Evaluating the added value of convection-permitting regional climate models in simulating hydrological extremes over basins in western Norway

Convection-permitting regional climate models (CPRCMs) are increasingly recognized for their ability to improve extreme precipitation predictions, yet their application to hydrological modeling in complex terrains remains uncertain. This study evaluates the performance of CPRCMs in predicting hydrological extremes in two basins in Western Norway: Røykenes, dominated by rainfall-induced floods, and Bulken, characterized by snowmelt-induced floods. We compare the capabilities of a high-resolution convection-permitting model (HCLIM3, 3 km resolution) with a coarser regional climate model (HCLIM12, 12 km resolution) in driving two hydrological models: the physically based Weather Research and Forecasting Model Hydrological system (WRF-Hydro) and the conceptual Hydrologiska Byråns Vattenbalansavdelning (HBV) model. Performance was evaluated based on precipitation, temperature, runoff, and hydrological extremes. We found that HCLIM3 exhibited significantly better performance in estimating annual maximum 1-day (Rx1d) and 1-hour (Rx1h) precipitation, with reduced biases compared to HCLIM12. It also showed added value in capturing the probability density distribution of daily and hourly precipitation, as quantified by the Distribution Added Value (DAV) metric. However, both HCLIM3 and HCLIM12 displayed cold biases, especially in mountainous areas. Besides, in the rainfall-dominated Røykenes basin, WRF-Hydro outperformed HBV in simulating extreme flood magnitudes across return periods (5, 10, 20, and 50 years). However, in the snowmelt-dominated Bulken basin, cold biases in HCLIM3 and HCLIM12 introduced uncertainties in snowmelt timing, leading to larger errors. The added value of HCLIM3 was observed in hourly discharge in the Røykenes basin. However, this benefit was less pronounced in the snowmelt-dominated Bulken basin, where temperature sensitivities significantly influenced snowmelt processes. Biases in HCLIM3 and HCLIM12 meteorological forcing propagated through hydrological models, leading to discharge errors, as highlighted by DAV metrics. This research highlights the importance of applying bias correction to CPRCM simulations to improve hydrological modeling of extreme events, especially in mountainous terrains where biases in temperature and precipitation critically affect hydrological processes.