Deforestation can impact precipitation through biophysical processes and such effects are commonly examined by models. However, previous studies mostly conduct deforestation experiments with a single model and the simulated precipitation responses to deforestation diverge across studies. In this study, eleven earth system models are used to robustly examine the biophysical impacts of deforestation on precipitation, precipitation extremes and the seasonal pattern of the rainy season through a comparison of a control simulation and an idealized global deforestation simulation with clearings of 20 million km2 of forests. The multimodel mean suggests decreased precipitation, reduced frequency and intensity of heavy precipitation and shortened duration of rainy seasons over deforested areas. The deforestation effects can even propagate to some regions that are remote from deforested areas, e.g., the tropical and subtropical oceans and the Arctic Ocean. Nevertheless, the eleven models do not fully agree on the precipitation changes almost everywhere. In general, the models exhibit higher consistency over the deforested areas and a few regions outside the deforested areas (e.g., the subtropical oceans) but lower consistency over other regions. Such intermodel spread mostly results from divergent responses of evapotranspiration and atmospheric moisture convergence to deforestation across the models. One of the models that has multiple simulation members also reveals considerable spread of the precipitation responses to deforestation across the members due to internal model variability. This study highlights the necessity of robustly examining precipitation responses to deforestation based on multiple models and each model with multiple simulation members.

The Loess Plateau is a highland area covering approximately 640,000 km 2 in north central China. Historical longterm deforestation and crop expansion, as well as the semiarid background climate, cause severe soil erosion and land degradation (Fu et al., 2011;He et al., 2006). Moreover, the plateau serves as the ecological safety barrier of the upper reach of the Yellow River. Soil erosion augments the sediment discharge of the Yellow River, increasing the flooding risk in densely populated areas downstream (Fu et al., 2017). To address these issues, China has carried out a series of large-scale ecological programs over the plateau, including the most influential "Grain for Green Program" (GFGP) initiated in 1999 (Bryan et al., 2018). Benefiting from the GFGP, the plateau has shown a significant greening trend since then, with the vegetation fraction increasing from 31.6% in 1999 to 59.6% in 2013 (Chen et al., 2015).Greening is beneficial to soil erosion control and, on the other hand, can influence the terrestrial water balance through biophysical processes (Piao et al., 2020;Zeng et al., 2018). Greening can facilitate surface evapotranspiration due to the larger leaf area, aerodynamically rougher surface and higher canopy conductance for tran-

Afforestation can impact surface temperature through local and nonlocal biophysical effects. However, the local and nonlocal effects of afforestation in China have rarely been explicitly investigated. In this study, we separate the local and nonlocal effects of idealized afforestation in China based on a checkerboard method and the regional Weather Research and Forecasting (WRF) Model. Two checkerboard pattern–like afforestation simulations (AFF1/4 and AFF3/4) with regularly spaced afforested and unaltered grid cells are performed; afforestation is implemented in one out of every four grid cells in AFF1/4 and in three out of every four grid cells in AFF3/4. The mechanisms for the local and nonlocal effects are examined through the decomposition of the surface energy balance. The results show that the local effects dominate surface temperature responses to afforestation in China, with a cooling effect of approximately −1.00°C for AFF1/4 and AFF3/4. In contrast, the nonlocal effects warm the land surface by 0.14°C for AFF1/4 and 0.41°C for AFF3/4. The local cooling effects mainly result from 1) enhanced sensible and latent heat fluxes and 2) decreases in downward shortwave radiation due to increased low cloud cover fractions. The nonlocal warming effects mainly result from atmospheric feedbacks, including 1) increases in downward shortwave radiation due to decreased low cloud cover fractions and 2) increases in downward longwave radiation due to increased middle and high cloud cover fractions. This study highlights that, despite the unexpected nonlocal warming effect, afforestation in China still has great potential in mitigating climate warming through biophysical processes.

Pielke et al., 2011). The impacts of deforestation on surface temperature vary spatially and temporally depending on the background climate (de Winckler et al., 2017). Deforestation, on the one hand, cools the surface due to increased albedo and, on the other hand, warms the surface as a result of decreased aerodynamic roughness and evapotranspiration (