Karst aquifers host significant water supplies but are easily contaminated because highly conductive inlets can rapidly transmit water to depth. The epikarst, which is the region of vegetation, soil, and weathered bedrock above karst aquifers, is a critical zone as it regulates the quantity and quality of recharge to the aquifer. As the epikarst exhibits complex heterogeneity, characterization at the field scale can be challenging. The objective of this thesis was to develop a model of epikarst hydrodynamics using longterm field measurements. To meet this objective, continuous hydrologic data of precipitation, speleothem drips, and an underground stream in James Cave in Pulaski County, VA, were collected to delineate seasonal recharge patterns, estimate effective recharge and catchment areas, characterize the number and permeability of flow paths, and evaluate storativity in the epikarst. Results demonstrate that after significant seasonal recharge, which occurs in the late winter and early spring, the epikarst can temporarily store a significant portion of recharge in low permeability flow paths. Effective recharge was estimated to be approximately 30% of total precipitation (2008-2009). Hydrograph recession analysis aided in delineation of flowpaths in the epikarst, including quickflow, moderate flow, and baseflow components. Hydrograph shape analysis suggests flow restrictions at two of the drip sites that may reveal spatial differences in storage capacity and retention time. Results of this work are intended to aid karst aquifer management by providing a multitechnique approach that can be used to assess seasonal patterns of recharge, quantify flowpath and storage characteristics, and delineate recharge zones. / Master of Science

Principalement associés à des milieux carbonatés, les aquifères karstiques présentent des écoulements souterrains rapides via des réseaux de fractures ou de conduits. Ces écoulements rapides sont à l’origine de la vulnérabilité de ces milieux dans le cas de la propagation de contaminants. Dans le cadre de la maîtrise environnementale, l’enjeu de ce travail de thèse est d’améliorer la compréhension du fonctionnement hydrodynamique d’un aquifère karstique étudié par le CEA. L’objectif est d’évaluer les contributions respectives de chacune des deux porosités aux écoulements souterrains. Trois approches ont été menées afin d’atteindre cet objectif : (1) le traitement des données hydrodynamiques et physico-chimiques par analyses corrélatoires, (2) la modélisation « double milieu » des variations du niveau de nappe, et enfin (3) l’interprétation de traçages artificiels pour la modélisation double milieu du transport. Cette thèse a permis de mettre en évidence la présence de deux milieux au sein de l’aquifère karstique étudié, avec : une porosité primaire (porosité de matrice) supportant les écoulements lents et la capacité de stockage du milieu souterrain, et une porosité secondaire (porosité de fissures/fractures ou de matrice fissurée) responsable des écoulements rapides et de la fonction transmissive de l’aquifère. Ce caractère « double milieu » est interprété comme la présence d’une fissuration (ou fracturation) spatialement étendue à l’échelle de la surface du site d’étude (17 km²). Ce résultat démontre qu’une approche « double milieu » est adaptée à la modélisation des écoulements souterrains et du transport (avec un milieu matriciel et un milieu fissuré/fracturé). Cette approche « double milieu » a ainsi permis de reproduire les chroniques piézométriques marquées par des récessions à double pente, ainsi que les courbes de traçages marquées par un retard de la restitution du traceur, et ce grâce à la contribution du milieu matriciel. De plus, les paramètres ajustés de ce modèle double milieu témoignent d’une densité de fissuration/fracturation spatialement variable au sein du milieu souterrain. Enfin, ces travaux apportent une quantification des flux d’eau échangés entre le milieu matriciel et le milieu fissuré/fracturé, 38 % de l’infiltration est stockée dans la matrice à l’échelle d’un évènement pluvieux. Cette approche « double milieu » montre également la contribution significative du processus de diffusion aux transferts de soluté vers le milieu matriciel lors de la propagation du traceur au sein du milieu fissuré/fracturé. La diffusion de soluté depuis le milieu matriciel, permet ensuite de reproduire la restitution tardive du traceur. / Karst aquifers are characterized by complex flow mechanisms induced by multiple porosities (matrix, fracture, conduit) which are responsible of the karst aquifer vulnerability to pollutant infiltrations. In the context of the environmental management, this thesis aims at enhancing the hydrodynamic functioning understanding of a karst aquifer. More specifically, the objective is the assessment of the multiple porosities contributions to the groundwater flow. Three approaches have been led during this work: (1) signal processing of the hydrodynamic and physical-chemical data by correlation analysis, (2) double medium modelling for reproducing the groundwater level (GWL) variations, and (3) double medium transport modelling for interpreting tracer test data. The double medium behaviour of the studied karst aquifer has been highlighted in this thesis with: a primary porosity (matrix porosity) supporting slow flows and representing the storage function and a secondary porosity (porosity of a fissured or a fractured matrix) responsible of the rapid flows and the transmissive function of the aquifer. This double medium behaviour has been interpreted as a spatially extended fracturation at the scale of the studied site (17 km²). This result makes a double medium approach relevant for simulating groundwater flows and solute transport (considering a matrix and a fractured medium). This double medium approach has succeeded in reproducing groundwater level variations marked by double-slope recessions and tracer breakthrough curves marked by a strong tailing, thanks to the matrix medium contribution. Moreover, the adjusted parameters of the double medium model indicate spatial variations of the underground media fissure/fracture density. Finally, this work brings a quantification of the exchanged flow between the two media during a recharge event (38 % of the infiltration is stored in the matrix medium), and shows the significant contribution of diffusion process to solute transfers in the matrix medium during the tracer propagation in the fractured medium. Then, the solute diffusion from the matrix allows reproducing the tracer tailing observed during tracer tests.

