AAGO

AAGO (AI Augmented Geospatial Observation) is a fire area prediction model built with a transformer encoder. The core idea: given a fire's current state and environmental conditions like wind speed, direction, and humidity, predict how it spreads.

Training data comes from synthetic fire simulations inspired by real wildfire satellite imagery, converted into heat intensity maps paired with meteorological data. We generated 1,000 training and 100 test datasets using a grid-based probabilistic BFS simulation.