Flash flood forecasting in poorly gauged basins using neural networks: case study of the Gardon de Mialet basin (southern France)

A groundbreaking approach using neural networks demonstrates the feasibility of accurate flash flood forecasting in poorly gauged basins, such as the Gardon de Mialet basin in southern France. This method achieves reliable predictions up to two hours without relying on rainfall forecasts, leveraging existing discharge measurements from a feedforward model. Even when discharge gauges fail, a recurrent neural network model can maintain functionality. The study also reveals that these models are robust against uncertainties in rainfall data, with ensemble forecasting not significantly impacting accuracy. This innovation offers promising solutions for flood prediction in ungauged or poorly monitored basins, addressing critical gaps in our understanding of rainfall-discharge relationships in heterogeneous and steeply-sloped landscapes.

Artigue, G., Johannet, A., Borrell, V., & Pistre, S. (2012). Flash flood forecasting in poorly gauged basins using neural networks: case study of the Gardon de Mialet basin (southern France). Natural Hazards and Earth System Sciences, 12(11), 3307–3324. https://doi.org/10.5194/nhess-12-3307-2012