Understanding the potential for karst aquifer contamination by sediment-sorbed pesticides is important for cave conservation efforts in agricultural landscapes. Flow rate, water quality parameters and suspended sediment concentrations were measured in Logsdon River, a ~10km karst conduit within the Turnhole Spring Groundwater Basin of Mammoth Cave National Park to determine characteristics of storm-period transport of sediment-sorbed atrazine through a conduit-flow karst aquifer. Analysis of two independent precipitation events occurring in the Spring of 2008 from May 2-4 and May 27-29 demonstrated the rapid response of the Logsdon River to precipitation events with detections of atrazine increasing during the initial turbidity peak and decline in spC, indicating that the atrazine arrives with the initial flush of surface waters that enters the conduit. Distinct peaks of atrazine did not coincide with fine grained (silt and clay-sized) sediment peaks and concentrations of atrazine remained elevated on the falling limb of the hydrograph as turbidity declined. In addition, no systematic relation between filtered and unfiltered samples was evident. There was also exceedingly weak correlation between the concentration of atrazine and suspended sediment, suggesting that if atrazine is sorbed to fine sediment particles this sorption involves only the fractions finer than 0.22 μm.

Karst aquifers are characterized by highly conductive conduit flow paths embedded in a less conductive fissured and fractured matrix resulting in strong permeability contrasts with structured heterogeneity and anisotropy. Groundwater storage occurs predominantly in the fissured matrix. Hence, most karst models assume quasi steady-state flow in conduits neglecting conduit associated drainable storage (CADS). The concept of CADS considers storage volumes, where karst water is not part of the active flow system but rather hydraulically connected to conduits (for example karstic voids and large fractures). The disregard of conduit storage can be inappropriate when direct water abstraction from karst conduits occurs, e.g. large scale pumping. In such cases, CADS may be relevant. Furthermore, the typical fixed head boundary condition at the karst outlet can be inadequate for water abstraction scenarios because unhampered water inflow is possible. The objective of this paper is to analyze the significance of CADS and flow-limited boundary conditions on the hydraulic behavior of karst aquifers in water abstraction scenarios. To this end, the numerical hybrid model MODFLOW-2005 Conduit Flow Process Mode 1 (CFPM1) is enhanced to account for CADS. Additionally, a fixed-head limited-flow (FHLQ) boundary condition is added that limits inflow from constant head boundaries to a user-defined threshold. The affect and proper functioning of these modifications is demonstrated by simplified model studies. Both enhancements, CAD storage and the FHLQ boundary, are shown to be useful for water abstraction scenarios within karst aquifers. An idealized representation of a large-scale pumping test in a karst conduit is used to demonstrate that the enhanced CFPM1 is potentially able to adequately represent water abstraction processes in both the conduits and the matrix of real karst systems.

