This article appeared in a journal published by Elsevier. The attached
copy is furnished to the author for internal non-commercial research
and education use, including for instruction at the authors institution
and sharing with colleagues.
Other uses, including reproduction and distribution, or selling or
licensing copies, or posting to personal, institutional or third party
websites are prohibited.
In most cases authors are permitted to post their version of the
article (e.g. in Word or Tex form) to their personal website or
institutional repository. Authors requiring further information
regarding Elsevier’s archiving and manuscript policies are
encouraged to visit:
http://www.elsevier.com/copyright
Author's personal copy
Journal of Hydrology 386 (2010) 55–66
Contents lists available at ScienceDirect
Journal of Hydrology
journal homepage: www.elsevier.com/locate/jhydrol
Water exchange and pressure transfer between conduits and matrix
and their influence on hydrodynamics of two karst aquifers with sinking streams
Vincent Bailly-Comte a,b,*, Jonathan B. Martin b, Hervé Jourde a, Elizabeth J. Screaton b,
Séverin Pistre a, Abigail Langston b
a
b
Hydrosciences Laboratory, UMR 5569, University of Montpellier 2, Place E. Bataillon, 34095 Montpellier Cedex 5, France
Department of Geological Sciences, University of Florida, 241 Williamson Hall, PO Box 112120, Gainesville, FL 32611, United States
a r t i c l e
i n f o
Article history:
Received 9 May 2009
Received in revised form 15 January 2010
Accepted 7 March 2010
This manuscript was handled by P. Baveye
Editor-in-chief, with the assistance of
Michel Bakalowicz, Associate Editor
Keywords:
Karst aquifers
Hydrograph recession
Matrix–conduit exchange
Hydrodynamics
Causse d’Aumelas
Santa Fe River
s u m m a r y
Karst aquifers are heterogeneous media where conduits usually drain water from lower permeability volumes (matrix and fractures). For more than a century, various approaches have used flood recession
curves, which integrate all hydrodynamic processes in a karst aquifer, to infer physical properties of
the movement and storage of groundwater. These investigations typically only consider flow to the conduits and thus have lacked quantitative observations of how pressure transfer and water exchange
between matrix and conduit during flooding could influence recession curves.
We present analyses of simultaneous discharge and water level time series of two distinctly different
karst systems, one with low porosity and permeability matrix rocks in southern France, and one with
high porosity and permeability matrix rocks in north-central Florida (USA). We apply simple mathematical models of flood recession using time series representations of recharge, storage, and discharge processes in the karst aquifer. We show that karst spring hydrographs can be interpreted according to
pressure transfer between two distinct components of the aquifer, conduit and matrix porosity, which
induce two distinct responses at the spring. Water exchange between conduits and matrix porosity successively control the flow regime at the spring. This exchange is governed by hydraulic head differences
between conduits and matrix, head gradients within conduits, and the contrast of permeability between
conduits and matrix. These observations have consequences for physical interpretations of recession
curves and modeling of karst spring flows, particularly for the relative magnitudes of base flow and quick
flow from karst springs. Finally, these results suggest that similar analyses of recession curves can be
applied to karst aquifers with distinct physical characteristics utilizing well and spring hydrograph data,
but information must be known about the hydrodynamics and physical properties of the aquifer before
the results can be correctly interpreted.
Ó 2010 Elsevier B.V. All rights reserved.
Introduction
Classical hydrogeological surveys, such as well tests for estimating hydrodynamic parameters, tracer experiments, and speleological and geophysical observations provide only limited
information on karst groundwater behavior because of the heterogeneous distribution of flow and storage within karst aquifers.
Many methods using hydrodynamic or hydrochemical analyses
of karst springs have been developed to understand both mass
and pressure transfers within the aquifer. Primary among these
methods is analysis of hydrograph recession curves, which assumes that spring behavior integrates all processes that occur in
* Corresponding author at: Hydrosciences Laboratory, UMR 5569, University of
Montpellier 2, Place E. Bataillon, 34095 Montpellier Cedex 5, France.
E-mail address: [email protected] (V. Bailly-Comte).
0022-1694/$ - see front matter Ó 2010 Elsevier B.V. All rights reserved.
doi:10.1016/j.jhydrol.2010.03.005
the aquifer. Such recession analyses can be based on many characteristics of the spring discharge such as flow, water temperature,
chemical or isotopic composition as a ‘‘global response” of the
karst aquifer to input events (Kiraly, 2003). For example, the aim
of spring hydrograph separation, which is addressed in this paper,
is to infer relative contributions from different components of a
karst aquifer (Atkinson, 1977; Bonacci, 1993; Padilla et al.,
1994). Although critical to wise use of water resource and remediation of introduced contaminants, water provenances (e.g., infiltration, pre-storm water stored in vadose or phreatic zones, etc.)
requires measurements of tracers, typically natural or introduced
chemicals, and is not included here.
Spring hydrographs have been used to infer geometric and
hydrodynamic properties of the aquifer volumes that control
underground flows, for example, between flow in large channels
or large fissures and flow in thin fissures (Renault, 1959; Schoeller,
1965; Forkasiewicz and Paloc, 1967; Drogue, 1969), or conduit
Author's personal copy
56
V. Bailly-Comte et al. / Journal of Hydrology 386 (2010) 55–66
flow and diffuse (or matrix) flow (Atkinson, 1977; Baedke and
Krothe, 2001; Martin and Dean, 2001; Martin and Screaton,
2001). Following the model proposed by Maillet (1905), one or
more exponential functions have been used to describe
hydrograph recessions. For instance, Schoeller (1965) assumes that
successive exponential decreasing limbs on spring hydrographs
are due to changes of flow regimes (from turbulent to laminar).
Alternatively, Forkasiewicz and Paloc (1967) believe that the decline is due to depletion of various components of the aquifer
which have successively lower hydraulic conductivity. These interpretations suggest that karst systems consist of several parallel
reservoirs, all contributing independently to the discharge of the
spring. Spring hydrographs are thus described by a model of multiple reservoirs, each of which empties with a characteristic exponential flow rate. As pointed out by Ford and Williams (2007),
however, it seems unrealistic to consider different reservoirs to
be hydraulically isolated from each other. Changes in hydrographs
have also been explained to result from changes in catchment area,
discharge of temporary springs (overflows), or temporary flooding
of poljes or caves (Bonacci, 1993). Mangin (1975) considers that
flood flow occurs when the spring hydrograph is still influenced
by underground runoff and/or infiltration through the unsaturated
zone, while a unique exponential decrease according to the Maillet
(1905) law characterizes the base flow after infiltration stops. Consequently, an inflection point on the recession curve represents the
end of the infiltration and is used to separate flood flow (also
called quick flow), which demonstrates a non exponential decrease, from base flow, which demonstrates an exponential decrease (Mangin, 1975). In this conceptual model, groundwater
movement and storage in fractures or intergranular porosity of
the matrix is neglected, assuming that the strong contrast of several orders of magnitude in permeability between matrix and conduits limits exchange and makes the storage in the matrix
negligible (Bakalowicz, 2005). Following this concept, base flow
only originates from water that was previously stored in large
karst voids called Annex to Drains System (ADS, also called ‘‘karst
annexes” (Mangin, 1975; BRGM-CNRS, 1992; Mangin, 1994; Palmer, 2003; Bakalowicz, 2005)). These voids are supposed to be
poorly connected with the main karst conduits, resulting in high
hydraulic head loss and the observed exponential decrease of discharge with time (Mangin, 1975).
Numerical simulations have been used to test the interpretation
of karst spring hydrographs. These simulations suggest that transient phenomena near a high hydraulic conductivity network are
sufficient to yield the observed exponential decrease that occurs
between the flood flow and base flow phases (Kiraly and Morel,
1976; Kiraly, 2003; Kovacs et al., 2005; Kovacs and Perrochet,
2008). Eisenlohr et al. (1997) have shown that recession limbs of
spring hydrographs cannot be interpreted solely in terms of infiltration process and discharge from components of the aquifer that
are characterized by distinct hydraulic conductivities. The shape of
the recession limbs also depends on the geometry and density
of the high permeability channel network as well as on the form
of the recharge function (Geyer et al., 2008).
