Sie sind auf Seite 1von 22

Journal of Hydrology 274 (2003) 248269 www.elsevier.

com/locate/jhydrol

Evaluation of aquifer thickness by analysing recession hydrographs. Application to the Oman ophiolite hard-rock aquifer
B. Dewandela,1, P. Lachassagneb,*, M. Bakalowiczc, Ph. Wengb, A. Al-Malkid
b

ISTEEM, Tectonophysique, UM II, c.c. 049, 34095 Montpellier cedex 05, France BRGM, Water Division, Resource Assessment, Discontinuous Aquifers Unit, 1039, rue de Pinville, 34000 Montpellier, France c ISTEEM, Hydrosciences, UM II, c.c. MSE, 34095 Montpellier cedex 05, France d Research Department, Ministry of Water Resources, P.O. Box 2575 RUWI, Postal Code: 112, Muscat, Sultanate of Oman Received 29 August 2002; revised 5 December 2002; accepted 6 December 2002

Abstract For more than a century, hydrologists and hydrogeologists have been investigating the processes of stream and spring baseow recession, for obtaining data on aquifer characteristics. The Maillet Formula [Librairie Sci., A. Hermann, Paris (1905) 218], an exponential equation widely used for recession curve analysis, is an approximate analytical solution for the diffusion equation in porous media whereas the equation proposed by Boussinesq [C. R. Acad. Sci. 137 (1903) 5; J. Math. Pure Appl. 10 (1904) 5], that depicts baseow recession as a quadratic form, is an exact analytical solution. Other formulas currently used involve mathematical functions with no basis on groundwater theory. Only the exact analytical solutions can provide quantitative data on aquifer characteristics. The efciency of the two methods was compared on the basis of recession curves obtained with a 2D cross-sectional nite differences model that simulates natural aquifers. Simulations of shallow aquifers with an impermeable oor at the level of the outlet show that their recession curves have a quadratic form. Thus, the approximate Maillet solution largely overestimates the duration of the inuenced stage and underestimates the dynamic volume of the aquifer. Moreover, only the Boussinesq equations enable correct estimates of the aquifer parameters. Numerical simulations of more realistic aquifers, with an impermeable oor much deeper than the outlet, proves the robustness of the Boussinesq formula even under conditions far from the simplifying assumptions that were used to integrate the diffusion equation. The quadratic form of recession is valid regardless of the thickness of the aquifer under the outlet, and provides good estimates of the aquifers hydrodynamic parameters. Nevertheless, the same numerical simulations show that aquifers with a very deep oor provide an exponential recession. Thus, in that conguration, the Maillet formula also provides a good t of recession curves, even if parameter estimation remains poor. In fact, the recession curve appears to be closer to exponential when ow has a very important vertical component, and closer to quadratic when horizontal ow is dominant. As a consequence, aquifer permeability anisotropy also changes the recession form.

* Corresponding author. E-mail addresses: p.lachassagne@brgm.fr (P. Lachassagne), baka@msem.univ-montp2.fr (M. Bakalowicz), dewandel@ngri.res.in (B. Dewandel), p.weng@brgm.fr (Ph. Weng). 1 Present address: BRGM, Water Division, Resource Assessment, Discontinuous Aquifers Unit, Indo-French Center for Groundwater Research, Uppal Road, 500 007 Hyderabad, India. 0022-1694/03/$ - see front matter q 2003 Elsevier Science B.V. All rights reserved. doi:10.1016/S0022-1694(02)00418-3

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269

249

The combined use of the two tting methods allows one to quantify the thickness of the aquifer under the outlet. q 2003 Elsevier Science B.V. All rights reserved.
Keywords: Recession hydrograph; Baseow; Aquifer geometry; Exponential recession; Quadratic recession; Inuenced stage; Boussinesq; Maillet

1. Introduction Baseow rates and recession mechanisms of streams and springs have been extensively investigated for more than a century (Boussinesq, 1877; Maillet, 1905). The characterization of baseow rates and their prediction during drought are necessary in hydrology for determining the storage and exploitation possibilities of surface-water resources for drinking, irrigation, and industrial purposes, and for determining the impact of contamination, e.g. in areas downstream from wastewater discharge (Thomas and Cervione, 1970; Tasker, 1972; Parker, 1977; Vogel and Kroll, 1992). The study of stream recession, where the terminal recession parts correspond to aquifer discharge, provides data on the structure and functioning of the aquifer and, more particularly, on its hydrodynamic parameters such as permeability and storage coefcient (Mijatovic, 1974; Brutsaert and Nieber, 1977; Troch et al., 1993; Szilagyi et al., 1998). The analysis of karst spring recession is also one of the most commonly used approaches for characterizing this type of aquifer (Skolov, 1965; Mangin, 1970, 1975; Soulios, 1991; Padilla et al., 1994). In this paper, we focus only on porous, or equivalent porous, aquifers. Hydrograph recession curves are usually described

in two distinct stages (Fig. 1): the inuenced stage during which quickow (runoff and/or inltration towards the saturated zone) predominates and the baseow stage when the saturated zone only discharges. The quantitative analysis of such curves is generally through methods derived from the work of Maillet (1905) who showed that the recession of a spring or a river can be represented by an exponential formula, implying a linear relationship between hydraulic head and owrate. Even though models of this type are widely used (Mangin, 1975; Nathan and McMahon, 1990; Soulios, 1991; Vogel and Kroll, 1992; Gheorghe and Rotaru, 1993; Sugiyama, 1996; Shevenell, 1996; Vasileva and Komatina, 1997; Lastennet and Mudry, 1997; Long and Derickson, 1999; etc.), it is also known from the work of Boussinesq (1903, 1904) that the discharge of aquifer systems is characterized by a non-linear behavior. So which model (linear or non-linear), and thus which recession formula is the most suitable for analysing recession curves? Here we briey review and assess existing methods used for hydrograph recession analysis. The two that are most often used, i.e. tting the recession curve with exponential (Maillet, 1905) and quadratic (Boussinesq, 1903) formulas, are then compared using numerical simulations of realistic aquifer recessions.

2. Conceptual methods of hydrograph analysisa brief review Even though the overall shapes of recession curves are generally similar (reduction of ow decrease rate with time), differences are observed from one aquifer (or spring) to another. Curve shape is affected by the hydrodynamic properties of the aquifer, such as hydraulic conductivity, storage coefcient, hydraulic gradient (Schoeller, 1948, 1967; Tison, 1960; Forkasiewicz and Paloc, 1965;

Fig. 1. Use of a recession curve to separate the different components of a hydrograph.

