West Africa exhibits decadal patterns in the behaviour of droughts and floods, creating challenges for effective water resources management. Proposed drivers of prolonged shifts in hydrological extremes include the impacts of land-cover change and climate variability in the region. However, while future land-degradation or land-use are highly unpredictable, recent studies suggest that prolonged periods of high-flows or increasing flood occurrences could be predicted by monitoring sea-surface temperature (SST) anomalies in the different ocean basins. In this study, we thus examine: i) what ocean basins would be the most suitable for future seamless flood-prediction systems; ii) how these ocean basins affect high-flow extremes (hereafter referred as extreme streamflow); and iii) how to integrate such nonstationary information in flood risk modelling. We first use relative importance analysis to identify the main SST drivers modulating hydrological conditions at both interannual and decadal timescales. At interannual timescales, Pacific Niño (ENSO), tropical Indian Ocean (TIO) and eastern Mediterranean (EMED) constitute the main climatic controls of extreme streamflow over West Africa, while the SST variability in the North and tropical Atlantic, as well as decadal variations of TIO and EMED are the main climatic controls at decadal timescales. Using regression analysis, we then suggest that these SST drivers impact hydrological extremes through shifts in the latitudinal location and the strength of the Intertropical Convergence Zone (ITCZ) and the Walker circulation, impacting the West African Monsoon, especially the zonal and meridional atmospheric water budget. Finally, a nonstationary extreme model, with climate information capturing regional circulation patterns, reveals that EMED SST is the best predictor for nonstationary streamflow extremes, particularly across the Sahel. Predictability skill is, however, much higher at the decadal timescale, and over the Senegal than the Niger catchment. This might be due to stronger impacts of land-use (-cover) and/or catchment properties (e.g. the Inner Delta) on the Niger River flow. Overall, a nonstationary framework for floods can also be applied to drought risk assessment, contributing to water regulation plans and hazard prevention, over West Africa and potentially other parts of the world.

西非在干旱和洪水的行为方面表现出十年的模式，为有效的水资源管理带来了挑战。水文极端事件长期变化的拟议驱动因素包括该地区土地覆盖变化和气候变化的影响。然而，虽然未来的土地退化或土地利用是高度不可预测的，但最近的研究表明，可以通过监测不同海洋盆地的海面温度 (SST) 异常来预测长时间的高流量或洪水发生的增加。在这项研究中，我们因此检查：i) 哪些洋盆最适合未来无缝洪水预测系统；ii) 这些洋盆如何影响极端高流量（以下称为极端流量）；iii) 如何将此类非平稳信息整合到洪水风险建模中。我们首先使用相对重要性分析来确定在年际和年代际时间尺度上调节水文条件的主要 SST 驱动因素。在年际时间尺度上，太平洋尼诺（ENSO）、热带印度洋（TIO）和东地中海（EMED）构成了西非极端流量的主要气候控制，而北大西洋和热带大西洋的海温变化以及年代际变化TIO 和 EMED 是十年时间尺度上的主要气候控制。使用回归分析，我们然后建议这些 SST 驱动因素通过纬度位置的变化和热带辐合带 (ITCZ) 和沃克环流的强度影响水文极端事件，影响西非季风，尤其是纬向和经向大气水收支. 最后，具有捕捉区域环流模式的气候信息的非平稳极端模型表明，EMED SST 是非平稳水流极端事件的最佳预测因子，尤其是在整个萨赫勒地区。然而，在十年时间尺度上，塞内加尔的可预测性技能比尼日尔流域高得多。这可能是由于土地利用（覆盖）和/或集水区属性（例如内三角洲）对尼日尔河流量的更大影响。总体而言，洪水的非平稳框架也可用于干旱风险评估，有助于西非和世界其他地区的水调节计划和灾害预防。尤其是整个萨赫勒地区。然而，在十年时间尺度上，塞内加尔的可预测性技能比尼日尔流域高得多。这可能是由于土地利用（覆盖）和/或集水区属性（例如内三角洲）对尼日尔河流量的更大影响。总体而言，洪水的非平稳框架也可用于干旱风险评估，有助于西非和世界其他地区的水调节计划和灾害预防。尤其是整个萨赫勒地区。然而，在十年时间尺度上，塞内加尔的可预测性技能比尼日尔流域高得多。这可能是由于土地利用（覆盖）和/或集水区属性（例如内三角洲）对尼日尔河流量的更大影响。总体而言，洪水的非平稳框架也可用于干旱风险评估，有助于西非和世界其他地区的水调节计划和灾害预防。