The various conceptualizations of karst spring hydrographs
lead to interpretations that can differ from each other or even
be contradictory, largely because they depend strongly on
characteristics of each study location, such as climate, dynamics
of recharge, geology, regional hydraulic gradient, degree of karstification, and respective volumes of unsaturated and saturated
materials. These studies often differentiate a flood flow component related to fast infiltration and groundwater flows within
aquifer volumes of high hydraulic conductivity (conduits), from
a slow component often called base flow, related to the slow
depletion of the low hydraulic conductivity volumes or low permeability volumes (LPV, (Jeannin, 1996)), which represents the
more or less porous matrix of the karst aquifer (intergranular
porosity, joints and/or fractures).
Physical interpretation of recession curves, specifically the
separation between flood and base flow, can be complicated
for any single aquifer, or between aquifers, because the spatial
and temporal distribution of recharge is generally unknown. Furthermore, karst aquifers often differ in the relative importance of
matrix porosity within the aquifer, which depends on burial
depths and resulting alteration of the carbonate. Karst aquifers
found in largely unaltered carbonate rocks has been labeled as
eogenetic karst systems, while highly altered, dense, and low
porosity karst aquifers have been labeled as telogenetic karst
systems (Vacher and Mylroie, 2002). This distinction of two
types of karst aquifers is important because the matrix porosity
greatly influence the hydrodynamics of the aquifers (Florea and
Vacher, 2006). Most hydrodynamic studies of karst aquifers use
telogenetic karst systems as examples and consequently may neglect the influence of intergranular matrix porosity on spring
discharge (White, 2002). However, exchange of water between
conduits and matrix occurs during floods (Martin and Dean,
2001), when hydraulic gradients between the conduit and matrix
are reversed (Atkinson, 1977; Drogue, 1980; Jeannin, 1996; Bailly-Comte et al., 2009). These conditions result in base flow that
is zero (or negative) (Kiraly, 2003). Indeed, the widely accepted
assumption that base flow contributes to spring flow at the
beginning of the flood event (see for example Atkinson (1977)
or Bonacci (1993)) can only be true if the initial water level in
the LPV is high enough to exceed that within the conduits
throughout the flood.
Some karst aquifers provide the opportunity to simultaneously
monitor inflows to conduits from allogenic inputs (i.e. sinking
stream) and outflows from conduits at springs. When conduits
are full of water, pressure is transmitted from the sinking stream
to the spring, allowing direct measurement of allogenic recharge.
This recharge differs from autogenic recharge where both diffuse
and local infiltration through soils and epikarst (Mangin, 1975) distributes groundwater flow in the vadose zone into two distinct
flow paths. One flow path contributes to flood flow and occurs in
high permeability zones and the other is slow flow through the
matrix porosity and contributes to base flow (Kiraly, 2003; Perrin
et al., 2003).
In this study, only flow from the LPV to conduits will be referred to as base flow, which means that a negative flow will be
considered as a recharge to the LPV from the conduits, and consequently as a period without base flow. With these definitions,
spring hydrograph separation should thus allow better understanding of karst aquifer flow, in particular the relative influence
of high and low permeability volumes on karst spring flows, the
consequences of flow inversions between these volumes, and the
origin and evolution of base flow during and following floods,
regardless of the porosity or permeability of the LPV. This paper
is divided into two parts, the first of which briefly reviews previous
work dealing with hydrodynamic analysis of karst spring hydrographs and describes mathematical models for drainage of the
conduits and matrix. The second part focuses on two karst systems, one eogenetic and one telogenetic, in which conduit flows
(presence of large cave systems), recharge and water level representative of conduits and LPV are well characterized. The two systems differ greatly in their head gradients and amounts of matrix
porosity and permeability. These case studies allow application
and evaluations of mathematical flow models to discuss the origin,
the evolution and the relative importance of flood flow and base
flow during and following recharge in two different karst contexts,
and development of conceptual models to show the influence of
exchange between matrix and conduits on flood hydrographs of
karst springs.
Author's personal copy
V. Bailly-Comte et al. / Journal of Hydrology 386 (2010) 55–66
where QCFR(t) [L3 T1] is the discharge at time t at the exit of R. The
integration of Eq. (2c) between the flood peak time tmax [T] and t [T]
indicates there is a linear decrease of the discharge according to
time:
Methods
Simple mathematical models for karst spring recession
Drainage of low permeability volumes (porous matrix) – Matrix
Restrained Flow Regime (MRFR)
Boussinesq (1904) and Maillet (1905) were among the first
authors to propose simple mathematical models for spring recessions. Others have expanded these initial models in attempts to
improve mathematical fits to field measurements (Drogue, 1969,
1972; Mangin, 1975). Dewandel et al. (2003) and Ford and Williams (2007) review these different recession models.
In this study, the slow depletion of the LPV is assumed to control the flow regime at the spring during the base flow period on
the spring hydrograph similar to groundwater flow in porous media. An approximate solution of the groundwater flow equation
(without recharge) in one (Boussinesq, 1904) or two (Kovacs
et al., 2005) dimensions for a porous, homogeneous, isotropic
and unconfined aquifer gives a single exponential decrease. Kovacs
et al. (2005) term this flow regime the Matrix Restrained Flow Regime period (MRFR):
Q MRFR ðtÞ ¼ Q 0 eat
Drainage of high permeability volumes (conduits) – Conduit Flow
Regime (CFR)
The first part of a recession curve can be strongly influenced by
recharge through the vadose zone modulated by its transfer
through the phreatic zone (Mangin, 1975), which explains why
flood flow is not easily described by mathematical functions.
Therefore, we initially assume that springs respond only to an impulse in hydraulic head elevation in conduits, there is no diffuse recharge through the vadose zone, and there is no exchange between
conduits and matrix. These strong assumptions and their applicability to real karst systems will be discussed subsequently.
Under these conditions, the whole conduit network is considered as a single reservoir (R) with infinite hydraulic conductivity,
and the Bernoulli incompressible flow equation of fluid dynamics
gives (Eq. (2a)):
pffiffiffiffiffiffiffiffi
2gh
ð2aÞ
where VR is the flow velocity at the exit of R [L T1], g is the gravitational constant [L T2], and h is the hydraulic head in R [L]. Considering that the horizontal area of R is constant, the mass
conservation in R during dt gives (Eq. (2b)):
AR dh ¼ aR V R dt
ð2bÞ
where AR is the area of R [L2] and aR is the flow section at the exit of
R [L2].
Combining and rearranging Eqs. (2a) and (2b) gives Eq. (2c)
dQ CFR ¼ 2g
a2R
dt
AR
(
a2
b ¼ 2g ARR
Q CFR ðtÞ ¼ Q max b t
ð3Þ
where Qmax [L3 T1] is the maximum discharge and b [L3 T2] is the
slope of the linear decrease of the spring discharge according to
time expressed in m3 s1 d1.
We refer to this simple model (Eq. (3)) as Conduit Flow Regime
(CFR). It has been shown that b will take high values for a large
flow section at the exit of R, which can be related to a high degree
of karstification (or a high conduit frequency) (Bailly-Comte, 2008).
Inversely, b will be low for a high horizontal section in R, meaning
that a small amount of water is stored in the conduit system. It can
also be demonstrated that quadratic head loss in R, which conceptually represents conduit constrictions, does not modify the linear
relationship in Eq. (3). In this case, b is found to be inversely proportional to the quadratic head loss coefficient (Bailly-Comte,
2008).
ð1Þ
where QMRFR(t) [L3 T1] is the discharge at time t, Q0 [L3 T1] is the
spring discharge when the MRFR period starts and a is the recession
coefficient [T1] usually expressed in days. This solution is referred
to as model MRFR; Eq. (1) supposes that no delayed recharge
occurs.
Boussinesq (1904) and Kovacs et al. (2005) expressed a according to physical parameters using an analytical approach. They
show that the recession coefficient (a) is proportional to the ratio
TL
, where L is the length of the aquifer domain [L], T is the hydrauS
lic transmissivity [L2 T1] and S is the storativity [–] of the equivalent porous, homogeneous, isotropic and unconfined media.
V R ðtÞ ¼
57
ð2cÞ
Base flow evolution
The difference between inflow at the swallet and outflow (DQ)
at the spring of a sink/rise karst system yields the groundwater discharge draining from (or injected to) the LPV to (or from) the conduit (Eq. (4)):
DQ ¼ Q S Q A
ð4Þ
where QS is the spring discharge and QA is the allogenic recharge.