Groundwater vulnerability maps are important decision support tools for water resource protection against pollution and helpful in minimizing environmental damage. However, these tools carry a high subjectivity along the multiple steps taken for the development of such maps. Additionally, the theoretical model on which they are based does not consider other important parameters, such as pollutant concentration or pollutant residence time in a given section of the aquifer, solely focusing on the theoretical travel time of a pollutant particle from a release point towards a target. In this work, an integrated methodology for the evaluation of potential (intrinsic) and actual vulnerability is presented. This integrated method, named Integrated Karst Aquifer Vulnerability (IKAV), was developed after the analysis of several study cases around the world and the application of multiple intrinsic groundwater vulnerability methods in a selected study area. Also, a solute transport model served as the basis to define additional parameters for vulnerability analysis for areas severely affected by anthropogenic practices. However, the focus of the transport model must not be mistaken to be hazards and risk mapping. A defined workflow and several criteria for parameters and attributes selection, rating and weighting, and vulnerability classification are presented here. The first application of the IKAV method was carried out in the Yucatán karst, demonstrating to be a reliable method for vulnerability estimation. Results demonstrated the scope of the IKAV method in highlighting important regional conditions, minimizing the subjectivity, and expanding the analysis of vulnerability.

La côte ouest irlandaise située dans les comtés de Clare et Galway est depuis récemment affectée par une augmentation de la fréquence des inondations et une montée du niveau marin. L'impact de ces changements climatiques se fait particulièrement ressentir au niveau des aquifères karstiques côtiers de ces régions, tel que le bassin karstique de Bell Harbour (~50 km²) sélectionné dans le cadre de ce projet: le positionnement d'un biseau salé dans cet aquifère est clairement dû à l'interaction entre sa recharge liée aux événements pluvieux et la forte influence de la marée. Des données (niveaux d'eau et conductivités spécifiques) ont été collectées à l'aide de sondes, à deux sources côtières, six forages, trois lacs et au milieu de la baie durant des périodes prolongées. Trois types d'environnement hydrodynamiques (conduits, fissures et matrices) ont été définis pour chacun des six forages en comparant deux méthodes qui se sont avérées cohérentes entre-elles. Les variations des conductivités spécifiques combinées à la chimie des eaux des six forages et de lacs ont permis d'évaluer l'étendue du biseau salé dans l'aquifère en fonction de la recharge de l'aquifère et des cycles de marée haute/basse et vives eaux/basses eaux: l'étendue du biseau salé dépend des propriétés hydrodynamiques de l'aquifère mais aussi de l'influence relative de la recharge et de la marée sur les niveaux piézométriques, induisant chacun des comportements opposés. Ainsi, la compétition entre la recharge et la marée contrôle l'intrusion d'eau de mer dans l'aquifère, ce qui explique les changements spatio-temporels du biseau salé dans l'aquifère. Les fortes amplitudes de marée semblent être le moteur des brèves intrusions salines dans l'aquifère proche de la rive alors que la position du niveau piézométrique semble influencer l'intensité de cette augmentation de la salinité. La recharge karstique de la région est suffisamment importante tout au long de l'année pour que le biseau salé ne s'étende pas plus loin qu'un kilomètre de la rive. La diminution des précipitations durant l'été (~10 %) ainsi que l'augmentation du niveau marin prévues pour les prochaines décennies, contribueront certainement à une intrusion plus importante du biseau salé dans l'aquifère. / Low-lying coastal areas in the west of Ireland, such as the Galway and Clare coasts, have seen recent increases in flooding frequencies coupled with overall increases in sea level. The impacts of these changes are most strongly felt in coastal karst catchments and this study focuses on one such area, the Bell Harbour catchment (~50 km²), where there is a clear interaction between rainfall inputs and tidal influences, to create a terrestrial salt water wedge. Data (specific conductivity and water levels) have been collected at two coastal springs, six boreholes, three lakes, and from discrete locations in the middle of the bay, using dedicated loggers over extended periods. Two approaches that provided consistent results were used to explain the different hydrodynamic behaviours identified in the boreholes (conduits, fissures and matrix flows). Specific conductivity variations and water chemistry measured inland allowed for assessment of the spatial extent of the saltwater wedge into the aquifer as a function of both karst recharge and tidal movements at high/low and neap/spring tidal cycles. The extent of the saltwater wedge depends on the intrinsic properties of the aquifer but also on the relative influence of the recharge and the tide on groundwater levels, which induce opposite behaviours. This dynamic between recharge and the tide thus controls the seawater inputs, hence explaining temporal and spatial changes in the saltwater wedge in this coastal karst aquifer. Strong tidal amplitudes seems to be the motor of sudden saltwater intrusion observed in the aquifer near the shore while the relative elevation of the groundwater appears to influence the intensity of the salinity increase. The magnitude of annual recharge in the area is high enough to limit saltwater intrusion to no more than about one kilometre inland from the shore. Given the anticipated decreases in summer precipitation (~10 percent), coupled with anticipated sea level rises, the extent of the saltwater wedge into the aquifer is likely to increase in coming decades.

