters monitored in
karst water can be divided into soil-related (e.g., CO2, DOC, and total organic carbon TOC), carbonate rock-
related (Ca2+, Mg2+, and HCO3), other rock type related (e.g., K+ from silicate weathering and SO42
from gypsum), and anthropogenic compounds or contaminants (e.g., pesticides) [Hunkeler and Mudry, 2007].
As an example, owing to the different solution kinetics of Mg and Ca minerals, the Mg/Ca ratio in karst spring
water is often used to obtain information on residence times in karst systems, advantageously combined with
other time tracers, such as organic carbon [Batiot et al., 2003]. Monitoring of stable and radioactive isotopes of
the water molecule and water constituents can also be used to obtain quantitative information on transit
times, mixing processes, water-rock interaction, and aquifer behavior [e.g., Criss et al., 2001; Erss et al., 2012;
Hartmann et al., 2013c; Roa-Garca and Weiler, 2010].
While spring observations only characterize the bulk response of the unsaturated and saturated zones,
karstic caves offer unique insights into unsaturated zone and recharge dynamics. Isotopic proles within
speleothems (mostly calcite) deposited from cave drips may characterize paleoclimate and climatic change
[McDermott et al., 2006]. Monitoring of the dripping water itself allows for the analysis of short-term

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 224
Reviews of Geophysics 10.1002/2013RG000443

dynamics. Thereby, drips can be classied


into distinct types [Arbel et al., 2010]:
(1) perennial drips discharge all year long,
(2) seasonal drips during the rainy season,
and (3) poststorm/overow drips respond
directly to rainfall events. This behavior can
characterize the duality of inltration,
recharge, and unsaturated ow in karstic
systems with the rst two types dominated
by the matrix and the third type dominated
by the conduit system [Lange et al., 2010].
More recent monitoring approaches also
include parameters such as dissolved gases
[Klump et al., 2008; Savoy et al., 2011], trace
metals [Vesper and White, 2003], turbidity,
Figure 7. Hydraulic conductivity-scale effect [after Kiraly, 1975], expanded
and particle size distribution [Pronk et al.,
by a temporal dimension that includes the evolution of a conduit
network (early stage of karstication: no karst conduit network devel- 2007], as well as microbiological variables
oped, hydraulic conductivity at catchment/aquifer scale not much [Pronk et al., 2008]. These variables yield
higher than at the borehole scale; system strongly karstied: devel- considerable information concerning karst
oped network of karst conduits, hydraulic conductivity at much larger aquifer dynamics and transport processes,
than at the borehole scale; see also section 3.1).
though this knowledge has not yet been
fully transferred to modeling.
4.3. Assessment of Karst Aquifer Properties
Direct and indirect information concerning the conduit network can be obtained by means of
speleological exploration and tracer tests as discussed earlier, while hydraulic methods and geophysical
surveys help to investigate the aquifer at other places. Hydraulic methods require pumping and/or
observation wells, while geophysical methods can be applied where drilling is not available (besides
borehole geophysics). In this context, the hydraulic conductivity-scale effect [Kiraly, 1975] is of primary
importance. Due to the spatial scale of observation, the conductivity of carbonate rock samples obtained
from laboratory tests (scale of micropores) is several orders of magnitude lower than the conductivity
obtained from hydraulic borehole tests (scale of ssures and small fractures), which is again much lower
than the conductivity of the conduit network at aquifer scale (scale of large fractures and conduits)
obtained from tracer tests (Figure 7).
Hydraulic methods, including pumping tests, slug tests, and multiple pressure permeability tests, deliver
quantitative hydraulic parameters, such as transmissivity values and storage coefcients, while also allowing
for the estimation of the basic hydraulic aquifer setting (conned, unconned, semiconned, etc.) and of the
boundary conditions (xed hydraulic head, no-ow boundary, etc.). A large number of analytical solutions are
available not only for a variety of aquifer types and settings, mostly for porous aquifers, but also for karst
aquifers [Kresic, 2007]. In karst aquifers, ow is often concentrated in a small number of dissolution-enlarged
bedding planes or fractures [Worthington et al., 2012]. This can lead to discontinuous, stepped drainage
characteristics of a pumped well (Figure 8a). Major karst conduits near the pumped well can have a similar
hydraulic effect as a nearby river or lake, i.e., a xed hydraulic head boundary that leads to quasi-stationary
conditions (Figure 8b) [Larsson, 1984].
In locations where boreholes are not available, or for interpolation between existing boreholes,
geophysical methods can be used to obtain indirect information on the internal geometry and the external
boundaries of the aquifer and on its hydraulic properties [Bechtel et al., 2007]. Seismic methods are most
suitable to delineate geological boundaries and, thus, the external geometry of the aquifer [Bechtel et al.,
2007]. Microgravity, a method that measures very small changes in the Earths gravity, can be used to
identify large voids, such as potential sinkholes, as well as caves and major karst conduits [e.g., Debeglia
et al., 2006]. Electromagnetic methods are useful to identify anomalies, such as fractures [e.g., Bosch and
Mueller, 2001]. Spontaneous potential methods that measure the electrical potential eld caused by
natural-occurring currents of the Earth can be used to identify downward movement of water and thus

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 225
Reviews of Geophysics 10.1002/2013RG000443

Figure 8. Examples of time-drawdown curves from pumping tests in karst aquifers: (a) stepped drawdown resulting from
consecutive drainage of a limited number of solutionally enlarged bedding planes and (b) a nearby karst conduit acts as
xed-head boundary and leads to quasi-stationary conditions (modied after Kresic [2006] and Larsson [1984]).

help to identify zones of preferential inltration. Ground-penetrating radar is most useful to characterize
epikarst structure and heterogeneity [Al-fares et al., 2002]. A wide range of borehole geophysical
methods can be combined with hydraulic borehole methods to obtain a more complete picture of
aquifer structure and hydraulics. More detailed information on geophysical methods can be found in
Bechtel et al. [2007].

5. Approaches to the Prediction of Karst Water Resources


Karst simulation models cover a wide range of application areas. They can be used for global analysis, i.e.,
the analysis of the integrated behavior of the karst system hydrodynamics [Kovacs and Sauter, 2007], for
the theoretical investigation of karst evolution [see Gabrovsek and Dreybrodt, 2001; Kaufmann and Braun,
2000; Liedl et al., 2003], for water quality and vulnerability simulations [Butscher and Huggenberger, 2008;
Charlier et al., 2012; Hartmann et al., 2013a], and nally for the simulation of karst hydrology and the
prediction of karst water resources. The simplest way to model karst hydrology is the application of black
box models. They transfer input to output without the explicit representation of any physical processes.
Often they are composed of analytical transfer functions [Jukic and Denic-Jukic, 2006, 2008] or neural
networks [Hu et al., 2008; Kurtulus and Razack, 2006]. Due to the lack of process representation, their results
lose their reliability outside the range of conditions they were calibrated for [Kuczera and Mroczkowski,
1998]. For this reason, estimates of future karst water resources likely require the application of process-
based models that adequately consider karst hydrology and how it might vary over longer time periods
(see section 4). There are two main groups of such process-based models: distributed and lumped karst
modeling approaches.
5.1. Distributed Karst Simulation Models
Karst systems cover areas from < 1 km2 up to thousands of km2. Distributed karst models discretize the
karst system in two- or three-dimensional grids and require the assignment of characteristic hydraulic
parameters and system states to each grid cell. For a two-dimensional grid, the grid resolution is usually
around 0.011 km2. Distributed karst models are further described in Kinzelbach [1986], Wang and Anderson

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 226
Reviews of Geophysics 10.1002/2013RG000443

[1982], and Huyakorn and Pinder [1983]. In distributed karst models, groundwater ow in the matrix is
commonly described using Darcys law:
H
Ss KH (3)
t
Where SS is the specic storage coefcient, K the hydraulic conductivity, H is the hydraulic head, is the Nabla
operator, and t is time. In the more dynamic conduits, ow is most often represented by the Darcy-Weisbach
equation [e.g., Liedl et al., 2003; Reimann et al., 2011a]:
H q2
 (4)
x 2gd
Where represents a friction coefcient, d is the conduit diameter, q is the mean ow velocity, and x is the
ow length along the ow path. There are different possibilities to include karst heterogeneity in distributed
models that we will elaborate on in the following sections.
5.1.1. Equivalent Porous Medium Approach
The equivalent porous medium approach (EPM) assumes that hydraulic heterogeneities can be represented
by average properties, i.e., an equivalent porous medium (Figure 9b). Previous studies showed that at
regional scales this assumption can be used for karst water resource estimation and predictions [e.g., Loaiciga
et al., 2000; Rodrguez et al., 2013; Scanlon et al., 2003], since local inuences of karst conduits may average
out over larger areas [Abusaada and Sauter, 2013]. However, since the EPM approach does not consider
rapid ow in the karst conduits, it loses realism in systems with a high degree of karstication [Worthington,
2009]. To apply this approach, average hydraulic properties and aquifer geometry have to be known
(see section 4.3). In addition, groundwater level time series that represent the average groundwater
dynamics over the whole aquifer area should be available to calibrate the model.
5.1.2. Double Continuum Approach
The Double Continuum approach (DC) considers the heterogeneity of karst systems through the denition of
two interacting continua, one for the matrix and another one for the karst conduits (Figure 9c). Using a linear
exchange term, both continua exchange water as a function of their states (water levels). That way the DC
approach can describe the dual behavior of karst aquifers as shown in Teutsch and Sauter [1998], Marchal
et al. [2008], and Kordilla et al. [2012]. Hydraulic properties of the matrix and the karst conduits, as well as the
aquifer geometry, should be known when applying this approach (see section 4.3). Possible evaluation
variables for this strategy include spring discharges, groundwater head observations, and tracer observations
that represent both the karst conduits behavior as well as the matrix behavior.
5.1.3. Combined Discrete-Continuum Approach
The combined discrete-continuum approach (CDC) models the matrix as a continuum in which the karst
conduits are embedded as discrete elements [Kiraly and Morel, 1976] (Figure 9d). In this way, the spatial
inuence of the karst conduits on groundwater levels in the matrix can be calculated across the whole extent of
the karst system. The CDC approach has been applied in many theoretical studies of karst processes [Reimann
et al., 2011a, 2011b] and of karst evolution [Bauer and Liedl, 2005; Liedl et al., 2003]. It is included in the widely
used groundwater model MODFLOW (MODFLOW-CFP) [Shoemaker et al., 2008]. For its application, hydraulic
properties of the matrix and the karst conduits should be known, as well as aquifer and conduits geometry and
location (see section 4.3). It can be evaluated using similar variables as the discussed for the DC approach above.
There have also been attempts to explicitly consider the geometry of the fracture network [Cacas et al.,
1990; Dverstorp et al., 1992], but the data required to achieve robust karst water resources predictions are
generally not available. Further information can be found in other reviews, which describe the capabilities
and applications of the different subtypes of distributive modeling approaches in greater detail [e.g.,
Ghasemizadeh et al., 2012; Kovacs and Sauter, 2007].
5.2. Lumped Karst Simulation Models
Lumped approaches conceptualize the physical processes at the scale of the whole karst system without
modeling spatial variability explicitly. They are mostly based on linear or nonlinear relationships between
storage and discharge:

dS
Q a  Sb (5)
dt

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 227
Reviews of Geophysics 10.1002/2013RG000443

2
Figure 9. Representation of (a) the real karst system (medium scale: 50200 km ) by the different distributed modeling
approaches: (b) equivalent porous medium approach EPM, (c) Double Continuum approach DC, and (d) combined
discrete-continuum approach CDC.

