Multicountry dataset and model improve wildfire spread forecasts over days
From Next Day Wildfire to Next Days Wildfire: A New Multi-Continent Dataset of Active-Fire Spread for Sequence-to-Sequence Forecasting
Computer Vision and Pattern Recognition
Summary
Predicting how wildfires will spread is important for safety and planning. Existing data and models focus on large fires lasting weeks and often come from similar places. The authors created a new, bigger dataset of smaller, shorter fires from many countries and a new model that predicts fire growth day by day, not just for the next day. Their model uses less computing power but makes better forecasts for several days ahead.
What this means in practice
- •For wildfire management teams: Use the new dataset and model to improve day-by-day forecasts of wildfire spread across different environments and fire sizes, enabling better resource allocation.
- •For environmental monitoring organizations: Incorporate the autoregressive forecasting method to generate longer-range wildfire spread predictions from satellite data, enhancing monitoring capabilities.
Authors
Jose Guillen, Renzo Zanca, Valentin Barriere
Abstract
Benchmarks such as WildfireSpreadTS (WSTS) have enabled data-driven forecasting of wildfire spread from multimodal satellite time series, but they suffer from limitations: size restricts them to large fires that burn for weeks, and they are from similar environments. We introduce AFSeq, a multimodal dataset of 1,194 recent fire events (2023 to 2026) from 8 different countries across three regions, dominated by the short, small fires that existing benchmarks underrepresent. Our dataset is the largest and most diverse to date for time-series-based wildfire spread prediction. We also introduce UTAE-AR, an extension of a classical U-Net with Temporal Auto Encoder, enhanced with recurrent modules before the decoder. This allows the model to not only predict the next day but the whole sequence in an autoregressive way. Moreover, our model can feed its own predicted fire mask back one day at a time, producing the day-by-day forecasts that single-step models cannot. Our results show that UTAE-AR surpasses state-of-the-art on both WSTS, WSTS+ and AFSeq for next-day accuracy while having 21.5% of their parameters, and being able to predict confidently at a longer horizon.