Estimating flow rates using open-source datasets to improve drainage systems in forest and rural roads

Résumé fourni par la source

Accurate streamflow estimation in ungauged catchments is vital for dimensioning drainage structures along rural and forest road networks, where local hydrometric data are often scarce. Addressing this research gap, this study integrates open-source topographic and climatic datasets with Global Runoff Data Centre (GRDC) records to model discharge in Italy’s Po River sub-region. Four ensemble learning algorithms – LightGBM, XGBoost, random forest, and CatBoost – were evaluated to benchmark predictive performance against traditional estimation challenges. CatBoost exhibited superior accuracy (R2 = 0.94) and robustness during sensitivity testing via controlled noise introduction. To assess long-term infrastructure resilience, hydrological responses were simulated using incremental climate perturbations (systematic precipitation and temperature shifts) over a 100-year horizon. The findings provide a quantitative framework for the preliminary dimensioning of culverts, enabling data-driven flood risk management and infrastructure protection in data-limited rural basins.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.