Where Q is discharge, S is the stored water volume, and a and b are model parameters that can be related
to system properties [Rimmer and Hartmann, 2012]. In the special case when b = 1, equation (5) becomes a
linear relationship. The model parameters represent average or effective values over the whole modeling
domain due to the lumped structure of the model. Such parameters are generally incommensurate with
eld measurements [Wagener and Gupta, 2005]. Instead, their specic values are estimated through a
calibration process, in which the model parameters are systematically varied until an acceptable t between
observations and simulations, e.g., discharge at the karst spring, is found. Therefore, lumped karst modeling
approaches require continuously monitored discharge data (see section 4.2). In contrast to distributed karst
models that focus mostly on the movement of groundwater in the karst aquifer (see section 5.1), lumped
karst models are used to simulate karst processes from the inltration into the soil to discharge at the
karst spring in different ways and detail. In the following sections, a selection of lumped karst prediction
models is presented that consider (1) internal and external runoff, (2) epikarst storage and ow processes,
(3) groundwater storage and ow in karst conduits and the matrix, (4) varying surface and subsurface
recharge areas, and (5) drainage through several springs.

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 228
Reviews of Geophysics 10.1002/2013RG000443

Figure 10. Representation of (a) the real soil/epikarst system by (b) a simple overow reservoir [Fleury et al., 2007b] and by
(c) separate soil epikarst storages with varying depth [Hartmann et al., 2013a].

5.2.1. Internal and External Runoff


In the majority of the lumped karst models, soil and epikarst are represented by a single overow storage
(Figure 10b). In such a model, actual evapotranspiration is a direct function of the soil moisture state, and
recharge only occurs when the inltrating water exceeds the water holding capacity of the soil compartment.
Using such an overow reservoir, Jukic and Denic-Jukic [2009] identied recharge area variations of a spring
that were partially attributed to allogenic recharge. However, the majority of lumped karst models considers
concentrated inltration without a distinction between internal and external runoff [e.g., Butscher and
Huggenberger, 2008; Fleury et al., 2007b; Le Moine et al., 2008].
5.2.2. Epikarst Storage and Flow Concentration
Tritz et al. [2011] showed how epikarst recharge dynamics can be included in a lumped karst model by
introducing a model structure that simulates diffuse recharge to a groundwater matrix and hysteresis effects
in the initiation and routing of concentrated recharge. In their model, a soil and epikarst reservoir provides
diffuse recharge via slow drainage. It only produces concentrated recharge when the stored water volume
exceeds an activation threshold. Concentrated recharge stops when the water volume falls below a second
threshold, which is lower than the activation threshold. In a similar but slightly simpler approach, Rimmer and
Salingar [2006] distinguished between diffuse and concentrated recharge. They dened a soil and epikarst
routine that provides concentrated recharge only when its capacity is exceeded. Diffuse recharge is provided
via slow drainage from a lower outlet of the reservoir (similar to Figure 10b but with a second outlet at the
bottom of the reservoir that accounts for diffuse recharge). Hartmann et al. [2012b] found a way to consider
the spatial variability of the soil and epikarst characteristics in karst systems without explicit parameterization
of spatial heterogeneity. By dening distribution functions of system properties, their semilumped model
provides a set of simulation time series that represents the spatial and temporal variability of recharge
statistically (Figure 10c). Models that do not consider ow concentration in the epikarst rely on recharge
separation factors that dene constant fractions of matrix and conduit recharge [e.g., Charlier et al., 2012;
Fleury et al., 2009].
5.2.3. Groundwater Flow and Storage in Conduits and Matrix
The duality of slow and diffuse groundwater ow and fast and concentrated ow in the karst conduits is an
important concept in karst hydrology (section 3.2). Therefore, many lumped karst models include this
process, for example, by dening two separate groundwater reservoirs, one representing the matrix and the
other representing the conduits (Figure 11c). Some models simply combine both reservoirs [e.g., Butscher and
Huggenberger, 2008; Fleury et al., 2007b], while the matrix recharges the conduit reservoir in others [Geyer
et al., 2008; Kiraly, 2003]. Yet others only consider a matrix reservoir and dene routing functions that transfer

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 229
Reviews of Geophysics 10.1002/2013RG000443

Figure 11. Representation of different components of (a) the real karstic groundwater system, (b) intercatchment ground-
water ow [Le Moine et al., 2007], (c) exchanging matrix and conduits [Hartmann et al., 2013b], and (d) overow springs
[Rimmer and Hartmann, 2012].

the concentrated recharge directly to the spring [e.g., Le Moine et al., 2008; Tritz et al., 2011] or assume a
bidirectional exchange between matrix and conduits (Figure 11c) [e.g., Cornaton and Perrochet, 2002;
Hartmann et al., 2013b; Rimmer and Hartmann, 2012].
5.2.4. Varying Recharge Area
Only few modeling studies consider varying recharge areas in lumped karst models. Le Moine et al. [2007]
used an empirical function to calculate intercatchment groundwater ow from storage lling of the matrix
reservoir (Figure 11b). This way, their approach considers effects of moving groundwater divides and piracy
routes (Figure 4), though allogeneic recharge from external runoff is not explicitly included. Jukic and Denic-
Jukic [2009] took allogeneic recharge into account, as well as other contributions from moving groundwater
divides and piracy routes (see section 5.2.1). However, because their inverse modeling strategy requires
discharge observations as model input to obtain recharge as model output, it cannot be used for prediction.
Hartmann et al. [2013a] allowed for variably saturated soil and epikarst compartments through distribution
functions of soil and epikarst properties in their model. In their model, a soil compartment only contributes to
recharge when it is saturated. Hence, the area contributing to recharge is expressed by the number of
saturated model compartments and therefore varies with time.
5.2.5. Drainage by Several Springs
Rimmer and Salingar [2006] simulated the discharge behavior of a large karst system draining to several
springs. They considered the fact that a fraction of the recharge is not appearing at any of the observed
springs but drains to other outlets. Inside the model this assumption is included by attributing a constant
fraction of recharge to each spring. These factors that dene this attribution do not sum up to one but include
a certain fraction of recharge as an unknown system loss. Tritz et al. [2011] included an overow spring in
their conceptual model. Recharge to an overow spring is initiated when the threshold is exceeded in their
combined soil and epikarst reservoir (section 5.2.1). An extra reservoir with its own dynamics simulates
the discharge of the overow spring. Other studies simply allowed a second outow from their conduit
reservoir (Figure 11c) [Rimmer and Hartmann, 2012].
5.3. Comparison and Benchmarking
5.3.1. Calibration of Karst Models
The parameters of karst models can rarely be measured directly in the eld and therefore typically have to
be determined by calibration. This is due to the simplication of karst processes within distributed and
lumped karst models, the incommensurability between observations and model parameters, or the lack
of information. Calibration can be done manually by considering discharge observations [e.g., Jukic and
Denic-Jukic, 2009; Rimmer and Salingar, 2006], though this process is time consuming and cumbersome. In
most cases automatic calibration routines are applied involving the denition of a measure for the goodness

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 230
Reviews of Geophysics 10.1002/2013RG000443

of t, like the Nash-Sutcliffe efciency [Nash and Sutcliffe, 1970] or modications thereof [e.g., Charlier et al.,
2012; Mazzilli et al., 2012a]. The reason most lumped karst models introduced above only consider some, but
not all important karst processes, is overparameterization. Typically no more than four to six parameters can
be estimated when calibrating a model with a particular goodness of t measure [Jakeman and Hornberger,
1993]. More complex models can show increased prediction uncertainty because different combinations of
their parameters provide similar model performance [Wheater et al., 1986; Ye et al., 1997], i.e., the parameters
lose their identiability [Beven, 2006; Perrin et al., 2001]. If this occurred, the uncertainty of lumped (and
also distributed) models strongly increases [Wagener et al., 2002] and both modeling approaches can lose
their reliability in predicting karst water resources.
In order to improve parameter identiability, recent studies tried to extract more information from available
data or to use additional information. For example, in addition to the Nash-Sutcliffe efciency, Moussu et al.
[2011] considered autocorrelation of discharges, which has widely been used for the characterization of
karst systems [Mangin, 1984]. In another study, Mazzilli et al. [2012b] included ground-based gravity
measurements to improve the identiability of model parameters controlling the volume of groundwater
stored in the matrix. The value of using hydrochemical data and tracer information to increase parameter and
model structure identiability was shown, for instance, by Hartmann et al. [2013a, 2013b].
5.3.2. Data Requirements of Distributed Karst Models
Distributed karst models provide spatial information about the temporal evolution of groundwater levels.
Except for single applications of the EPM approach (section 5.1.1) [see Brouyre et al., 2003; Loaiciga et al.,
2000; Scanlon et al., 2003], most studies with distributed karst models were performed at well-explored test
sites [e.g., Birk et al., 2005; Doummar et al., 2012], or they restricted themselves to theoretical calculations of
general behavior of karst hydrology [e.g., Birk et al., 2006; Reimann et al., 2011b]. This is due to generally high
data requirements of distributed karst models. The necessary spatial information about hydraulic properties
of the matrix and the karst conduits, aquifer geometry (EPM and DC approach), or conduit geometry (CDC
approach) must all be determined in the eld (see section 4.3). This is rarely feasible at the same resolution
with which the model was discretized, which is commonly around 0.011 km2 per grid cell. Even an
inverse calibration with several time series of groundwater levels at several wells will most probably yield
wrong or biased groundwater levels at other grid points [Refsgaard, 1997]. For this reason, applications of
distributed karst models for spatial predictions of karst water resources predictions are limited.
5.3.3. Synthesis
Karst simulation models nd their application in karst evolution studies, in karst system characterization, in
water quality modeling and vulnerability assessment, and in water resources prediction. The approaches
presented in the preceding sections are, in theory, useful for the prediction of future karst water resources
when coupled with projections of future climates or scenarios of land use change. However, karst simulation
models have to adequately represent karst processes to provide robust predictions, and the available
information has to be sufcient to determine the model parameters. Even though distributed modeling
approaches show a high sophistication in karst process representation, the lack of available data often
prohibits their application for water resources prediction. For the calibration of lumped karst models,
information is more often available because their simple structures result in a low number of parameters. But
the simple structures are often not sufcient to represent the karst processes of the system to be modeled.
Therefore, their prediction performance is rather limited. Introducing more complex, process-based model
structures will result in overparameterization because of the large number of model parameters. In addition,
the parameters of karst models may not be stable in time [Singh et al., 2014; Wagener et al., 2003], and they
are dependent on the period they were calibrated [Merz et al., 2011]. For these reasons, and despite the
importance of karst water resources for human water supply, the impact of climate or land use change
has rarely been addressed separately for karst regions. There is an urgent need to develop adequate
simulation tools that make better use of the available data and to apply these tools to larger scales to allow
sustainable water management and to avoid threats for water security.

