FURI | Spring 2026

Machine Learning enabled Future Wildfire Spread Prediction using a Unified Spatiotemporal Dataset

Data icon, disabled. Four grey bars arranged like a vertical bar chart.

This project develops a unified geospatial dataset and machine learning framework for next-day wildfire spread prediction across the contiguous United States. Using Google Earth Engine, the dataset integrates wildfire boundaries, terrain, land cover, meteorological variables, and MODIS active-fire observations into spatially aligned 96 × 96 km samples with five days of historical inputs and next-day target masks. These data support the development of deep learning models for forecasting fire growth patterns. This work advances data-driven wildfire forecasting for improved monitoring, early warning, and fire-response decision-making.