Advancing Hydrologic Modelling Through Bias Correcting Weather Radar Data: The Valgrosina Case Study in the Italian Alps

The urgency of understanding the intricate input–output relationships of the hydrologic cycle is amplified by the accelerating climate change impacts in mountain environments. This study focuses on optimising water resource management of a dammed valley in the Central Alps (Northern Italy). The research aims to integrate radar data and precipitation interpolation techniques (TIN, Copula, cumulative distribution function; CDF techniques, inverse distance weighting; IDW, thin plate spline; TPS, ordinary kriging; OK and detrended kriging; DK) into a semi-distributed hydrologic model, by utilising hourly precipitation data from 22 rain gauges and a composite weather radar product spanning 2010–2020. Two main objectives were pursued: (i) to develop and evaluate various radar precipitation correction methods against a benchmark dataset and (ii) to calibrate and assess the performance of the GEOFrame hydrologic model forced with corrected precipitation input. Point-based and spatial correction approaches were evaluated against ground measurements through leave-one-out tests. The former derives dependence functions between the biased radar series and those of the closest three rain gauges to the target point applying a triangular irregular network. The latter combines deterministic and geospatial interpolations to the rain gauge/radar residuals to derive a corrected surface by incorporating radar values as trends. Precipitation series exceeding the composite scaled score of the benchmark dataset were used as input for hydrologic modelling. The spatial method combining radar values with ordinary kriging provided the best results for both correction and modelling (hourly KGE > 0.75). The spatial approaches proved easier to apply than the point-based methods. In addition, correcting precipitation significantly improved low-flow simulation from negative hourly lnNSE to values greater than 0.25. As a further step, given the overall good performance of the spatial methods, they could be used operationally as an ensemble to analyse management scenarios.