6. Challenges in the Prediction of Karst Water Resources


Natural and anthropogenic changes increasingly impact our environment. Many of these changes have
strong impacts on the hydrological regime of the watershed in which they occur [Dry et al., 2009; Stahl et al.,
2012; Wagener, 2007]. Furthermore, these changes can have signicant negative impacts on water security

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 231
Reviews of Geophysics 10.1002/2013RG000443

for the population and ecosystems in many


regions of the world [Conway and
Toenniessen, 1999; Falkenmark, 2001;
Johnson et al., 2001]. The problems will
continue to increase through the growing
imbalances between freshwater supply,
water consumption, and population
growth [De Stefano et al., 2012; Vrsmarty
et al., 2000]. Therefore, a major challenge for
hydrological science lies in providing
predictive models to support water
resources management decisions
under different future environmental,
population, and institutional scenarios
[Wagener et al., 2010].
About one quarter of the global population
Figure 12. Comparison of percentage of carbonate outcrops on is at least partially depending on freshwater
total country area and contribution of karst water to total water
from karst aquifers [Ford and Williams, 2007],
supply for selected countries in Europe [Cooperation in Science and
Technology, 1995]. and many countries receive a large fraction
of their drinking water from karst regions
(sometimes more than 50%, Figure 12).
Many karst regions will experience strong natural and anthropogenic changes in the future (Figure 1).
Reduced precipitation in some areas and higher temperatures will result in reduced groundwater recharge,
and rising agricultural demands will aggravate the depletion of karstic groundwater resources and their
contamination by fertilizers and pesticides. Although there are many large-scale modeling studies that deal
with changes of water availability due to global change [Gleeson et al., 2012; Milly et al., 2005; Taylor et al.,
2012], only few studies consider karst water resources at a large scale or possible threats to their future
availability [e.g., Finger et al., 2013; Hartmann et al., 2012c]. The reason for this gap lies majorly in a lack of data.
In this section, we discuss methods and directions (1) to enhance the available information that is necessary
to apply prediction models for karst water resources, (2) to produce reliable future predictions, and (3) to
apply karst models on large scales.

6.1. Dealing With an Insufcient Database


The tremendous heterogeneity of karst system characteristics means that detailed local observations have to
be made to dene a catchment or to tune a local model. A range of experimental karst catchments with a
more or less intensive monitoring databases are available [Hartmann et al., 2013c; White, 2002], but for most
karst systems, the available data are insufcient for the application of mechanistic distributed karst models
(see section 4.3). Methods to transfer information from sites with better data availability, like those reviewed
for the Predictions in Ungauged Basins initiative [Blschl et al., 2013; Sivapalan, 2003], are still insufcient
given the specic local evolution each karst system undergoes [Hartmann et al., 2013c]. The urgent need for
karst water resources prediction forces us to consider alternative solutions applicable under current data
conditions. Recent studies showed that there is a way to reduce the problem of overparameterization (see
section 5.3.1) by extracting more information from discharge and other time series of relevant hydrologic
variables. Several studieswithout a specic karst focusshowed that it is possible to extract more
information from existing time series by (1) multiobjective calibration with additional output variables like
hydrochemical time series, water isotopes or SEC (e.g., in Kuczera and Mroczkowski [1998], Seibert and
McDonnell [2002], Son and Sivapalan [2007], and Vach and McDonnell [2006]), (2) time series broken into
multiple time periods (e.g., in Dunne [1999], Gupta et al. [1998], Madsen et al. [2002], and Wagener et al.
[2001]), and (3) by signatures, i.e., hydrologically relevant metrics extracted from hydrological time series
[Gupta et al., 2008, 1992; Wagener et al., 2007; Yadav et al., 2007; Yilmaz et al., 2008].
These methods offer a high potential for advancement in karst hydrology, especially since such system
characterization is usually performed through the integrated system behavior at one major karst spring alone
[Jeannin and Sauter, 1998]. If a larger part of continuously monitored data and tracers (see sections 4.1 and 4.2)

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 232
Reviews of Geophysics 10.1002/2013RG000443

Figure 13. Example for excessive pumping: large karst spring in the Middle East (Faria spring, Palestine) before and after
excessive pumping began (photos by Jens Lange); hydrological impact modeling of climate change indicated that even
under a drier and warmer climate, the spring would have yielded signicant amounts of water if pumping had been limited
[Hartmann et al., 2012c].

could be considered in model identication and calibration, more complex and hence more process-based
models could be developed and applied. Existing process knowledge obtained from various karst exploration
techniques (see section 4) should guide the development of new modeling concepts. Successful examples of
incorporating hydrochemical information for calibration and evaluation of conceptual karst models can be
found in Charlier et al. [2012] and Hartmann et al. [2013a, 2013b]. Many studies performed in nonkarstic areas
already showed the value of remotely sensed data for better identication of model parameters. For instance,
Werth et al. [2009] used satellite gravity measurements to assess temporal variations of the water storage to
better calibrate a large-scale water balance model. Miralles et al. [2011] constrained their estimates of actual
evaporation using satellite-derived soil moisture data. Parajka and Blschl [2008] improved their model
calibration by assimilating satellite snow cover data. The value of these data sources for constraining karst
model parameters is different for each case and has to be explored by uncertainty and sensitivity analysis (see
methods in Hartmann et al. [2012a] and Hartmann et al. [2013a].

6.2. Simulations of Future Karst Water Resources


Climate change (Figure 1), as well as excessive water consumption (Figure 13), can have strong impacts on
future availability of karst water resources [Gleeson et al., 2012; Taylor et al., 2012]. Large uncertainties in
climate change projections propagate through both numerical groundwater models [Crosbie et al., 2011; Stoll
et al., 2011] and lumped karst models [Hartmann et al., 2012c], thus limiting their value for long-term
management and planning. These uncertainties are further exacerbated by the difculty of predicting
land use changes, such as urbanization or an intensication of agriculture. Alternative and more robust
approaches to provide insight into potential implication of future change are needed to support
decision making.
One possibility is the model-based assessment of conditions under which critical stakeholder-dened
thresholds of water availability are exceeded [Lempert et al., 2008; Singh et al., 2014]. Rather than propagating
highly uncertain downscaled climate change projections forward, one starts with critical model outputs and
uses the modeled system to estimate which combination of input conditions (including climate and land use)

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 233
Reviews of Geophysics 10.1002/2013RG000443

produce unwanted outcomes. That way, it is possible to compare these combinations with climate change
projections to assess their probability.
Another possibility is the use of paleoclimatic records to estimate possible future changes in water
availability. In karst caves, drip water containing dissolved calcium and bicarbonate can precipitate calcium
carbonate to form speleothems that contain paleoclimatic records (see reviews of Lachniet [2009] and
McDermott [2004]. Even though much uncertainty goes along with the interpretation of speleothem records
[Bradley et al., 2010; Fairchild et al., 2006], there have already been studies that exemplify their strong scientic
value. For instance, Medina-Elizalde and Rohling [2012] used quantitative paleoclimatic records from karst
caves and lake sediments in a simulation model and found that the collapse of the classic Maya civilization
was likely caused by only a modest reduction in precipitation. They did not mention that their study region
(Yukatan Peninsula, Mexico) is mainly a karstic region but highlighted that paleoclimatic records have a
high potential to constrain projections of future water availability. In another speleothem study, Baker et al.
[2013] provided quantitative assessments of changes in temperature, precipitation, water balance, and
specically in the timing and amount of groundwater recharge during glacial transition phases in China, i.e.,
during time periods when temperature rapidly increased or decreased.

6.3. Large-Scale Simulation of Karst Water Resources


As stated above, existing large-scale global change impact studies do not consider the particularities of
karst hydrology. Therefore, the reliability of these predictions remains limited for karst regions. A priori
information, i.e., maps of soil types, geology, etc., is often incommensurable with model parameters because
of their different scales. In addition, there is usually no a priori information about the degree of karstication
present. Hence, the necessary information to apply karst models at large scales is not available.
In recent years, mathematical techniques have been developed that use local a priori information about soil
and other characteristics to constrain the parameters of distributed models using so-called regularization
strategies [e.g., Pokhrel et al., 2008]. There was also progress in the combined application of local exploration
techniques (section 4.3) and distributed modeling [Geyer et al., 2013]. Regionalization, i.e., transferring
information from measured to unmeasured sites, is another widely used strategy, though the regionalization
of model parameters is often highly uncertain [Merz and Blschl, 2004; Samaniego et al., 2010; Wagener
and Wheater, 2006]. The regionalization of hydrologic signatures (e.g., runoff ratio or base ow index) is
sometimes shown to be more reliable [Sawicz et al., 2011; Yadav et al., 2007]. Even though there have been
many attempts to regionalize system signatures in nonkarstic regions [Anderson and Goulden, 2011; Carrillo
et al., 2011; Harman and Sivapalan, 2009; Oudin et al., 2010], studies addressing the transfer of karst models to
ungauged basins are rarely found. Hartmann et al. [2013c] found correlations between hydrologic and
hydrochemical system signatures and the parameters of a process-based karst model. However, they did not
succeed in regionalizing the signatures with the available predictors (mean temperature, altitude difference,
and mean annual precipitation). Usually, size and location of the subsurface catchment are unknown for karst
systems, which is why metrics that include the catchment area, e.g., runoff ratio or specic runoff [see Sawicz
et al., 2011; Yadav et al., 2007], cannot be used. Instead, information about general geological properties and
the degree of karstication might be quantied, using, for instance, descriptors of initial porosity, fractures,
and age of the karst system derived from modeling studies [e.g., Bloomeld et al., 2005; Hubinger and Birk,
2011] or age dating of speleothems [e.g., Vaks et al., 2003; White, 2007]. Using such descriptors, regionalization
of the necessary information to apply karst models at larger scales might be feasible.
Another possibility is the large-scale estimation of karst recharge. Recharge determines the maximum
amount of water that can be sustainably extracted from the aquifer for water supply [Seiler and Gat, 2007],
although hydraulic transmissivity and ecological aspects, such as minimum ow rates in springs and rivers
also have to be considered [Bredehoeft, 2002]. Hence, future water supply in karst regions depends on future
karst recharge as much as on extractions. Global modeling studies already tried to determine recharge at
large scales [e.g., Bundesanstalt fr Geowissenschaften und Rohstoffe/United Nations Educational, Scientic and
Cultural Organization, 2008; Wada et al., 2010], but their results have a limited value for karst water resources
estimation since they do not consider the particularities of karst hydrology. Present modeling approaches
[Andreu et al., 2011; Ireson and Butler, 2013] face the same problems as the prediction models for entire karst
systems when they attempt to predict karst recharge (sections 5.3 and 6.1). However, since inltration
capacities are usually high in karst regions [Jeannin and Grasso, 1997], surface runoff may often be neglected,

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 234
Reviews of Geophysics 10.1002/2013RG000443

and the difference between precipitation and actual evaporation can be used to estimate recharge. This way
the determination of recharge reduces to measuring precipitation and to assessing actual evaporation, which
is often feasible at the relevant scale [e.g., Armbruster, 2002].

