Conference Agenda

Convection-permitting climate models (CPMs) can resolve convection-scale processes and improve estimates of short-duration, extreme precipitation events that are critical for urban drainage models. Their outputs, however, still require bias correction to match station-scale resolution. Quantile-mapping (QM), commonly used for bias-correction due to its simplicity, has limitations that need to be evaluated for CPMs. This study tests five QM variations to bias-correct and downscale simulations from a CPM for over 70 weather stations in Switzerland. For each station, ten years of data from the CPM, COSMO-CLM, is downscaled from a 2.2 km grid to the station scale at a 30-minute interval. Techniques to increase robustness of the QM approach were tested, including adding a temporal moving window, spatially pooling surrounding grid cells, and extending the observational record. Cross-validation shows that all QM methods reduce wet biases in the raw CPM output up to the 98^th quantile. Only the moving window technique and its combination with spatial pooling effectively reduce higher quantile biases. However, QM methods may distort the climate change signal, especially in rainfall frequency indices. Despite added computational effort, the moving window technique is recommended for robust CPM downscaling in urban drainage.