Underestimation of soil organic carbon by global spatially explicit soil datasets in Chinese urban greenspaces

Soil organic carbon (SOC) in urban greenspaces can provide important ecosystem services for a large number of urban population around the world. However, the distribution, stocks, and drivers of SOC change in urban greenspaces are relatively unknown compared with those of natural ecosystems. To address this knowledge gap, we utilized 744 SOC measurements (0-30 cm depth) collected from urban greenspaces across 148 cities in China to investigate the key the distribution and drivers of surface SOC stocks. Our analysis revealed weak correlations between in-field SOC measurements and estimates derived from widely used global SOC datasets (<i>e.g.,</i> SoilGrids250m v2.0 and HWSD v2.0), underscoring the necessity for improved characterization of SOC distribution and its determinants in urban greenspaces. By employing Random Forest (RF) and SHapley Additive exPlanations (SHAP), we found that the enhanced vegetation index (EVI), mean annual precipitation (MAP), and social-economic status indicated by night-time light density (NLD) and population density (POP) were the most important predictors of surface SOC in Chinese urban greenspaces. RF model-predicted surface SOC in urban greenspaces was higher in moderately developed cities than in underdeveloped or highly developed cities. SOC increased with EVI, but displayed a hump-shaped trend with MAP, NLD, and POP. Spatial prediction of SOC stocks from the RF model showed that there were 0.21 (0.20-0.22, 95% confidence intervals) Pg C in the top 30 cm of soil in urban greenspaces of China, which was 17% and 50% greater than estimated by SoilGrids250m v2.0 (0.18 Pg C) and HWSD v2.0 (0.14 Pg C), respectively. Our study provides valuable data support and scientific guidance for carbon neutrality and urban greenspaces ecosystem management in China.