Efficient and accurate active fire detection is crucial for timely firefighting and mitigating hazards. Geostationary satellites deliver high-frequency observations that offer valuable data for near-real-time fire monitoring. However, current operational fire detection algorithms often underutilize temporal information, failing to decouple fire-induced anomalies from inherent surface thermal heterogeneity, which results in frequent false alarms. To address this limitation, we constructed a ten-day historical background brightness temperature (BT) reference database from multi-year Himawari-8/9 data, serving as a stable, fire-undisturbed baseline. Based on this, an active fire detection algorithm was developed that integrates instantaneous spatial–contextual differences with historical deviations of these differences from the reference database. Evaluated against a robust dataset of over 55,000 fire pixels (cross-verified using 10 m Sentinel-2 burn-scar data), the proposed algorithm significantly outperforms the Himawari-8/9 Wildfire (WLF) product, achieving a commission error (CE) of 2.9%, an omission error (OE) of 37.5%, and an F1-score of 0.76. The framework demonstrated superior detection accuracy in challenging scenarios such as low-temperature, smoke-obscured, and early-stage fires, while maintained robust performance across diverse fire types. The approach enables rapid full-disk fire detection in less than one minute and can be adapted to other geostationary satellites, providing a technical foundation for building a globally coordinated fire monitoring system.
Liu et al. (Mon,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: