FireGPT: A Vision Language Model for Prescribed Burn Decision Support Using NASA Earth Science Data

OpalAI proposes to develop FireGPT, an innovative prescribed burn decision-making tool that integrates NASA Earth Science Data with commercial remote sensing datasets to revolutionize wildland fire management. The key aspects include: Purpose: To create a Vision Language Model (VLM) decision support tool for prescribed burn management and wildfire impact assessment. Intended use of funding: Develop quasi-real-time, on-demand fuel maps using VLMs Translate fuel maps into actionable prescribed burn decisions Evaluate the approach through partnerships with burn managers in the FASMEE program Target markets: State Government Fire Agencies (e.g., CALFIRE) U.S. Forest Service (USFS) and Fire and Smoke Model Evaluation Experiment (FASMEE) Insurance companies and real estate developers The proposal aims to leverage recent advancements in multi-modal large language models and new remote sensing datasets (e.g., NASA's NISAR, UAVSAR, Landsat, ECOSTRESS, and commercial SAR/EO data) to provide more accurate predictions of surface fuels. This will enable better-informed decisions on prescribed burns to prevent fire propagation. The innovation addresses the critical need for timely and accurate information in wildfire management, offering potential benefits in ecological preservation, economic savings, and public safety. The proposed solution integrates various data sources with advanced AI capabilities to provide comprehensive insights for wildfire risk mitigation and post-fire recovery planning. By the end of Phase I, OpalAI aims to deliver a comprehensive technical document and a fully functional software prototype, demonstrating the feasibility and practical application of their decision-making tool for wildfire management.