An Evaluation of CMIP6 Models in Representing the Biophysical Effects of Deforestation With Satellite‐Based Observations

Deforestation can impact surface temperature via biophysical processes. Earth system models (ESMs) are commonly used tools to examine biophysical effects of deforestation, but the model capacity to represent deforestation effects remains unclear. In this study, we comprehensively evaluate the performance of four ESMs of the Coupled Model Intercomparison Project Phase 6 (CMIP6) in representing deforestation effects with a satellite‐based benchmark. The results show that the ESMs can basically capture the sign of the temperature response but over‐ or underestimate the magnitude. Such biases are the consequence of biases in the simulated responses of albedo and sensible and latent heat fluxes. Specifically, the ESMs consistently overestimate the albedo response under snow‐covered conditions, for example, in the northern latitudes and in the cold season. The ESMs fail to fully reproduce the observed responses of sensible and latent heat fluxes, and the model bias depends on the model, region and season. The ESMs and observations even disagree on the sign of responses of sensible and latent heat fluxes in some cases. An attribution analysis further shows that biases in the simulated surface temperature response mainly result from biases related to the response of the surface energy partitioning. Biases related to the albedo response only play an important role under snow‐covered conditions. Given these model biases, we highlight that when the CMIP6 models are used to investigate deforestation effects, the simulated result should be interpreted with caution. Moreover, the identified model deficiency shown here also has implications for model improvement. Plain Language Summary Deforestation can alter energy and water exchanges between the land and atmosphere via changes in surface properties and consequently impact surface temperature. Earth system models (ESMs) are commonly used tools to examine such deforestation effects. However, whether ESMs can reasonably represent deforestation effects remains unclear. Benefiting from the rapid development of satellite‐based remote sensing technologies, increasing numbers of observational datasets on deforestation effects have been reported, which provide a benchmark for model evaluation. In this study, we comprehensively evaluate the performance of four ESMs in representing deforestation effects with satellite‐based observations. The results show that the ESMs can basically reproduce the observed deforestation effects on