Improvement of rainfall-runoff modelling with distributed radar rainfall data: a case study in the Lez, French Mediterranean, catchment

The Mediterranean catchments in the south of France are prone to intense rainfall leading to destructive flash floods. These rainfalls mainly occur in autumn and show a high spatial variability. This study aims to assess the quality and impact in hydrological modelling of the radar rainfall data, in the Lez catchment (114 km(2)) near Montpellier, France. Comparison of both the raingauges and radar data proved to be satisfactory for events at the beginning of autumn. In contrast, important differences appeared for events occurring at the end of autumn. This can be explained by the weak vertical extension of the clouds and the low altitude of the 0 degrees C isotherm in this period, which could affect the accuracy of radar measurements due to the distance between the basin and the radar (similar to 60 km). To take advantage of the spatial variability of the radar rainfall data, the flood simulations were performed through a distributed event-based rainfall runoff model. The model was calibrated using a sample of 21 floods observed from 1994 to 2008 where both recording raingauge and radar rainfall data were available. When the radar rainfalls were reliable, they led to: (i) an improvement of the optimal flood simulation at the outlet, and (ii) an improvement of the relationship between the calibrated initial condition of the model and external predictors such as piezometric level, baseflow and Hu2 index from the Meteo-France SIM model. Installation of an X-band radar near the study area could improve rainfall estimation at the end of the autumn for the Lez catchment and the Montpellier agglomeration.