Increasing Situational Awareness Through Real-Time Classification of Amazon Fire Events

Abstract:

Each year, satellite sensors detect millions of active fires in the Amazon and surrounding biomes. However, the number of active fire detections provides an incomplete understanding of the type, size, duration, and behavior of fires in the Amazon region needed to guide fire management and estimate the Earth system and air quality impacts of Amazon fire emissions. Here, we cluster VIIRS 375 m active fire detections from the Suomi-NPP and NOAA-20 satellites into individual fire events, and classify each event based on metrics of fire behavior and ancillary land cover information. Our approach separates deforestation fires, understory forest fires, small clearing and agricultural fires, and savanna fires in near-real time (https://globalfiredata.org/pages/amazon-dashboard/). Tracking fire events, rather than fire detections, yields three important advances. First, the main fire types in the Amazon have distinctive characteristics in terms of fire radiative power (FRP), fire persistence, fire spread rates, fractional tree cover, and association with deforestation in previous years. Given diverse fire management responsibilities in the Amazon region, this rapid separation of fire events by fire type can help the responsible agencies to target specific fire events. Second, by tracking fire events over time, it is possible to separate daily active fire detections into new fire starts and ongoing fire events. By August 2019, 80-85% of daily VIIRS active fire detections in the Amazon were associated with ongoing fire events, aiding the real-time classification. Rapid identification of new fire events is also critical to mitigate damages before fires grow large. Third, the behavior of individual fire events offers new constraints on fire carbon emissions from the Amazon region. Fire type provides a strong constraint on fuel loads, and metrics of fire behavior (size, persistence, spread rate, and duration) influence the total fuel consumption over the lifetime of the fire event. Together, these advances support more rapid and more informed decision-making regarding fire suppression, greater transparency regarding drivers of fire activity, and improved estimates of fire carbon emissions from the Amazon region.