Statistical model for compound flood hazard to separate local and upstream flow inputs on the Niger River at Niamey

In recent years, West Africa has experienced an increasing number of flood disasters, urging governments and decision makers to take adaptation measures, particularly in cities where the population is growing very fast. The present study focuses on the hydrological hazard of the city of Niamey (capital of Niger) along the Niger River. The regime of the Niger river at Niamey is marked by two successive floods: a first flood (July-August) resulting from the contributions of local Sahelian catchment (Sahelian red flood) and a second flood (February-March) resulting from the contribution of upstream catchments (Guinean black flood). Past studies have shown how the 1st flood, which was almost non-existent in the 1950s and 1960s, has predominated since the 1980s causing the most damaging floods due to a combined effect of climate and land use/cover changes. However, there is no in-depth study on statistical flood modelling, which is an essential step in any flood management strategy. The aim of the study is to evaluate how Niger discharge regime has changed since the 1950s with a specific focus on floods with the aim of deciphering the impact of Guinean and local flows in a statistical model. To do so, an original method is first proposed to isolate local flows from Guinean flows in the Niamey discharge data series. This separation allows us to distinguish three typologies of maximum annual floods according to whether they are generated by Guinean flows only, local flows only or a mix of the two. This observation leads us to propose different models of extreme discharge in Niamey based on the non-stationary extreme value theory able to consider both the typology of the floods and their temporal evolution.