Session: Using Machine Learning to Quantify and Improve Earth System Predictions
PROcess-guided deep learning and DAta-driven modelling (PRODA) to uncover key patterns and mechanisms in global soil carbon cycle
Wednesday, August 4, 2021
Link To Share This Presentation: https://cdmcd.co/gQ6Gg3
Feng Tao and Xiaomeng Huang, Department of Earth System Science, Tsinghua University, Beijing, China, Yuanyuan Huang, CSIRO Oceans and Atmosphere, Aspendale, Australia, Bruce A. Hungate, Center for Ecosystem Science and Society, Northern Arizona University, Flagstaff, AZ, Xingjie Lu, Sun Yat-sen University, Guangzhou, OK, China, Toby D. Hocking, Northern Arizona University, Flagstaff, AZ, Umakant Mishra, Sandia National Laboratories, Livermore, CA, Gustaf Hugelius, Department of Physical Geography and Quaternary Geology, Stockholm University, Stockholm, Sweden, Yiqi Luo, Center for Ecosystem Science and Society (ECOSS), Northern Arizona University, AZ
Department of Earth System Science, Tsinghua University Beijing, China
Background/Question/Methods Soil carbon storage is a vital ecosystem service. Yet mechanisms that regulate global soil organic carbon (SOC) dynamics remain elusive. Here we explicitly retrieve the spatial patterns of key biogeochemical mechanisms and their regulation pathways on SOC storage using the novel PROcess-guided deep learning and Data-driven modelling (PRODA) approach. PRODA integrates data assimilation, deep learning, big data with 54,073 globally distributed vertical SOC profiles, and the Community Land Model version 5 (CLM5) to best represent and understand global soil carbon cycle. Results/Conclusions The PRODA-optimised CLM5 can represent 56±2% spatial variation of SOC across the world. Among all the mechanisms we explored in this study, microbial carbon use efficiency (CUE) emerges as the most critical regulator of global SOC storage, distribution, and turnover time. High CUE favours global SOC accrual by increased turnover time. Conversely, increasing environmental modifiers (i.e., increased temperature and more favourable soil moisture) relaxes restriction on SOC decomposition, diminishing global SOC storage. Both carbon input distributing and substrate recalcitrancy play minor roles in regulating SOC than CUE. All the key mechanisms present significant dependence on local environments. We further identify soil texture as the most important environmental variable, followed by climate, in determining these biogeochemical mechanisms in the soil carbon cycle. We conclude that how soil microbes utilize organic carbon is central to SOC stabilization. Understanding detailed processes underlying CUE and its environment dependence will be critical in accurately describing global soil carbon dynamics under climate change.