Frequency analysis of extreme flows using an Artificial Neural Network (ANN) model case Western High Atlas - Morocco

The definition of flood protection plans in semi-arid mountain environments requires awareness of the hazard, its assessment and the analysis of its probabilities of occurrence. The study carried out made it possible to highlight the use of artificial neural networks (ANN) in the frequency analysis of extreme flows. This is made possible by a comparative approach between 14 Probability Distributions and the ANN model. The results of the frequency analysis according to the two models shows a predictive priority of the ANN with an average performance criterion of Nash that exceeds 0.9 in the different basins studied. The study reveals that the specificities of the use of ANN compared to the deterministic models in the frequency analysis is that ANN highlights the sensitivity of the extreme river flood to the physiographic variations of each watershed. This feature allows to include the predictions of extreme events in the different basins studied into a single model and makes them more accurate.