Browsing by Author "Canwat, Emmanuel"

Now showing 1 - 1 of 1
  • No Thumbnail Available
    Item
    Comparative Assessment of Hec-Hms and Hec-Hms_Xgboost Models to Simulate Streamflow under Climate Change for Flood Control Structure Design in River Sipi Catchment, Eastern Uganda
    (Busitema University, 2026) Canwat, Emmanuel
    Climate change has significantly intensified the frequency and magnitude of flood events in data-scarce tropical mountainous catchments such as the River Sipi catchment on the slopes, posing serious risks to infrastructure, agriculture, and livelihoods. Conventional physically-based models such as HEC-HMS exhibit critical limitations in capturing the nonlinear, complex dynamics of high-flow peak events under non-stationary climate conditions, compromising the reliability of flood control structure design in such areas. The study comparatively assessed the performance of the conventional HEC-HMS model and a hybrid HEC-HMS_XGBoost model in simulating streamflow under present and future climate scenarios for improved flood control structure design in the Sipi River catchment. The HEC-HMS and the hybrid models utilized historical hydrological data from 2010-2024 for calibration and validation. Future streamflow projections from 2025-2100 were generated by running both calibrated models with bias-corrected and downscaled CMIP6 climate data. Flood frequency analysis using the Gumbel extreme value distribution was applied to derive design discharges Q_10, Q_25, Q_50, and Q_100, which were subsequently used to size four flood control structures: culverts, spillways, bridge waterway openings, and detention basins. The hybrid model substantially outperformed HEC-HMS, achieving NSE values of 0.985 and 0.967 during calibration and validation, with KGE values of 0.934 and 0.893 respectively, compared to unsatisfactory HEC-HMS NSE values of 0.311 and 0.400. Future simulations revealed significantly divergent behavior under SSP5-8.5, HEC-HMS projected a Q_100 of 30.32 m³/s, a 38.5% increase above its Q_10 of 21.89 m³/s, while the hybrid model produced a more stable Q_100 of 21.51 m³/s, only 20.7% above its Q_10 of 17.82 m³/s. Notably, under SSP2-4.5 the hybrid model projected higher discharges than HEC-HMS, providing a physically corrected rather than uniformly lower estimates. Model divergence translated directly into significant structural sizing differences; culvert diameters differed by 14%, spillway crest lengths by 41%, bridge waterway spans by 36%, and detention basin volumes by 59% under SSP5-8.5 Q_100. The study concludes that the HEC-HMS_XGBoost hybrid model provides more reliable, physically plausible, and climate-resilient design discharges than standalone HEC-HMS, and recommends its adoption as the primary basis for flood control structure design in data-scarce tropical mountainous catchments. The standalone HEC-HMS projections under SSP5-8.5 are best reserved as upper-bound safety checks.