Table 1 Review of different conceptual methods used for baseow recession analysis Method Boussinesq (1877) Model Formula Qt Q0 e2at Q0 p=2KHlhm =L; a p2 KH =4wL2 Exact solution Approximate solution Mathematical t Interpretation Aquifer and river recession

250

Maillet (1905)

Qt Q0 e2at

Aquifer and river recession

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269

Boussinesq (1903)

Qt Q0 =1 at2 Q0 1:724Kh2 m l=L; a 1:115Khm =wL2

Aquifer and river recession

Schoeller (1948); Barnes (1939)

Qt

Pn

i 1

Q0i e2ai t

Entire recession (including inuenced stage)

Horton (1933)

Qt Q0 e2at

Entire recession (including inuenced stage)

Coutagne (1948)

Qt Q0 1 n 2 1a0 tn=12n at a0 1 n 2 1a0 t21

Entire recession (including inuenced stage)

Drogue (1972)

Qt Q0 =1 atn

Entire recession (including inuenced stage)

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269 Entire recession: inuenced stage aquifer recession Entire recession: inuenced stage aquifer recession Entire recession (including inuenced stage)

251

Berkaloff, 1967; Castany, 1967; Drogue, 1967; Mijatovic, 1974; Kiraly and Morel, 1976; Brutsaert and Nieber, 1977; Moore, 1992; Eisenlohr et al., 1997), and by the geomorphologic characteristics of the catchment basin, such as geometry, slope, length of perennial streams or of conduits for the karst, etc. (Horton, 1945; Farvolden, 1963; Knisel, 1963; Carlston, 1966; Comer and Zimmerman, 1969; Kiraly and Morel, 1976; Eisenlohr et al., 1997). Also important are the roles played by (a) climate and season, and (b) the nature, thickness, spatial distribution, and saturation level of the soil horizons (Clark, 1956; Tison, 1960; Kunkle, 1962; Riggs, 1964; Carlston, 1966; Comer and Zimmerman, 1969; Beran and Gustard, 1977; Lorup et al., 1998; Lacey and Grayson, 1997). These observations explain why hydrologists and hydrogeologists are well aware that the study of stream and spring discharge can provide information on the structure and functioning of aquifers. Hall (1968, 1969) and Tallaksen (1995) have published excellent reviews of quantitative-analysis methods for recession curves. Several conceptual models have been developed (Table 1), most of which are based either on empirical relations to provide a mathematical t, or on the diffusion equation (or its approximation), describing ow in porous media. Boussinesq (1877) was one of the rst to conduct theoretical work on spring owrates and the mechanisms of aquifer drainage. For this he used the diffusion equation that describes ow through a porous medium:   h K h h 1 t w x x where K is the hydraulic conductivity (Darcy permeability), w the effective porosity (or free-aquifer storage coefcient) of the aquifer, h the hydraulic head, and t the time. He integrated this differential equation by introducing simplifying assumptions (Table 1)i.e. (i) the aquifer is porous, free, homogeneous and isotropic, with a width (perpendicular to the stream) of L and a length of l (parallel to the stream), (ii) capillarity effects above the water table are neglected, (iii) the aquifer oor is concave with a depth H under the outlet level, and (iv) variations of h are negligible

Qt Q0 2 Qc 1 n 2 1a0 tn=12n Qc

Samani and Ebrahimi (1996)

Padilla et al. (1994)

Mangin (1975)

ut Q0 2 q0 1 n 2 1a0 tn=12n ft q0 e2at

ct q0 1 2 ht=1 1t wt qr0 e2at

252

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269

compared to the depth H. However, Boussinesq only obtained the approximate analytical solution (linearization) given by Eqs. (2a) (2c), i.e.: Qt Q0 e where Q0 and p2 KH a 4wL2 p h KHl m 2 L 2b
2at

quadratic form. Qt where Q0 0:862Kl and h2 m L 3b Q0 1 at 2 3a

2a

a
2c

1:115Khm wL2

3c

and in which Q0 is the initial ow rate, Qt the ow rate at time t, a the recession coefcient characterizing the aquifer, L the width of the aquifer, H the depth of the aquifer under the outlet, hm the initial hydraulic head at the distance L, and l the length of the perennial stream. Eq. (2a) is now well-known as the Maillet formula because Maillet (1905), approximated the aquifer recession curve using an analogue model of a water-lled reservoir emptying through a porous plug (Table 1); this small-scale model generates a owrate vs. time that Maillet (1905) assimilated to an exponential function comparable to Eq. (2a). Alternative forms of this formula can be found in the work of Barnes (1939), Farvolden (1963), Knisel (1963), Riggs (1964), Vogel and Kroll (1992), Troch et al. (1993), Szilagyi and Parlange (1998, 1999), Szilagyi et al. (1998) and Szilagyi (1999). Meanwhile, Boussinesq (1903, 1904) developed an exact analytical solution of the diffusion Eq. (1) by considering the simplifying assumptions (Table 1) of a porous, free, homogeneous, isotropic, aquifer with no capillary effect, that is limited by an impermeable horizontal layer at the level of the outlet that is considered to be localized. The initial free surface is curvilinear and its shape is an incomplete inverse beta function (see Polubarinova-Kochina, 1962, pp. 515 517, or Troch et al., 1993); and all velocities within the aquifer are horizontal and parallel to one another in the same vertical plane (Dupuits assumptions; see Marsily de (1986)). This gives the nonlinear solution of Eqs. (3a) (3c) that we here term the

In spite of its simplifying assumptions, this model has the advantage of being a non-approached analytical solution for porous media and of providing a simple expression where the parameters a and Q0 have a physical meaning (Eqs. (3b) and (3c)). Eq. (3c) also shows that a varies with hm and thus increases with Q0 ; whereas a in Eq. (2c) is a constant. Practice has conrmed these variations of a when tting different recession curves of a river, a spring or an aquifer, even with the exponential form (cf. Tison, 1960; Kunkle, 1962; Knisel, 1963; Riggs, 1964; Beran and Gustard, 1977; Dewandel, 2002; among others). However, the mathematical properties of Eq. (3a) are not as convenient as those of the exponential solution Eq. (2a) in which the derivative is also an exponential, and a is independent of Q0 (for the linearization of Eq. (3a) see, for example, Drogue, 1972). This possibly explains why the Maillet formula is extensively used as compared to the quadratic form. Nevertheless, since the quadratic form is an exact analytical solution, and not an approximate one, it is a priori less erroneous than the exponential solution Eq. (2a). Since the exponential form Eq. (2a) is only applicable to the last part of a recession curve (i.e. the most distant in time from the last precipitation), many authors (Maillet, 1905; Barnes, 1939; Schoeller, me nie ras, 1960; 1948; Werner and Sundquist, 1951; Re Forkasiewicz and Paloc, 1965; Shevenell, 1996; Vasileva and Komatina, 1997; etc.) have proposed more complex models that consider the cumulative effect of several reservoirs, each with an exponential recession, in order to interpret the entire recession

