Downscaling and uncertainty analysis of future concurrent long-duration dry and hot events in China

Using fourteen global climate models (GCMs) from phase 5 of the Coupled Model Intercomparison Project (CMIP5) downscaled by four statistical downscaling methods, future changes and the associated uncertainty in concurrent long-duration dry and hot (LDDH) events are investigated in China during summer. The downscaling methods include BCSD (bias-correction and spatial downscaling), BCCI (bias-correction and climate imprint), BCCAQ (bias correction constructed analogues with quantile mapping reordering), and CDF-t (cumulative distribution function transform). The downscaling methods can efficiently improve the accuracy over the driving GCMs in terms of spatial variability, bias, and inter-annual variability of LDDH characteristics. Overall, the three quantile mapping based techniques (BCSD, BCCI, and BCCAQ) outperform CDF-t in simulating the spatial and temporal features of LDDH events. In the twenty-first century, all downscaling methods project a consistent increasing tendency for the frequency, magnitude, and total days of LDDH events over most parts of China, with higher increases under RCP8.5 compared to RCP4.5. A substantial increase in spatially contiguous regions simultaneously experiencing LDDH events is seen by mid-century under both scenarios. Changes in the frequency, magnitude, and total days of LDDH events are predicted with high confidence. For most indices, model uncertainty dominates throughout the century and does not change much over time. However, for the projection of temperature magnitude of LDDH events, the dominant role of GCM related uncertainty in the early twenty-first century declines as scenario uncertainty becomes more important towards the end of the century.