7. Conclusions and Summary


Karst regions represent a large part of the Earths continental area and provide drinking water to almost
a quarter of the worlds population. This review highlights the relevance of karst regions for human
water supply and the importance of projections of future karst water availability. Existing experimental
methods to explore and characterize karst systems are presented followed by a review of the most
important approaches to karst modeling for water resources predictions. We show that karst systems
are distinct from other hydrological systems both in terms of their evolution and their hydrological
behavior. They are characterized by strong heterogeneities in their hydraulic characteristics and porosities.
They therefore need specic exploration techniques and modeling approaches. However, (1) difculties
in collecting sufcient information about the karst system properties, i.e., to capture its entire spatial
heterogeneity, and (2) the limited information content of observed discharge make the parameterization
and hence the application of presently available model approaches highly uncertain. Consequently,
there is still limited knowledge about how karst systems might respond to future changes of climate
and land use.
We will likely experience an increase of temperature and a decrease in precipitation in many karst areas in the
worldthe Mediterranean region serves as a prominent example. We can also expect continuous population
growth, paired with industrial expansion and improving (and more water-intensive) lifestyles. This has and
will put pressure on karst water resources in terms of water quantity and water quality. Hence, there is a
denite need for modeling approaches to provide reliable projections of future water availability in karst
regions. Such projections must include all relevant karst processes while reconciling data requirements
with data availability and existing exploration methods. Furthermore, we need better methods to apply karst
models at large scales.
However, heterogeneity of karst systems might also help to face the challenge of future predictions:
1. Due to their complex hydrological behavior, most databases of monitored karst springs also include a
large variability of water quality variables for a better understanding of the spring behavior and to
secure drinking water quality. This combined data can be used to enhance the information content of
observed spring discharge time series and may allow for the calibration of more realistic lumped karst
models with better prediction skills.
2. Many karst systems provide their own climate archives stored in the speleothems of their cavesa
unique advantage of karst hydrology. This information can be used to understand the hydrological
response to past changes and to provide better estimates of how the karst system and its water resources
will respond to future changes.
3. There is also considerable knowledge about the evolution of karst systems. Since the duality of preferen-
tial ow in the conduits and diffuse ow in the matrix is dominating karst hydrology, knowledge about
the evolution of conduit networks may help to transfer karst prediction models to sites with no or only
little information. Hence, we could make progress in the large-scale application of karst models to
provide estimates of karst water resources at relevant scales.
We are living in a changing world. Many of the expected changes force us to develop adaptation
strategies, especially for the management of our water resources. But we can only develop adaptation
strategies if we can anticipate the expected changes. As demonstrated in this review, achieving
projections of such expected changes is far from straightforward for karst water resources. The directions
we elaborated on show that it is feasible to improve our tools for karst water resources prediction and
projection and to provide the necessary information for decision makers though. In addition, the
processes that govern karst hydrology are not as special as they may appear. For instance, similar
dynamics of preferential and diffuse ow have also been found in soils and hillslopes where plant and
animals form and generate preferential pathways [Weiler and Naef, 2003]. Transferring knowledge from
modeling approaches in these elds to karst modeling and vice versa might open new research
perspectives that have not yet been explored.

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 235
Reviews of Geophysics 10.1002/2013RG000443

Glossary

Allogenic recharge is run off from neighboring or overlying nonkarstic rocks that drain into the karst aqui-
fer. Diffuse allogenic recharge percolates through overlaying layers before it slowly recharges the karst
aquifer. Concentrated allogenic recharge is run off from neighboring catchments that recharge the karst
aquifer by draining into fractures or swallow holes.
Articial tracers are substances that are added to the water in well-dened hydrologic situations in space
and time. They travel through a hydrological system, and their concentration is measured at the outlet of
the investigated system (see BTC). The ideal tracer is highly soluble, detectable at extremely low concentra-
tions (thus requiring only low injection quantities), behaves conservative (i.e., no degradation or adsorption),
and is nontoxic to humans and the environment [Kss, 1998; Leibundgut et al., 2009].
Autogenic recharge originates from precipitation directly on the karst landscape. Diffuse autogenic
recharge is produced by water slowly percolating vertically through the soil and the unsaturated matrix,
while concentrated autogenic recharge results from lateral ow concentration at the surface and in the
epikarst toward fracture and swallow holes that feed the karst conduits.
Breakthrough curves (BTC) are the primary result of a tracer test. They show the concentration of
the injected tracer over time at the outlet of the investigated system.
Carbonate rock is a type of sedimentary rock that is composed primarily of carbonate minerals. The two
major subtypes are limestone, which is composed of calcite or aragonite CaCO3, and dolostone, which is
composed of the mineral dolomite CaMg(CO3)2.
Calibration (of hydrological models): The process of model calibration is the systematic variation of
model parameters within a certain predened parameter range to nd an optimum agreement between
observed and modeled discharge (or other modeled and observed variables). It can be done manually, but
in most cases it is done by automatic calibration routines [e.g., Beven and Binley, 1992; Duan et al., 1992; Vrugt
et al., 2003] using measures for the goodness of t like the Nash-Sutcliffe efciency [Nash and Sutcliffe, 1970].
Distributed karst models are hydrologic models that discretize the karst system into a two- or three-
dimensional grid. To each grid cell characteristic hydraulic parameters are assigned that allow the
numerical solution of groundwater ow equations. That way, they provide spatial information about the
temporal evolution of groundwater levels.
Dolines or sinkholes are natural depressions or holes in the ground caused by surface dissolution or a
collapse of the surface layer due to karstication in the subsurface.
Dry valleys develop on carbonate rocks. They do not hold surface water because the karstied surface
has too high inltration capacities.
Epikarst is the uppermost layer of the carbonate rock that develops due to higher availability of CO2
originating from the precipitation and the soil [Williams, 2008]. It can store and further concentrate the
downward ow before it is routed to the karst conduits [Aquilina et al., 2006; Williams, 1983].
Hydrologic models are prediction tools that transform meteorological time series (precipitation, tempera-
ture, etc.) into discharge or groundwater level time series. Depending on their spatial resolution and
degree of process representation, they provide information about the llings of the different hydrological
storages (e.g., the soil or the karst aquifer) and different locations.
Karren develop on the surface due to karstication. When they form under soil cover, they tend to be
rounded, when formed on exposed limestone, they tend to be sharped rimmed. Hence, rounded karren
on exposed rock surfaces give indications for recent soil erosion [Goldscheider, 2012; Liu et al., 2007].
Karst: The term Karst comes from a Slavic word Kras or Krs, which means bleak waterless place. There is
also a region called Kras in Slovenia. Karst shows characteristic surface and subsurface karst features. One of
the rst persons exploring it was Jovan Cvijic in 1893, and the slightly modied name of the region was
used later on to refer to this type of landscape and its hydrological behavior.
Karst aquifers are groundwater bodies that store the water in a karst system. The water is stored in
both the karst conduits and the matrix (see their denitions in the glossary).
Karst conduits are dissolution-enlarged ssures or fractures. In a developed karst system they form a network
that rapidly conducts water to the spring on the one hand and exchanges water with the matrix on the other hand.
Karstication is the process of the formation of karst aquifers and caves by dissolution kinetics.
Lumped karst models hydrologic models that use mathematical expressions that conceptualize the
physical processes in the scale of the whole karst system, mostly based on linear or nonlinear relations

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 236
Reviews of Geophysics 10.1002/2013RG000443

between storage and discharge. They transform input time series (precipitation, temperature, etc.) into
system output (spring discharge).
Model parameters represent the characteristics of the modeled hydrological system that are commonly
considered to be constant over time. Mostly, they are effective or average representations of physical
properties of the system at the simulation scale of the model (lumped models: entire karst system, distributed
models: grid resolution). In many cases, when the measurement scale is not in accordance with the modeling
scale, they have to be estimated by calibration.
Natural tracers or environmental tracers are inherent components of the water cycle. Instead of being
added to the water at a well-dened time and location like articial tracers, they are naturally abundant
and added spatially and continuously to the hydrological system, for instance, heavy isotopes in the preci-
pitation or dissolution of components of the rock.
Preferential ow: Water owing in areas of higher hydraulic conductivities than their surroundings is
called preferential ow. It can take place at the surface (rills), within the soil (wormholes, voids created by
roots), and in the rock (karst conduits, large fractures).
Poljes are large at plains formed by karstication in regions with carbonate rocks. Supercial deposits
tend to accumulate on their oor. They are mostly drained by swallow holes or by episodic streams.
Swallow holes are places where water drains to the subsurface in a carbonate rock area. They are similar to
dolines, but they can also be the locations where entire streams from allogenic areas inltrate submerge
into the karst system.
Speleothem is the general term for all cave mineral deposits. In karst caves calcium-saturated water can
precipitate again and form speleothems (e.g., stalactites and stalagmites), mainly controlled by CO2 level in
the water that reaches the karst caves. They also contain paleoclimatic records (see reviews of Lachniet
[2009], and McDermott [2004]).