This method allows computing base flow as the positive value
of DQ during and following floods and the amount of loss from
negative values of DQ.
A comparison of case studies
Study areas
The Vène Spring in the Aumelas-Thau karst system
Near Montpellier, Southern France, the Vène Spring (Fig. 1A),
Station Number 10162X0033) is a temporary spring partly fed by
the sinking of the ephemeral Coulazou River (Bonnet and Paloc,
1969). The Coulazou River crosses the Aumelas Causse where the
karst aquifer is formed in Jurassic limestones that are exposed at
the surface (Fig. 1A).
This karst system, known as the Aumelas-Thau karst system, is
an example of a telogenetic karst system, and feeds the ephemeral
Vène Spring in high water conditions. A gauging station is used to
estimate discharge a few meters downstream from the Vène Spring
(QS, Fig. 1A). In low flow, karst groundwater discharges farther to
the southwest by other permanent karst springs below a Tertiary
basin capped by impervious siliciclastic rocks. A monitoring network has been constructed in the region to assess the karst/river
hydrodynamic interactions (Jourde et al., 2007; Bailly-Comte
et al., 2008a,b). As part of this network, allogenic recharge is monitored using a gauging station upstream from the karst aquifer (QA,
Fig. 1A). Water level is recorded in a cave called Puits de l’aven
(Fig. 1A) and in a monitoring well PZ2 (Fig. 1). Well PZ2 is a
120 m deep, uncased well which has been drilled into the Jurassic
limestone; no large karst voids were identified during the drilling,
though fractures and small karst drains were encountered. Previous studies show that this well represents hydrodynamics of the
LPV influenced by the Coulazou River (Bailly-Comte et al., 2008b,
2009). Puits de l’aven is a cave that is at least 1300 m long with
Author's personal copy
58
V. Bailly-Comte et al. / Journal of Hydrology 386 (2010) 55–66
A
B
Fig. 1. Location and monitoring network of the A. Aumelas-Thau karst system in South of France (left) and B. the Santa Fe River sink/rise system within the upper Floridan
aquifer (right).
perennial water in its base (Douchet, 2007). During floods it acts as
a spring and/or a sinkhole along the riverbed and thus is formally
an estavelle.
The Santa Fe River sink/rise system within the upper Floridan aquifer
In north-central Florida, the River Rise Spring is the main outlet
of a sink/rise system sourced by the Santa Fe River (Fig. 1B). Similar
to the Aumelas-Thau system, a monitoring network has been constructed to assess the exchange of water between matrix and conduit (Martin and Dean, 2001; Martin and Screaton, 2001; Screaton
et al., 2004; Martin et al., 2006).
River Sink stage is recorded each day by the staff of the O’Leno
State Park and converted to runoff based on a rating curve developed by the Suwannee River Water Management District (QA,
Fig. 1B, Rating No. 3 for Station Number 02321898, Santa Fe River
at O’Leno State Park). Water levels are recorded at a 10 min interval
at the River Rise and converted to discharge using the rating curve
produced by Screaton et al. (2004) (Fig. 1B). Eight wells have been
screened close to the depth of the main conduit and four additional
wells have been drilled and screened across the water table within
a few meters of four of the deep wells. The deep wells are designated by a single digit (1–8) and the shallow wells are designated
by the letter A (4A, 5A, 6A, 7A). Most of these wells are instrumented so that water level, conductivity, and temperature are
monitored within the LVP at a 10 min time interval. These monitoring data provide information about pressure (water level) and
mass (temperature and electrical conductivity) transfer in the matrix during floods.
In high flow conditions, flood pulses from the Santa Fe River are
transmitted to the River Rise through large karst conduits (Screaton
et al., 2004). Increases in the hydraulic head of the conduits lead to
flow from the conduits to the porous matrix (Martin and Dean,
2001; Martin and Screaton, 2001). At this site, the matrix permeability is greater than at the Aumelas-Thau site, but is still signifi-
cantly lower than the conduit permeability. Thus, even though
Santa Fe River is an eogenetic karst system, the matrix represents
an LPV relative to the conduit. Moore et al. (2009) have shown that
the water chemistry in the Well 4 is relatively constant, though
this well is within a few hundred meters of the main karst conduits
(Fig. 1B), which suggests little exchange of mass between the conduit and Well 4. Consequently, we consider water level fluctuation
in Well 4 as representative of hydrodynamics of the LPV. The complementary shallow Well 4A is compared to Well 4 to assess vertical hydraulic gradient within LPV. Finally, conduit head is
characterized by measurements of water level at the River Rise.
Comparisons
Table 1 summarizes the main characteristics of the groundwater/surface water interactions and conduits/matrix hydrodynamic
properties for these two karst systems, focusing on their genesis,
their hydrologic regime, their hydrodynamics and especially the
boundary condition that applies in the conduit system during
flood.
The fundamental characteristics of the karst aquifers as well as
the recharge condition in their conduits (see boundary conditions
in Table 1) differ greatly in these two areas (Table 1), providing
ideal settings for comparative studies of fundamental processes.
Indeed, differences in topography between the two regions create
differences in water table gradients, resulting in differences in flow
rates through the aquifers, as well as responses to extreme precipitation events. In addition to topographic dissimilarities, physical
properties of the aquifers differ because of their geologic histories
(telogenetic vs. eogenetic). As a result of the age and deformation,
the French karst aquifers in Languedoc contain little primary
porosity and most water is stored within fractures and conduits.
The lack of deformation of the Florida karst is reflected in the high
primary porosity of the matrix rocks, up to 20% (Table 1, (Martin
et al., 2006)). This high primary porosity allows for extensive water
Author's personal copy
59
V. Bailly-Comte et al. / Journal of Hydrology 386 (2010) 55–66
Table 1
Comparisons of the two case studies showing the main characteristics of the groundwater/surface water interactions and conduits/matrix hydrodynamic properties.
Aumelas-Thau system
Santa Fe sink/rise system
Telogenetic
Eogenetic
5.3 km between the cave Puits de l’aven and the
Vène Spring
4.7 km from River Sink to River Rise
No surface flows in the River and no flows at the
spring
Base flow at the spring due to sinking surface
streams and release of groundwater storage
Flood transfer through open channel and karst
conduits in both vadose and phreatic zone
Flood transfer through karst conduits in the
phreatic zone
Discontinuous recharge (intermittent river).
Sinkholes are filled by surface water in few hours
during flash floods, coming from non-karst terrains
More than 15
Continuous recharge, large flood events from
upstream confined terrains that source the aquifer
over few days to months
Only River Sink drains the sinking stream
Low (few %)
During flood: Riverbed elevation = head boundary
condition in karst conduits; then, water level in
sinkholes decreases with no or few influence of
allogenic recharge
High (around 20%, Martin et al., 2006)
Sinking surface flows = flow boundary condition in
karst conduits
Type of karst aquifer
Dimensions
Distance (straight line) between
the monitored sinkhole and the spring
Hydrologic conditions before a flood
Flood propagation between
QA and QS (Fig. 1)
Allogenic recharge
Time scale
Number of sinkholes between QA and QS
Hydrodynamic properties
Matrix porosity
Boundary conditions in conduits
storage and influences flow paths through the Floridan aquifer.
These fundamental differences allow assigning typical flow behavior to each component of the karst spring hydrograph by different
approaches:
– In the Aumelas-Thau system, flood flow and base flow at Vène
Spring can be described according to time-series measurements
of water level in Puits de l’aven and Well PZ2 when concentrated
recharge along the riverbed ceases.
– In the Santa Fe sink/rise system, variations of base flow
during the entire recession period can be computed using the difference between outflow and inflow, assuming that autogenic
recharge can be neglected compared to inflows from the sinking
stream.
