Identifying robust bias adjustment methods for extreme precipitation in a pseudo-reality setting

Abstract. Severe precipitation events occur rarely and are often localized in space and of short duration; but they are important for societal managing of infrastructure. Therefore, there is a demand for estimating future changes in the statistics of these rare events. These are usually projected using Regional Climate Model (RCM) scenario simulations combined with extreme value analysis to obtain selected return levels of precipitation intensity. However, due to imperfections in the formulation of the physical parameterizations in the RCMs, the simulated present-day climate usually has biases relative to observations. Therefore, the RCM results are often bias-adjusted to match observations. This does, however, not guarantee that bias-adjusted projected results will match future reality better, since the bias may change in a changed climate. In the present work we evaluate different bias adjustment techniques in a changing climate. This is done in an inter-model cross-validation setup, in which each model simulation in turn plays the role of pseudo-reality, against which the remaining model simulations are bias adjusted and validated. The study uses hourly data from present-day and RCP8.5 late 21st century from 19 model simulations from the EURO-CORDEX ensemble at 0.11° resolution, from which fields of selected return levels are calculated for hourly and daily time scale. The bias adjustment techniques applied to the return levels are based on extreme value analysis and include analytical quantile-matching together with the simpler climate factor approach. Generally, return levels can be improved by bias adjustment, compared to obtaining them from raw scenarios. The performance of the different methods depends of the time scale considered. On hourly time scale, the climate factor approach performs better than the quantile-matching approaches. On daily time scale, the superior approach is to simply deduce future return levels from observations and the second best choice is using the quantile-mapping approaches. These results are found in all European sub-regions considered.