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269

253

hydrograph (Table 1). These give the expression Qt


n X i1

Thus, for n 1; 4a Qt Q0 1 n 2 1a0 tn=12n 5a 5b

Q0i e

2ai t

at a0 1 n 2 1a0 t21

which is not an exact analytical solution. Barnes (1939), in dealing with streamow, attributed the different components of the recession curve to runoff, interow, and aquifer discharge. Schoeller (1967), in dealing with karst aquifers, considered that the highest recession coefcient corresponds to the permeability of the conduits, and the lowest to that of blocks and ssures, and that when an intermediary coefcient exists, it is characteristic of an intermediate-permeability medium (Forkasiewicz and Paloc, 1965; Drogue, 1967). Although this method could be used for characterizing independent reservoirs in the same catchment, all outowing at the same location, Mangin (1970) showed that their assumptions do not generally agree with reality. Kiraly and Morel (1976), and later Eisenlohr et al. (1997), built nested numerical models of karst systems with two different permeability media conduits and blocks. Like Schoeller (1948, 1967) and Forkasiewicz and Paloc (1965), they were able to separate the simulated recession curves into three components. However, they were not able to attribute each of the three exponential portions of recession to different permeability media, as their model only included two media. Horton (1933) proposed the sum of two exponentials Eq. (4b), sometimes called the Horton double exponential (Table 1): Qt Q0 e2at
m

where n is a constant that depends on reservoir geometry, and where the coefcient at increases with time if n , 1; and decreases if n . 1. When n 1, we arrive at the exponential form of Eq. (2a). However, the expression of Eq. (5a) does not provide information on the hydrodynamic parameters of aquifers. Padilla et al. (1994) showed that although the formulas of Eqs. (5a) and (5b) can be applied to the recession of a spring or stream, where n generally varies from 0 to 2, they do not perfectly describe karst spring hydrographs. They therefore added a constant discharge Qc to Eq. (5a) so that the equation now reads (Table 1): Qt Q0 2 Qc 1 n 2 1a0 tn=12n Qc 6

According to Padilla et al. (1994), Qc results from the recession of aquitards neighboring the karst aquifer, or from the recession of low-transmissivity compartments of the karst; the authors do not, however, rule out that this parameter may have no physical meaning. Following the work of Forkasiewicz and Paloc (1965), Dauty (1967) and Drogue (1972) proposed to t the entire hydrograph of a karst spring with a hyperbolic function (Table 1): Qt Q0 1 at n 7

4b

where m is a coefcient. Similar solutions (sum of n exponentials) are proposed by Brutsaert (1994) and Brutsaert and Nieber (1977) to, respectively, describe (i) the outow from a hillslope aquifer, and (ii) Boussinesq (1903) model taking into account a water depth in the channel different from 0 (Table 1). All these expressions are approximate analytical solutions. With the objective of interpreting the recession curve, Coutagne (1948) proposed an equation that generalized the discharge from a reservoir (Table 1).

where Q0 is the initial ow rate, a the recession coefcient, and n a constant (generally between 1.5 and 2) that expresses the concavity of the curve and, depending on the author, reects a state of initial saturation of soils and micro-ssures. Mangin (1970, 1975) supposed that the discharge of a karst system is represented by two components (Table 1): one linear, wt; represented by the Maillet formula, to represent the recession of the saturated zone, and the other non-linear, represented by a homographic empirical formula ct; which represents the ow in the unsaturated zone (inuenced stage): Qt ct wt 8a

254

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269

wt qr0 e2at c t q 0
1 2 ht 1 1t

8b 8c

where q0 is the maximal inltration owrate q0 qr0 2 Q0 ; Q0 being the owrate of the hydrogram peak, and qr0 being the baseow extrapolated at the onset of the recession), h characterizes the duration of inltration (h 1=ti ; with ti being the duration of inltration), and 1 the concavity of the recession curve resulting from inltration heterogeneity. Samani and Ebrahimi (1996) propose a similar approach in adding a Coutagne owrate, ut ; to the Maillet formula to represent the contribution of the unsaturated zone (Table 1): Q t ut f t 9a
n=12n

provide more information. Moreover, since it never reaches zero, it is necessary to stop the inltration articially. To summarize, we nd two approaches with these methods: (i) those attempting to separately characterize recession of the saturated zone of the aquifer and the mechanisms of its recharge by inltration (inuenced stage), and (ii) those that consider only the drainage processes of the saturated zone. Where the recession of the saturated zone is concerned, only the equation proposed by Boussinesq (1903, 1904) is an exact analytical solution of the diffusion equation in porous media. The other formulas correspond either to approximate solutions of the diffusion equation, such as the Maillet (1905) formula, or to mathematical functions with no hydrodynamic signicance.

ut Q0 2 qr0 1 n 2 1a0 t ft qro e2at

9b 9c 3. Comparison of the Boussinesq quadratic solution and the Maillet approximate solution in analysing recession curves For an initial comparison of the Boussinesq and Maillet analytical methods, we used hydrograph recession curves of an ophiolite hard-rock aquifer

in which the parameter at Eq. (5b) describes ow from the unsaturated zone to the saturated zone, i.e. the inltration. Although the inltration formula ut is somewhat more analytical than that of Mangin (1975), it does not

Fig. 2. Analysis of a recession curve at Wadi Khafah beginning on 26/01/1997. Squares: measured ows. Dotted line: Maillet (1905) method with an exponential t: of Q0 0.028 m3/s, a 0.016 j21, VD 1.52 105 m3. Bold line: Boussinesq (1903) method with a quadratic t of Q0 0.066 m3/s, a 0.031 j21, VD 1.84 105 m3. Insert: Wadi Khafah owrate chronicle.

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269

255

measured in Wadi Khafah, Sultanate of Oman (Fig. 2, recession starting on 26/01/1997). Other works showed that this ssured aquifer could be considered at this scale as an equivalent porous medium. We found that the quadratic form enables a correct t of the hydrograph from the second day of the recession up to its end (43 days), whereas the t with the Maillet method is only valid from the 22nd daythus, the two methods provide very different estimations of the duration of the inuenced stage (2 days as against 22 days). One consequence is that the dynamic volumes VD ; calculated by integrating either Eqs. (2a) or (3a) (0 1) are signicantly different (about 20%) 2 1.5 105 m3 using the Maillet method Eq. (2a) and 1.8 105 m3 using the Boussinesq method Eq. (3a) although of the same order of magnitude. The better t obtained with the quadratic form is related to the fact that this method uses an exact analytical solution of the diffusion equation. However, as the Boussinesq method relies on certain simplifying assumptions (see Section 2), the reliability of the obtained t cannot be ensured. We therefore decided to analyze the efciency of the two methods on the basis of recession curves obtained with a gridded numerical model that simulates the behavior of natural aquifers; i.e. this model does not use all the Boussinesq or Maillet simplifying assumptions (such as the Dupuit assumption). Our approach comprised of the following steps: 1. a sensitivity analysis of the numerical model to

check that the recession curves provided by the model are not altered by numerical errors related to the choice of discretization, time step, etc.; 2. simulation of the recession curves of an aquifer with similar properties to those of the Boussinesq analytical solution of the diffusion equation. Steps 1 and 2 are performed together, comparison of numerical model results to those of an analytical solution being a classical method to validate numerical models; 3. simulation of the recession curves of several hydrogeologic congurations that differ from the Boussinesq simplifying assumptions; and 4. comparison of the Maillet and Boussinesq methods applied to the obtained recession curves. 3.1. Description of the numerical model The numerical simulations were carried out with the 3D hydrodynamic modeling code MARTHE (Modeling Aquifers with Rectangular grids in Transitory regime for Hydrodynamic ow calculation) developed at BRGM (French Geological ry (1993, 1994). This code allows Survey) by Thie the 3D (or pseudo 3D) and 2D (in plan or in vertical section) hydrodynamic and hydrodispersive modeling of groundwater ow in porous media. The hydrodynamic calculations are carried out using the nite-difference method with an implicit schemehere we used the method of conjugate gradients with Cholesky decomposition. Conver-

Fig. 3. Grid area used for the numerical model (2D cross section) here shown with 20 m-side square meshes. The black mesh is the outlet and i is the initial recharge to simulate the initial hydraulic head prole.

