Past, Present, and Future of the Monitoring Trends in Burn Severity (MTBS) Project

Beginning in 2006, the Monitoring Trends in Burn Severity (MTBS) program has generated over 17,000 Landsat-based maps and assessments of wildland and prescribed fires that occurred in the United States since 1984. These products support local, regional, and national assessments of fire effects, emissions, risks, regimes, and post fire fuels inventories. Currently, MTBS relies upon state and federal fire occurrence data (FOD) to guide Landsat scene selection. Analysts manually review Landsat low resolution browse images to select the best scenes to assess each fire. This process is labor intensive and the FOD suffers from spatial and omission errors.  To address these issues, we developed a process to automatically identify burned areas on Landsat imagery. From these burned/no burn (BNB) classifications we create fire perimeters and label them with a start date.  BNB images are created using CUBIST derived models that use MTBS training data to classify burned areas in Landsat scenes. Hazard Mapping System (HMS) fire detection data is incorporated by using a series of geospatial queries in PostgreSQL to assign fire start dates to the burn perimeter based its proximity to the HMS fire detection.  Subsequent automated processes 1) use the fire start date to determine and order many candidate Landsat pre- and post-fire scenes, 2) clip the Landsat scenes to the area around fire perimeter, and 3) bundle all the clipped imagery into easily viewable package.  These high resolution image clips are reviewed by analysts to determine the best pre- and post-fire scenes.  Successive automated procedures will create differenced Normalized Burn Ratio (dNBR), relativized dNBR (RdNBR), and thresholded burn severity images. These outputs will then be reviewed by analysts to determine their validity.  Overall, these automations will enable MTBS to map more fires.

Read more >