Session summary
This GeoField 2026 research session presents papers connected to a special issue of the Journal of Development Economics on geospatial impact evaluation. The session explores how remotely sensed data can extend conventional evaluation designs by measuring outcomes that are difficult, dangerous, or expensive to observe directly.
Moderator: Karen Macours
Using Remotely Sensed Data to Assess War-Induced Damage to Agriculture
Klaus Deininger presents a study of war-induced damage to agricultural cultivation in Ukraine. Using freely available satellite imagery, hand-labeled Sentinel-2 data, machine learning, and village-level cultivated area measures, the paper estimates how conflict affected agricultural land use from 2019 to 2024. The study compares satellite-based damage measures with media-based conflict indicators and finds that remotely sensed data capture agricultural damage in less populated and harder-to-access areas that media-based indicators may miss.
Agricultural Technology Adoption and Deforestation
Jeff Bloem presents a study linking a randomized trial on urea super granule adoption in Nigeria to Earth observation outcomes. The paper examines whether a climate-smart rice fertilizer technology affected productivity and deforestation. Using pixel-level deforestation data, vegetation indices, and a probabilistic measure of treatment exposure, the study finds heterogeneous effects consistent with agricultural intensification: productivity gains appear stronger in less densely forested areas, while tree canopy loss falls in more densely forested areas.
Irrigation Infrastructure and Satellite-Measured Land Cultivation Impacts
Joel Ferguson presents an evaluation of irrigation infrastructure along the Senegal River. Using administrative shapefiles from SAED and Landsat imagery from 1988 to 2019, the study tracks cultivation before and after irrigation project completion. The results show large average increases in cultivation, but also substantial heterogeneity: some projects are consistently productive, while others are never used or are used intermittently. Follow-up farmer surveys point to water-access, engineering, drainage, salinization, finance, and producer-organization constraints as key reasons for underuse.
Together, the session demonstrates how geospatial data can enrich research designs that start from administrative records, randomized trials, or policy rollouts. Satellite imagery makes it possible to measure conflict damage in inaccessible areas, track forest loss around agricultural technology interventions, evaluate land acquisition outcomes, and study decades of irrigation performance. The broader lesson is that Earth observation is most powerful when paired with credible institutional data, careful treatment measurement, and follow-up evidence that explains why impacts vary across places.