Combining convection permitting modeling results with CMIP6 global climate model results to produce scenarios for local precipitation extremes

Abstract

Convection permitting climate models (CPMs) display much improved present-day rainfall statistics at local scales as compared to common regional and global climate models. Yet, because CPMs are computationally very demanding, runs are short — typically covering 10 to 20 years  only  —  which makes it hard to distinguish the changes due to global warming from the noise due to internal variability. In addition, runs cover a limited set of changes at larger scales as only few global climate models have been downscaled so far. This challenges the representativeness of the results. Here, we discus these issues within the context of the production of the Dutch climate scenarios issued in fall 2023. We use spatial pooling of information to improve signal to noise. To produce scenarios for local rainfall extremes, we combined information from the CPMs with information from CMIP6 and one RCM (RACMO) using a simple scaling framework. From the CPMs we derived sensitivities of changes in rainfall intensity to surface dew point temperature change. By using spatial pooling and by taking out rain frequency change (using wet conditional statistics) a reasonable collapse of the data of 7 CPM simulations could be obtained, with typical dependencies between 1 and 2 times the Clausius Clapeyron relation. The change in rain frequency and the dew point temperature are derived from a  set of RACMO simulations using pseudo-global warming perturbations derived from CMIP6 combined with a simple perturbed physics method. With these RACMO simulations we covered a range in large-scale conditions compatible with CMIP6.  Subsequently, rain intensity change and frequency change are combined using a transformation of the observed rainfall distribution.  In this way, we could produce a set of climate scenarios for daily and hourly precipitation extremes covering a wide range in global change conditions. Besides these changing rainfall statistics, we also analyzed the spatial temporal characteristics of showers in order to investigate whether showers become larger in scale in the future climate.