Assessing climate change impacts on design of small hydropower projects in Nepal using a process-based hydrological model with machine learning post-processor

Small hydropower (SHP) projects are rapidly expanding in Nepal as a clean energy solution, yet their design typically only relies on historical streamflow records and overlooks climate-driven changes in extreme floods. This creates a risk of underestimating future flood magnitudes and infrastructure damage. This study evaluates projected changes in 25-, 50-, and 100-year design floods across four basins containing SHP in Nepal. Process-based Hydrologiska Byråns Vattenbalansmodell (HBV) models were first calibrated for each basin, and a machine learning post-processor was developed to correct systematic residuals, improving streamflow simulation performance. The hybrid model was then forced with CMIP6 climate scenarios (historical, SSP245, and SSP585) to simulate baseline and future streamflow. Annual flood maxima were fitted using Generalized Extreme Value distributions, and percentage changes in design floods were estimated using parametric bootstrapping. Results show substantial increases in design flood magnitudes: about 10–30% under SSP245 and 30–65% under SSP585, with upper estimates exceeding 100% in some cases. Larger increases and greater uncertainty are associated with higher return periods and higher emission scenarios. Central basins show higher sensitivity compared with eastern basins. These findings highlight the urgent need to integrate climate-informed flood estimates into SHP design to ensure resilient hydropower development in Nepal.