Flash floods forecasting without rainfalls forecasts by recurrent neural networks. Case study on the Mialet basin (Southern France)

Abstract

The feasibility of flash flood prediction without rainfall forecasts nor previous discharge is considered. After a presentation of the important stakes involved in this task (23 fatalities in the Var event in June 2010, France) the important stage of variable and complexity selection is addressed for the small basin of Mialet (a part of the Gardon d'Anduze basin, in Southern France). Considering two architectures inspired from the multilayer perceptron, both designs and performances are presented and the model considering the linear and non-linear behaviors independently is proved to be the better. Generalization properties are assessed for four predictions up to two hours ahead thereby allowing an early warning of the population.

DOI: 10.1109/NaBIC.2011.6089612

11 Figures and Tables

Cite this paper

@article{Artigue2011FlashFF, title={Flash floods forecasting without rainfalls forecasts by recurrent neural networks. Case study on the Mialet basin (Southern France)}, author={G. Artigue and Anne Johannet and Valerie Borrell Estupina and Severin Pistre}, journal={2011 Third World Congress on Nature and Biologically Inspired Computing}, year={2011}, pages={303-310} }