Soil AI Lab
Soil AI Lab

Geospatial GPT for Enhanced Disaster Response and Agricultural Resilience

This project is developing a chat-based “Geospatial GPT” that helps farmers, extension agents, and other non-experts quickly access and interpret Earth Observation (EO) information after extreme events (e.g., hurricanes and flooding). Instead of navigating complex GIS platforms, users will be able to ask natural-language questions (e.g., “Where is flooding affecting my fields?” or “How is crop recovery trending compared to past seasons?”) and receive maps, statistics, and clear explanations tailored to decision-making.

Technical approach

The platform integrates (1) a cloud-based EO data cube / processing cyberinfrastructure to automate ingestion and analysis of satellite time series, (2) LLM-driven geospatial query translation (with retrieval-augmented workflows) that converts user questions into reproducible analytics, and (3) AI super-resolution to enhance spaceborne imagery toward field-scale (target ~1 m) decision support where feasible. The system will support robust validation and iterative improvement using user feedback and targeted ground-truth/field observations, ensuring outputs are reliable and usable for real-world response and recovery planning.

Expected outcomes

  • A conversational interface that democratizes EO analytics for non-experts and reduces “data-to-decision” latency.
  • Two core decision-support services: (i) flood extent/likelihood insights and (ii) crop damage and recovery benchmarking against prior seasons.
  • AI-enhanced satellite products (including higher-resolution outputs where appropriate) to improve field-level assessment and targeting of follow-up actions.

Impact

By lowering technical barriers and accelerating geospatial analytics, this project enables faster, more targeted disaster response and more resilient agricultural management at scale.

Outreach materials