Résumé
Our work deals with the simulation of karstic spring hydrographs in order to: (i) test the likelihood of the conceptual model of functioning inferred from hydrodynamic and chemical data analysis; (ii) evaluate the water storage in the karstic system; (iii) predict the discharge and quality variations according to rainfall events. Our approach shows that, although each karst aquifer is characterized by a complex structural heterogeneity, it is possible to simulate its functioning with a common simple reservoir model. The daily time step model is based on three reservoirs: one for the production and two for the transfer function (for slow flow and rapid discharge respectively). The model has three fitted parameters and two parameters assessed from the hydrograph analysis (the recession coefficient of the slow discharge reservoir and the initial level in the production reservoir). The three linked reservoirs feed and empty according to simple discharge equations, transforming rainfall to flow rate and keeping in memory the previous water height in each reservoir. We applied this model to three Mediterranean karstic systems: Fontaine de Vaucluse (catchment area of 1100 km (super 2) ), Durzon (Larzac Plateau, 110 km (super 2) ) and Lez springs (Montpellier City, 380 km (super 2) ), all located in the south of France. This study shows that as long as the rainfall time series is still stationary, the reservoir models are particularly well suited to simulate both high and low flow rates of karstic spring hydrographs at the daily scale. It shows also that the basic structure of the model can be easily adapted to the specificity of each system by adding non-linearity like thresholds in the reservoirs. With its small number of fitted parameters, this approach appears as a new tool to forecast and to improve the sustainable management of the karst groundwater resources.