State of Wildfires 2025-26: Population Exposure to Wildfire

We developed a fire-level population exposure dataset by overlaying individual fire boundaries from the Global Fire Atlas (GFA; version 20260408) with gridded population maps from WorldPop at approximately 100 m resolution. For each fire, we used the WorldPop population layer corresponding to the fire start year to estimate direct population exposure, defined as the population located within the burned-area boundary of each individual fire. For fires that occurred during 2002–2020, we used the population map from the same year as the fire start date, while for fires that occurred during 2021–2026, we used the 2020 WorldPop population map because later years were not available at the time of analysis. To reduce impacts from non-wildfire records, we applied a filtering procedure designed to remove agricultural burns, persistent industrial heat sources, urban false detections, and low-vegetation hot-desert artifacts. This procedure used MODIS-based annual land cover, daily active fire detections, and vegetation indices to exclude fire records with more than 40% urban or agricultural land cover within the fire perimeter; records with very low average active-fire activity, indicating likely false burned-area detections, or unusually persistent active-fire activity, indicating potential industrial hot spots; and records with very low annual vegetation index values, indicating unvegetated or low-productivity areas where heat signatures may be misclassified as burned areas. For more information, refer to Seyd Teymoor Seydi et al. Increasing global human exposure to wildland fires despite declining burned area.Science389,826-829(2025).DOI:10.1126/science.adu6408 Wildfire exposure estimates are aggregated to spatial units, such as counties, IPCC regions, and biomes, based on the month of each fire’s start date for the period 2002–2026. For 2026, the dataset includes only fires available in the specified GFA version.