Flood flow and base flow at the Vène Spring
Spring hydrographs analysis according to water level evolution in
matrix and conduits
In Aumelas-Thau, 19 flood events have been recorded at Vène
Spring since October 2002; 10 have been described using simultaneous water level measurements in Well PZ2 and Puits de l’aven
since April 2004. Three distinct types of recession curves have been
identified at Vène Spring according to the initial (pre-storm) water
level in Well PZ2 (Fig. 2):
During low water table conditions, i.e. after the summer
drought, floods recorded at Vène Spring are relatively short, lasting
only a few days. At these times, spring hydrographs shows a linear
decrease with time (Fig. 2A) as long as the water level in conduits
(Puits de l‘aven) is higher than the water level in the matrix (Well
PZ2). Water level in Well PZ2 slightly decreases at the beginning
of the recession, indicating a recharge probably due to a low pressure transfer from conduit to LPV and eventually the influence of
local fast infiltration, but this water level always stays lower than
the water level in conduits. Moreover, it quickly reaches a constant
value, which means that the water level evolution in LPV cannot
explain the discharge evolution at the spring. Discharge at the
spring during the whole recession is thus primarily influenced by
the water level change in conduits. As a result, flood hydrographs
only show the flood flow component influenced by conduit flow;
it is an ideal example of the CFR condition. Bailly-Comte et al.
(2009) showed that the low water table condition is characterized
by hPZ2 < 3500 cmasl prior to the flood.
During medium to high water table conditions, floods recorded
at Vène Spring are longer than during low water conditions. Spring
hydrographs (Fig. 2B) initially show a similar evolution consisting
of a linear decrease through time, while the water level in Well PZ2
increases slightly. This increase cannot explain the flood recession;
it shows that conduits simultaneously feed the spring and recharge
the LPV. An inflection point appears in the hydrograph as soon as
the water level in conduits equals the water level in matrix (Well
PZ2). This inflection point is thus interpreted as a flow regime
modification within the karst drainage network, which abruptly
changes from the CFR condition to the MRFR condition.
Under very high water table conditions, the LPV is already
drained by an upper karst drainage system (which occurs when
hPZ2 > 4700 cmasl, see (Bailly-Comte, 2008) for details). The spring
hydrograph in Fig. 2C perfectly represents the MRFR condition
since only the hydrodynamic properties of the matrix control the
flow at the spring, assuming that the discharge capacity of the
karst conduit system is not exceeded (which occur on this site
when QS > 5.5 m3/s (Bailly-Comte et al., 2009)).
As a result, these three types of recession curves show that
water level comparisons between Puits de l’aven and Well PZ2 allow an understanding of the conditions in the conduits or matrix
that control the flow at the spring. These results also show that
water level measurements in a well representing the LPV (PZ2)
prior to the flood allow forecasting of the spring behavior (see
the initial water level in PZ2 shown in Fig. 2). The CFR and MRFR
mathematical models (Eqs. (1) and (3)) should only be used to fit
recession curves of spring discharge when delayed recharge is negligible. This assumption is reasonable when allogenic recharge
from an ephemeral stream constitutes the primary recharge, producing the quick response at the spring. In the three types of recession curves exemplified by Fig. 2 allogenic recharge ceased quickly
after the river flooded, and most surface flow was captured by
numerous sinkholes in less than 1 h. Consequently, QA was negligible when the recession started at the spring (Fig. 2). The recession
can thus be interpreted as the spring response to an impulse recharge in conduits. Moreover, as a first approximation, water exchange between conduit and matrix can be neglected during the
conduit flow period since the contrast of permeability between
conduits and matrix is high and water movements through
Author's personal copy
60
V. Bailly-Comte et al. / Journal of Hydrology 386 (2010) 55–66
A
A
B
C
B
C
Fig. 3. Evolution of discharge and water level at the Santa Fe River sink/rise system
according to recharge estimated by Ritorto et al. (2009).
Evolution of groundwater flows between LPV and conduits at the
Santa Fe sink/rise system
Fig. 2. Recession curve analysis at Vène Spring according to water level measurements in the aquifer during low (A), medium to high (B) and very high (C) water
table conditions prior to the flood. The horizontal dashed line shows the mean level
of the upper karst drainage network which controls the karst flows from Puits de
l’aven to the Vène Spring during floods. The vertical dashed line shows the transition
between conduit flow and base flow periods. () Water level in Well PZ2 prior to the
flood is shown on the right axes.
conduits towards the spring is enhanced by a strong hydraulic gradient through the conduits. These two factors explain why the CFR
model (Eq. (3)) agrees with measurements.
Influence of the type of recharge on the flood propagation
A direct estimate of flow between conduits and LPV using Eq.
(4) only applies when autogenic recharge is negligible so that recharge dynamics of the Santa Fe sink/rise system are known.
Fig. 3C shows how the discharge at the River Sink (Fig. 1B) and
the River Rise (Fig. 1B) varies during three large floods, two in
2005 and one in 2006. At this site, recharge to the aquifer was estimated at a daily time step by Ritorto et al. (2009) who used a massbalance method similar to Dripps and Bradbury (2007).
In 2005, the River Sink discharge (QA) is always less than the
River Rise discharge (QS), reflecting additional flow at the spring
and continuous drainage of autogenic water even during flood
periods. Initial water level in Well 4 was relatively high
(1050 cmasl), but this level remains much lower than water level
at the spring during the flood peak (Fig. 3B). Thus, autogenic recharge through the LPV cannot explain a higher discharge in QS
(Fig. 3C). The positives values of DQ also show great variations dur-
Author's personal copy
61
Case of simple allogenic recharge
Fig. 4 shows a flood that occurred at the end of summer 2008 for
which the rainfall recorded from 08/16/2008 to 09/02/2008 in the
confined catchment area (236 mm),1 upstream of the karst aquifer,
is much higher than the rainfall recorded on the karst aquifer
(156 mm).2 No recharge estimates are available, but because flooding occurred in the summer, medium to high evapotranspiration
(ET) would be expected, reducing the relative influence of the autogenic recharge as compared to the allogenic recharge.
Cumulative flows are the same in River Sink and River Rise
(120 106 m3 from 08/15/08 to 12/05/08), which could be misinterpreted as low matrix/conduit exchange during the flood transfer. Indeed, hydrographs are greatly modified from River Sink to
River Rise (Fig. 4), which can be explained using water level time
series at River Rise, and Well 4 and Well 4A (Fig. 1B). Synchronous
measurements of water level are used to describe vertical groundwater flows by computing vertical hydraulic gradients between
Well 4 and Well 4A (Fig. 4, see the gray line), while evolution of
hydraulic gradient between Well 4 and River Rise (Fig. 4, see the
black line) characterizes flow from LPV to conduits. Positive values
of hydraulic gradients indicate downward flow from Well 4A to
Well 4 and from matrix to conduit respectively.
The gradients between the conduit (measured at the River Rise)
and Well 4 and between Well 4 and Well 4A become negative
simultaneously with the DQ becoming negative at the start of
the flood. This timing indicates that allogenic water flows to the
LPV in all directions around the conduit and that recharge from
the surface through the LPV is negligible at that time. Subsequently
in the recession curve, DQ and the hydraulic gradient between
Well 4 and River Rise become positive at the same moment. Thus,
hydraulic gradients within the LPV both between Well 4 and Well
4A and between the LPV and conduits (W4 and River Rise) reflects
DQ, suggesting that DQ can be considered as a measure of ex1
Average value of cumulated values of stations 234 (Louis Hill Tower), 235 (New
River Tower) and 263 (Union Tower).
2
Average of cumulated values of stations 240 (O’Leno State Park) and 02321975
(Santa-Fe River at US441 near High Springs)
1%
0%
between W4 and River Rise
-1%
between W4a and W4
-2%
20
150
40
River Sink discharge (QA)
River Rise discharge (QS)
100
60
Rainfall (mm/d)
0
Daily discharge (m 3/s)
ing the flood peak (Fig. 3B, see t1). These fluctuations and the positives values of DQ reflect fast infiltration through the thin and high
permeability unsaturated zone, sinkholes and vertical shafts (i.e.
fast autogenic recharge) which amplifies the flood peak transfer
from River Sink to River Rise.