256

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269

gence of the calculations is checked using several criteria such as difference of hydraulic head between two successive iterations, residual error ow, etc. A vertical cross section of the aquifer was simulated using the following grid geometry for the numerical model (Fig. 3): 2D-ow simulation of a cross section of a rectangular aquifer, grid with regular square meshes, width L 1000 m; length l 1 m; thickness B 440 m (see Fig. 3), the outlet is simulated by a mesh with an imposed potential equal to the outlets elevation; this mesh is located at the base level of the aquifer, the other limits of the aquifer are no-ow boundaries. The recession curves were computed as follows: 1. a steady-state simulation was performed with a constant recharge rate, i, all along the upper boundary of the model (Fig. 3) in order to generate a realistic steady-state piezometric prole within the aquifer, 2. the recession curve was then computed with a transient-state simulation starting from the abovedescribed steady state with a 0 recharge rate (i) condition. 3.2. Model sensitivity analysis 3.2.1. Sensitivity to discretization Various simulations were carried out, progressively reducing the square mesh sides from 40 to 10 m, with a resultant increase in the total number of grid cells from 275 to 4400, in order to maintain the same model size and shape (B 440 m, L 1000 m). A permeability of 1027 m/s and an effective porosity (storage coefcient in unconned aquifer) of 1023 were assigned to all the grid cells of the model. Recession curves of 200 days were computed for the different grid systems using one-day time steps, and compared to the analytical solution of Boussinesq Eqs. (3a) (3c) as follows:

1. the simulated discharge was normalized with respect to the analytical solution of Boussinesq, from which the apparent bias R Qsim =QBouss is computed (Fig. 4(a)), 2. the deviation from this solution (in %) is represented by Error % 100 Qsim 2 QBouss QBouss Fig:4b:

The ratios and deviations of the simulated ows compared to the Boussinesq solution show that the discretization of the 275-grid cells model (square grid cells with 40 m sides) is insufcient, in particular near the outlet where the discretization is not ne enough to establish a regular hydraulic head prole: the ratios vary from 1.42 to 1.80 and deviations from 39 to 89.5%. Although the ratios and deviations tend to decrease with time as a result of the reduction of hydraulic gradients, they nonetheless remain very high (1.42 and 39.6% at t 200 d). The 1100-grid cells model (grid cells with 20 m sides) provides results that are closer to the Boussinesq solution (Fig. 4(a) and (b)): the mean ratio is 1.04 and the mean deviation (calculated from the mean of absolute values) is 4.48%. There is nevertheless signicant oscillation of the ratio from 1.00 to 1.12 and of the deviation from 0.07 to 12.1% with time. As with the 275- grid cells model, this leads to the occurrence of small parasite oods that decay in about 10 20 days. The numerical errors are still too high to accurately simulate the owrate at the outlet of the model. The 4400-grid cells model (grid cells with 10 m sides) provides a much better resolution (Fig. 4(a) and (b)). The mean ratio is 1.00 and the mean deviation is 2.0%. The oscillation of ratios and deviations (R 0.97 1.05 and E 0.01 5.30%) are acceptable. It was thus this model grid (1000 440 m with 10 m-side square grid cells) that was chosen for simulating the recession curves. 3.2.2. Sensitivity to the calculation time step Simulations were also carried out to determine whether or not an increase or decrease in the calculation time step would change the numerical

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269

257

Fig. 4. Choice of mesh: comparison of simulated ows to the exact analytical solution of Boussinesq Eqs. (3a)(3c) for different grid systems (2754400 square meshes). (a) R Qsim =QBouss ; apparent bias. (b) Error% 100 Qsim 2 QBouss =QBouss ; percent difference.

results. Three simulations at time steps of 0.2, 1 and 10 days were compared for a recession period of 200 days (Fig. 5(a) and (b)). We found that decreasing the calculation time step from 1.0 to 0.2 days does not improve the numerical calculation; in fact, the numerical error is even a bit higher during certain periods because of the increased number of calculations that temporarily causes a greater divergence of the results. The simulation run with a 10-day time step (which involved one initial step at 2.5 days) provided

results of a higher quality than those obtained with the 0.2-day time step and of a similar quality to those obtained with the 1.0-day time step. CPU time, with a 10-day time step, is signicantly reduced, but the simulated recession curves show a higher sensitivity to initial conditions: . 4% deviation at the early time of the recession, although progressively decreasing.

258

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269

Fig. 5. Choice of time step: comparison of simulated ows to the exact analytical solution of Boussinesq Eqs. (3a)(3c) for different time steps (0.2, 1 and 10 days). (a) R Qsim =QBouss ; apparent bias. (b) Error% 100 Qsim 2 QBouss =QBouss ; percent difference. s.d.: standard deviation.

3.2.3. Discretization and time step adopted for the simulations As a result of the sensitivity analyses, it was decided to carry out the simulations with a model containing 4400 square grid cells of 10 m sides and using a 10-day time step over a recession period of 200 days (Fig. 6). Flowrates shown on the recession curves were normalized with respect to the initial owrate in order to facilitate interpreting the gures. Subsequently, we were to consider as signicant only those deviations that differ by at least 3% from the

Boussinesq solution. Moreover, the structure of the numerical error, generally in the form of small parasite oods and not a general trend, makes its identication relatively easy. 3.3. Numerical simulation of recession curves from aquifers conforming to the Boussinesq reductive assumptions The sensitivity analysis not only proves the validity of the numerical tool (cf. Fig. 6), but also,

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269

259

Fig. 6. Analysis of the recession curve obtained from the model with an impermeable oor at outlet level (H 0 m) and ows normalized to initial ow. Dotted lines: Maillet (1905) method with an exponential t of Q0 0.934 m3/s, a 0.0016 j21, VD 5.05 107 m3. Bold lines: Boussinesq (1903) method with a quadratic t of Q0 1 m3/s, a 0.0011 j21, VD 7.68 107 m3. Insert: semi-logarithmic representation of ow and t of the exponential form (Maillet). Ln: natural logarithms.

