Afforestation Surpasses Abandonment in the Recovery of Post‐Agricultural Soil Organic Carbon in China as Estimated by Machine Learning Models

The surface soil organic carbon (SOC) dynamics typically follow a trend of initial loss followed by subsequent accumulation after cropland abandonment. However, the timing of SOC stock increase (referred to as the threshold in this study) remains insufficiently explored in spatial terms. While afforestation incentives are considered an effective nature-based strategy, the benefits of SOC recovery still require further investigation. Here, we integrated 3502 SOC observation datasets from 211 publications, 41 spatially explicit environmental covariates, and interpretable machine learning models to simulate early-stage SOC dynamics at the pixel level across China's abandoned cropland. We then identified the threshold and quantified surface SOC stock loss (0-30 cm). Our simulation results show that the SOC threshold of cropland abandonment is 3.77 ± 0.91 (standard deviation) years, during which 3.13 ± 0.52 Mg C ha^-1 of SOC is lost. Afforestation has the potential to advance this threshold by 47.7% (approximately 1.97 ± 1.36 years), though with significant spatial variation. Further analysis indicates that afforestation also has the potential to mitigate SOC loss by 1.75 ± 0.50 Mg C ha^-1 (approximately 10.55 ± 3.01 Tg C across all of China's abandoned cropland) during the initial stage of the post-agricultural period. Our findings highlight afforestation incentives as an effective strategy for accelerating SOC recovery and preventing SOC loss in the post-agricultural period.