Potential for seasonal flood forecasting in West Africa using climate indexes

Abstract Floods are among the most devastating natural disasters and are expected to become more severe with changing climate and population growth. Flood forecasting is one of the key components of flood risk reduction. The potential for seasonal flood forecasting through climate indexes has not been studied for West Africa so far. This work investigates how climate indicators can be used to predict in advance, one to several months ahead of the flood season, above or below normal flood discharge in West Africa. Six global and regional climate indexes were screened for their potential to predict flood discharge of 56 river gauging stations across West Africa. Forecasting models are developed, based on simple and multiple linear regressions between climate indexes and annual maximum discharge, and evaluated using the relative operating characteristics and the relative operating levels scores. The western dipole mode index is the most skillful individual climate index for above normal flood prediction. Combining climate indexes via multiple linear regressions outperforms individual climate indexes for both above and below normal flood prediction. The models show forecasting skills for up to 4 months prior to the flood season. Hence, this study opens promising possibilities for seasonal flood forecasting in West Africa. This may help alert disaster reduction agencies of entering a period of an increased chance of flooding and may trigger adequate mitigation measures.