Acknowledgments References
This article belongs to a series of
Abusaada, M., and M. Sauter (2013), Studying the ow dynamics of a karst aquifer system with an equivalent porous medium model, Ground
Reviews in Karst Hydrogeology pro-
Water, 51(4), 641650.
moted by the IAH Karst Commission
Aeschbach-Hertig, W., and T. Gleeson (2012), Regional strategies for the accelerating global problem of groundwater depletion, Nat. Geosci.,
(www.iah.org/karst) with the goal to
5(12), 853861.
collect and evaluate current knowledge
Al-fares, W., M. Bakalowicz, R. Gurin, and M. Dukhan (2002), Analysis of the karst aquifer structure of the Lamalou area (Hrault, France) with
in different elds of karst hydrogeology
ground penetrating radar, J. Appl. Geophys., 51(24), 97106.
and make it available to the scientic
Anderson, R. G., and M. L. Goulden (2011), Relationships between climate, vegetation, and energy exchange across a montane gradient,
community. In addition, we want to
J. Geophys. Res., 116(G1), G01026, doi:10.1029/2010JG001476.
thank Juergen Strub from the Chair of
Andreu, J. M., F. J. Alcal, . Vallejos, and A. Pulido-Bosch (2011), Recharge to mountainous carbonated aquifers in SE Spain: Different
Hydrology, Freiburg, Germany, for
approaches and new challenges, J. Arid Environ., 75(12), 12621270.
designing large parts of the gures
Aquilina, L., B. Ladouche, and N. Doeriger (2006), Water storage and transfer in the epikarst of karstic systems during high ow periods,
presented in this review. This work was
J. Hydrol., 327, 472485.
supported by a fellowship within the
Arbel, Y., N. Greenbaum, J. Lange, and M. Inbar (2010), Inltration processes and ow rates in developed karst vadose zone using tracers in
Postdoc Programme of the German
cave drips, Earth Surf. Processes Landforms, 35(14), 16821693.
Academic Exchange Service (DAAD).
Armbruster, V. (2002), Grundwasserneubildung in Baden-Wrttemberg, Freiburger Schriften zu Hydrologie, Band 17, Freiburg im Breisgau, Germany.
Bakalowicz, M. (2005), Karst groundwater: A challenge for new resources, Hydrogeol. J., 13, 148160.
The Editor on this paper was Fabio
Baker, A., C. Bradley, and S. J. Phipps (2013), Hydrological modeling of stalagmite 18O response to glacial-interglacial transitions, Geophys.
Florindo. He thanks two anonymous
Res. Lett., 40, 32073212, doi:10.1002/grl.50555.
reviewers for their review assistance on
Barber, J. A., and B. Andreo (2011), Functioning of a karst aquifer from S Spain under highly variable climate conditions, deduced from
this paper.
hydrochemical records, Environ. Earth Sci., 65(8), 23372349.
Batiot, C., C. Emblanch, and B. Blavoux (2003), Total Organic Carbon (TOC) and magnesium (Mg2+): Two complementary tracers of residence
time in karstic systems, C. R. Geosci., 335(2), 205214.
Bauer, S., and R. Liedl (2005), Modeling the inuence of epikarst evolution on karst aquifer genesis: A time-variant recharge boundary
condition for joint karst-epikarst development, Water Resour. Res., 41, W09416, doi:10.1029/2004WR003321.
Bechtel, T., F. Bosch, and M. Gurk (2007), Geophysical methods in karst hydrogeology, in Methods in Karst Hydrogeology, edited by
N. Goldscheider and D. Drew, pp. 171199, Taylor and Francis/Balkema, London, U. K.
Berner, R. A., and J. W. Morse (1974), Dissolution kinetics of calcium carbonate in sea water: IV. Theory of calcite dissolution, Am. J. Sci., 274(2),
108134.
Beven, K. J. (2006), A manifesto for the equinality thesis, J. Hydrol., 320(12), 1836.
Beven, K. J., and A. Binley (1992), The future of distributed models: Model calibration and uncertainty prediction, Hydrol. Processes, 6,
279298.
Bundesanstalt fr Geowissenschaften und Rohstoffe/United Nations Educational, Scientic and Cultural Organization (BGR/UNESCO) (2008),
Groundwater Resources of the World 1: 25 000 000, BGR/UNESCO, Paris, Hannover.
Birk, S., R. Liedl, and M. Sauter (2006), Karst spring responses examined by process-based modeling, Ground Water, 44(6), 832836.
Birk, S., T. Geyer, R. Liedl, and M. Sauter (2005), Process-based interpretation of tracer tests in carbonate aquifers, Ground Water, 43,
381388.
Bloomeld, J. P., J. A. Barker, and N. Robinson (2005), Modeling fracture porosity development using simple growth laws, Ground Water, 43(3),
314326.

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 237
Reviews of Geophysics 10.1002/2013RG000443

Blschl, G., M. Sivapalan, T. Wagener, A. Viglione, and H. Savenije (2013), Runoff Prediction in Ungauged Basins: Synthesis Across Processes,
Places and Scales, Cambridge Univ. Press, Cambridge, U. K.
Bonacci, O., and T. Roje-Bonacci (2000), Interpretation of groundwater level monitoring results in karst aquifers: Examples from the Dinaric
karst, Hydrol. Processes, 14(14), 24232438.
Bosch, F., and I. Mueller (2001), Continuous gradient VLF measurements: A new possibility for high resolution mapping of karst structures,
First Break, 19(6), 343350.
Bradley, C., A. Baker, C. N. Jex, and M. J. Leng (2010), Hydrological uncertainties in the modelling of cave drip-water 18O and the implications
for stalagmite palaeoclimate reconstructions, Quat. Sci. Rev., 29(1718), 22012214.
Bredehoeft, J. D. (2002), The water budget myth revisited: Why hydrogeologists model, Ground Water, 40(4), 340345.
Brouyre, S., G. Carabin, and A. Dassargues (2003), Climate change impacts on groundwater resources: Modelled decits in a chalky aquifer,
Geer basin, Belgium, Hydrogeol. J., 12(2), 123134.
Buhmann, D., and W. Dreybrodt (1985), The kinetics of calcite dissolution and precipitation in geologically relevant situations of karst areas:
1. Open system, Chem. Geol., 48(1), 189211.
Butscher, C., and P. Huggenberger (2008), Intrinsic vulnerability assessment in karst areas: A numerical modeling approach, Water Resour.
Res., 44, W03408, doi:10.1029/2007WR006277.
Cacas, M. C., E. Ledoux, G. de Marsily, B. Tilie, A. Barbreau, E. Durand, B. Feuga, and P. Peaudercerf (1990), Modeling fracture ow with a
stochastic discrete fracture network model: Calibration and validation, 1. The ow model, Water Resour. Res., 26, 479489.
Carrillo, G., P. A. Troch, M. Sivapalan, T. Wagener, C. Harman, and K. Sawicz (2011), Catchment classication: Hydrological analysis of catch-
ment behavior through process-based modeling along a climate gradient, Hydrol. Earth Syst. Sci., 15(11), 34113430.
Charlier, J.-B., C. Bertrand, and J. Mudry (2012), Conceptual hydrogeological model of ow and transport of dissolved organic carbon in a
small Jura karst system, J. Hydrol., 460461, 5264.
Christensen, J. H., et al. (2007), Regional Climate Projections, in Climate Change 2007: The Physical Science Basis. Contribution of Working Group
I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, edited by S. Solomon et al., 996 pp., Cambridge Univ.
Press, Cambridge, U. K., and New York.
Clark, I. D., and P. Fritz (1997), Environmental Isotopes in Hydrogeology, Crc Pr Inc., Boca Raton, Fla.
Conway, G., and G. Toenniessen (1999), Feeding the world in the twenty-rst century, Nature, 402, C55C58.
Cornaton, F., and P. Perrochet (2002), Analytical 1D dual-porosity equivalent solutions to 3D discrete-continuum models. Application to
karstic spring hydrograph modelling, J. Hydrol., 262, 165176.
Cooperation in Science and Technology (COST) (1995), COST 65: Hydrogeological aspects of groundwater protection in karstic areas, Final
report (COST action 65) Rep., 446 pp., Bruessel, Luxemburg.
Criss, R. E., S. A. Fernandes, and W. E. Winston (2001), Isotopic, geochemical and biological tracing of the source of an impacted karst spring,
Weldon Spring, Missouri, Environ. Forensics, 2(1), 99103.
Crosbie, R. S., W. R. Dawes, S. P. Charles, F. S. Mpelasoka, S. Aryal, O. Barron, and G. K. Summerell (2011), Differences in future
recharge estimates due to GCMs, downscaling methods and hydrological models, Geophys. Res. Lett., 38, L11406, doi:10.1029/
2011GL047657.
De Stefano, L., J. Duncan, S. Dinar, K. Stahl, K. M. Strzepek, and A. T. Wolf (2012), Climate change and the institutional resilience of interna-
tional river basins, J. Peace Res., 49(1), 193209.
De Waele, J., L. Plan, and P. Audra (2009), Recent developments in surface and subsurface karst geomorphology: An introduction,
Geomorphology, 106(12), 18.
Debeglia, N., A. Bitri, and P. Thierry (2006), Karst investigations using microgravity and MASW: Application to Orlans, France, Near Surf.
Geophys., 4(4), 215225.
Dry, S. J., K. Stahl, R. D. Moore, P. H. Whiteld, B. Menounos, and J. E. Burford (2009), Detection of runoff timing changes in pluvial, nival, and
glacial rivers of western Canada, Water Resour. Res., 45, W04426, doi:10.1029/2008WR006975.
Doummar, J., M. Sauter, and T. Geyer (2012), Simulation of ow processes in a large scale karst system with an integrated catchment model
(Mike She)Identication of relevant parameters inuencing spring discharge, J. Hydrol., 426427, 112123.
Dreybrodt, W. (1990), The role of dissolution kinetics in the development of karst aquifers in limestone: A model simulation of karst evolu-
tion, J. Geol., 98(5), 639655.
Duan, Q. Y., S. Sorooshian, and H. V. Gupta (1992), Effective and efcient global optimization for conceptual rainfall-runoff models, Water
Resour. Res., 28(4), 10151031, doi:10.1029/91WR02985.
Dunne, S. M. (1999), Imposing constraints on parameter values of a conceptual hydrological model using baseow response, Hydrol. Earth
Syst. Sci., 3(2), 271284.
Dverstorp, B., J. Andersson, and W. Nordqvist (1992), Discrete fracture network interpretation of eld tracer migration in sparsely fractured
rock, Water Resour. Res., 28, 23272343, doi:10.1029/92WR01182.
Erss, A., J. Mdl-Sznyi, H. Surbeck, . Horvth, N. Goldscheider, and A. . Csoma (2012), Radionuclides as natural tracers for the charac-
terization of uids in regional discharge areas, Buda Thermal Karst, Hungary, J. Hydrol., 426427, 124137.
Fairchild, I. J., C. L. Smith, A. Baker, L. Fuller, C. Sptl, D. Mattey, F. McDermott, and E.I.M.F (2006), Modication and preservation of environ-
mental signals in speleothems, Earth Sci. Rev., 75(14), 105153.
Falkenmark, M. (2001), The greatest water problem: The inability to link environmental security, water security and food security, Int. J. Water
Resour. Dev., 17(4), 539554.
Field, M. S., and P. F. Pinsky (2000), A two-region nonequilibrium model for solute transport in solution conduits in karstic aquifers, J. Contam.
Hydrol., 44(34), 329351.
Finger, D., et al. (2013), Identication of glacial meltwater runoff in a karstic environment and its implication for present and future water
availability, Hydrol. Earth Syst. Sci., 17(8), 32613277.
Fleury, P., M. Bakalowicz, and G. de Marsily (2007a), Submarine springs and coastal karst aquifers: A review, J. Hydrol., 339(12), 7992.
Fleury, P., V. Plagnes, and M. Bakalowicz (2007b), Modelling of the functioning of karst aquifers with a reservoir model: Application to
Fontaine de Vaucluse (South of France), J. Hydrol., 345, 3849.
Fleury, P., B. Ladouche, Y. Conroux, H. Jourde, and N. Driger (2009), Modelling the hydrologic functions of a karst aquifer under active
water managementThe Lez spring, J. Hydrol., 365(34), 235243.
Foley, J. A., et al. (2011), Solutions for a cultivated planet, Nature, 478(7369), 337342.
Ford, D. C., and P. W. Williams (2007), Karst Hydrogeology and Geomorphology, Wiley, Chichester.
Gabrovsek, F., and W. Dreybrodt (2001), A model of the early evolution of karst aquifers in limestone in the dimensions of length and depth,
J. Hydrol., 240(34), 206224.