In 2006, a high flow event occurred with a lower initial water
level in Well 4 prior to the flood (995 cmasl). The estimated autogenic recharge is higher than for the summer rainfall event
(Fig. 3A), but fast autogenic infiltration appears to be relatively
low since DQ exhibits negative peaks during the transfer of the
flood peak (Fig. 3B, see t2). During the flood recession, water level
in Well 4 becomes higher than at the River Rise (Fig. 3B, see t3),
indicating that conduits drain the LPV following its recharge by diffuse infiltration. Consequently, t3 characterizes a flood recession
influenced by delayed infiltration which could result from discharge from wetlands and perched water table lakes. This delayed
infiltration explains the positive and smoothed values of DQ
(Fig. 3B, compare t3 to t1). From June 2006 to December 2006
(Fig. 3B, see t4), no recharge occurs (no flows in River Sink and no
or little recharge from rainfall on Fig. 3A). Delayed recharge
through a few meter thick vadose zone (Ritorto, 2007) may thus
be neglected. LPV and conduits reach pressure equilibrium during
the t4 period during which time a single exponential decreasing
limb perfectly describes the base flow recession curve, as proposed
by the MRFR model (Eq. (1)) without recharge. However, quick
flows between matrix and conduit cannot be described using DQ
since autogenic recharge is found to influence the flood transfer.
Hydraulic gradient
V. Bailly-Comte et al. / Journal of Hydrology 386 (2010) 55–66
50
0
-50
ΔQ
-100
15-Aug
25-Aug
4-Sep
14-Sep
24-Sep
Fig. 4. Flood event at the River Rise mostly influenced by allogenic recharge.
Rainfall data are provided by the O’Leno State Park station.
change between LPV and conduits. A positive value for DQ can
be considered the onset of baseflow to the spring, even though
there is no inflection point on the spring hydrograph at this time.
This flood event shows how allogenic water is initially stored in
the LPV (DQ < 0) and subsequently is released and drained by conduits (DQ > 0). In other words, DQ shows how exchange of water
between matrix and conduit modifies the quick flow transfer,
although cumulative flow at the sink and rise over the time of
the flood does not reflect this exchange of water. Based on the
DQ evolution shown in Fig. 4 between 08/25/08 and 09/04/08,
there was clear exchange during this flood of around 30% of allogenic recharge. This amount of water was stored in the porous matrix during the flood before being slowly drained by conduits to the
spring during the recession. Although the flow inversion between
matrix and conduits is clearly shown by the discharge volumes,
it does not generate a clear change in the spring hydrograph.
Comparisons of Aumelas-Thau and Santa Fe River sink/rise systems
Fig. 2A and B shows that the hydrograph recession of Vène Spring
perfectly fits Eq. (3) when water level is higher in the cave (conduit)
than in the well (LPV), which characterizes the CFR condition. Flood
flows at Vène Spring thus reflects discharge from a conduit network
previously recharged by a pulse of allogenic recharge. Linear decreases of flood flows are usually not observed on hydrographs of
karst springs, even in karst systems where matrix porosity is very
low, because of the influence of fast and delayed infiltration
through the epikarst and the vadose zone. Indeed, the dynamics
of recharge has such a strong influence on the flood flow transfer
that it often becomes impossible to physically interpret the flood
flow period, i.e. to assign a flow behavior to each flow period. At
the Vène Spring, short and focused allogenic recharge easily allows
separating the CFR component of the spring hydrograph.
Using numerical methods based on two-dimensional analytical
solutions describing the hydraulic response of a theoretical karst
aquifer with simple geometries, Kovacs et al. (2005) have shown
Author's personal copy
62
V. Bailly-Comte et al. / Journal of Hydrology 386 (2010) 55–66
that both flood and base flow recessions should be characterized
with exponential decreases, but with two different recession coefficients. If the recession coefficient is low, however, an exponential
decrease is close to a linear decrease, but the hydrograph is consequently nearly flat, which is not consistent with the field observations in Fig. 2B and C. Another explanation of the linear decrease of
flood flows observed at the Vène Spring (Fig. 2A and B) could be
found by considering both the contrast of permeability between
conduits and matrix, and the hydraulic gradient in conduits: In
the Aumelas-Thau system, this contrast is high and hydraulic gradients between sinkholes in the riverbed and Vène Spring can exceed
1%, suggesting that water exchange between conduits and matrix
is low compared to flow in the conduits. Limited exchange means
that the conduit/matrix boundary can be approximated by a noflow boundary, as proposed by the CFR mathematical model (Eq.
(3)). In case of higher matrix permeability (fractures or intergranular porosity) and lower hydraulic gradients in conduits, exchange
of water between the conduit and matrix could explain an exponential evolution of spring discharge as proposed by theoretical
models with simple geometry (Kovacs et al., 2005).
When water level in Well PZ2 is high prior to the flood, it remains higher than the water level in conduits during the flood
(Fig. 2C). Thus, karst conduits only drain the matrix and never restrain groundwater flow as long as the discharge capacity of karst
conduits is not exceeded. This process explains why spring hydrographs perfectly follow a single exponential decrease related to the
relatively slow depletion of the LPV volume (Eq. (1)), which characterizes the MRFR condition (Fig. 2B after 5/7/07, and Fig. 2C).
Influence of two aquifer volumes of different hydrodynamic
properties is also identified at the Santa Fe sink/rise system, highlighting the concept of double porosity of karst aquifers. Both continuous recharge at the sink (QA) and water exchange between a
highly permeable matrix and conduits drive the flood flow transfer,
which prevents the use of Eq. (3) to describe the hydrodynamic
behavior of the River Rise. Exchange of water between matrix and
conduit can be compared to stream/aquifer interactions (or bank
storage) which control the surface flow routing in open channels
(Cooper and Rorabaugh, 1963; Pinder and Sauer, 1971; Hunt,
1990; Moench and Barlow, 2000; Chen et al., 2006). There are however some non negligible differences since surface flow routing in
open channels and water flow routing in a conduit system under
pressure are governed by different processes, and thus represented
by different equations. Moreover, pressure in conduit flow can enhance the loss of water from conduits to matrix porosity.
Maillet’s law (1905) can also explain an exponential decrease of
discharge from the spring as due to the emptying of large karst
voids poorly connected to the main karst conduit (karst annexes
(Mangin, 1975)). Such karst voids or cavities were not encountered
when drilling Well PZ2 (Aumelas-Thau) or Well 4 (Santa Fe River
sink/rise), suggesting that groundwater is actually released from
matrix porosity (intergranular porosity and/or from dissolved fractures and joints) during base flow according to Eq. (1) for both the
eogenetic and telogenetic studied karst systems.
Conceptual model and discussion
These observations can be used to develop a conceptual model
of groundwater flow through conduits embedded in matrices with
variable amounts of porosity and hydraulic conductivity in response to a sudden increase of hydraulic head in the conduit system due to focused recharge. The evolution of spring
hydrographs is exemplified by the Aumelas-Thau system for systems with high hydraulic gradients within conduits and contrasting permeability between conduits and matrix (Fig. 2). In this
case conduits and LPV behave as two parallel reservoirs with two
different hydraulic heads during quick flow recession. This condition continues for as long as the hydraulic head in conduits is higher than the hydraulic head in the LPV (from t0 to ti on Fig. 5). Once
pressure equilibrium between the conduit and LPV is reached, conduit flow regime (CFR) ends and Matrix Restrained Flow Regime
(MRFR) begins. This modification to the flow regime occurs when
the recession curve changes from a linear decrease to an exponential decrease at the Vène Spring (ti on Fig. 5 – Site A) in the case of
medium to high initial water level in the LPV. After matrix restrain
flow regime (MRFR) begins, karst conduits drain groundwater
stored in the LPV and the spring behavior only depends on the matrix hydrodynamic properties (e.g. after ti on Fig. 5 – Site A). This
condition induces an exponential decrease of the spring discharge,
as described by Eq. (1).
This conceptual model can also be used to understand hydrodynamics at the Santa Fe sink/rise system (Fig. 5 – Site B). During
flood flows, water exchange between conduits and the LPV imply
that conduits and matrix are two dependent subsystems. The links
between these subsystems means that the CFR condition will be
more complex than in the previous case of low matrix porosity,
and that groundwater flow will slowly change from the CFR to
the MRFR condition without a visible inflection point on the spring
hydrograph (Fig. 5 – Site B). Moreover, a low contrast of permeability between conduits and matrix and a high effective porosity are
not consistent with the low initial water level in the LPV shown
in Fig. 5. Thus, no schematic spring hydrograph is given for Site B
in the case of low antecedent condition in the matrix.
