Multidecadal mapping of status and trends in annual burn probability over Canada's forested ecosystems
Christopher Mulverhill,Nicholas C. Coops,Michael A. Wulder,Joanne C. White,Txomin Hermosilla,Christopher W. Bater
DOI: https://doi.org/10.1016/j.isprsjprs.2024.02.006
IF: 12.7
2024-02-20
ISPRS Journal of Photogrammetry and Remote Sensing
Abstract:Globally, wildfires burn an average of approximately 5.5 Mha of forest per year. Deriving a detailed inventory of forest fuel conditions is critical to managing resources both before and during a fire. However, data products that form the basis of these inventories often come from disparate sources, may not be subject to update, or may not capture information relevant to fuels and fire behaviour. Satellite-derived data products present a valuable solution to these concerns due to exhaustive spatial coverage and capacity for frequent temporal revisit. In this study, we modeled annual burn probability in Canadian forests for a 27-year period (1994–2020) using satellite-derived data products informing on topographic, biotic, and climatic variables. While not predicting the presence or absence of a wildfire, a high burn probability as output from the models showed strong correspondence with actual burned areas. Results indicated that biotic variables representing canopy cover and stand volume were among the most important to estimate burn probability, with the most relevant climatic variables being spring precipitation and the annual range of temperature. Analysis of the temporal trends in burn probability highlighted that 15% of tiles analyzed had significantly declining trends in mean burn probability across the study period. These decreasing trends did not always correspond to reductions in total burned area, potentially indicating a shift in forest composition of some regions and a change in the conditions which burn in others. This study highlights the importance and potential of annual burn probability mapping in order to capture these longer-term trends as well as changes in climatic conditions, which are more dynamic than other biotic or topographic drivers of burn probability. This study demonstrates the use of freely available satellite-derived datasets to generate maps of burn probability across a large, structurally and compositionally diverse forested extent, aimed to inform effective fire management.
imaging science & photographic technology,remote sensing,geography, physical,geosciences, multidisciplinary