and above all, conrms the validity of the Boussinesq quadratic form for analysing recession curves (cf. Szilagyi et al., 1998), even under conditions where some of the simplifying assumptions formulated by Boussinesq are not fullled: e.g. two-dimensional ow (Dupuits assumption is not invoked in the numerical model), non-isolated outlet, etc. Determination of a and Q0 (Eq. (3a)) enables K and w Eqs. (3b) and (3c) to be calculated at 1.05 1027 m/s and 9.01 1024, respectively, i.e. very close to the values introduced in the model (see Table 3). The exponential formula of Maillet (Eq. (2a)) does not describe the recession as well as the quadratic function (Fig. 6)the t begins only after 100 days of recession with the Maillet formula, whereas it is valid from the start with the Boussinesq quadratic form. As a consequence, the dynamic volume VD (5.1 107 m3 using the Maillet method and 7.7 107 m3 using the Boussinesq method) is underestimated (52%) and leads to a wrong estimate of the duration of the inuenced stage. 3.4. Inuence of aquifer diffusivity Using Eqs. (3a) (3c), recession curves for aquifers having different diffusivities (K/w 1025, 5 1025, 10 24, 102 3, 1022 and 1021 m/s; in all cases

w 1023) have been calculated. For all these curves (Fig. 7), the exponential ts are only valid after 100 days, regardless of the diffusivity, and so the difference between the Boussinesq and Maillet ts remains constant. Consequently, the Boussinesq quadratic form should be preferred to the Maillet exponential formula for tting recession curves of a relatively shallow aquifer with an impermeable boundary at the level of the outlet.
3.5. Inuence of the depth of the impermeable oor of the aquifer In natural catchments, the impermeable oor of the aquifer does not systematically coincide with the outlet level; generally, groundwater ow occurs under the outlet. To simulate these conditions we modied our numerical model by increasing the depth H of the impermeable boundary from outlet level (0 m; Boussinesq) to 50, 80, 160, 310 and 540 m under outlet level (Fig. 8). The model was thus increased from 4400 to 10,000 grid cells, although retaining the same hydrodynamic parameters as in the previous simulation: K 1027 m/s, w 1023. All the recession curves (Fig. 9) have been started with an initial hydraulic head prole obtained in the steady state for an identical recharge rate (i). This means that the initial hydraulic head proles differ according to the aquifer geometry because the

260

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269

Fig. 7. Stability of the difference of the t period between the two methods (Maillet and Boussinesq) for aquifers with different diffusivities K =w and with ow rates calculated from Eqs. (3a)(3c) (H 0) and normalized to initial ows. (a) K =w 1025, 5 1025 and 1024 m/s; (b) K =w 1023, 1022 and 1021 m/s. Ln: natural logarithms.

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269

261

Fig. 8. Flow schemes for different aquifer oor depths under outlet level, H. (a) H 0 m (Boussinesq, 1903, 1904), (b) H 160 m, (c) H 540 m.

hydraulic gradient (Table 2) is greater in congurations with a shallow impermeable oor than with a deep impermeable oor. Similarly, because the equivalent transmissivity of the model T KH increases roughly in proportion to the depth of the impermeable oor, the aquifer drains more rapidly in congurations with a deep impermeable oor, H. The quadratic form (Table 2 and Figs. 6 and 10(a) (e)) gives good ts for all the simulated recession curves; the regression coefcients and the Nash criterion calculated for each simulation are high (Table 2). The use of the quadratic form of recession Eq. (3a) is valid regardless of the thickness of the aquifer under the outlet. The exponential formula provides satisfying

results only when the depth of the substratum is equal to or more than about 160 m under the outlet. In these conditions, the quality of the t is somewhat higher than that obtained with the quadratic form. Although hydrograph analysis with the exponential formula is valid if the impermeable oor is deep, the Eqs. (2a) (2c) cannot be used to deduce hydrodynamic parameters from the recession curves. For example, the exponential approached analytical solution corresponding to a deep concave oor (Boussinesq, 1877; Table 1) was used to calculate the recession curve obtained by the numerical model with a 540 m-deep oor (Fig. 10(e))the two results were very different. The recession curves show that the deeper the aquifer oor, the more rapidly the aquifer is drained because increasing the depth of the impermeable substratum results in an increase of the equivalent transmissivity of the aquifer. Nevertheless this trend is asymptotic. In the tested conguration, the effect of deepening the level of the impermeable substratum was less and less apparent below a depth of 150 200 m under the outlet. Here, the increase of aquifer transmissivity due to the increase of its thickness at constant permeability, is balanced by longer ow lines and thus by a smaller mean hydraulic gradient along the ow lines. This results from the fact that most of the ow circulates in the upper part of the aquifer to a depth of 100 200 m under the outlet. Moreover, the depth of circulation is highly inuenced by the width of the aquifer, L; it would be less deep for a narrower aquifer. The values of equivalent permeability and storage for the different simulations (H 0, 50, 80, 160, 310 and 540 m) were calculated with the Boussinesq Eqs. (3a) (3c), adopting the value of the initial hydraulic head Dh provided by the numerical model at the start of the recession (Table 3). The values deduced for the Boussinesq conguration (H 0 m) are not very different from those assigned to the numerical model, which conrms both the validity of the numerical model (cf. Section 3.3) and the robustness of the Boussinesq formula in this conguration. In addition, the calculated aquifer permeability increases with increasing depth of the substratum (up to a factor of 9.7 for H 540 m), which is the logical consequence of increasing the aquifer transmissivity (see above). The storage coefcient also increases with the depth of

262

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269

Fig. 9. Comparison of recession curves obtained for aquifers with different geometries and owrates normalized to initial ows. H (depth of aquifer oor under outlet level) 0, 50, 80, 160, 310, and 540 m. The error bars for H 0 m signify a 3% difference.

the impermeable oor, but to a lesser extent than the permeability (up to a factor of 1.5 for H 540 m). 3.6. Application to the evaluation of aquifer thickness (H) under the outlet Our results justify the use of combined quadratic and exponential processing of a recession for characterizing aquifer geometry, i.e. the thickness of an aquifer under its outlet, H. In the case of a very deep aquifer with a large thickness/width ratio H =L; the quadratic and exponential formulas provide similar ts. In the case of a shallow aquifer, however, the exponential t erroneously suggests the existence of a longer inuenced stage than in reality and, in

any event, a longer inuenced stage than with the quadratic form. In this case, the geometry of the studied aquifer can be characterized by an additional parameter, tDL (Fig. 11), which is the difference in the duration of the inuenced stage between the quadratic t and the exponential t (i.e. the lapse time between the quadratic t and the exponential t). According to our simulations, H can be described by the following empirical relationship: H L20:08 LntDL 0:37 10

where L is the width of the aquifer and ln the natural logarithms. We applied this formula to the Wadi Khafah perennial stream recession curve (cf. Fig. 2) of a hard-

Table 2 Simulations with various depths of impermeable oor (H). Linear regression coefcient (R) and Nash criterion obtained for both quadratic form (quadra.) and exponential form (expo.) ts. D(h) initial (steady-state) maximum hydraulic head (see Fig. 3) H (m) 0 50 80 160 310 540 Dh (m) 100.6 78.5 57.5 44.1 36.2 33.7 R 2expo. 0.987 0.969 0.997 0.992 0.997 0.995 R12quadra. 0.994 0.991 0.995 0.992 0.997 0.991 Crit. Nashexpo. 99.74 99.6 99.08 99.56 99.7 99.62 Crit. Nashquadra. 99.87 99.71 99.74 99.6 99.85 99.48

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269

263

Fig. 10. Analysis of recession curves obtained from different geometries with owrates normalized to initial ows. Dotted curve: Maillet (1905) methodexponential form. Bold curve: Boussinesq (1903) methodquadratic form. Insert: semi-logarithmic representation of the ow and t of the exponential form (Maillet). (a) H 50 m, (b) H 80 m, (c) H 160 m, (d) H 310 m, (e) H 540 m. Ln: natural logarithms.