Les aquifères karstiques sont parmi les milieux souterrains les plus difficiles à caractériser. Ils sont connus pour avoir des réponses complexes aux événements pluvieux en raison de l’hétérogénéité de la porosité allant de la porosité matricielle (micro-porosité) à la porosité de conduits (macro-porosité). La porosité de conduits génère des vitesses de transfert élevées au sein de l’aquifère, à l'origine de la vulnérabilité des aquifères karstiques. L'objectif de cette thèse est de définir les modalités du transfert de l'eau et des solutés au sein d'un aquifère karstique du Dogger situé en Bourgogne. Deux approches complémentaires ont été utilisées: (i) l'analyse de chroniques piézométriques et débitmétriques à partir d'outils de traitement du signal et (ii) une caractérisation géochimique et isotopique des eaux souterraines basée sur l'utilisation d'une large palette de traceurs, notamment des traceurs de datation. Ces deux approches ont été appliquées à différentes échelles temporelles: un suivi haute fréquence (jusqu'à 1h) lors de pics de crue pour étudier les transferts rapides et un suivi à long terme (pluri-annuel) pour caractériser les transferts lents et étudier la variabilité saisonnière des modalités d'écoulement. Ce travail a permis de mettre en évidence une très large distribution des temps de résidence de l’eau souterraine, de quelques heures, lors d’événements de crue souligné par des temps de réponse pluie-niveau piézométrique faible (inférieurs à 70 h) à quelques années pour l’écoulement de base mis en évidence par l’utilisation de traceurs de datation (39Kr, CFC-12 et SF6). En réponse aux précipitations, différentes masses d'eau ont pu être décrites: (i) une eau à vitesse d’écoulement lente, qui correspond à l’écoulement de base mis en évidence par l'utilisation des traceurs de datation; (ii) une eau de recharge rapide dont le temps de résidence au sein de l'aquifère varie, selon les porosités empruntées, de quelques heures à quelques jours soulignés grâce au suivi en continu de la conductivité électrique, des éléments majeurs et du tritium; et (iii) des eaux stockées au sein de la zone non saturée, remobilisées lors de pics de crue mises en évidence par la variation des compositions géochimiques en tritium et Ca2+ des eaux souterraines lors de pics de crue. Malgré une forte hétérogénéité des processus d'écoulement, une logique spatiale des écoulements rapides apparaît. Le temps de réponse du niveau piézométrique à un événement pluvieux et le temps de résidence augmentent vers l’aval du dôme hydrogéologique. L'étude temporelle de la réponse impulsionnelle par la méthode du corrélogramme croisé coulissant montre que le temps de réponse varie de façon saisonnière, et est plus court en été. La variation du temps de réponse pour un forage entre été et hiver est significative et peut atteindre une centaine d’heures. Cette variabilité temporelle s'explique en partie par une variabilité de l'intensité des pluies (plus importantes en été) qui induit une variation des mécanismes d'écoulement au sein de l'épikarst. Lors d'événements de pluie intense, la saturation de l'épikarst est plus importante provocant des transferts latéraux en son sein et permettant de transférer les eaux vers les conduits les plus ouverts. / Karst aquifers are among the most difficult to characterize. They are known for their complex hydrological responses to rainfall events due to the high heterogeneity of their porosity, going from matrix conduit porosity. The porosity of conduit induces a high transfer velocity in the aquifer which is the cause of the karst aquifer vulnerability. The objective of this study is to define the modes of water transfer and solute in a karst aquifer of Dogger in Burgundy. Two complementary methods are used: (i) the analyse of piezometric and debitmetric time series, and (ii) the geochemical and isotopic characterization of groundwater based on the using of large panel of tracers, notably some dating tracers. These two methods are applied at different temporal scales: a high frequency monitoring (1 hour time step) during discharge event to study fast water transfer and a long time monitoring to characterize slow water transfer and study the seasonal variability of water transfer processes. This study highlights a large distribution of residence time of groundwater, from several hours during a recharge event, underlined by short response time rainfall-piezometric level (less to 70 hours) to several years for the baseflow showed by the using of dating tracers (39Kr, CFC-12 et SF6). In response to rainfall events, several water bodies are described. The first one, with slow transfer velocity, corresponds to the baseflow and is showed by dating tracers. The second one, with a direct recharge and a residence time in the karst aquifer that varies with the porosity from several hours to several days, is described by a continous monitoring of electrical conductivity, majors elements and tritium. Finally, the water stored in the unsaturated zone, remobilized during recharge events, is revealed by the variation of geochemical composition in tritium and Ca2+ in groundwater during recharge events. Despite of the high heterogeneity of water transfer processes, the fast water transfer is spatially organized. The response time of piezometric level to rainfall events and the residence time increase toward the hydrogeological downhill. The temporal study of the impulse response using the method of the sliding-window cross-correlation shows that the response time varies seasonally, being shorter during the summer. The variability of the response time is significant, up to a hundred hours. This temporal variability could be explained in part by a variability of the rainfall intensity which induces a variability of water flow processes in the epikarst. During intensive rainfall, the saturation of the epikarst is higher which induces lateral transfers and allows transferring water toward large conduits.

