Methane Emission Estimation from CAFOs with Machine Learning
PI and co-PIs: Daniel Ho (Stanford University, United States); Elena Eneva (Stanford University, United States); Evan Cook (Stanford University, United States); Victoria Hollingshead (Stanford University, United States)
Funding amount: $150,000
Project overview: Mitigating agricultural methane has been identified as the strongest lever for slowing climate change over the next 25 years. Concentrated Animal Feeding Operations (CAFOs), also referred to as “intensive livestock farms”, contribute over a third of U.S. agricultural methane emissions, but no national agency maintains accurate data on the number, size, or location of CAFOs, leaving existing emissions estimates incomplete and under-reported. This project aims to address such critical data gaps by building a computer vision pipeline that analyzes satellite imagery to automatically detect livestock barns, estimate animal populations, and calculate facility-level methane emissions, beginning with hog farms in Iowa. Working in partnership with environmental advocacy organizations and communities disproportionately impacted by CAFO pollution, this work will equip regulators with tools for effective enforcement, enable advocates to hold facilities accountable, and provide communities with evidence to demand environmental protections. The project will also release an open-source codebase, trained models, and a comprehensive methane emissions dataset that can be adapted to other regions and livestock types, establishing a scalable framework for nationwide agricultural methane monitoring. Such work stands to transform how we measure and mitigate agricultural methane, enabling evidence-based climate policy in the food system.
Agriculture & Food Earth Observation & Monitoring Land Use Computer Vision & Remote Sensing Uncertainty Quantification & Robustness