Soil organic and inorganic carbon pools in the Loess Plateau: New estimation, change and trade-offs

Soil carbon pool accurate estimation is important for land use management and coping with future climate change. However, the size, distribution, and drivers of soil organic (SOC) and inorganic (SIC) carbon lack precise quantification regionally. We developed ensemble machine learning models to quantify SOC, SIC, and soil total carbon (STC) stocks across the Loess Plateau (LP) using a novel, spatially extensive dataset of 241 sites with 723 paired SOC–SIC measurements. The LP’s soils stored 6.5 Pg C as SOC, 13.4 Pg C as SIC, and 19.9 Pg C as soil total carbon (STC) within the top 2 m, with 22%, 16%, and 18% of these respective carbon pools distributed in the top 30 cm. Changes in SIC stocks are crucial for carbon accounting. Soil pH drives trade-offs between SOC and SIC, thereby playing a critical role in determining the contribution of SOC and SIC to net change in STC. Under future scenarios, soil acidification associated with climate change to terrestrial ecosystems will reduce LP’s STC (2 m) by 26~42 Tg C by 2100. Our study underscored the necessity of concurrently estimating changes in both SOC and SIC stocks to accurately assess soil carbon pool and formulate effective climate change mitigation strategies.