Assessment of pre and post fire remotely sensed data products across spatial and temporal scales for prescribed fire operations in rangelands

Authors: WOODRUFF, CRAIG - Wilfrid Laurier University; Clark, Patrick; Olsoy, Peter; ENTERKINE, JOSH - Boise State University; BYRNE, AKIRA - Boise State University; SCLAFANI, MICHELLE - Idaho State University; HUBER, DAVID - University Of Texas - El Paso; PARK, JOHN - Orise Fellow

Submitted to: Fire Ecology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 7/28/2026
Publication Date: 8/4/2026
Citation: Woodruff, C., Clark, P., Olsoy, P.J., Enterkine, J., Byrne, A., Sclafani, M., Huber, D., Park, J. 2026. Assessment of pre and post fire remotely sensed data products across spatial and temporal scales for prescribed fire operations in rangelands. Fire Ecology. 22. Article 113. https://doi.org/10.1186/s42408-026-00565-6.
DOI: https://doi.org/10.1186/s42408-026-00565-6

Interpretive Summary: Medium resolution remote sensing products (e.g., Landfire) are widely available but do not adequately meet the needs of rangeland prescribed fire planners and fuels managers who require fine-scale, spatial information on fuel type composition, burned area, and burn severity to ensure personnel safety, resource conservation, and effective fire treatments. For a fall 2023 prescribed fire in southwestern Idaho on western juniper-encroached big sagebrush rangelands we 1) contrasted fuel type mapping accuracy of Landfire with maps derived from high-resolution WorldView-2 imagery, 2) contrasted burned area estimation from burn indices derived from medium-resolution Sentinel 2A imagery with that from an ash map derived with high-resolution WorldView-3 imagery, and 3) developed and evaluated machine learning (ML) models to predict burned area using only prefire WorldView-2 fuel type maps and topographic variables. We found the Worldview-2 map had more accurate fuel type classifications (83.0% accuracy) that Landfire fuel type maps (45.1%); the Worldview-3 ash map more accurately estimated burned area (90.4% accuracy) than Sentinel derived indices (76%); and ML gradient boosted trees model indicated elevation was a more important variable for predicting burn areas than prefire fuel type. This research demonstrates that the fine-scale fuels and burn assessment information needed by rangeland prescribe fire planners, fuels managers, and wildland firefighters of the mountainous West can be derived from high-resolution satellite imagery like WorldView-2 and -3 which in many cases can be freely obtained by Federal Government employees from the Geospatial Intelligence Agency, and that readily available topographic data like elevation can be very useful for predicting burned area on rangelands of diverse topography.

Technical Abstract: Western Juniper (Juniperus occidentalis) covers 40 million hectares in the Western United States. Prescribed fire is an effective method to control Juniper encroachment into sagebrush ecosystems. Remotely sensed data is a valuable option for prescribed fire planning in remote rangelands, but many available products are annually based and too spatially coarse to represent the dynamic fuels of rangelands which limits their effective use. Our study focused on a prescribed fire that took place in a small sagebrush dominated watershed in Southwestern Idaho for juniper control on October 6, 2023. Our goal was to predict burned areas with pre-fire remotely sensed datasets and compare the tradeoffs between pre-fire fuel type datasets and post fire burned area datasets at different spatial resolutions. We first compared the pre-fire fuel datasets and demonstrated a high-resolution 0.5-m WorldView-2 pre-fire fuel map more accurately mapped fuel types (83.0% accuracy) than the commonly used 30-m fuel type dataset Landfire (45.1%). We also demonstrate a post fire 0.5 m WorldView-3 modeled ash map more accurately mapped burned area (90.4% accuracy) than two Sentinel 2A derived indices: the relative differenced normalized burn ratio (76.0% accuracy) and the differenced normalized burn ratio (68.7% accuracy). We developed machine learning models to predict what areas will burn given only pre-fire datasets of fuel type, topographic variables, and derivatives of both. Elevation was the most important variable across all 12 models and fuel type was the least important. Initially the coarser resolution dataset appeared to produce more accurate predictions of burned areas, 76.8% accuracy versus 72.8%. We found that when mapped the gradient boosted trees model trained with high resolution pre-fire data produced predicted more realistic maps at a high resolution of burned area, and the mapping accuracy was 4.1% greater for the high-resolution model than the coarse resolution model. Safe planning and execution of prescribed fire for juniper control requires accurate remotely sensed data. Our work demonstrates the significant advantage to high resolution datasets over coarser products including rapidly available maps and more accurate descriptions of the conditions that provide managers with the data needed to make informed decision when planning prescribed fire.