Assessing the performance of convection-permitting climate model in reproducing basin-scale hydrological extremes: A western Norway case study

Abstract

Convection-permitting regional climate models (CPRCMs) have been shown to improve the representation of extreme precipitation compared to coarser resolution regional climate models (RCMs). Their benefits for hydrological extremes, such as floods, remains uncertain. This study evaluates the performance of a 3-km resolution convection-permitting model (HCLIM3) against a coarser 12-km resolution climate model (HCLIM12) from the HARMONIE-Climate (HCLIM) model, focusing on precipitation, temperature, and floods in two basins over Western Norway: Røykenes basin (dominated by rainfall-generate flood) and Bulken basin (dominated by snowmelt-generate flood). In the study, we use both a physically-based, distributed Weather Research and Forecasting Model Hydrological system (WRF-Hydro) and a conceptual, lumped Hydrologiska Byråns Vattenbalansavdelning (HBV) model to assess flood simulations. The results show: (1) HCLIM3 better captures spatial variability of annual maximum 1-day and 1-hour precipitation compared to HCLIM12, but both HCLIM models exhibit cold biases which are more pronounced at lower elevation areas, particularly in HCLIM12. (2) HCLIM3-driven simulations do not show benefit in flood simulations across the two basins, except for severe flood peaks, compared to HCLIM12, the choice of hydrological model has a large impact on the results. The HBV model underestimates flood peaks and frequency, while WRF-Hydro more accurately simulates them in the Røykenes but overestimates them in the Bulken likely due to the biases of forcing data, particularly when driven by HCLIM3. The study concludes that CPRCMs improve the simulation of extreme precipitation and temperature but not show clear added value for flood simulations, especially in Bulken. This highlights the critical need for bias correction to ensure accurate flood predictions, even when driven by CPRCMs.