Potential of satellite rainfall products to predict Niger River flood events in Niamey

International audienceA dramatic increase in the frequency and intensity of flood events in the city of Niamey, Niger, has been observed in the last decade. The Niger River exhibits a double outflow peak in Niamey. The first peak, is due to the rainfall occurring within about 500 km of Niamey. It has reached high values in recent years and caused four drastic flood events since 2000. This paper analyses the potential of satellite rainfall products combined with hydrological modelling to monitor these floods. The study focuses on the 125,000 km2 area in the vicinity of Niamey, where local runoff supplies the first flood. Six rainfall products are tested : a gauge only product — the Climate Prediction Centre (CPC); two gauge adjusted satellite products — the Tropical Rainfall Measurement Mission (TRMM) Multi-Platform Analysis (TMPA 3B42v7) and the CPC regional product African Rainfall Estimate (RFE version 2); and three satellite only products, 3B42RT, the CPC Morphing method (CMORPH) and the Precipitation Estimation from Remotely Sensed Information using Artificial Neural Network (PERSIANN). The products are first inter-compared over the region of interest. Differences in terms of rainfall amount, number of rainy days, spacial extension of the rainfall events and frequency distribution of the rain rates are highlighted. The satellite only products provide more rain than the gauge adjusted ones. The hydrological model ISBA-TRIP is forced with the six products and the simulated discharge is analysed and compared with the discharge observed in Niamey over the period 2000 to 2013. The simulations based on the satellite only rainfall produce an excess in the discharge. For flood prediction, the problem can be overcome by a prior adjustment of the products – as done here with probability matching – or by analysing the simulated discharge in terms of percentile or anomaly. All tested products exhibit some skills in detecting the relatively heavy rainfall that preceded the flood and in predicting that the 95th percentile of the discharge (i.e., the flood alert level in Niamey) will be exceeded. These skills are however variable among products and the best overall results are obtained with the TMPA 3B42 products