Flood Risk Reduction through the Assessment of Flood Forecasting System in the Upper Niger River Basin in Mali.

Abstract

This work was based on the issue of flood risk reduction through the assessment of a flood forecasting system. First of all, an overview of the flood forecasting system already implemented in Africa, especially in West Africa is carried out. Secondly, the trend analysis of flood events as well as their relationship with extreme rainfall is performed in Bamako. Thirdly, the hydrological models are calibrated and validated using the observed rainfall data. And lastly, the flood forecasting system is implemented in order to mitigate the coming floods in the study area. It allows the population to take necessary advances against floods, reducing considerably the flood impacts. HBV, HEC-HMS, SWAT, WRF-Hydro are the hydrological models the most used for the flood forecasting issues in Africa. From 1982 to 2019, the flood events historical trend in Bamako has increased as well as the flood victims. The extreme rainfall indices (RX1 day, RX5 day, R99P, PRCPTOT, CWD) show the same trend increasing during the same period 1982 to 2019. Therefore, a strong relationship between flood occurrences and extreme rainfall is identified. Most of the years where floods occurred in Bamako correspond to the years of high rainfall value, especially for the R99P. However, it has been observed that the rainfall responsible for flood in Bamako is approximatively 47mm, which is far from being exceptional. It is in the range of normal to severely abnormal classes; meaning that there are other causes that should be added to the extreme rainfall leading to flood events. Non-maintenance of channels, building houses on the way of the river-bed in addition to the soil destruction, could be considered as the major causes to be linked to the extreme rainfall. The flood forecasting system was explored in the study area through two hydrological models HBV and HEC-HMS. Both of the models are well calibrated and validated with satisfactory model efficiency values for the four days ahead NCEP rainfall data. Nash values are respectively 0.74, 0.76, 0.71, 0.62 for the 24h, 48h, 72h, and 96h ahead precipitation during the calibration and 0.82, 0.84, 0.77 and 0.76 during the validation for the HEC-HMS model respectively for the 24h, 48h, 72h and 96h ahead precipitation. For the HBV hydrological model, Nash values for the calibration are respectively 0.67, 0.86, 0.83, 0.80 for 24h, 48h, 72h and, 96h. During the validation, Nash values are 0.78, 0.78, 0.83, 0.80 for respectively 24h, 48h, 72h and 96h ahead precipitation. This interesting finding confirms the suitability of these models to be used for the flood forecasting system in the area. PoD and FAR indices calculated, gave satisfactory results of NCEP center data, and more success than false alarm rate was obtained. The knowledge of the forecasted rainfall can allow a better management of the Selingue dam, reducing floods impact and producing sufficient energy for the population.