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 238
Reviews of Geophysics 10.1002/2013RG000443

Geyer, T., S. Birk, R. Liedl, and M. Sauter (2008), Quantication of temporal distribution of recharge in karst systems from spring hydrographs,
J. Hydrol., 348, 452463.
Geyer, T., S. Birk, T. Reimann, N. Driger, and M. Sauter (2013), Differentiated characterization of karst aquifers: Some contributions,
Carbonates Evaporites, 28(12), 4146.
Ghasemizadeh, R., F. Hellweger, C. Butscher, I. Padilla, D. Vesper, M. Field, and A. Alshawabkeh (2012), Review: Groundwater ow and
transport modeling of karst aquifers, with particular reference to the North Coast Limestone aquifer system of Puerto Rico, Hydrogeol. J.,
20(8), 14411461.
Gleeson, T., Y. Wada, M. F. Bierkens, and L. P. van Beek (2012), Water balance of global aquifers revealed by groundwater footprint, Nature,
488(7410), 197200.
Goeppert, N., N. Goldscheider, and H. Scholz (2011), Karst geomorphology of carbonatic conglomerates in the Folded Molasse zone of the
Northern Alps (Austria/Germany), Geomorphology, 130(34), 289298.
Goldscheider, N. (2009), Delineation of spring protection zones, in Groundwater Hydrology of Springs - Engineering, Theory, Management, and
Sustainability, edited by N. Kresic and Z. Stevanovic, pp. 305338, Elsevier, Oxford, U. K.
Goldscheider, N. (2012), A holistic approach to groundwater protection and ecosystem services in karst terrains, AQUA mundi, 3(2), 117124.
Goldscheider, N., and D. Drew (2007), Methods in Karst Hydrogeology, 264 pp., Taylor and Francis Group, Leiden, Netherlands.
Goldscheider, N., J. Meiman, M. Pronk, and C. Smart (2008), Tracer tests in karst hydrogeology and speleology, Int. J. Speleol., 37(1), 3.
Gupta, H. V., S. Sorooshian, and P. Yapo (1998), Toward improved calibration of hydrologic models: Multiple and noncommensurable
measures of information, Water Resour. Res., 34(4), 751763, doi:10.1029/97WR03495.
Gupta, H. V., T. Wagener, and Y. Liu (2008), Reconciling theory with observations: Elements of a diagnostic approach to model evaluation,
Hydrol. Processes, 22(18), 38023813.
Gupta, S. K., W. L. Darnell, and A. C. Wilber (1992), A parameterization for longwave surface radiation from satellite data: Recent improve-
ments, J. Appl. Meteorol., 31(12), 13611367.
Hao, Y., G. Liu, H. Li, Z. Li, J. Zhao, and T.-C. J. Yeh (2012), Investigation of karstic hydrological processes of Niangziguan Springs (North China)
using wavelet analysis, Hydrol. Processes, 26(20), 30623069.
Harman, C., and M. Sivapalan (2009), Effects of hydraulic conductivity variability on hillslope-scale shallow subsurface ow response and
storage-discharge relations, Water Resour. Res., 45, W01421, doi:10.1029/2008WR007228.
Hartmann, A., M. Kralik, F. Humer, J. Lange, and M. Weiler (2012a), Identication of a karst systems intrinsic hydrodynamic parameters:
Upscaling from single springs to the whole aquifer, Environ. Earth Sci., 65(8), 23772389.
Hartmann, A., J. Lange, M. Weiler, Y. Arbel, and N. Greenbaum (2012b), A new approach to model the spatial and temporal variability of
recharge to karst aquifers, Hydrol. Earth Syst. Sci., 16(7), 22192231.
Hartmann, A., J. Lange, . Viv Aguado, N. Mizyed, G. Smiatek, and H. Kunstmann (2012c), A multi-model approach for improved simulations
of future water availability at a large Eastern Mediterranean karst spring, J. Hydrol., 468469, 130138.
Hartmann, A., J. A. Barber, J. Lange, B. Andreo, and M. Weiler (2013a), Progress in the hydrologic simulation of time variant recharge areas of
karst systemsExemplied at a karst spring in Southern Spain, Adv. Water Resour., 54, 149160.
Hartmann, A., T. Wagener, A. Rimmer, J. Lange, H. Brielmann, and M. Weiler (2013b), Testing the realism of model structures to identify karst
system processes using water quality and quantity signatures, Water Resour. Res., 49, 33453358, doi:10.1002/wrcr.20229.
Hartmann, A., et al. (2013c), Process-based karst modelling to relate hydrodynamic and hydrochemical characteristics to system properties,
Hydrol. Earth Syst. Sci., 17(8), 33053321.
Hu, C., Y. Hao, T.-C. J. Yeh, B. Pang, and Z. Wu (2008), Simulation of spring ows from a karst aquifer with an articial neural network,
Hydrol. Processes, 22, 596604.
Hubinger, B., and S. Birk (2011), Inuence of initial heterogeneities and recharge limitations on the evolution of aperture distributions in
carbonate aquifers, Hydrol. Earth Syst. Sci., 15(12), 37153729.
Hunkeler, D., and J. Mudry (2007), Hydrochemical tracers, in Methods in Karst Hydrogeology, edited by N. Goldscheider and D. Drew, pp. 93121,
Taylor and Francis/Balkema, London, U. K.
Huyakorn, P. S., and G. F. Pinder (1983), Computational Methods in Subsurface Flow, Academic Press, London, U. K.
Ireson, A. M., and A. P. Butler (2013), A critical assessment of simple recharge models: Application to the UK Chalk, Hydrol. Earth Syst. Sci.,
17(6), 20832096.
Jakeman, A. J., and G. M. Hornberger (1993), How much complexity is warranted in a rainfall-runoff model?, Water Resour. Res., 29, 26372649,
doi:10.1029/93WR00877.
Jeannin, P.-Y., and D. A. Grasso (1997), Permeability and hydrodynamic behavior of karstic environment, in Karst Waters Environmental
Impact, edited by G. Gunay and A. I. Johnson, pp. 335342, A.A. Balkema, Rotterdam, Netherlands.
Jeannin, P.-Y., and M. Sauter (1998), Analysis of Karst Hydrodynamic Behaviour Using Global Approach: A Review, Universit de Neuchatel, Neuchatel.
Johnson, N., C. Revenga, and J. Echeverria (2001), Managing water for people and nature, Science, 292(5519), 10711072.
Jukic, D., and V. Denic-Jukic (2006), Nonlinear kernel functions for karst aquifers, J. Hydrol., 328, 360374.
Jukic, D., and V. Denic-Jukic (2008), Estimating parameters of groundwater recharge model in frequency domain: Karst springs Jadro and Z
rnovnica, Hydrol. Processes, 22, 45324542.
Jukic, D., and V. Denic-Jukic (2009), Groundwater balance estimation in karst by using a conceptual rainfall-runoff model, J. Hydrol., 373(34),
302315.
Kss, W. (1998), Tracing Technique in Geohydrology, Balkema, Rotterdam, Netherlands.
Kaufmann, G., and J. Braun (2000), Karst Aquifer evolution in fractured, porous rocks, Water Resour. Res., 36(6), 13811391, doi:10.1029/
1999WR900356.
Kinzelbach, W. (1986), Groundwater Modelling, Elsevier, Amsterdam, N. Y.
Kiraly, L. (1975), Rapport sur ltat actuel des connaissances dans le domaine des caractres physiques des roches karstiques, in
Hydrogeology of Karstic Terrains, edited by A. Burger and L. Dubertret, pp. 5367, International Union of Geological Sciences, Paris, France.
Kiraly, L. (1998), Modelling karst aquifers by the combined discrete channel and continuum approach, Bull. dHydrogologie, 16, 7798.
Kiraly, L. (2003), Karstication and groundwater ow, Speleogenesis Evolution Karst Aquifers, 1(3), 124.
Kiraly, L., and G. Morel (1976), Etude de regularisation de lAreuse par modle mathmatique, Bull. du Centre dHydrogologie, Neuchatel, 1,
1936.
Klump, S., O. A. Cirpka, H. Surbeck, and R. Kipfer (2008), Experimental and numerical studies on excess-air formation in quasi-saturated
porous media, Water Resour. Res., 44, W05402, doi:10.1029/2007WR006280.
Kordilla, J., M. Sauter, T. Reimann, and T. Geyer (2012), Simulation of saturated and unsaturated ow in karst systems at catchment scale using
a double continuum approach, Hydrol. Earth Syst. Sci., 16(10), 39093923.

