Building hazard maps of extreme daily rainy events from PDF ensemble, via REA method, on Senegal River Basin

The Sudano-Sahelian zone of West Africa, one of the poorest of the Earth, is characterized by high rainfall variability and rapid population growth. In this region, heavy storm events frequently cause extensive damage. Nonetheless, the projections for change in extreme rainfall values have shown a great divergence between Regional Climate Models (RCM), increasing the forecast uncertainty. Novel methodologies should be applied, taking into account both the variability provided by different RCMs, as well as the non-stationary nature of time series for the building of hazard maps of extreme rainfall events. The present work focuses in the probability density functions (PDFs)-based evaluation and a simple quantitative measure of how well each RCM considered can capture the observed annual maximum daily rainfall (AMDR) series on the Senegal River basin. Since meaningful trends have been detected in historical rainfall time series for the region, non-stationary probabilistic models were used to fit the PDF parameters to the AMDR time series. In the development of PDF ensemble by bootstrapping techniques, Reliability Ensemble Averaging (REA) maps were applied to score the RCMs. The REA factors were computed using a metric to evaluate the agreement between observed -or best estimated- PDFs, and that simulated with each RCM. The assessment of plausible regional trends associated to the return period, from the hazard maps of AMDR, showed a general rise, owing to an increase in the mean and the variability of extreme precipitation. These spatial-temporal distributions could be considered by local stakeholders in such a way as to reach a better balance between mitigation and adaptation.