Assessing the mid-term weather forecasts in hydrological modelling

Accurate hydrological forecasting is critical for water resource management, flood prediction, hydropower generation, and risk assessment and mitigation but is fundamentally challenged by uncertainties arising from meteorological forcing, model structure, and initial conditions. This study provides an evaluation of medium range weather forecasting in hydrological modelling from European Centre for Medium-Range Weather Forecasts (ECMWF) to reduce these uncertainties for two mountainous basins. Using twenty hydrological models within the HydrOlOgical Prediction LAboratory (HOOPLA) framework, this research systematically compares four forecasting configurations: Open-Loop (OL) and Data Assimilated (DA) for both deterministic and ensemble forecasts. The Ensemble Kalman Filter (EnKF) was used for data assimilation, and multi-model (MM) combinations were generated using a Simple Averaging Method (SAM). Performance, assessed by the Kling-Gupta Efficiency (KGE), revealed a clear hierarchy. Ensemble forecasts consistently outperformed deterministic ones, and DA significantly enhanced forecast skill over OL simulations by correcting initial model states. The study demonstrates that an integrated approach combining all techniques yields the most reliable results. The DA multi-model ensemble proved superior, effectively mitigating multiple uncertainty sources and maintaining high accuracy across the 10-day forecast horizon. These findings offer a robust, evidence-based framework for improving operational hydrological forecasting systems.