Enhancing whole-profile soil organic carbon predictions in croplands through a depth-resolved modelling approach

Soil organic carbon (SOC) enrichment in agricultural soils plays a vital role in supporting climate-smart sustainable crop production. Process-based agricultural system models are key tools for assessing the whole-profile SOC dynamics to help identify proper agricultural management practices. However, the depth-dependent characteristics of SOC turnover are often overlooked in these models, leading to substantial uncertainties in SOC predictions. Here, we evaluated the capabilities of the Agricultural Production System sIMulator (APSIM) to predict multi-layer SOC dynamics using data from five long-term field experiments across the main wheat and maize producing regions in China. Our results suggested that incorporating a depth-modifier for SOC decay rates significantly improved APSIM's performance in predicting the vertical distribution and temporal dynamics of SOC, with the coefficient of determination (R 2 ) being increased from 0.75 to 0.93 and relative root mean square error being decreased from 0.2 to 0.07. Specifically, the maximum SOC decay rates were predicted to decrease with increasing soil depth, though the decreasing rate varied widely across the experimental sites. This depth-resolved modelling approach has implications for predicting whole-profile SOC dynamics in response to nitrogen fertilization, tillage and residue management scenarios. Our findings demonstrate the importance of depth-resolved modelling approach to enhance the reliability of whole-profile SOC predictions, thereby informing effective management strategies. • Modelling vertical gradients in SOC decay improves the accuracy of depth-specific SOC predictions. • Depth-resolved SOC modelling modified the simulated effects of agricultural practices. • Subsoil SOC dynamics influence crop-soil interactions, thus agricultural management practices. • Deep tillage may promote subsoil SOC sequestration but further investigation is warranted.