264

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269

Fig. 10 (continued )

Table 3 Simulations with various depths of impermeable oor (H). Equivalent permeability (K-equ) and equivalent storage coefcient (w-equ) calculated with Eqs. (3a) (3c) (quadratic form) compared to real values introduced into the numerical model (K-model 1027 m/s, wmodel 1023) H (m) 0 50 80 160 310 540 Dh (m) 100.6 78.5 57.5 44.1 36.2 33.7 K-equ 1.05E07 1.71E07 3.23E07 5.47E07 8.35E07 9.72E07

w-equ
9.01E04 1.05E03 1.24E03 1.32E03 1.45E03 1.46E03

K-equ/K-model 1.05 1.71 3.23 5.47 8.35 9.72

w-equ/w-model
0.90 1.05 1.24 1.32 1.45 1.46

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269

265

Fig. 11. Relationship between aquifer geometry H =L and the difference of the inuenced stage between the Maillet and Boussinesq methods (tDL). Ln: natural logarithms.

Fig. 12. Analysis of recession curves obtained for different permeability anisotropy conditions Kv =Kh :1, Kv =Kh :10 and Kv =Kh :0.1; Kh 1027 m/s) for the model where H 540 m. Dotted line: Maillet (1905) methodexponential form. Bold line: Boussinesq (1903) methodquadratic form.

266

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269

Table 4 Simulations with various permeability anisotropies. Kv vertical permeability; Kh horizontal permeability. Linear regression coefcient (R) and Nash criterion obtained for both quadratic form (quadra.) and exponential form (expo.) ts. D(h) initial (steady-state) maximum hydraulic head (see Fig. 3) H 540 (m) Kv =Kh : 1 Kv/Kh: 0.1 Kv =Kh : 10 Dh (m) 33.7 19.0 64.8 R 2expo. 0.995 0.969 0.992 R 2quadra. 0.991 0.996 0.994 Crit. Nashexpo. 99.62 99.68 99.75 Crit. Nashquadra. 99.48 99.42 99.80

rock aquifer in an ophiolite catchment basin of the Ibra region, Sultanate of Oman. The basin is composed exclusively of harzburgite, which represents the mantle part of the ophiolite (fragments of oceanic lithosphere obducted onto the continent), and is drained by about 40 km of perennial streams, l, with no alluvium (l was determined with an opisometer on a 1:100,000-scale topographic map). The area of the basin, A, is 37 km2, mean slope is about 30% and mean aquifer width, L, about 470 m (L A/2l). Parameter tDL was estimated at 20 days (Fig. 2). Using relationship (10) gives us an estimation of aquifer development over a mean depth of about 60 m under the base level that is the river. Parameter tDL was evaluated on three other recession curves (44 127 days of recession) and the results gave a mean tDL value of 27 ^ 7 (standard deviation) days, i.e. an estimated depth H of 51 ^ 10 m. This estimate agrees very well with that obtained from (i) a statistical analysis of borehole data, showing that the most transmissive horizons are located in the rst 50 m underground (Dewandel, 2000, 2002; Dewandel et al., 2001) (ii) geochemical data of the water in this catchment basin that show the shallow origin of the river water (Stanger, 1987; Dewandel, 2002), and (iii) seismic measurements showing the rapid closure of fractures within a depth of several tens of meters (Dewandel et al., 2003).

3.7. Inuence of permeability anisotropy A permeability anisotropy with a 10-fold horizontal and then vertical tensor (Kv =Kh 0.1 and Kv =Kh 10 for Kh 1027 m/s) was simulated with the deep-ow (H 540 m) model (Fig. 12). Analysis of the recession curves with the exponential and quadratic forms shows that the quality of both ts is good in the case of horizontal anisotropy, but that the exponential t is better than the quadratic t in the case of vertical anisotropy (Fig. 12; Table 4). This is consistent with what we have shown earlier; i.e. the recession form is closer to exponential when the ow contains a signicant vertical component. The existence of a horizontal permeability anisotropy (K v/ Kh 0.1, Kgeom 3.2 10 28 m/s, Table 5) leads to a more efcient drainage of the aquifer and thus increases its bulk transmissivity. The calculation of the equivalent permeability using the Boussinesq quadratic t conrms this observation (Kequ 3.2 1026 m/s, Table 5). With a vertical anisotropy, the opposite is obtained (Kequ 2.6 1027 m/s, Kgeom 3.2 1027 m/s, Table 5). Permeability anisotropy emphasizes or reduces the effects related to the depth of the impermeable oor of the aquifer. A permeability tensor with a principal

Table 5 Simulations with various permeability anisotropies. Equivalent permeability (K-equ) and equivalent storage coefcient (w-equ) calculated with Eqs. (3a) (3c) (quadratic form) compared to the storage introduced into the numerical model (K-model 1027 m/s, w-model 1023) and the geometrical mean of permeability for each model (K-geom: Kv =Kh : 0.1 3.2 1028 m/s and Kv=Kh: 10 3.2 1027 m/s) Anisotrophy (H 540) Kv =Kh : 1 Kv/Kh: 0.1 Kv =Kh : 10 Dh (m) 33.7 19.0 64.8 K-equ 9.72 1027 3.19 1026 2.60 1027

w-equ
1.46 1023 1.06 1023 1.30 1023

K-equ/K-geom 9.72 100.84 0.82

w-equ/w-model
1.46 1.06 1.30

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269

267

horizontal component will favor horizontal ow and reduce the effects related to the depth of this substratum by limiting the occurrence of deep ow lines-recession tends to a quadratic form. Conversely, vertical anisotropy will increase the vertical ow component, generate longer and deeper ow lines, and produce an exponential recession that is even more pronounced than in the isotropic case. The logical consequence of a permeability anisotropy should be the under- or overestimation of parameter tDL, depending on whether the anisotropy is horizontal or vertical. The Wadi Khafah recession curve (Fig. 2) is from a basin in which the dominant fracturing is vertical (Dewandel et al., 2001b). Parameter t DL is thus probably underestimated, which may mean an overestimation of the mean groundwater circulation depth, about 50 m, determined in Section 3.6.

