Soil is the largest carbon pool of the terrestrial ecosystem, yet the mechanisms of soil organic carbon stabilization are not well characterized nor incorporated in Earth system models. We applied a novel approach that integrated data assimilation, deep learning, a big dataset with more than 100,000 vertical soil organic carbon profiles, and the Community Land Model version 5 (CLM5) to optimize the model representation of soil organic carbon.
Results/Conclusions
We found that global heterogeneity of processes representing carbon use efficiency, vertical mixing, and decomposition exist and are jointly controlled by climatic, edaphic and vegetation variables. Among all the processes investigated in this study, carbon use efficiency related processes contribute most to the variation of soil carbon storage and its turnover across the world. Without considering the heterogeneity of carbon use efficiency, a strong bias occurs in the model simulation. Our findings support the globally heterogeneous processes and the critical role of carbon use efficiency in regulating soil carbon stabilization. Incorporating spatially and potentially temporally varying soil carbon cycle processes will help to improve the Earth system models in predicting soil carbon sequestration potential in response to climate change.