Scalable Assessment of Soil Organic Carbon for Carbon Incentive Programs

This project developed and tested a machine learning framework to estimate soil organic carbon (SOC) in dryland agricultural systems using satellite imagery and environmental data. The work responds to a major limitation in carbon incentive programs, which typically rely on practice-based payments rather than direct measurement of soil carbon outcomes.

Using multispectral satellite data combined with soil, weather, and management variables, the team demonstrated that SOC in the top 30 cm can be predicted with meaningful accuracy. They also implemented uncertainty quantification methods that provide prediction intervals and guide more efficient data collection. Results showed that targeted, uncertainty-informed sampling can achieve similar model performance with roughly half the number of samples, reducing monitoring costs.

The project provides early evidence that scalable, outcome-based SOC monitoring is feasible. These findings directly support the development of more reliable carbon markets and incentive programs, while also strengthening future proposals for extramural funding.