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 239
Reviews of Geophysics 10.1002/2013RG000443

Kovacs, A., and M. Sauter (2007), Modelling karst hydrodynamics, in Methods in Karst Hydrogeology, edited by N. Goldscheider and D. Drew,
pp. 6591, Taylor and Francis/Balkema, London, U. K.
Kresic, N. (2006), Hydrogeology and Groundwater Modeling, 2nd ed., 828 pp., CRC Press, Boca Raton, Fla., 2 edition (October 26, 2006).
Kresic, N. (2007), Hydraulic methods, in Methods in Karst Hydrogeology, edited by N. Goldscheider and D. Drew, pp. 6591, Taylor and Francis/
Balkema, London, U. K.
Kresic, N., and Z. Stevanovic (2009), Groundwater Hydrology of Springs: Engineering, Theory, Management and Sustainability, Butterworth-
Heinemann, Oxford, U. K.
Kuczera, G., and M. Mroczkowski (1998), Assessment of hydrologic parameter uncertainty and the worth of multiresponse data, Water
Resour. Res., 34(6), 14811489, doi:10.1029/98WR00496.
Kurtulus, B., and M. Razack (2006), Evaluation of the ability of an articial neural network model to simulate the inputoutput responses of a
large karstic aquifer: The La Rochefoucauld aquifer (Charente, France), Hydrogeol. J., 15, 241254.
Lachniet, M. S. (2009), Climatic and environmental controls on speleothem oxygen-isotope values, Quat. Sci. Rev., 28(56), 412432.
Lange, J., Y. Arbel, T. Grodek, and N. Greenbaum (2010), Water percolation process studies in a Mediterranean karst area, Hydrol. Processes,
24(13), 18661879.
Larsson, I. (1984), Ground Water in Hard Rocks. Project 8.6 of the International Hydrological Programme, UNESCO, Paris, France.
Le Moine, N., V. Andrassian, C. Perrin, and C. Michel (2007), How can rainfall-runoff models handle intercatchment groundwater ows?
Theoretical study based on 1040 French catchments, Water Resour. Res., 43, W06428, doi:10.1029/2006WR005608.
Le Moine, N., V. Andrassian, and T. Mathevet (2008), Confronting surface- and groundwater balances on the La Rochefoucauld-Touvre
karstic system (Charente, France), Water Resour. Res., 44, W03403, doi:10.1029/2007WR005984.
Leibundgut, C., P. Maloszewski, and C. Kulls (2009), Tracers in Hydrology, 1st ed., 415 pp., Wiley-Blackwell, Chichester, U. K.
Lempert, R., J. Benjamin, P. Bryant, and S. C. Bankes (2008), Comparing algorithms for scenario discovery, Rep., Santa Monica, Calif.
Liedl, R., M. Sauter, D. Hckenhaus, T. Clemens, and G. Teutsch (2003), Simulation of the development of karst aquifers using a coupled
continuum pipe ow model, Water Resour. Res., 39(3), 1057, doi:10.1029/2001WR001206.
Liu, Z., Q. Li, H. Sun, and J. Wang (2007), Seasonal, diurnal and storm-scale hydrochemical variations of typical epikarst springs in subtropical
karst areas of SW China: Soil CO2 and dilution effects, J. Hydrol., 337(12), 207223.
Loaiciga, H. A., D. R. Maidment, and J. B. Valdes (2000), Climate-change impacts in a regional karst aquifer, Texas, USA, J. Hydrol., 227(14),
173194.
Long, A. J., and L. D. Putnam (2004), Linear model describing three components of ow in karst aquifers using 18O data, J. Hydrol., 296, 254270.
Madsen, H., G. Wilson, and H. C. Ammentorp (2002), Comparison of different automated strategies for calibration of rainfall-runoff models,
J. Hydrol., 261(14), 4859.
Maloszewski, P. (1994), Mathematical Modelling of Tracer Experiments in Fissured Aquifers, Professur fur Hydrologie Univ. Freiburg i. Br.,
Freiburg, Germany.
Maloszewski, P., R. Benischke, T. Harum, and H. Zojer (1998), Estimation of solute transport parameters in a karstic aquifer using articial
tracer experiments, Int. Contrib. Hydrogeol., 18, 177190.
Maloszewski, P., W. Stichler, A. Zuber, and D. Rank (2002), Identifying the ow systems in a karstic-ssured-porous aquifer, the Schneealpe,
Austria, by modelling of environmental 18O and 3H isotopes, J. Hydrol., 256, 4859.
Mangin, A. (1984), Pour une meilleure connaissance des systmes hydrologiques partir des analyses corrlatoire et spectrale, J. Hydrol.,
67(14), 2543.
Marchal, J. C., B. Ladouche, N. Doeriger, and P. Lachassagne (2008), Interpretation of pumping tests in a mixed ow karst system,
Water Resour. Res., 44, W05401, doi:10.1029/2007WR006288.
Mauna Loa Observatory (2013), Carbon cycle gases, Mauna Loa, Hawaii, United States, time series, in Manua Loa Obervatory, Global
Monitoring Division, Earth System Research Laboratory, National Oceanic and Atmospheric Administration, Mauna Loa. [Available at
http://www.esrl.noaa.gov/gmd/obop/mlo/.]
Mazor, E. (2004), Chemical and Isotopic Groundwater Hydrology, Marcel Dekker, New York.
Mazzilli, N., V. Guinot, and H. Jourde (2012a), Sensitivity analysis of conceptual model calibration to initialisation bias. Application to karst
spring discharge models, Adv. Water Resour., 42, 116.
Mazzilli, N., H. Jourde, T. Jacob, V. Guinot, N. Moigne, M. Boucher, K. Chalikakis, H. Guyard, and A. Legtchenko (2012b), On the inclusion of
ground-based gravity measurements to the calibration process of a global rainfall-discharge reservoir model: Case of the Durzon karst
system (Larzac, southern France), Environ. Earth Sci., 68(6), 16311646.
McDermott, F. (2004), Palaeo-climate reconstruction from stable isotope variations in speleothems: A review, Quat. Sci. Rev., 23(78),
901918.
McDermott, F., H. Schwarcz, and P. J. Rowe (2006), Isotopes in Speleothems, Springer, Dordrecht, Netherlands.
Medina-Elizalde, M., and E. J. Rohling (2012), Collapse of classic Maya civilization related to modest reduction in precipitation, Science,
335(6071), 956959.
Merz, R., and G. Blschl (2004), Regionalisation of catchment model parameters, J. Hydrol., 287(14), 95123.
Merz, R., J. Parajka, and G. Blschl (2011), Time stability of catchment model parameters: Implications for climate impact analyses,
Water Resour. Res., 47, W02531, doi:10.1029/2010WR009505.
Milly, P. C. D., K. A. Dunne, and A. V. Vecchia (2005), Global pattern of trends in streamow and water availability in a changing climate,
Nature, 438(7066), 347350.
Miralles, D. G., T. R. H. Holmes, R. A. M. De Jeu, J. H. Gash, A. G. C. A. Meesters, and A. J. Dolman (2011), Global land-surface evaporation
estimated from satellite-based observations, Hydrol. Earth Syst. Sci., 15(2), 453469.
Moussu, F., L. Oudin, V. Plagnes, A. Mangin, and H. Bendjoudi (2011), A multi-objective calibration framework for rainfall-discharge models
applied to karst systems, J. Hydrol., 400(34), 364376.
Mudarra, M., and B. Andreo (2011), Relative importance of the saturated and the unsaturated zones in the hydrogeological functioning of
karst aquifers. The case of Alta Cadena (Southern Spain), J. Hydrol., 397, 263280.
Nash, J. E., and J. V. Sutcliffe (1970), River ow forecasting through conceptual models. Part I: A discussion of principles, J. Hydrol., 10, 282290.
Oudin, L., A. Kay, V. Andrassian, and C. Perrin (2010), Are seemingly physically similar catchments truly hydrologically similar?, Water Resour.
Res., 46(11), W11558, doi:10.1029/2009WR008887.
Parajka, J., and G. Blschl (2008), The value of MODIS snow cover data in validating and calibrating conceptual hydrologic models, J. Hydrol.,
358(34), 240258.
Perrin, C., C. Michel, and V. Andrassian (2001), Does a large number of parameters enhance model performance? Comparative assessment
of common catchment model structures on 429 catchments, J. Hydrol., 241, 275301.

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 240
Reviews of Geophysics 10.1002/2013RG000443

