A Dissection of the Surface Temperature Biases in the Community Earth System Model

Based on the climate feedback-responses analysis method (CFRAM), a quantitative contribution analysis is performed for the annul-mean surface temperature bias in Community Earth System Model version 1 (CESM1). Local surface temperature biases are decomposed into partial temperature biases associated with albedo, water vapor, cloud, sensible/latent heat flux, surface (oceanic) dynamics, and atmospheric dynamics. Global-averaged cold bias of CESM1 (-1.22 K) is contributed largely to cold biases from albedo (-0.80 K), cloud (-0.48 K), and oceanic (surface) dynamics (-0.68 K) and partly compensated by warm biases from sensible (0.30 K) and latent (0.67 K) heat fluxes. The albedo bias contributes to most of cold bias of the land surface temperature. Most of cold bias over the ocean is determined by biases related to sensible/latent heat flux and oceanic dynamics. The strong cold bias over the Arctic sea-ice is mainly attributed to biases in albedo, cloud, and atmospheric dynamics. Albedo bias serves as a primary contributor to the surface temperature bias in high-latitude land and the Arctic/Antarctic regions, which is associated with the CESM1 biases for snow cover and sea-ice. The bias from water vapor, dominating its longwave feedback (i.e., greenhouse) effect, provides warm bias in the subtropics through more specific humidity while provides cold bias over the equatorial region through less specific humidity near surface. The dominant shortwave cloud forcing over low/mid-latitude and longwave cloud forcing over high-latitude induce a cloud bias, leading to the surface temperature bias. The sensible and latent heat flux bias gives the opposite bias to the corresponding partial temperature through the heat release by conduction and evapotranspiration. The change in heat storage associated with the ocean mixed layer depth bias is responsible for the ocean dynamics-related warm and cold bias. Compared with non-radiative energy perturbation (i.e., atmospheric dynamics), bias from physical processes (albedo, water vapor, and cloud processes) are more important for two thirds of global grid-points. In particular, biases over the Arctic sea-ice and the Antarctica are mostly subjected to the physical biases for more than 90% and 70%, respectively, of the total bias. Over ocean, dynamical biases are comparable to those of physical ones.