4. Conclusion Recession curve analysis is a valuable tool, among others, for obtaining data on aquifer characteristics, e.g. permeability, storage coefcient, and geometry. The two main approaches in recession curve analysis are (i) mathematical functions used to t the entire recession curve, but without any link to groundwater ow equations, and (ii) methods based on approximate or exact analytical solutions of the diffusion equation. The literature review shows that the Maillet (1905) formula a widely used exponential equation, is an approached analytical solution for porous media whereas the quadratic form proposed by Boussinesq (1903 and 1904) is an exact analytical solution. Only the exact analytical solutions can provide quantitative data on aquifer characteristics. Numerical simulations of shallow aquifers with an impermeable oor at the level of the outlet show that their recession curves have a quadratic form. Thus, the Boussinesq formula should be preferred to the Maillet exponential form. The quadratic form ts the entire recession, whereas the approached Maillet solution largely overestimates the duration of the inuenced stage and thus underestimates the dynamic volume of the aquifer. Moreover, only the Boussinesq equations enable correct estimates of the quantitative aquifer parameters (permeability, storativity). Numerical simulations of more realistic aquifers,

with an impermeable oor much deeper than the outlet, proves the robustness of the Boussinesq formula even under conditions far from the simplifying assumptions that were used to integrate the diffusion equation. The quadratic form of recession is valid regardless the thickness of the aquifer under the outlet. Thus, we recommend tting all recession curves with the quadratic form rather than the exponential formula, particularly when the objective is to determine the aquifers hydrodynamic parameters or the duration of the inuenced stage. Nevertheless, the same numerical simulations show that aquifers with a very deep oor provide an exponential recession. Then, the Maillet formula provides a good t with such recession curves, even if parameter estimation remains poor. Thus, the combined use of the two tting methods allows to estimate the thickness of the aquifer under the outlet through determining a new parameter, tDL: the lapse time between the Boussinesq quadratic t and the Maillet exponential t of the recession. In fact, the recession curve appears to be closer to exponential when ow has a very important vertical component, and closer to quadratic when horizontal ow is dominant. As a consequence, aquifer permeability anisotropy also changes the recession form.

Acknowledgements This work was nancially supported by BRGM (French Geological Survey), the Languedoc Roussillon Regional Council, and the French Ministry for Foreign Affairs (French Embassy in Muscat). It also beneted from the logistical support of and the data collected by the Ministry of Water Resource of the Sultanate of Oman. A. Mangin, an other anonymous reviewer, and the Editor-in-Chief M. Sophocleous are thanked for their fruitful comments and suggestions. H.M. Kluijver and P. Skipwith (BRGM) are thanked for improving the English. We also thank A. Qatan (MWR) for his collaboration during eld measurements and chal (BRGM) and B. Ambroise (IMFS, J.C. Mare Strasbourg) for fruitful discussions during the writing and revision of this paper.

268

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269 tention par chenaux et ssures. Chronique dHydroge ol. 10, de re 43 47. crues Drogue, C., 1972. Analyse statistique des hydrogrammes de de des sources karstiques. J. Hydrol. 15, 4968. Eisenlohr, L., Kiraly, L., Bouzelboudjen, M., Rossier, Y., 1997. Numerical simulation as a tool for checking the interpretation of karst spring hydrographs. J. Hydrol. 193, 306 315. Farvolden, R.N., 1963. Geologic control on groundwater storage and baseow. J. Hydrol. 1, 219249. gime de tarissement de la Forkasiewicz, J., Paloc, H., 1965. Le re liminaire. Proc. Foux de la Vis (GardFrance). Etude pre es, Dubrovnik, 213 226. Hydrol. des Roches Fissure ` res karstiques Gheorghe, A., Rotaru, A., 1993. Evaluation des aquife ment a ` lanalyse de lhydrogramme des de bits de la conforme ` re. Application aux bassins de Motru et de Tismana rivie (Roumanie). Theor. Appl. Karstol. 5, 101108. Hall, F.R., 1968. Base-ow recessiona review. Water Resour. Res. 4, 973 983. Hall, F.R., 1969. Addendum to base-ow recession review. Water Resour. Res. 5, 914. Horton, R.E., 1933. The role of inltration in hydrologic cycle. Trans. Am. Geophys. Union 14, 446 460. Horton, R.E., 1945. Erosion development on streams and their drainage basins: hydrophysical approach for quantitative morphology. Geol. Soc. Am. Bull. 56, 275 370. Kiraly, L., Morel, G., 1976. Remarques sur lhydrogramme des par mode ` les mathe matiques. Bull. sources karstiques simule ol.Univ. de Neuchatel (Suisse) 1, 37 60. Centre dHydroge Knisel, W.G., 1963. Baseow recession analysis for comparison of drainage basins and geology. J. Geophys. Res. 68, 36493653. Kunkle, G.R., 1962. The baseow-duration curve, a technique for the study of groundwater discharge from a drainage basin. J. Geophys. Res. 67, 15431554. Lacey, G.C., Grayson, R.B., 1997. Relating baseow to catchment properties in south-easthern Australia. J. Hydrol. 204, 231250. Lastennet, R., Mudry, J., 1997. Role of karstication and rainfall in the behavior of a heterogeneous karst system. Environ. Geol. 32, 114 123. Long, A.J., Derickson, R.G., 1999. Linear system analysis in a karst aquifer. J. Hydrol. 219, 206 217. Lorup, J.C., Refsgaard, J.C., Mazvimavi, D., 1998. Assessing the effect of land use change on catchment runoff by combined use of statistical tests and hydrological modelling: case studies from Zimbabwe. J. Hydrol. 205, 147 163. Maillet, E., 1905. Essais dhydraulique souterraine et uviale. Librairie Sci., A. Hermann, Paris, 218. ` le tude des aquife ` res karstiques a ` Mangin, A., 1970. Contribution a crue et tarissement. Ann. partir de lanalyse des courbes de de le ol. 25, 581610. Spe ` le tude hydrodynamique des Mangin, A., 1975. Contribution a ` res karstiques. Institut des sciences de la Terre de aquife de Dijon, Moulis, The ` se de Doctorat e ` s Sciences luniversite Naturelles. Marsily De, G., 1986. Quantitative Hydrogeology, Groundwater Hydrology for Engineers, Academic Press, New York, p. 440. termination de la transmissivite et du Mijatovic, B., 1974. De

