Analyzing hydrological model versatility to simulate flash floods in Mediterranean catchments

Extreme flood events in the Mediterranean region are known for being natural hazards with heavy consequences for the population. They are often the result of climate and topographic characteristics of the region. Therefore, the way rainfall-runoff generation processes are represented in hydrological models is a key factor to simulate and forecast floods. This work presents how an adapted hydrological model (the GRSDi model; Peredo et al., 2021) is able to better simulate flash floods occurring after long dry periods, while also simulating well other floods occurring during or after wet periods in Mediterranean catchments in France (Aude, Gardon, Ardèche). The GRSDi model is the result of the adaptation of an existing semi-distributed hydrological model (the GRSD model). The adaptation consisted in introducing a dependency on the rainfall intensity in the production rate, which is calculated before the water flows into the model’s production store. A new parameter to be calibrated is introduced (parameter i₀). Differences in runoff production between the two models were quantified at each catchment over the modelling units of the semi-distributed modelling framework and along the hourly time steps of a 10-year study period (2008-2018). A visualization of model versatility is proposed to detect the time steps when the new model structure tends to behave more similarly or differently from the original model structure in terms of runoff production. The results show that the highest number of occurrences of high differences in runoff production between models is observed when modelling units have lower values of parameter i₀, at time steps with high rainfall intensity and low filling rate of the production reservoir. The modification introduced in the original model structure is activated when the model needs to adapt to specific conditions, illustrating the added-value of the versatility provided by the modified model structure, despite its simplicity.

 

Peredo, D., Ramos, M-H., Andréassian, V., Oudin, L. Investigating hydrological model versatility to simulate extreme flood events. Hydrological Sciences Journal, 2021, submitted.