Statistical modeling of the flood hazard in a complex and non-stationary hydrological system: case study on the Niger River in Niamey

Droughts are a recurring long-term problem in West Africa, but in recent years the region has also experienced a significant increase in damaging floods. Increasing trends in extreme discharges have been shown, reflected by an increase in return level magnitude since 1980s in Sahelian rivers. Our study targets flood prediction within the regional catchment of the Niger River at Niamey, the capital of Niger. The river hydrograph has changed significantly since the drought between the 1970s and 1980s, evolving from a single peak to a two-hump hydrograph: a first flood, coming mainly from three direct Sahelian tributaries of the right bank of the Niger River, and a second one, coming from the more remote Guinean basin. Predicting floods in Niamey is not straightforward because of the complexity of the hydrological system, which combines non-stationarity and the difficulty of deconvoluting the two floods and separate their own trends.

The objective of the study is to quantitatively assess the hydrological hazard on the Niger River at Niamey based on a flow data set covering the period 1950-2020 by developing a statistical modeling approach that allows to integrate both the hydrological complexity and the non-stationarity of floods. An important question addressed by the proposed approach is to evaluate the contribution of considering hydrological complexity in non-stationary statistical modeling. This is achieved by defining several flood samples and proposing different non-stationary models adapted to their complexity.