Both systems studied show MRFR conditions which can be characterized by a single exponential decrease on the spring hydrograph, but values of the recession coefficient are much lower in
the Santa Fe sink/rise system (a < 0.01 d1, Fig. 5) than in the Aumelas-Thau system (a > 0.1, Fig. 2B and C). This difference can be
T
ratio in the Santa Fe sink/rise
interpreted by considering a lower SL
system, i.e. a lower hydraulic diffusivity TS since both karst aquifer
lengths (L) are of the same order of magnitude (Fig. 1). A lower
hydraulic diffusivity in the Floridan aquifer is consistent with the
strong inertial behavior of eogenetic karst aquifers (Florea and
Vacher, 2006). This characteristic is mainly due to the high porosity
of the limestone (thus a high storativity and a low hydraulic diffusivityTS), which explains why this aquifer is one of the most productive aquifers in the world. In contrast, the matrix porosity of
Jurassic limestones of the Aumelas-Thau system is low (thus a
low storativity and a high hydraulic diffusivity TS because of the
low fracture porosity).
Base flow at River Rise, as defined as the positive value of DQ, is
relatively low and constant as long as the discharge simultaneously
decreases at River Sink and River Rise. Consequently, discharge variations at River Rise (QS) are strongly influenced by the recharge in
River Sink (QA). Similar to this behavior in karst aquifers, low flows
in surface streams are also controlled by base flow due to bank
storage effects, which are also characterized by an exponential decrease (Cooper and Rorabaugh, 1963). As a result, identifying an
exponential decrease on karst spring hydrographs does not necessarily mean that groundwater is released from the LPV to conduits.
In a general way, dynamics of recharge has to be well identified before analyzing recession curves.
Consequences for karst spring modeling
Although the examples shown are systems dominated by allogenic recharge, we suggest that the conceptual model has applicability to many karst systems. Most karst aquifers are connected to
surface streams, which can be perennial, intermittent or ephemeral. These streams are drained by sinkholes (Zhou, 2007) or by focused recharge at particular points (sinkhole, swallow hole, ponor)
Author's personal copy
63
V. Bailly-Comte et al. / Journal of Hydrology 386 (2010) 55–66
Antecedent condition in the matrix
Low
ti
CFR
Transition
MRFR
QMC<0
QMC 0
QMC 0
C
QS
=Spring discharge
Medium to
high
C
M
C
M
M
QS
M
C
M
C
M
Spring hydrograph
Site A
Site B
QMC low
QMC high
QS=0
C
hM
C
M
QS 0
hM<Spring altitude
QS
Very
high
Increasing time since the
beginning of the flood
t0
Linear decrease
QS
C
M
ti
ti
C
Exponential
decrease
Exponential
decrease
M
hM
QS
QS
QS
ti t 0
Increasing initial GW
level in matrix (hM)
ti t 0
C = Conduit
M = Matrix
QMC = Flow from matrix to conduit
Fig. 5. Schematic representation of high (C = conduit) and low (M = matrix) permeability volumes during flood and its consequences on flood hydrographs of the studied
karst springs in response to a sudden hydraulic head elevation in conduits. Low, medium to high and very high initial water levels are depicted by the initial water level in
matrix (hM) when conduits are filled with water (t0, see the left part of the figure); ti is the transition time from CFR to MRFR. Aumelas Cause represents Site A hydrographs
and Santa Fe River Rise represents Site B hydrographs.
of endorheic area as a consequence of flooding of karst depressions
like dolina and polje (Salvayre, 1964; Mijatovic, 1988; Currens
et al., 1993; Bruxelles and Caubel, 1996; Lopez-Chicano et al.,
2002; Bradley and Hileman, 2006). Recharge through the epikarst
(Mangin, 1975) also occurs as temporary storage and fast infiltration through vertical shafts or paleo-conduits. This fast infiltration
drains towards the conduit system in the vadose zone thereby
increasing hydraulic head in conduits as compared to the LPV. Fast
infiltration is also shown in cases of highly permeable matrix in
Fig. 3B where DQ exhibits strong positive variations during the
2005 flood transfer, as discussed previously. Furthermore, empirical (Drogue, 1969, 1980; Tripet, 1972; Jeannin, 1996) and theoretical (Jeannin, 1996 and references herein) approaches showed that
flow reversals from conduits to LPV occur for various types of karst
aquifers during flood, whatever the recharge processes.
We show that groundwater flows from conduits to LPV even following the main recharge event. This flow decreases as the contrast of permeability between conduits and LPV increases. In
cases of low matrix permeability, loss of water from conduit to matrix is negligible as compared to conduit flows. In these settings,
spring discharge represents conduit flow that can be described
by the emptying of a reservoir with infinite hydraulic conductivity
(Fig. 2A before 5/7/07 and Fig. 2B). In case of high matrix permeability, loss of water from conduit to matrix is part of the aquifer
recharge. It thus strongly modifies the flood wave propagation to
the spring by storing water during flood that is slowly released
during base flow.
Since most aquifers have fast recharge, results of our study
should represent many types of karst aquifers. Similar behaviors
should occur in karst aquifers with a combination of allogenic
and autogenic recharge, although the processes should be less
prominent. Allogenic recharge in the two systems used as case
studies here provide sufficient constraints to assess the influence
of each aquifer volume, thereby allowing a physical explanation
of the two different components on a karst spring hydrograph
through the CFR and MRFR models.
Considering conduits and matrix as two linked aquifer volumes
that are interdependent in terms of hydraulic head may have great
consequences on karst spring conceptual modeling. Indeed, simple
conceptual models commonly state that karst spring behavior can
be simulated using a model with two reservoirs connected in series
(see for instance (Kiraly, 2003; Geyer et al., 2008)). Recharge is split
into low permeability and a high permeability reservoirs; the low
permeability reservoir is slowly drained by the high permeability
reservoir connected to the spring (Fig. 6A).
The depletion of the reservoir representing the low permeability volume in Fig. 6A also gives the antecedent recession, which is
added to the fast infiltration during quick flow and is used to show
the volume of water that is stored and released within the aquifer
(see for instance (Atkinson, 1977)). Following this assumption,
Fig. 6A shows that: (i) the LPV and conduits always contribute to
spring flows, (ii) quick flows at the spring are considered as the
sum of conduit flows and matrix flows and (iii) pressure equilibrium between the LPV and conduits are not considered. This hydrodynamic behavior is not consistent with our observations at both
the Aumelas-Thau and Santa Fe systems, two systems with distinctly different aquifer properties. Our observations have important consequences for interpretations of recession curves and
modeling of karst spring flows, particularly for interpretations of
relative importance of base flow and quick flow from karst spring
Author's personal copy
64
V. Bailly-Comte et al. / Journal of Hydrology 386 (2010) 55–66
A
B
Recharge
Delayed
infiltration
Recharge
Delayed
infiltration
Fast infiltration
Low
permeability
volume
Low
permeability
volume
Fast infiltration
High
permeability
volume
Q
High
permeability
volume
Q
Q
Q
Flow from low to
high permeability
volumes
t
t
CFR
MRFR
MRFR
CFR
Fig. 6. Conceptual models of karst springs hydrodynamics with two reservoirs connected in series (1) or in parallel (2) showing relative influence of low and high
permeability volumes on a karst spring hydrograph. CFR: Conduit Flow Regime – MRFR: Matrix Restrained Flow Regime. The solid black line represents the spring hydrograph
and the dashed line represents flow between the conduit and matrix (DQ).
hydrographs (e.g. (Bonacci, 1993; Padilla et al., 1994)). Our results
indicate that karst spring hydrographs should be interpreted
according to both aquifer subsystems which control the flow regime at the spring, and to water exchange caused by hydraulic
head difference between the LPV and conduits (Fig. 6B). These
observations reflect karst systems as ensembles of conduits of relatively low storage capacity embedded in a low-permeability matrix where groundwater is stored and slowly drained by conduits,
but that the low-permeability matrix can buffer flood pulses to
varying degrees depending on its physical characteristics, primarily its porosity and permeability.
