Earth observation embeddings improve wildfire burned area mapping accuracy
Annual Earth-observation embeddings encode wildfire disturbance and support simplified burned area mapping
Computer Vision and Pattern Recognition
Summary
Tracking the areas burned by wildfires using satellite images is important but can be complicated and slow. The authors found that special yearly summaries of satellite data, called embeddings, keep enough information to identify burned areas without needing detailed fire-specific images or lots of time-series data. These summaries allowed simple models to detect burned regions as well or better than traditional methods. They even worked well across different continents and helped estimate when fires started.
What this means in practice
- •For environmental monitoring teams: Map burned areas from wildfires more efficiently without needing detailed fire-specific imagery or dense satellite time series.
- •For natural disaster response units: Rapidly detect and estimate timing of wildfire events even in regions without local fire data for model training.
Authors
Jovana Knezevic, Clement Atzberger, Zhengpeng Feng, Adam F. A. Pellegrini, Srinivasan Keshav, David Coomes
Abstract
Medium-resolution (10-30 m) burned area mapping is vital for monitoring wildfires and their impacts, but remains difficult to scale. Existing methods require either curated fire-specific imagery or dense time-series analysis. Here, we tested whether annual Earth-observation embeddings retain wildfire disturbance signals sufficiently to map burned areas without either requirement. Using Tessera and AlphaEarth embeddings, we tested individual burn-scar delineation, mapping of all same-year fires within an area, regional wall-to-wall mapping, cross-continental transfer, and intra-annual fire timing. Tessera strongly encoded wildfire disturbance, allowing even linear models to separate burned from unburned pixels; the signal was weaker in AlphaEarth. Models trained on a single Tessera embedding matched or exceeded equivalent models using paired pre- and post-fire HLS imagery, and outperformed post-fire imagery alone. The same approach mapped all same-year fires within benchmark scenes (F1 = 0.90). Applied across California, with no California fire data used for downstream training, it recovered 97% of reference burned area and detected substantially more small and medium-sized fires than GABAM or MCD64A1. Separately, a model trained on 2018-2021 US fires transferred without retraining to 88 European fires from 2024-2025 (F1 = 0.88). For well-detected fires, ignition timing was recovered with a mean absolute error of 13 days. Performance declined for fires ignited near the end of the calendar year, and wall-to-wall deployment produced systematic false positives in some unseen landscapes. Annual embeddings nevertheless achieve high segmentation accuracy while moving the burden of dense time series processing upstream, providing a promising path towards simpler regional burned area mapping.