Research
I build environmental measurements that do not yet exist, using satellite data and machine learning, and use them to answer cause-and-effect questions about policy. Below are some of the research areas I work on.
Floods
I built Global Flood Mapper for SAR-based flood mapping, and use deep learning to separate natural flooding from managed inundation. My dissertation studies how flood risk is perceived versus actually assessed, and its causal impacts on agricultural livelihoods, in India. This work is supported by a NASA FINESST award.
Agriculture
I generate high-resolution agriculture maps in smallholder, data-scarce settings, working on field boundary delineation and crop type classification using geospatial foundation models, including one of India's first high-resolution crop type maps.
Impact Evaluation
Most policy questions turn on what would have happened otherwise, and in the real world that counterfactual is almost never observed. I build machine learning models to generate it. One is a physics-aware transformer that predicts the deforestation the Amazon would have seen without protection, which is what makes it possible to measure what protected areas actually achieved.
Urban Analysis
Earlier in my career I worked on urban growth modeling, building height estimation, road network analysis, and air pollution drivers. Urban systems remain a lens I bring to new research questions.
Funding
- NASA FINESST ($150,000), 2025–2028. High-Quality Flood Mapping Using SAR Sensors: Evaluating Uncertainty in Existing Global Flood Exposure Estimates.
- Goodchild Fellowship in GIS ($3,400), 2026. Hidden Pitfalls in Geospatial AI.
- UCSB Department of Geography Summer Research Fellowship ($8,130), 2024.
- Schmidt Family Foundation Research Accelerator ($8,000), 2023. Mapping Smallholder Agriculture Field Boundaries in the Indo-Gangetic Plain, India.
- Benioff Applied Environmental Science Summer Fellowship ($6,000), 2023. Using Satellites and AI to Map Smallholder Agriculture Field Boundaries With Limited Training Data.
- GEO – Microsoft Planetary Computer ($120,000), 2021. India Flood Monitoring and Impact Assessment Portal, as Co-Investigator.
I am always looking to collaborate with people working in these areas, especially my current research interests in floods and agriculture, geospatial foundation models, and causal analysis. If this sounds like you, feel free to reach out to ptripathy@ucsb.edu.
