article · PLOS Climate
An ensemble machine learning approach has been developed to predict climate change impacts on reservoir inflows and hydropower output across seven river basins in West Africa. The method combines multi-lag precipitation and temperature data with stacked, refined machine learning algorithms. Projections using climate models under three emission scenarios indicate temperature increases of up to 4.5 degrees Celsius alongside highly varied regional precipitation patterns. As a result, reservoir inflows could decline by up to 24 percent at Buyo and 13 percent at Nangbeto, leading to power generation drops of up to 58 percent at Taabo and 19 percent at Nangbeto. In contrast, locations such as Manantali and Taabo may experience increased inflows, while Bagre could achieve hydropower output gains of up to 42 percent under high emissions. These contrasting projections emphasise the necessity of flexible management and diversified regional energy systems.
West African electricity grids rely heavily on hydroelectric dams that are highly sensitive to climatic shifts. By forecasting long-term changes in river flow and power output, this research assists regional planners in anticipating localized water shortages and energy deficits. Understanding these varied regional impacts helps authorities plan energy diversification and strengthen infrastructure resilience against changing climate conditions.
The predictive framework is an applied research tool aimed at hydropower managers, grid utilities, and water resource authorities. It enables operators to evaluate climate risks and plan long-term energy and water resource management strategies. The methodology has been validated across seven basins using historical data and future climate projections, placing it at an applied research stage that requires software packaging or integration into existing grid operations before commercial use.
AI-generated from the published abstract. Always read the original work before citing.
This study presents a novel five-step ensemble machine learning approach to improve predictive accuracy in assessing climate change impacts on inflow patterns and hydropower generation across seven dam basins in West Africa. The methodology integrates precipitation and temperature using the multi-lag approach. An initial pool of fifteen machine learning models were evaluated, and top-performing models were selected for further refinement through iterative ensemble stacking and weak learner elimination. Historical analysis (1983–2014) utilized CHIRPS and CHIRTS datasets. Future projections employed twelve bias-adjusted CMIP6 models and their ensemble mean (EnsMean) under SSP1-2.6, SSP2-4.5, and SSP5-8.5 for the near (2036–2067) and far (2068–2099) futures. Results showed notable improvements in model accuracy and efficiency across layers, with R² and NSE exceeding 0.6 for all inflow simulations and for selected energy simulations (Bagre, Nangbeto, and Taabo). Projections indicated warming up to 4.5°C and spatially heterogeneous precipitation changes across basins and scenarios, with SSP5-8.5 projecting the most pronounced shifts. Inflow reductions are projected to reach up to 24% at Buyo and 13% at Nangbeto, while hydropower output may decline by up to 19% at Nangbeto and 58% at Taabo in the future. Conversely, Manantali and Taabo are projected to experience inflow increases, and Bagre may see energy gains of up to 42% under SSP5-8.5. These findings highlight heightened vulnerability, as well as contrasting opportunities across the region, underscoring the urgent need for adaptive management strategies, such as enhancing hydropower system resilience, diversifying energy portfolios, and integrating renewable sources to mitigate climate risks. Hydropower managers and policymakers must prioritize proactive measures to ensure energy security and sustainable resource management amid changing climatic conditions.
This page summarises published work. The authoritative version sits with the publisher.
DOI: 10.1371/journal.pclm.0000986
Is something wrong with this record? Report it or request removal.
Discussion
Have you built on this work, tried to replicate it, or seen it applied in practice? Share what you know. Verified researchers and MARATTO™ domain experts can open a discussion, and any member can reply. Contributions are reviewed before they appear.
No discussion yet. Open the first thread.
New to MARATTO™? Create a free account.