Groundwater vulnerability maps are useful for environmental planning and decision-making. They are usually produced by applying vulnerability assessment methods using overlay and index techniques. On the basis of a review of the vulnerability assessment and mapping methods, new research challenges in aquifers vulnerability assessment are identified. Operations like the parameter quantification, the vulnerability index computing, and the final classification, are affected by an empirical character which of course affects also the final product: the vulnerability map. In consequence, the validity of the resulted vulnerability maps must be evaluated in function of the objectives of the survey and in function of the specific characteristics of each studied zone. Analysing their uncertainty can represent the base for their validation. Uncertainty can be investigated through sensitivity analysis or through comparisons between vulnerability maps created using different methods. Both these strategies are developed in this study and illustrated from applications on practical case studies of vulnerability mapping. Applying the EPIK parametric method, a vulnerability assessment has been made for a small karstic groundwater system in southern Belgium. The aquifer consists in a karstified limestone of Devonian age. A map of intrinsic vulnerability of the aquifer shows three vulnerability areas. A parameter-balance study and a sensitivity analysis were performed to evaluate the influence of single parameters on aquifer vulnerability assessment using the EPIK method. This approach provides a methodology for the evaluation of vulnerability mapping and for more reliable interpretation of vulnerability indices for karst groundwater resources. Five different methods for assessing the intrinsic vulnerability were tested on a case study for comparison of their results. The test area consists in a slightly karsified basin located in the Condroz region (Belgium). The basin covers about 65 km² and the karstic aquifer provides a daily water supply of about 28000 m³ in drainage galleries. Several campaigns of measurements consisting in morpho-structural observations, shallow geophysics, pumping and tracer tests have provided useful data. The tested methods were: EPIK (Doerfliger and Zwahlen, 1997), DRASTIC (Aller et al., 1987), German methods (von Hoyer & Söfner, 1998), GOD (Foster, 1987), and ISIS (Civita and De Regibus, 1995). DRASTIC and GOD represent classic approaches in vulnerability assessment. ISIS is a development based on DRASTIC, SINTACS (Civita, 1994), and GOD methods, where more importance is given to the recharge. EPIK was developed specifically for karstic geological contexts and the German methods was developed in Germany for a broad range of geological contexts. Compared results are shown and commented. It seems that despite the fact that the EPIK method can better outline the karstic features about 92% of the studied area is assessed by this technique as low vulnerable. In contrast, the other four methods are considering extended zones of high or moderate vulnerability. From the analysis, it seems also that reducing the number of considered parameters is not ideal when adaptation to various geological contexts is needed. Reliability and validity of groundwater analysis strongly depend on the availability of large volumes of high quality data. Putting all data in a coherent and logical structure supported by a computing environment helps ensure a validity and availability, and provides a powerful tool for hydrogeological studies. A hydrogeological GIS database that offers facilities for groundwater vulnerability analysis and hydrogeological modelling has been designed in Belgium, for the Walloon Region. Data from five river basins, chosen for their contrasted applications that have been developed allow now further advances. However the basic concept of the database is represented by the commonly accepted Georelational model developed in the 1970s, the database concept presents a distinctive character. There is a growing interest in the potential for integrating GIS technology and groundwater simulation models. Between the mentioned spatial database schema and the groundwater numerical model interface GMS (Groundwater Modelling System) a loose-coupling tool was created. Following time and spatial queries, the hydrogeological data stored in the database can be easily used within different groundwater numerical models. This development can represent also a solid base for the physical processes integration within the quantification of the vulnerability methods parameters. The fundamental aim of this work was to help improving the aquifers protection strategy using vulnerability mapping and GIS. The results are offering the theoretical and practical basis for developing a strategy for protecting the groundwater resources.

