UCI-NASA Wildfire Event Tracking Using Satellites Fire Detections
This website describes some of the collaborative research between UCI and NASA focused on wildfire tracking and prediction using satellite-based observations of active fires and burned areas. This platform provides an overview of the different fire tracking algorithms we have developed, the datasets they have generated, and the scientific studies that have leveraged these data to advance our understanding of wildfire behavior and impacts.
Fire Event Data Suite (FEDS)
The FEDS algorithm tracks wildfires using 375-m resolution data from the Visible Infrared Imaging Radiometer Suite (VIIRS) active fire detections. At each half-daily time step, fire pixels are clustered together into groups based on spatial proximity and appended to an existing active fire event or assigned to a new one. This automated process continuously updates each fire’s attributes, outlines the fire perimeter, and identifies the active fire front in near real-time, following the most recent satellite data acquisition.
A detailed description of the FEDS algorithm is provided in Chen et al., 2022. The FEDS algorithm applies an alpha hull algorithm to encapsulate a cluster of active fire observation within a given shape (Hantson et al., 2022; Pateiro-Lopez et al., 2019). The algorithm also decides where and when to merge different fire objects, depending on ecosystem-specific parameters. Systematic tracking of the growth of fire perimeters for large wildfires in California has allowed our team to link fire behavior (i.e., spread rate) to post-fire impacts. This work has revealed that faster moving fires release more energy and kill more trees (Hantson et al., 2022).
FEDS California
FEDS California is optimized to monitor wildfire-prone regions across the state. Using this system, we mapped the wildfire history of California from 2012 to 2020, providing a comprehensive view of fire activity over nearly a decade.
For a detailed description of the FEDS California dataset, see Chen et al., 2022. The dataset and the initial version of the FEDS Python code are available in this Figshare repository.
A large fire dataset (2012-2023) derived using an updated version of FEDS (FEDS2.5) is also available for download here. Main improvements in FEDS2.5 encompass a better data structure for large fires, and the integration of fire spread history of merged fires.
Arctic–Boreal Fire Atlas
We modified the FEDS algorithm to map the sub-daily progression of all circumpolar Arctic–boreal fires from 2012 to 2023. We used this information to classify the Arctic–boreal domain into seven distinct ‘pyroregions’ based on fire behavior and ecosystem characteristics. This analysis revealed a stark variability in Arctic-boreal fire regimes and their response to a warming climate, which was driven by human and environmental influences (Scholten et al., 2024).
Key changes to the FEDS algorithm for the Arctic-boreal domain from the algorithm described by Chen et al. (2022) included the implementation of boreal-specific land cover thresholds for merging fire clusters, a peat fire module that allowed for extended smoldering periods of up to 30 days and the exclusion of surface water features to minimize commission errors in the fire perimeters. For ABFA, the fire tracking was performed in a projected coordinate system for accurate distance calculations, and observations were aggregated into AM and PM using local solar time instead of the VIIRS day/night filter. The Arctic–boreal fire atlas data for 2012–2023 is available via the Pangaea repository. The fire tracking code used to generate the Arctic–boreal Fire Atlas is freely accessible on Zenodo.
CONUS fire perimeter API
The FEDS algorithm has been implemented in NASA’s MAP system by the Earth Information System-Fire (EIS) project to create fire perimeters in CONUS for historical fires and for near real-time (NRT) fire analysis. The NASA Wildfire Tracking Lab made an OGC API that allows users to access the NRT fire perimeters data through online request.
The access point for the CONUS fire perimeter API is https://firenrt.delta-backend.com. Here is a tutorial for using this API to explore and filter data.
GOFER
The Geostationary Operational Environmental Satellites (GOES) provide complementary information about wildfire spread. Although GOES active fire data are coarser in spatial resolution (2 km at the equator) compared to VIIRS (375 m), GOES provides dense temporal information at a frequency of 10-15 min across the full disk of North and South America. GOES can therefore provide insights into extreme fires during periods of rapid spread.
We adapted and optimized a fire tracking algorithm originally developed by Google for California using the GOES-R satellites, GOES-16/West and GOES-17/West. Key improvements that we made included correcting for terrain parallax effects in GOES-East and GOES-West, optimizing the thresholds to delineating the fire boundary, incorporating an early perimeter adjustment to anchor the fire progression close to ignition, and deriving active fire lines and fire spread rates. The resulting GOES-Observed Fire Event Representation (GOFER) algorithm enables hourly fire tracking of large wildfires and creates a product of hourly fire perimeters, active-fire lines, and fire spread rates.
Please refer to Liu et al., 2024 for more information. The GOFER product of the 28 fires in California from 2019 to 2021 is available on Zenodo. The code for the GOFER algorithm is available at Zenodo. Online visualization of the GOFER product is available on Google Earth Engine Apps.
FEDS Hourly
By combining FEDS 12-hourly data and the hourly FRP data from GOES, we performed spatial interpolation to derive the best guess hourly fire perimeters. This work is still in the experimental stage and the hourly perimeters are only available for limited large fires in California.
Global Fire Atlas (GFA)
In parallel to the FEDS algorithm, we also developed a GFA approach to tracks the dynamics of individual fires. The GFA approach was initially implemented using 500m MODIS burned-area data, and a global GFA dataset was created. Later, we implemented the GFA approach using VIIRS active fire detections to track and classify Amazon fire events in near real time.
Please refer to Andela et al., 2019 for the description of the GFA method. The global GFA product for 2003-2016 is available on ORNL DAAC. The GFA application in the Amazon using VIIRS active fires was described in Andela et al., 2022, and the data for 2019-2020 are available in this Zenodo repository. The updated data are also available for interactive exploration in the Amazon Fire Dashboard.
References
Andela, N., Morton, D. C., Giglio, L., Paugam, R., Chen, Y., Hantson, S., ... & Randerson, J. T. (2019). The Global Fire Atlas of individual fire size, duration, speed and direction. Earth System Science Data, 11(2), 529-552.
Chen, Y., Hantson, S., Andela, N., Coffield, S. R., Graff, C. A., Morton, D. C., ... & Randerson, J. T. (2022). California wildfire spread derived using VIIRS satellite observations and an object-based tracking system. Scientific Data, 9(1), 249.
Hantson, S., Andela, N., Goulden, M. L., & Randerson, J. T. (2022). Human-ignited fires result in more extreme fire behavior and ecosystem impacts. Nature Communications, 13(1), 2717.
Liu, T., Randerson, J. T., Chen, Y., Morton, D. C., Wiggins, E. B., Smyth, P., ... & Nevo, O. (2023). Systematically tracking the hourly progression of large wildfires using GOES satellite observations. Earth System Science Data, 16, 1395–1424.
Pateiro-Lopez, B., Rodriguez-Casal, A., & Pateiro-Lopez, M. B. (2019). alphahull: Generalization of the Convex Hull of a Sample of Points in the Plane. R Package Version 2.2.
Scholten, R. C., Veraverbeke, S., Chen, Y., & Randerson, J. T. (2024). Spatial variability in Arctic–boreal fire regimes influenced by environmental and human factors. Nature Geoscience, 17, 866–873.