Abstract
In karst catchments where only sparse temporal monitoring is performed, it is generally difficult to correctly assess the overall hydrodynamics of the basin. However, sparse temporal spring-discharge data may contain information of major importance for the characterization of such catchments, especially when sparse spring-discharge data over a long period are available and combined with higher frequency discharge and/or piezometric-level data. This paper proposes a methodology for the characterization and hydrodynamic modeling of karst catchments by coupling sparse temporal data of discharge at a karstic spring over a 30-year measurement period, with higher frequency (i.e. hourly) data of hydrodynamic (piezometry, discharge), physicochemical (temperature, electrical conductivity) and meteorological data over a short monitoring period of 21 months. The study area is the Oeillal spring catchment, one of the main outlets of the Fontfroide-Montredon limestone aquifer located at the border of the Narbonne-Sigean sedimentary basin, southern France. The present study focuses on the use of numerical tools such as time-series analysis (recession analysis, auto-correlation and cross-correlation analysis) coupled with a lumped-parameter modeling approach, to assess the hydrodynamic behaviour of the karst system. The main results of the study highlight the necessity to couple the results from lumped-parameter rainfall-runoff modeling with results from high-resolution time-series analysis to evaluate the physical significance of the model, since classical numerical performance criteria such as the Nash-Sutcliff efficiency, Kling-Gupta efficiency and balance error, can be poorly estimated when only subsampled time series exist for model calibration.