Conclusion
We have shown that for karst aquifers with low matrix porosity,
quick flows at the spring correspond to a flow regime influenced by
the hydrodynamic properties of the conduit network. In contrast,
for karst aquifer with high matrix porosity, quick flow depends
ultimately on the magnitude of water exchange between matrix
and conduits. When conduits are filled with water without delayed
infiltration and conduit/matrix exchange is low, a linear decrease is
observed in the spring discharge, which can be interpreted as the
emptying of the conduit network. These hydrograph characteristics are specific to karst aquifer with both low matrix permeability
and focused recharge, exemplified by a sinking stream, inducing an
instantaneous recharge of the conduit system. In this case, a unique inflection point on the decreasing limb shows a change of flow
regime from CFR to MRFR. In most other cases, no simple mathematical fit can be used to analyze quick flows at the spring since
infiltration dynamics and/or matrix/conduits exchanges strongly
influence conduits flow.
We also demonstrate that a single exponential decrease characterizes the base flow when recharge ceases or become negligible,
which can be explained by the drainage of the LPV by conduits.
The recession coefficient can thus be used to estimate some hydrodynamic properties of the matrix component itself. These observations allow consideration of a different conceptual model for karst
spring hydrograph analysis where discharge evolution at the
spring results from pressure equilibrium between two aquifer subsystems of different hydrodynamic properties. This model requires
recharge and water level time series in different subsystems of a
karst aquifer to interpret the spring hydrograph. Each of these variables is easily obtained through monitoring pressure head at wells,
springs and conduits when a sinking stream constitutes the main
recharge of the karst system.
As a consequence of the strong spatial heterogeneity of karst
aquifers, water level time series in wells are often considered as
useless or hard to interpret at the karst system scale, particularly
when compared to karst springs hydrographs. However, our observations show that ‘‘well-chosen” water level time series allow
understanding hydrodynamics of low (matrix with intergranular,
joints and/or fracture porosity) and high (conduits) permeability
volumes, which together with simultaneous discharge time series
at the spring are essential to understand the whole karst aquifer
behavior.
As a result, considering porosity of the LPV and hydraulic gradients within conduit systems should improve the environmental
impact assessment of pollutants in surface waters drained by karst
sinkholes. Indeed, when matrix/conduit exchange is low, drainage
of pollutants towards the spring should be fast and limited to the
quick flow response at the spring, but also with little dilution. Inversely, high matrix/conduit exchanges should induce a delayed
but diluted discharge of the pollutants, with potential sorption/
desorption phenomena within the porous carbonate rocks. Differences in contaminant behavior in these two systems will be critical
for remediation strategies. This idea could be reinforced by the use
of natural tracers of the infiltration. For example high resolution
Author's personal copy
V. Bailly-Comte et al. / Journal of Hydrology 386 (2010) 55–66
time series of temperature (Martin and Dean, 1999; Screaton et al.,
2004) or specific electrical conductivity and regular sampling of
majors ions and total organic carbon (e.g. Batiot et al., 2003).
Chemical tracers require knowledge of the variations of the selected tracer in the surface water and the unsaturated zone, as well
as the characterization of pre-event waters in conduits.
Finally, we show how two karst systems that have distinct characteristics and are observed with similar field methods can be described by the same conceptual model. The unifying aspect of these
karst aquifers is their double porosity behavior, regardless of the
origin, extent and dynamics of the karstification process.
Acknowledgments
The authors would like to thank Schlumberger Water Services
for this productive collaboration and for the use of numerous data
loggers, the ‘‘Conseil Général de l’Hérault” for its financial support
and the ‘‘Syndicat intercommunal d’adduction d’eau du Bas
Languedoc’’ for the use of wells, M.-G. Tournoud of the HydroSciences Montpellier Laboratory for the use of discharge time series at the Vène Spring, the Florida Department of Environmental
Protection and the staff of O’Leno State Park for their cooperation.
Santa Fe Sink/rise research was funded by the National Science
Foundation grants EAR-003360 and EAR-0510054 and by the Florida Department of Environmental Protection Grants S00060,
S0141, and S0181. Valuable suggestions made by three anonymous
reviewers are greatly appreciated.
References
Atkinson, T.C., 1977. Diffuse flow and conduit flow in limestone terrain in the
Mendip Hills, Somerset (Great Britain). Journal of Hydrology 35, 93–110.
Baedke, S.J., Krothe, N.C., 2001. Derivation of effective hydraulic parameters of a
karst aquifer from discharge hydrograph analysis. Water Resources Research 37
(1), 13–19.
Bailly-Comte, V., 2008. Interactions hydrodynamiques entre les eaux de surface et
les eaux souterraines en milieu karstique –Approche descriptive, analyse
fonctionnelle et modélisation hydrologique appliquées au bassin versant
expérimental du Coulazou, Causse d’Aumelas, France,PhD Thesis, Université
Montpellier 2, 223 pp. <http://tel.archives-ouvertes.fr/tel-00319965/fr/>.
Bailly-Comte, V., Jourde, H., Roesch, A., Pistre, S., 2008a. Mediterranean flash flood
transfer through karstic area. Environmental Geology 54 (3), 605–614.
Bailly-Comte, V., Jourde, H., Roesch, A., Pistre, S., Batiot-Guilhe, C., 2008b. Time
series analyses for karst/river interactions assessment: case of the Coulazou
river (southern France). Journal of Hydrology 349 (1–), 98–114.
Bailly-Comte, V., Jourde, H., Pistre, S., 2009. Conceptualization and classification of
groundwater–surface water hydrodynamic interactions in karst watersheds:
case of the karst watershed of the Coulazou river (Southern France). Journal of
Hydrology 376 (3-4), 456–462.
Bakalowicz, M., 2005. Karst groundwater: a challenge for new resources.
Hydrogeology Journal 13 (1), 148–160.
Batiot, C., Emblanch, C., Blavoux, B., 2003. Total Organic Carbon (TOC) and
magnesium (Mg2+): two complementary tracers of residence time in karstic
systems. Carbone organique total (COT) et magnésium (Mg2+): deux traceurs
complémentaires du temps de séjour dans l’aquifère karstique. Comptes
Rendus Geosciences 335 (2), 205–214.
Bonacci, O., 1993. Karst springs hydrographs as indicators of karst aquifer.
Hydrological Sciences, Journal des Sciences Hydrologiques 38 (1), 51–62.
Bonnet, A., Paloc, H., 1969. Les eaux des calcaires jurassiques du bassin de
Montbazin-Gigean et de ses bordures (Pli de Montpellier et massif de la
Gardiole, Hérault). Bull. BRGM 2 (3), 1–12.
Boussinesq, J., 1904. Recherches théoriques sur l’écoulement des nappes d’eau
infiltrées dans le sol et sur le débit des sources. Journal de mathématiques pures
et appliquées 10, 5–78.
Bradley, M.W., Hileman, G.E., 2006. Sinkhole Flooding in Murfreesboro, Rutherford
County, Tennessee, 2001-02, US Department of the Interior, US Geological
Survey, Reston, Virginia.
BRGM-CNRS, 1992. Test de pompage en aquifère karstique dans le gouffre de la
Peyrère (09), Rapport BRGM-CNRS 35 924 MPY 4S 92.
Bruxelles, L., Caubel, A., 1996. Lacs temporaires et circulations de surface sur le
causse de l’hospitalet du Larzac (12) en 1996: fonctionnement et implications
geomorphologiques – temporary lakes and surface circulation on the Hospitalet
du Larzac karst plateau in 1996: functional and geomorphological implications.
Bulletin de la Société languedocienne de géographie 30 (3–4), 269–288.
Chen, X., Chen, D.Y., Chen, X.-H., 2006. Simulation of baseflow accounting for the
effect of bank storage and its implication in baseflow separation. Journal of
Hydrology 327 (3–4), 539–549.
65
Cooper, H.H., Rorabaugh, M.I., 1963. Ground-water movements and bank storage
due to flood stages in surface streams. Ground-Water Hydraulics 1536-J, 343–
366.
Currens, J.C., Douglas, C., Graham, R., 1993. Flooding of the sinking creek karst area
in Jessamine and Woodford counties, Kentucky, Kentucky geological survey,
university of kentucky, Lexington.
Dewandel, B., Lachassagne, P., Bakalowicz, M., Weng, P., Al-Malki, A., 2003.
Evaluation of aquifer thickness by analysing recession hydrographs.
Application to the Oman ophiolite hard-rock aquifer. Journal of Hydrology
274 (1–4), 248–269.
