Precipitation variability in CMIP6 climate models across the North AtlanticEuropean region

Long-term changes in climate variability are an important aspect of the climate change with various impacts on society and environment. In contrast to numerous studies which evaluated projected changes in mean values and extremes of precipitation amount, intensity and/or frequency, studies on changes in precipitation variability have been relatively scarce. To understand whether and how the precipitation variability will change in the future, projections of climate models are utilized. However, accurate simulation of this precipitation characteristic by current climate models is pivotal. In our study we analyze outputs from 13 CMIP6 GCMs across the North AtlanticEuropean region focusing on winter and summer seasons separately. We classify days with a total precipitation amount exceeding 1 mm as wet days, while the remaining days are considered as dry days. Precipitation probability denote the mean probability of a wet day, and precipitation variability is represented by the tendency to cluster wet/dry days into sequences. To quantify this, we use the persistence parameter defined as the 1-lag autocorrelation of a discrete two-state Markov chain. Firstly, we evaluate whether precipitation variability is simulated correctly over the historical period (19802010) by comparing model outputs against the ERA5 reanalysis. Subsequently, we analyse projected changes in the future period (20702100) using simulations forced by two Shared Socio-economic Pathways (SSP585 and SSP245). This allows for a comparison of possible future climate changes under different climate policies. We identify biases common to all models, notably an overestimated precipitation probability across much of Europe in winter, while its underestimation in summer, and a general tendency of models toward higher autocorrelation of wet/dry days. Projected changes in precipitation characteristics are more pronounced for the more pessimistic SSP585 scenario. We find that the changes in precipitation variability are independent on the changes in precipitation probability. Our findings also indicate that the model biases and simulated changes in precipitation probability and variability can be linked to the biases and changes in synoptic-scale atmospheric circulation.