Detecting Changes in Precipitation Extremes in China and Their Connections With the Sea Surface Temperature of the Eastern Tropical Pacific Ocean

To detect the frequency and intensity of precipitation extremes in China for the middle 21st century, simulations were conducted with the regional climate model RegCM4 forced by the global climate model GFDL_ESM2M under the middle emission scenario (RCP4.5). Compared with observed precipitation extremes for the reference period from 1982 to 2001, RegCM4 generally performed better in most river basins of China relative to GFDL. In the future period 2032-2051, more wet extremes will occur relative to the present period in most study areas, especially in southeast China while significantly less dry extremes will occur in arid and semiarid areas in northwest China. In contrast, areas in northwest China showed an increase in the trend of dry extremes (CDD) and a decrease in the trend of wet extremes (R95p and Rx5day), which might result in more drought in the future. Finally, we discuss in detail the possible reason of these processes, such as zonal wind, vertical wind, and water vapor. In the Huaihe river basin (HU), reduced south winds in summer (June-August) and a decrease of the upward vertical p velocity cause less future precipitation and might lead to changes of extreme events. We also completed correlation analysis between the precipitation extreme indices and the climate factors and found that the precipitation extremes were more sensitive to the annual and seasonal mean precipitation, total water vapor, and upward vertical wind relative to the geopotential height and 2 m temperature over most river basins in China. At the end, We investigated the connections between the present precipitation extremes in the dry and wet regions of China and the sea surface temperature (SST) in the eastern tropical Pacific Ocean (ETP; including EI Nio-Southern Oscillation 1.2, 3, and 3.4 regions) based on two sets of observations, 18 CMIP5 models, and nine regional climate models (RCMs), where the results were downscaled by RCM RegCM4.