Floods forecasting on karstic basin – Agly river, south of France

Flood forecasting is particularly complicated in karstic Mediterranean area. Forecasters are faced to complex relationships between surface and ground water. During low-flow periods, when water table is low, karst can store a significant quantity of water in its vadose zone. Conversely, during periods of high water, its storage capacity is limited, leading to an increase in runoff due to a refusal to infiltrate and also to rapid underground transfers that can amplify floods downstream of karst basins (Fleury et al., 2010 ; 2013).

Agly river is located in the department of Pyrénées-Orientales in the South of France, its catchment is over 1000 km². It flows partly over the karst terrains of the Corbières. The hydrogeological functioning of this system has been studied by the BRGM (Dorfliger, 2008). This river is known for its flash floods which can cause significant damages to the population. This river is monitored by the SPC (Service de Prévision des Crues, the French flood forecasting office), which undergoes forecasting difficulties. In this context, BRGM carried out a study in order to improve flood process understanding and to make a simple and robust forecasting tool on the area.

These studies made it possible to identify an observation well in the karst that is representative of the saturation level of the karst. A semi-distributed numerical model including five sub-catchments was used to simulate the water level in the karst as well as the flow at the outlet of each of these sub-catchments.

At the end of this modelling work, the model was used in forecasting mode using numerous climatic scenarios. A double abacus type tool was built. The first allows the characterisation of the storage capacity of the karst from the initial piezometric level measured, the second provides the response in terms of flow following a rainfall episode (total rainfall and intensity). The tool has been used for two years in real time by warning managers during heavy rainfall events. Results in terms of vigilance (green, orange, red) were satisfactory.