Developing an object-oriented near-real time fire tracking system using VIIRS active fire detections

Abstract:

Active fires recorded by satellite remote sensors have been used to monitor fire development, evaluate fire risk, and estimate near-real time fire emissions. However, current systems mainly use aggregated or statistical information of the thermal anomaly associated with fire pixels, but ignore the internal spatiotemporal connections between them. Here we develop a novel object-oriented system that tracks the dynamics of individual fires, leveraging the newly available dual Visible Infrared Imaging Radiometer Suite (VIIRS) data streams from Suomi-NPP and NOAA-20. We aggregate daily 375 m VIIRS active fire detections into fire clusters according to their spatial proximity. Then each cluster is either appended to an existing fire object or forms a new object. This allows us to derive the fire shape and track the fire perimeter expansion on a daily time scale. The system also combines various datasets of active fires, land surface, and meteorology in order to automatically track a suite of fire object attributes, including pixel-level attributes of fire and surface properties, vector attributes related to the fire shape, and meta attributes corresponding to the whole fire object. In doing so, the system (1) provides a daily overview of global and regional fire situations and (2) enables dynamically tracking of each fire's spatiotemporal evolution. This also makes it possible to estimate daily emissions associated with each fire using high-resolution vegetation type, fuel load, and moisture information in the expanding area. The near-real time datasets created by this system will be particularly useful for simulating the transport of pollutants from specific fires. Lastly, this effort provides the scientific basis for dynamic emissions forecasts, in contrast to most current approaches that holds emissions constant over the meteorological forecast interval.