Karst aquifers are characterized by highly conductive conduit flow paths embedded in a less conductive fissured and fractured matrix resulting in strong permeability contrasts with structured heterogeneity and anisotropy. Groundwater storage occurs predominantly in the fissured matrix. Hence, most karst models assume quasi steady-state flow in conduits neglecting conduit associated drainable storage (CADS). The concept of CADS considers storage volumes, where karst water is not part of the active flow system but rather hydraulically connected to conduits (for example karstic voids and large fractures). The disregard of conduit storage can be inappropriate when direct water abstraction from karst conduits occurs, e.g. large scale pumping. In such cases, CADS may be relevant. Furthermore, the typical fixed head boundary condition at the karst outlet can be inadequate for water abstraction scenarios because unhampered water inflow is possible. The objective of this paper is to analyze the significance of CADS and flow-limited boundary conditions on the hydraulic behavior of karst aquifers in water abstraction scenarios. To this end, the numerical hybrid model MODFLOW-2005 Conduit Flow Process Mode 1 (CFPM1) is enhanced to account for CADS. Additionally, a fixed-head limited-flow (FHLQ) boundary condition is added that limits inflow from constant head boundaries to a user-defined threshold. The affect and proper functioning of these modifications is demonstrated by simplified model studies. Both enhancements, CAD storage and the FHLQ boundary, are shown to be useful for water abstraction scenarios within karst aquifers. An idealized representation of a large-scale pumping test in a karst conduit is used to demonstrate that the enhanced CFPM1 is potentially able to adequately represent water abstraction processes in both the conduits and the matrix of real karst systems.

Ce manuscrit de thèse présente une nouvelle approche pour caractériser qualitativement et quantitativement la localisation et les propriétés des structures dans un aquifère fracturé et karstique à l’échelle décamétrique. Cette approche est basée sur une tomographie hydraulique menée à partir de réponses à une investigation de pompages et interprétée avec des méthodes d’inversions adaptées à la complexité des systèmes karstiques. L’approche est appliquée sur un site karstique d’étude expérimental en France, une première fois avec des signaux de pompage constants, et une deuxième fois avec des signaux de pompage harmoniques. Dans les deux cas, l’investigation a fourni des réponses de niveaux d’eau de nappe mesurés pendant des pompages alternés à différentes positions. L’interprétation quantitative de ces jeux de réponses consiste à les reproduire par un modèle avec un champ de propriété réaliste adéquat généré par inversion. Les méthodes d’inversions proposées dans ce manuscrit permettent de reconstruire un champ de propriétés hydrauliques réaliste en représentant les structures karstiques soit par un réseau généré par automates cellulaires, soit par un réseau discrétisé. Les résultats d’interprétations obtenus sur le site d’étude expérimental permettent d’imager les structures karstiques sur une carte et de « lire » leur localisation. De plus, les résultats obtenus avec les réponses à des pompages harmoniques tendent à montrer le rôle de la fréquence du signal sur les informations portées par les réponses. En effet, les fréquences plus élevées caractérisent mieux les structures les plus conductrices, alors que les fréquences plus faibles mobilisent des écoulements également dans des structures karstiques moins conductrices. / This thesis manuscript presents a novel approach to characterize qualitatively and quantitatively the structures localization and properties in a fractured and karstic aquifer at a decametric scale. This approach relies on a hydraulic tomography led from responses to a pumping investigation and interpreted with inversion methods adapted to the complexity of karstic systems. The approach is applied on a karstic experimental study site in France, a first time with constant pumping signals, and a second time with harmonic pumping signals. In both applications, the investigation resulted in groundwater level responses measured during alternated pumping tests at different locations. The quantitative interpretation of these sets of responses consists in reproducing these responses through a model with an adequate realistic property field generated by inversion. The inversion methods proposed in this manuscript permit to reconstruct a realistic hydraulic property field by representing the karstic structures either through a network generated by cellular automata, or through a discretized network. The interpretation results obtained on the experimental study site permit to image the karstic structures on a map and to‘read’ their localization. Furthermore, the results obtained with the responses to harmonic pumping tests tend to show the role of the signal frequency on the information carried by the responses. In fact, higher frequencies better characterize the most conductive structures, while lower frequencies mobilize flows also in less conductive karstic structures.