This project tackles a persistent and under-measured environmental challenge: agricultural residue burning, a major source of air pollution and greenhouse gas emissions. Existing satellite monitoring systems are not precise enough to detect fires at the scale of individual fields, limiting the effectiveness of both enforcement and policy responses. This research addresses this gap by developing an interpretable artificial intelligence framework that integrates high-resolution satellite imagery with on-the-ground validation data.
The innovation lies in combining advanced machine learning with transparency and real-world validation. Unlike traditional “black box” models, this approach allows fire detections to be traced back to visible environmental indicators such as burn scars and smoke patterns. It also builds scalable geospatial data infrastructure, creating a reusable system that can support broader environmental monitoring efforts beyond the initial study region.
The broader significance of this work is substantial. More accurate and trustworthy detection of agricultural burning can improve air quality management, support climate mitigation strategies, and strengthen emissions tracking systems. It also lays the groundwork for more reliable climate data systems that could inform policy, enable carbon accounting, and support emerging climate finance mechanisms. Ultimately, the project demonstrates how interpretable artificial intelligence (AI) can be applied in a practical, policy-relevant way to address complex environmental challenges with global implications.
