Hyperspectral Remote Sensing of Soil Organic Carbon Using Machine Learning

Faculty Lead: Sherrie Wang, assistant professor in the MIT department of Mechanical Engineering; Institute for Data, Systems, & Society (IDSS); and Laboratory for Information & Decision Systems (LIDS). Principal Investigator of the Earth Intelligence Lab.

Soils are the largest terrestrial carbon reservoir, containing approximately 2,500 gigatons of carbon–more than three times the atmospheric carbon and four times that in all plants and animals. Monitoring and managing soil organic carbon (SOC) is essential for enhancing carbon sequestration and reducing greenhouse gas emissions. Although accurate SOC mapping is challenging due to spatial heterogeneity and the high cost of traditional soil sampling, hyperspectral satellite imagery offers a promising solution as it provides detailed information on soil properties over large areas with high spatial, temporal, and spectral resolution. This study will use hyperspectral satellite data and machine learning to map SOC in the United States, India, and Kenya. By combining data from government and private sources, we aim to identify spectral features for SOC prediction and assess their generalizability across regions. Our goal is to develop a robust predictive algorithm for global SOC mapping and enable the quantification of SOC changes over time to better understand soil carbon dynamics and inform climate change mitigation efforts. This project’s outcomes have the potential to drive significant advances in digital MRV technologies, contributing to the broader aim of decarbonizing agriculture and improving ecosystem resilience.