Global climate response to idealized deforestation in CMIP6 models

Changes in forest cover have a strong effect on climate through the alteration of surface biogeophysical and biogeochemical properties that affect energy, water and carbon exchange with the atmosphere. To quantify biogeophysical and biogeochemical effects of deforestation in a consistent setup, nine Earth system models (ESMs) carried out an idealized experiment in the framework of the Coupled Model Intercomparison Project, phase 6 (CMIP6). Starting from their pre-industrial state, models linearly replace 20 x 10(6) km(2) of forest area in densely forested regions with grasslands over a period of 50 years followed by a stabilization period of 30 years. Most of the deforested area is in the tropics, with a secondary peak in the boreal region. The effect on global annual near-surface temperature ranges from no significant change to a cooling by 0.55 degrees C, with a multi-model mean of -0.22 +/- 0.21 degrees C. Five models simulate a temperature increase over deforested land in the tropics and a cooling over deforested boreal land. In these models, the latitude at which the temperature response changes sign ranges from 11 to 43 degrees N, with a multi-model mean of 23 degrees N. A multi-ensemble analysis reveals that the detection of near-surface temperature changes even under such a strong deforestation scenario may take decades and thus longer than current policy horizons. The observed changes emerge first in the centre of deforestation in tropical regions and propagate edges, indicating the influence of non-local effects. The biogeochemical effect of deforestation are land carbon losses of 259 +/- 80 PgC that emerge already within the first decade. Based on the transient climate response to cumulative emissions (TCRE) this would yield a warming by 0.46 +/- 0.22 degrees C, suggesting a net warming effect of deforestation. Lastly, this study introduces the forest sensitivity (as a measure of climate or carbon change per fraction or area of deforestation), which has the potential to provide lookup tables for deforestation-climate emulators in the absence of strong non-local climate feedbacks. While there is general agreement across models in their response to deforestation in terms of change in global temperatures and land carbon pools, the underlying changes in energy and carbon fluxes diverge substantially across models and geographical regions. Future analyses of the global deforestation experiments could further explore the effect on changes in seasonality of the climate response as well as large-scale circulation changes to advance our understanding and quantification of deforestation effects in the ESM frameworks.