Active fire detection with Meteosat Third Generation satellite

Fire detection and monitoring are critical for mitigating environmental and societal damage. In recent decades, Earth Observation satellites have played a role in fire detection, with polar-orbiting sensors like VIIRS offering global coverage and fire detections twice a day in each region. However, their limited revisit time constrains time at which fire detection occur. To address this, geostationary satellites have emerged as an alternative due to their high temporal resolution, enabling observations every few minutes. This advantage comes at the cost of spatial detail; for instance, Meteosat Second Generation (MSG) satellites provide 15 minute imagery but with a coarse resolution of ∼ 3 km at nadir, reducing their effectiveness for fire detection. This study explores the fire detection capabilities of Meteosat Third Generation (MTG) through two algorithms. MTG is a geostationary satellite with 1 km spatial resolution and 10-minute temporal resolution. Leveraging both the satellite’s spatial and temporal resolution can enhance the efficiency of fire monitoring. The first fire detection algorithm, "MTG FRP-pixel", provides a map of active fire detections and their Fire Radiative Power (FRP) rates. This approach is based on single-timestamp spectral and spatial image processing to derive anomalies in the infrared 3.8 μm band. The second fire detection method employs multi-temporal analysis to extract fire events as temporal anomalies, taking advantage of the high frequency of MTG data acquisition. The results show that The MTG FRP-pixel algorithm demonstrates high precision of approximately 0.9 compared to VIIRS satellite fire detection which has a spatial resolution of 375 m. In addition, the algorithm successfully detects 50% of VIIRS fire events at a power of 20 MW and begins detection at low FRP values around 5 MW in Europe. Concerning the multi-temporal fire detection algorithm, the study shows that it increases the number of fire detections compared to FRP-pixel, improving recall from approximately 0.25 to 0.39 for a specific region, with only a small decrease in precision from 0.95 to 0.91, thus providing an alternative for enhanced detection capabilities. However, it requires significant storage capacity to extend it to larger spatial coverage. On the whole, MTG provides excellent capabilities for fire detection, either with the FRP-pixel algorithm for single timestamp analysis with memory-efficient processing, or with an increased number of fire detections using multi-temporal analysis but with higher processing requirements.