Douchet, M., 2007. Puits de l’aven. Le Fil, Bulletin de la liaison de la commission
nationale Plongée Souterraine 17, 15–16. <http://souterraine.ffessm.fr/>.
Dripps, W., Bradbury, K., 2007. A simple daily soil–water balance model for
estimating the spatial and temporal distribution of groundwater recharge in
temperate humid areas. Hydrogeology Journal 15 (3), 433–444.
Drogue, C., 1969. Contribution à l’étude quantitative des systèmes hydrologiques
karstiques d’après l’exemple de quelques karst périméditerranéens, Thèse de
Docteur en Sciences Naturelles, Montpellier, 482 pp.
Drogue, C., 1972. Analyse statistique des hydrogrammes de decrues des sources
karstiques statistical analysis of hydrographs of karstic springs. Journal of
Hydrology 15 (1), 49–68.
Drogue, C., 1980. Essai d’identification d’un type de structure de magasins
carbonatés fissurés: application à l’interprétation de certains aspects du
fonctionnement hydrogéologique. Mémoire hors série de la Société.
géologique de France 11, 101–108.
Eisenlohr, L., Király, L., Bouzelboudjen, M., Rossier, Y., 1997. Numerical simulation as
a tool for checking the interpretation of karst spring hydrographs. Journal of
Hydrology 193 (1–4), 306–315.
Florea, L.J., Vacher, H.L., 2006. Springflow hydrographs: eogenetic vs. telogenetic
karst. Ground Water 44 (3), 352–361.
Ford, D., Williams, P., 2007. Karst Hydrogeology and Geomorphology. John Wiley &
Sons, Chichester, West Sussex, England. 562 pp.
Forkasiewicz, J., Paloc, H., 1967. Le régime de tarissement de la Foux-de-la-Vis.
Etude préliminaire. Chronique d’Hydrogéologie. BRGM 3 (10), 61–73.
Geyer, T., Birk, S., Liedl, R., Sauter, M., 2008. Quantification of temporal distribution
of recharge in karst systems from spring hydrographs. Journal of Hydrology 348
(3–4), 452–463.
Hunt, B., 1990. An approximation for the bank storage effect. Water Resources
Research 26 (11), 2769–2775.
Jeannin, P.-Y., 1996. Structure et comportement hydraulique des aquifères
karstiques. PhD Thesis, Université de Neuchâtel, 237 pp.
Jourde, H., Roesch, A., Guinot, V., Bailly-Comte, V., 2007. Dynamics and contribution
of karst groundwater to surface flow during Mediterranean flood.
Environmental Geology 51 (5), 725–730.
Kiraly, L., 2003. Karstification and groundwater flow. Speleogenesis and Evolution of
Karst Aquifers 1 (3), 26.
Kiraly, L., Morel, G., 1976. Remarques sur l’hydrogramme des sources karstiques
simulé par modèles mathématiques. Bulletin d’Hydrogéologie de Neuchâtel 1,
37–60.
Kovacs, A., Perrochet, P., 2008. A quantitative approach to spring hydrograph
decomposition. Journal of Hydrology 352 (1–2), 16–29.
Kovacs, A., Perrochet, P., Kiraly, L., Jeannin, P.-Y., 2005. A quantitative method for the
characterisation of karst aquifers based on spring hydrograph analysis. Journal
of Hydrology 303 (1–4), 152–164.
Lopez-Chicano, M., Calvache, M.L., Martin-Rosales, W., Gisbert, J., 2002.
Conditioning factors in flooding of karstic poljes – the case of the Zafarraya
polje (South Spain). CATENA 49 (4), 331–352.
Maillet, E., 1905. Essais d’hydraulique souterraine et fluviale. Herman et Cie, Paris
(218 pp).
Mangin, A., 1975. Contribution à l’étude hydrodynamique des aquifères karstiques.
3ème partie. Annales de Spéléologie 30 (1), 21–124.
Mangin, A., 1994. Karst hydrogeology. In: Gilbert, J. (Ed.), Ground-Water Ecology.
Elsevier, New York, pp. 43–67.
Martin, J.B., Dean, R.W., 1999. Temperature as a natural tracer of short residence
times for groundwater in kart aquifer. In: Institutes, K.W. (Ed.), Karst Modeling.
Charlottesville, Virginia, pp. 236–242.
Martin, J.B., Dean, R.W., 2001. Exchange of water between conduits and matrix in
the Floridan aquifer. Chemical Geology 179 (1–4), 145–165.
Martin, J.B., Screaton, E., 2001. Exchange of matrix and conduit water with examples
from the floridan aquifer. In: US Geological Survey Karst Interest Group
Proceedings, Water-Resources Investigations Report, vol. 1(4011), pp. 38–44.
Martin, J.M., Screaton, E.J., Martin, J.B., 2006. Monitoring well responses to karst
conduit head fluctuations: implications for fluid exchange and matrix
transmissivity in the Floridan aquifer. Geological Society of America, Special
Paper 404, pp. 209–217.
Mijatovic, B., 1988. Catastrophic flood in the polje of Cetinje in February 1986, a
typical example of the environmental impact of Karst. Environmental Geology
12 (2), 117–121.
Moench, A.F., Barlow, P.M., 2000. Aquifer response to stream-stage and recharge
variations. I. Analytical step-response functions. Journal of Hydrology 230 (3–4),
192–210.
Moore, P.J., Martin, J.B., Screaton, E.J., 2009. Geochemical and statistical evidence of
recharge, mixing, and controls on spring discharge in an eogenetic karst aquifer.
Journal of Hydrology 376 (3–4), 443–455.
Padilla, A., Pulido-Bosch, A., Mangin, A., 1994. Relative importance of baseflow and
quickflow from hydrographs of karst spring. Ground Water 32 (2), 267–277.
Author's personal copy
66
V. Bailly-Comte et al. / Journal of Hydrology 386 (2010) 55–66
Palmer, A.N., 2003. Dynamics of cave development by allogenic water.
Speleogenesis and Evolution of Karst Aquifers 1 (1), 11.
Perrin, J., Jeannin, P.-Y., Zwahlen, F., 2003. Epikarst storage in a karst aquifer: a
conceptual model based on isotopic data, Milandre test site, Switzerland.
Journal of Hydrology 279 (1–4), 106–124.
Pinder, G.F., Sauer, S.P., 1971. Numerical simulation of flood wave modification due
to bank storage effects. Water Resources Research 7 (1), 63–69.
Renault, P., 1959. Réseau de fentes et réseau de conduits en région karstique.
Compte rendu sommaire des séances de la Société Géologique de France, pp.
16–17.
Ritorto, M., 2007. Impacts of diffuse recharge on transmissivity and water budget
calculations in the unconfined karst aquifer of the Santa-Fe River basin. M. Sc.
Thesis, University of Florida, Gainesville.
Ritorto, M., Screaton, E., Martin, J., Moore, P., 2009. Relative importance and
chemical effects of diffuse and focused recharge in an eogenetic karst aquifer:
an example from the unconfined upper Floridan aquifer, USA. Hydrogeology
Journal 17 (7), 1687–1698.
Salvayre, H., 1964. Les lacs temporaires du Larzac et de son avant causse. Spelunca,
Mémoires 4, 79–103.
Schoeller, 1965. Hydrodynamique dans le karst, Hydrologie des roches fissurées. coedition IAHS/UNESCO, Dubrovnik.
Screaton, E., Martin, J.B., Ginn, B., Smith, L., 2004. Conduit properties and
karstification in the unconfined Floridan aquifer. Ground Water 42 (3), 338–
346.
Tripet, J.-P., 1972. Etude hydrogéologique du bassin de la source de l’Areuse (Jura
neuchâtelois). Thèse de doctorat Thesis, Faculté des Sciences de l’Université de
Neuchâtel, Neuchâtel, 183 pp.
Vacher, H.L., Mylroie, J.E., 2002. Eogenetic karst from the perspective of an
equivalent porous medium. Carbonates and Evaporites 17 (2), 182–196.
White, W.B., 2002. Karst hydrology: recent developments and open questions.
Engineering Geology 65 (2–3), 85–105.
Zhou, W., 2007. Drainage and flooding in karst terranes. Environmental Geology 51
(6), 963–973.