Plummer, L. N., and T. M. L. Wigley (1976), The dissolution of calcite in CO2-saturated solutions at 25C and 1 atmosphere total pressure,
Geochim. Cosmochim. Acta, 40(2), 191202.
Plummer, L. N., E. Busenberg, J. B. McConnell, S. Drenkard, P. Schlosser, and R. L. Michel (1998), Flow of river water into a karstic limestone
aquifer: 1. Tracing the young fraction in groundwater mixtures in the Upper Floridan Aquifer near Valdosta, Georgia, Appl. Geochem., 13(8),
9951015.
Pokhrel, P., H. V. Gupta, and T. Wagener (2008), A spatial regularization approach to parameter estimation for a distributed watershed model,
Water Resour. Res., 44, W12419, doi:10.1029/2007WR006615.
Pronk, M., N. Goldscheider, and J. Zop (2005), Dynamics and interaction of organic carbon, turbidity and bacteria in a karst aquifer system,
Hydrogeol. J., 14(4), 473484.
Pronk, M., N. Goldscheider, and J. Zop (2007), Particle-size distribution as indicator for fecal bacteria contamination of drinking water from
karst springs, Environ. Sci. Technol., 41(24), 84008405.
Pronk, M., N. Goldscheider, and J. Zop (2008), Microbial communities in karst groundwater and their potential use for biomonitoring,
Hydrogeol. J., 17(1), 3748.
Pronk, M., N. Goldscheider, J. Zop, and F. Zwahlen (2009), Percolation and particle transport in the unsaturated zone of a karst aquifer,
Ground Water, 47(3), 361369.
Ravbar, N., I. Engelhardt, and N. Goldscheider (2011), Anomalous behaviour of specic electrical conductivity at a karst spring induced by
variable catchment boundaries: The case of the Podstenjek spring, Slovenia, Hydrol. Processes, 25(13), 21302140.
Refsgaard, J. C. (1997), Parameterisation, calibration and validation of distributed hydrological models, J. Hydrol., 198(14), 6997.
Reimann, T., C. Rehrl, W. B. Shoemaker, T. Geyer, and S. Birk (2011a), The signicance of turbulent ow representation in single-continuum
models, Water Resour. Res., 47(9), W09503, doi:10.1029/2010WR010133.
Reimann, T., T. Geyer, W. B. Shoemaker, R. Liedl, and M. Sauter (2011b), Effects of dynamically variable saturation and matrix-conduit coupling
of ow in karst aquifers, Water Resour. Res., 47, W11503, doi:10.1029/2011WR010446.
Rimmer, A., and Y. Salingar (2006), Modelling precipitation-streamow processes in karst basin: The case of the Jordan River sources, Israel,
J. Hydrol., 331, 524542.
Rimmer, A., and A. Hartmann (2012), Simplied conceptual structures and analytical solutions for groundwater discharge using reservoir
equations, in Water Resources Management and Modeling, edited by D. P. C. Nayak, pp. 217238, InTech, Kakinada, India.
Roa-Garca, M. C., and M. Weiler (2010), Integrated response and transit time distributions of watersheds by combining hydrograph separation
and long-term transit time modeling, Hydrol. Earth Syst. Sci., 14(8), 15371549.
Rodrguez, L., L. Vives, and A. Gomez (2013), Conceptual and numerical modeling approach of the Guarani Aquifer System, Hydrol. Earth Syst.
Sci., 17(1), 295314.
Samaniego, L., R. Kumar, and S. Attinger (2010), Multiscale parameter regionalization of a grid-based hydrologic model at the mesoscale,
Water Resour. Res., 46, W05523, doi:10.1029/2008WR007327.
Savoy, L., H. Surbeck, and D. Hunkeler (2011), Radon and CO2 as natural tracers to investigate the recharge dynamics of karst aquifers,
J. Hydrol., 406(34), 148157.
Sawicz, K., T. Wagener, M. Sivapalan, P. A. Troch, and G. Carrillo (2011), Catchment classication: Empirical analysis of hydrologic similarity
based on catchment function in the eastern USA, Hydrol. Earth Syst. Sci., 15(9), 28952911.
Scanlon, B. R., R. E. Mace, M. E. Barrett, and B. Smith (2003), Can we simulate regional groundwater ow in a karst system using equivalent
porous media models? Case study, Barton Springs Edwards aquifer, U.S.A, J. Hydrol., 276(14), 137158.
Schnegg, P.-A. (2002), An inexpensive eld uorometer for hydrogeological tracer tests with three tracers and turbidity measurement, in
Groundwater and Human Development, edited by E. Bocanegra, D. Martinez, and H. Massone, pp. 14841488, Leiden, Netherlands.
Seibert, J., and J. J. McDonnell (2002), On the dialog between experimentalist and modeler in catchment hydrology: Use of soft data for
multicriteria model calibration, Water Resour. Res., 38(11), 1241, doi:10.1029/2001WR000978.
Seiler, K. P., and J. R. Gat (2007), Groundwater Recharge From Run-Off and Inltration, Springer Verlag, Dordrecht, Netherlands.
Sheffer, N. A., E. Dafny, H. Gvirtzman, S. Navon, A. Frumkin, and E. Morin (2010), Hydrometeorological daily recharge assessment model
(DREAM) for the Western Mountain Aquifer, Israel: Model application and effects of temporal patterns, Water Resour. Res., 46, W05510,
doi:10.1029/2008WR007607.
Shoemaker, W. B., E. L. Kuniansky, S. Birk, S. Bauer, and E. D. Swain (2008), Documentation of a Conduit Flow Process (CFP) for MODFLOW-2005,
US Department of the Interior, US Geological Survey, Reston, Va.
Shuster, E. T., and W. B. White (1971), Seasonal uctuations in the chemistry of lime-stone springs: A possible means for characterizing
carbonate aquifers, J. Hydrol., 14(2), 93128.
Singh, R., T. Wagener, R. Crane, M. E. Mann, and L. Ning (2014), A vulnerability driven approach to identify 1 adverse climate and land use
change combinations for critical hydrologic indicator thresholdsApplication to a watershed in Pennsylvania, U.S.A., Water Resour. Res.,
50, 34093427, doi:10.1002/2013WR014988.
Sivapalan, M. (2003), Prediction in ungauged basins: A grand challenge for theoretical hydrology, Hydrol. Processes, 17(15), 31633170.
Smart, C. (1988), Articial tracer techniques for the determination of the structure of conduit aquifers, Ground Water, 26(4), 445453.
Son, K., and M. Sivapalan (2007), Improving model structure and reducing parameter uncertainty in conceptual water balance models
through the use of auxiliary data, Water Resour. Res., 43(1), W01415, doi:10.1029/2006WR005032.
Stahl, K., L. M. Tallaksen, J. Hannaford, and H. A. J. van Lanen (2012), Filling the white space on maps of European runoff trends: Estimates
from a multi-model ensemble, Hydrol. Earth Syst. Sci., 16(7), 20352047.
Stoll, S., H. J. Hendricks Franssen, M. Butts, and W. Kinzelbach (2011), Analysis of the impact of climate change on groundwater related
hydrological uxes: A multi-model approach including different downscaling methods, Hydrol. Earth Syst. Sci., 15(1), 2138.
Taylor, R. G., et al. (2012), Ground water and climate change, Nat. Clim. Change, 3(4), 322329.
Teutsch, G., and M. Sauter (1998), Distributed parameter modelling approaches in karst hydrological investigations, Bull. dHydrogologie, 16, 99109.
Tritz, S., V. Guinot, and H. Jourde (2011), Modelling the behaviour of a karst system catchment using non-linear hysteretic conceptual model,
J. Hydrol., 397(34), 250262.
Vach, K. B., and J. J. McDonnell (2006), A process-based rejectionist framework for evaluating catchment runoff model structure, Water
Resour. Res., 42, W02409, doi:10.1029/2005WR004247.
Vaks, A., M. Bar-Matthews, A. Ayalon, B. Schilman, M. Gilmour, C. J. Hawkesworth, A. Frumkin, A. Kaufman, and A. Matthews (2003),
Paleoclimate reconstruction based on the timing of speleothem growth and oxygen and carbon isotope composition in a cave located in
the rain shadow in Israel, Quat. Res., 59(2), 182193.
Vesper, D. J., and W. B. White (2003), Metal transport to karst springs during storm ow: An example from Fort Campbell, Kentucky/
Tennessee, U.S.A, J. Hydrol., 276(14), 2036.

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 241
Reviews of Geophysics 10.1002/2013RG000443

Vrsmarty, C. J., P. Green, J. Salisbury, and R. B. Lammers (2000), Global water resources: Vulnerability from climate change and population
growth, Science, 289(5477), 284.
Vrugt, J. A., H. V. Gupta, W. Bouten, and S. Sorooshian (2003), A shufed complex evolution metropolis algorithm for optimization and
uncertainty assessment of hydrologic model parameters, Water Resour. Res., 39(8), 1201, doi:10.1029/2002WR001642.
Wada, Y., L. P. H. van Beek, C. M. van Kempen, J. W. T. M. Reckman, S. Vasak, and M. F. P. Bierkens (2010), Global depletion of groundwater
resources, Geophys. Res. Lett., 37, L20402, doi:10.1029/2010GL044571.
Wagener, T. (2007), Can we model the hydrological impacts of environmental change?, Hydrol. Processes, 21(23), 32333236.
Wagener, T., and H. V. Gupta (2005), Model identication for hydrological forecasting under uncertainty, Stoch. Environ. Res. Risk Assess., 19(6),
378387.
Wagener, T., and H. S. Wheater (2006), Parameter estimation and regionalization for continuous rainfall-runoff models including uncertainty,
J. Hydrol., 320(12), 132154.
Wagener, T., D. P. Boyle, M. J. Lees, H. S. Wheater, H. V. Gupta, and S. Sorooshian (2001), A framework for development and application of
hydrological models, Hydrol. Earth Syst. Sci., 5(1), 1326.
Wagener, T., M. J. Lees, and H. S. Wheater (2002), A toolkit for the development and application of parsimonious hydrological models,
Math. Models Large Watershed Hydrol., 1, 87136.
Wagener, T., N. McIntyre, M. J. Lees, H. S. Wheater, and H. V. Gupta (2003), Towards reduced uncertainty in conceptual rainfall-runoff modelling:
Dynamic identiability analysis, Hydrol. Processes, 17(2), 455476.
Wagener, T., M. Sivapalan, P. Troch, and R. Woods (2007), Catchment classication and hydrologic similarity, Geography Compass, 1(4), 901931.
Wagener, T., M. Sivapalan, P. A. Troch, B. L. McGlynn, C. J. Harman, H. V. Gupta, P. Kumar, P. S. C. Rao, N. B. Basu, and J. S. Wilson (2010),
The future of hydrology: An evolving science for a changing world, Water Resour. Res., 46, W05301, doi:10.1029/2009WR008906.
Wang, H. F., and M. P. Anderson (1982), Introduction to Groundwater Modeling, W. H. Freeman, San Francisco.
Weiler, M., and F. Naef (2003), An experimental tracer study of the role of macropores in inltration in grassland soils, Hydrol. Processes, 17(2),
477493.
Werth, S., A. Gntner, S. Petrovic, and R. Schmidt (2009), Integration of GRACE mass variations into a global hydrological model, Earth Planet.
Sci. Lett., 277(12), 166173.
Wheater, H. S., K. H. Bishop, and M. B. Beck (1986), The identication of conceptual hydrological models for surface water acidication,
Hydrol. Processes, 1(1), 89109.
White, W. B. (2002), Karst hydrology: Recent developments and open questions, Eng. Geol., 65, 85105.
White, W. B. (2003), Conceptual models for karstic aquifers, Speleogenesis, 1(1), 16.
White, W. B. (2007), Cave sediments and paleoclimate, J. Cave Karst Studies, 69(1), 7693.
Williams, P. W. (1983), The role of the Subcutaneous zone in karst hydrology, J. Hydrol., 61, 4567.
Williams, P. W. (2008), The role of the epikarst in karst and cave hydrogeology: A review, Int. J. Speleol., 37(1), 110.
Williams, P. W., and D. C. Ford (2006), Global distribution of carbonate rocks, Z. Geomorphol., Suppl. 147, 12.
Worthington, S. R. (1991), Karst Hydrogeology of the Canadian Rocky Mountains, 227 pp., McMaster Univ., Hamilton, Ont., Canada.
Worthington, S. R. (2009), Diagnostic hydrogeologic characteristics of a karst aquifer (Kentucky, USA), Hydrogeol. J., 17(7), 16651678.
Worthington, S. R., and D. C. Ford (2009), Self-organized permeability in carbonate aquifers, Ground Water, 47(3), 326336.
Worthington, S. R., C. Smart, and W. Ruland (2012), Effective porosity of a carbonate aquifer with bacterial contamination: Walkerton, Ontario,
Canada, J. Hydrol., 464465, 517527.
Yadav, M., T. Wagener, and H. Gupta (2007), Regionalization of constraints on expected watershed response behavior for improved
predictions in ungauged basins, Adv. Water Resour., 30(8), 17561774.
Ye, W., B. C. Bates, N. R. Viney, M. Sivapalan, and A. J. Jakeman (1997), Performance of conceptual rainfall-runoff models in low-yielding
ephemeral catchments, Water Resour. Res., 33(1), 153166, doi:10.1029/96WR02840.
Yilmaz, K. K., H. V. Gupta, and T. Wagener (2008), A process-based diagnostic approach to model evaluation: Application to the NWS
distributed hydrologic model, Water Resour. Res., 44, W09417, doi:10.1029/2007WR006716.

Erratum

In the originally published version of this article, the Introduction to this manuscript began with a reference
to the current global abstraction of groundwater which has since been removed. These errors have been
corrected and this version may be considered the authoritative version of record.

HARTMANN ET AL. 2014. American Geophysical Union. All Rights Reserved. 242

Vous aimerez peut-être aussi