Spatial distribution, drivers, and future variation of soil organic carbon in China’s ecosystems: A meta-analysis and machine-learning assessment

The characteristics of soil carbon pools across various ecosystems remain uncertain under different Shared Socioeconomic Pathways (SSPs) in China. Here, we conducted a meta-analysis of existing data and integrated machine-learning models to project spatiotemporal changes in soil organic carbon density (SOCD) by 2050 and 2100 under three Coupled Model Intercomparison Project Phase 6 (CMIP6) climate scenarios. For China’s terrestrial ecosystems, national average SOCD was 4.08 kg C m2 in the 0–20 cm soil layer and 9.42 kg C m2 in the 0–100 cm layer, with corresponding carbon stocks of 39.18 Pg C and 90.46 Pg C, respectively. Wetlands exhibited the highest SOCD but contributed minimally to total carbon stock due to their limited area, while forests and grasslands served as the dominant carbon reservoirs, particularly in deeper soils. Spatially, SOCD was highest in northeastern China and the eastern Qinghai–Tibet Plateau, and lowest in northwestern arid regions. Climate was the most critical determinant of SOCD in both soil depth, although its explanatory power was relatively weaker for deep SOCD, where soil factors gained prominence. Human activities significantly reduced surface SOCD in forests and grasslands. Future climate change would exacerbate the decline of surface SOCD, particularly in wetlands, posing substantial challenges to achieving the 4 per mil initiative goals.