Including Small Fires in Global Historical Burned Area Products: We Should Dig More into the Landsat Archive

Despite advancements in satellite-derived fire analysis, reconstructing the past remains constrained by historical data limitations. Algorithms for extracting burned area information at the global scale continue to evolve, but complex landscapes and small fires are often excluded, mainly due to sensor spatial resolution. We evaluated the burned area detection capabilities of two recent global products, MODIS FireCCI51 and the Landsat-based GABAM, in a challenging mountainous region over the period 2001-2019. Overall, the spatio-temporal distribution of burn counts within 10 km mesh grid covering the study area correlated well for GABAM (Spearman R: 0.76 and 0.74) and discretely for FireCCI51 (R: 0.53 and 0.45) against two benchmark datasets. Fire event detection performance in the sampled squares in this specific landscape was limited for FireCCI51, despite its reported improvements globally (User’s Accuracy: 0.83, Producer’s Accuracy: 0.08). Conversely, GABAM exhibited relatively strong detection capabilities with reduced commission errors (User’s Accuracy: 0.85, Producer’s Accuracy: 0.68). This evaluation highlights the importance of Landsat-based approaches for global burned area assessments. Landsat's long, consistent time-series and higher spatial resolution offer significant advantages, as reported by the numerous local and regional applications. Yet, its potential for global assessments has been underutilized, hindered by difficulties in handling big amounts of data and the scarcity of analyzable images within a fire year. Nonetheless, the development of GABAM may serve as a proof of concept, demonstrating how the Landsat archive and powerful cloud computing can enhance global burned area mapping, improving accuracy, including small fires, and extending time-series. We encourage researchers to integrate Landsat into global fire extraction routines for comprehensive past fire history reconstruction.