Understanding of CMIP6 surface temperature cold bias over the westerly and monsoon regions of the Tibetan Plateau

The Tibetan Plateau (TP) directly heats the middle tropospheric atmosphere, and accurate simulation of its surface temperature is of great concern for improving climatic prediction and projection capabilities, but climate models always exhibit a cold bias. Based on the Coupled Model Intercomparison Project Phase 6 (CMIP6) models and in-situ observations during 1981-2014, this study elucidates the impact of the snow overestimation on the temperature simulation over the TP in CMIP6 from the perspective of local radiation processes and atmospheric circulation. On the one hand, more snow in the CMIP6 models not only directly cools the surface more, but also makes the surface receive less shortwave radiation due to the higher surface albedo, and thus has lower ground surface temperature (GST), and the more snow/albedo-low temperature process is particularly evident in the westerly region due to more uncertainty at high elevations. This process contributes 87% to the annual GST cold bias. Lower GST corresponds to less latent heat transfer and thereby lower surface air temperature (SAT). In addition, the more snow in the CMIP6 models leads to the weaker the South Asian summer monsoon and the westerlies, and brings less warm and moist air (less integrated water vapor flux), as well as less clear-sky downward longwave radiation (less water vapor amount and lower tropospheric air temperature) to the TP (contributing 58% to the annual GST cold bias). These processes will result in less both precipitation and surface latent heat loss, which offsets a 35% annual GST cold bias. Besides, the physical mechanism of snow on GST and SAT differs with season over the westerly and monsoon regions of the TP. The research highlights the importance of topography and snow parameterization schemes for optimizing CMIP6 models.