Skip to Content

On Improvement of Soil Organic Carbon Modeling and Simulation viaIntegrated Deep Learning and Data Assimilation Approaches

Active Dates 9/1/2022-6/30/2024
Program Area Atmospheric System Research
Project Description
Microbes acquire energy and nutrients from soils and release CO2 and other gases to atmosphere. The next generation earth system model strives to integrate microbes to improve its accuracy. The increasing available dataset can help reduce the model uncertainty by a Bayesian Markov Chain Monte Carlo (MCMC) data assimilation approach. Despite the progress, to deepen the use and value of large dataset, this project is proposed to combine a deep learning technique and MCMC data assimilation to further improve the model performance. Based on a former success in integrating dataset of Harvard Forest (HF), this project will integrate other covariates such as environmental and edaphic dataset at HF that have not been considered formerly. The current proposed study will integrate deep learning, data assimilation, a highly comprehensive monitoring dataset of plants, soils and microbes, and a microbial model to optimize the model representation of soil carbon over two decades at HF. This project is expected to identify more accurate microbial models and will likely provide alternative model structures to scale microbial functions to regional and global scales. Through the collaborations with ORNL scientists, the HBCU faculty member and students will promote their research capacity. Ultimately, work force from minority groups can be prepared to help advance soil model development.
Award Recipient(s)
  • Tennessee State University Nashville (PI: Li, Jianwei)