Global changes in extreme precipitation linked with rapid flood simulation tools

The consensus of climatic research indicates that the likelihoods of extreme precipitation events are going to change significantly, but specific trends depend on the type of dominant weather system, and regional landscape and climate details. Despite the effort of increasingly accurate and converging global circulation models, the uncertainties in the ensemble of CMIP6 models, and the often course spatial resolution make translation of climate models to actionable information of flood forecasts complex and uncertain. We carried out an analysis of a large ensemble of Global Circulation Models (GCMs) of the CMIP6 ensemble that were downscaled statistically as part of the NASA NEX-GDDP-CMIP6 dataset. The analysis looked at segmented windowed return period analysis using the method of l moments to fit general extreme value distributions to global climate models. With analysis of 1, 3 and 7 day duration, median, 15 and 85 percent quantiles, between 5 and 100 year return period, and global spatial coverage, the results show variations in how precipitation events of various return periods and durational are predicted to change in GCMs, and what the associated uncertainty is for various regions of the world. Intermediate analysis outputs show artifacts in yearly extreme precipitation due to the applied statistical downscaling, but relative factors to be used in precipitation scaling under climate change resolves these. Average increases in precipitation extremes of percent are observed globally (+5.1%), with many local outliers for the SSP585 scenario in 2050 (e.g. regions such as the Himalayan region (+23.4 percent median), the Sahel region(+21.6%) or South-Western Spain (-3.9%)). The other SSP scenarios change the global average factors to +3.75% and +4.32% for SPP245 and SSP370 Respectively. Very low variability in the changes is observed for return periods, indicating that the intensity probability curves shift uniformly in the model output. Precipitation events duration does more significantly alter the analysis outputs, and various areas show differences here that correlate with flash and fluvial flood susceptibility. Finally, we open-source the analysis code and link the output as a built-in dataset in the fastflood.org rapid flood simulation platform. Here, automatically derived extreme precipitation events from era5 datasets can be rescaled under climate change conditions by applying the scaling factors derived in this work.