Do CMIP6 climate models capture rapid shifts between dry and wet extremes?

Abstract

Do CMIP6 climate models capture rapid shifts between dry and wet extremes?Authors: Rong Gan1, Yuting Yang1,*Affiliations: 1State Key Laboratory of Hydroscience and Engineering, Department of Hydraulic Engineering, Tsinghua University, Beijing, China*Correspondence to: Yuting Yang (yuting_yang@tsinghua.edu.cn)Keywords: CMIP6, climate extremes, compound eventsAbstract:Rapid shifts between dry and wet extremes may impose higher socioeconomic and environmental pressure than single extremes. Whether the sixth phase of the Coupled Model Intercomparison Project (CMIP6) models are capable of capturing the abrupt alternations between dry and wet periods remain elusive. Here we examine such compound events simulated by CMIP6 models based on the state-of-the art reanalysis datasets, namely ERA5, NCEP-NCAR and MERRA-2. The 1-month Standard Precipitation-Evapotranspiration Index (SPEI) were first calculated to identify dry spells (SPEI≤1) followed by wet spells (SPEI≥1), and vice versa. Event characters including frequency, duration and intensity were then evaluated across all CMIP6 models and reanalysis datasets spanning 1980-2014. We find the following:CMIP6 multimodel-ensemble median and reanalysis ensemble give close estimates of event characters on global average, with frequency being about 4.1 and 3.67 (No. events/20-year), duration of 2.50 and 2.55 (months), and intensity around 3 (SPEI mean) for dry-wet events, respectively. Similar values were found for wet-dry events. During 1980-2014, CMIP6 and reanalysis indicate roughly 10% increase in event frequency comparing the first and last 20-year periods, and less than 1% increase in duration and intensity for both dry-wet and wet-dry events. Spatial distribution for event frequency tends to overlap for dry-wet and wet-dry events, as shown by both CMIP6 models and reanalysis. Hot spots were found in North-eastern America, Europe, Eastern Asia, South-western America, and Middle Africa. Higher latitude regions were shown to experience more events. Despite general spatial agreement between CMIP6 and reanalysis, discrepancies can be seen on finer scales within each region. Common spatial patterns for duration were also found between the two types of events based on CMIP6 models, where the events tend to last longer in middle and southern Eurasia, Eastern Africa, northwest of South America and west of Northern and Central America. However, reanalysis indicates longer events also happened in Middle Africa and eastern Australia. Both CMIP6 models and reanalysis indicate longer event duration roughly around the equator. CMIP6 models give much higher dry-wet intensity compared to wet-dry, especially in Australia and Southern and Western Asia. Reanalysis agrees well on this pattern, yet greater magnitude differences were found in eastern South America. Overall, CMIP6 models are capturing the variations of abrupt dry and wet alternations well when compared to reanalysis. The models are more skilful in simulating event frequency than duration and intensity in general. Caution should be paid assessing such compound events especially on smaller spatial scales and sensitive regions such as Africa for frequency and Australia for duration and intensity. Our results can be further employed to support climate risk adaptation and mitigation.