Greenhouse gas and aerosol contributions to the observed global and regional changes in extreme temperature changes

This study carried out an updated detection and attribution analysis of extreme temperature changes for 1951-2015. Four extreme temperature indices (warm extremes: annual maximum daily maximum/minimum temperatures; cold extremes: annual minimum daily maximum/minimum temperatures) were used considering global, continental (6 domains), and subcontinental (33 domains) scales. HadEX3 observations were compared with CMIP6 multi-model simulations using an optimal fingerprinting technique. Response patterns of extreme indices (fingerprints) to anthropogenic (ANT), greenhouse gas (GHG), anthropogenic aerosol (AA), and natural (NAT) forcings were estimated from corresponding CMIP6 forced simulations. Pre-industrial control simulations (CTL) were also used to estimate the internal variability. Results from two-signal detection analysis where the observations are simultaneously regressed onto ANT and NAT fingerprints reveal that ANT signals are robustly detected in separation from NAT in global and most continental regions for all extreme indices. At subcontinental scale, ANT detection occurs especially in warm extremes (more than 60% of regions). Results from three-signal detection analysis where observations are simultaneously regressed onto GHG, AA, and NAT fingerprints show that GHG signals are detected and separated from other external forcings over global, most continental, and several subcontinental (more than 60%) domains in warm extremes. In addition, AA influences are jointly detected in warm extremes over global, Europe and Asia. The detected GHG forcings are found to explain most of the observed warming while AA forcings contribute to the observed cooling for the early decades over globe, Europe, and Asia with a slight warming over Europe during recent decades. Overall, improved detection occurs compared to previous studies, especially in cold extremes, which is due to the use of extended period which increases signal-to-noise ratios.