The Improvements of Sea Surface Temperature Simulation over China Offshore Sea in Present Climate from CMIP5 to CMIP6 Models

By using the 43 Historical experiments from phase 6 of the Coupled Model Intercomparison Project (CMIP6) and 45 Historical experiments from phase 5 of CMIP (CMIP5) for the period of 1950–2005, we comprehensively assess the improvements in simulating the spatial pattern, warming trend, climatology and interannual variation of sea surface temperature (SST) in China offshore sea (COS) from CMIP5 to CMIP6 models. Both CMIP6 multi-model ensemble mean (CMIP6 MME) and CMIP5 multi-model ensemble mean (CMIP5 MME) well simulated the spatial pattern of climatological-mean COS SST, but they tend to underestimate the warming trends of COS SST at both seasonal and interannual timescales, which is due to the low estimations of SST warming rate before the late 1970s, particularly for the CMIP6 models. Nevertheless, both the simulated trend biases and inter-model uncertainties are reduced from CMIP5 to CMIP6 models during the period 1979–2005. Compared to the simulated annual-mean and seasonal-mean COS SST in the CMIP5 models, the inter-model uncertainties and cold biases of SST simulated by the CMIP6 models have been significantly reduced, particularly for the autumn-mean and summer-mean SST. Similarly, the CMIP6 models perform better than the CMIP5 models in simulating the interannual variation of COS SST, as evidenced by a much lower interannual variability skill score over the South China Sea and Huang Bo China Sea. Furthermore, more than 60