References
Barnes, B.S., 1939. The structure of discharge recession curves. Trans. Am. Geophys. Union 20, 721725. Beran, M.A., Gustard, A., 1977. A study into low ow characteristics of British rivers. J. Hydrol. 35, 147157. des formules courantes de Berkaloff, E., 1967. Limite de validite bit. Chronique dHydroge ol. 10, 3141. tarissement du de orie des eaux courantes du Boussinesq, J., 1877. Essai sur la the mouvement nonpermanent des eaux souterraines. Acad. Sci. Inst. Fr. 23, 252 260. coulement des Boussinesq, J., 1903. Sur un mode simple de ` lit horizontal, avec nappes deau dinltration a rebord vertical tout autour lorsquune partie de ce rebord e depuis la surface jusquau fond. C. R. Acad. Sci. est enleve 137, 5 11. oriques sur le coulement des Boussinesq, J., 1904. Recherches the es dans le sol et sur le de bit des sources. nappes deau inltre J. Math. Pure Appl. 10, 578. Brutsaert, W., 1994. The unit response of ground water outow from a hillslope. Water Resour. Res. 30, 27592763. Brutsaert, W., Nieber, J.L., 1977. Regionalized drought ow hydrographs from a mature glaciated plateau. Water Resour. Res. 34, 233240. Carlston, C.W., 1966. The effect of climate on drainage density and streamow. Bull. Int. Assoc. Sci. Hydrol. 11, 62 69. ` le tude des courbes de tarisseCastany, G., 1967. Introduction a ol. 10, 2330. ments. Chronique dHydroge Clark, W.E., 1956. Forecasting dry-weather ow of Pond Creek, Oklahoma, a progress report. Trans. Am. Geophys. Union 37, 442450. Comer, G.H., Zimmerman, R.C., 1969. Low ow and basin characteristics of two streams in Vermont. J. Hydrol. 7, 98108. te orologie et hydrologie. Etude ge ne rale Coutagne, A., 1948. Me bits et des facteurs qui les conditionnent. 2e ` me partie: des de bit en pe riode noninuence e par les les variations de de cipitations. Le de bit dinitration (corre lations uviales pre internes). La Houille Blanche (1948), 416 436.Septembre Octobre. ` le de de crue suivant la structure Dauty, J., 1967. Choix dun mode servoir simple sche matique. EDF Dir. Hydro. Rapport 44. dun re Dewandel, B., 2000. Groundwater potential of the Oman ophiolitic complex. Report no. 2University of Montpellier IIBRGM, Montpellier, France.. ologique Dewandel, B., 2002. Structure et fonctionnement hydroge ` re discontinu: lophiolite dOman. PhD Thesis, dun aquife University of Montpellier II, France. 640 p. Dewandel, B., Al-Khamisi, S., Lachassagne, P., Boudier, F., 2001. Hydrogeological properties of the Oman ophiolite. Abstr. Int. Conf. Geol. Oman, Muscat January, 2001. Dewandel, B., Boudier, F., Kern, H., Warsi, W.K., Mainprice, D., 2003. Seismic wave velocity and anisotropy of serpentinized peridotite in the Oman ophiolite. Tectonophysics In press. termination des composantes de Drogue, C., 1967. Essai de de coulement des sources karstiques. Evaluation de la capacite le

B. Dewandel et al. / Journal of Hydrology 274 (2003) 248269 coefcient demmagasinement par la courbe de tarissement dans ` res karstiques. Int. Assoc. Hydrogeol. 10 (1), 225230. les aquife Moore, G.K., 1992. Hydrograph analysis in a fractured rock terrane. Ground Water 30 (3), 390395. Nathan, R.J., McMahon, T.A., 1990. Evaluation of automated techniques for baseow and recession analysis. Water Resour. Res. 26, 14651473. Padilla, A., Pulido-Bosh, A., Mangin, A., 1994. Relative importance of baseow and quickow from hydrographs of karst spring. Ground Water 32, 267 277. Parker, G.W., 1977. Methods for determining selected ow characteristics for streams in Maine. US Geol. Surv. OpenFile Rep., 78871. Polubarinova-Kochina, P.Y., 1962. Theory of Groundwater Movement, Princeton University Press, New Jersy, p.355. me nie ras, G., 1960. Hydrologie de linge nieur. Collection du Re laboratoire national dhydraulique. Riggs, H.C., 1964. The baseow recession curve as an indicator of groundwater. Int. Assoc. Sci. Hydrol. 63, 352363. Samani, N., Ebrahimi, B., 1996. Analysis of spring hydrographs for hydrogeological evaluation of a karst aquifer system. Theor. Appl. Karstol. 9, 97 112. gime hydroge ologique des calcaires Schoeller, H., 1948. Le re oce ` nes du Synclinal du Dyr el Kef (Tunisie). Bull. Soc. Ge ol. e Fr. 5 (18), 167 180. Schoeller, H., 1967. Hydrodynamique dans le karst. Chronique ol. 10, 721. dhydroge Shevenell, L., 1996. Analysis of well hydrographs in a karst aquifer: estimates of specic yields and continuum transmissivities. J. Hydrol. 174, 331355. Skolov, D.S., 1965. Hydrodynamic zoning of karst waters. A.I.H.S. conference on Hydrology of ssured rocks, Dubrovnik, 1965. Ed: AIHS, Paris. ` le tude des courbes de re cession Soulios, G., 1991. Contribution a nique. J. Hydrol. des sources karstiques: exemples du pays Helle 124, 29 42. Stanger, G., 1987. The hydrogeology of the Oman mountains. PhD Thesis, 355 p. Sugiyama, H., 1996. Analysis and extraction of low ow recession characteristics. Water Resour. Bull. 32, 491 497.

269

Szilagyi, J., 1999. On the use of semi-logarithmic plots for base ow separation. Ground Water 37, 660 662. Szilagyi, J., Parlange, M.B., 1998. Baseow separation based on analytical solutions of the Boussinesq equation. J. Hydrol. 204, 251 260. Szilagyi, J., Parlange, M.B., 1999. A geomorphology-based semi-distributed watershed model. Water Resour. 23, 177 187. Szilagyi, J., Parlange, M.B., Albertson, J.D., 1998. Recession ow analysis for aquifer parameter determination. Water Resour. Res. 37, 18511857. Tallaksen, L.M., 1995. A review of baseow recession analysis. J. Hydrol. 165, 349370. Tasker, G.D., 1972. Estimating low ow characteristics of streams in southeastern Massachusets from maps of groundwater availability. US Geol. Survey Prof. Pap. 800, 217220. ry, D., 1993. Calage automatique des mode ` les hydrodynamiThie s. De termination des zones ge ographiques homoques maille ` nes et des parame ` tres optimaux associe s. Application a ` cinq ge ` mes aquife ` res. Rev. Hydroge ol. 4, 281 291. syste ry, D., 1994. Calibration of groundwater models by optimizThie ation of parameters in homogeneous geological zones. In: Hipel, K.W., (Ed.), Stochastic and Statistical Methods in Hydrology and Environmental Engineering, Kluwer, Dordrecht, pp. 281 292. Thomas, M.P., Cervione, M.A., 1970. A proposed streamow data program for Connecticut. Conn. Water Resour. Bull., 23. coulement et Tison, G., 1960. Courbe de tarissement, coefcient de abilite du bassin. Me m. IAHS, Helsinski, 229243. perme Troch, P., De Troch, F., Brusaert, W., 1993. Effective water table depth to describe initial conditions prior to storm rainfall in humid regions. Water Resour. Res. 29, 427 434. Vasileva, D., Komatina, M., 1997. A contribution of the alpha recession coefcient investigation in karts terrains. Theor. Appl. Karstol. 10, 45 54. Vogel, R.M., Kroll, C.N., 1992. Regional geohydrologic-geomorphic relationships for the estimation of low-ow statistics. Water Resour. Res. 28, 24512458. Werner, P.W., Sundquist, K.J., 1951. On the groundwater recession curve for large watersheds. IAHS Publ. 33, 202212.

Das könnte Ihnen auch gefallen