Fuel load and potential fire behavior vary seasonally and across grassland plant communities

Abstract Grassland fuels can be patchy across space and time, especially on landscapes where multiple grassland plant communities co-occur. However, the standard fuel models inadequately capture the spatiotemporal variability in grassland fuels for local, strategic planning of fuel treatments. We aimed to better understand the seasonal variation in grassland fuels and fire behavior across grassland types. Here, we developed remote sensing-based grassland fuel maps using multispectral satellite data (10-m resolution) from Sentinel-2, compared seasonal fire behavior across grassland types, and performed a sensitivity analysis of fire behavior models to changes in litter, dead 1 h, and live herbaceous loads, and fuelbed depth estimates. We were able to best predict live herbaceous load (R2 = 0.85), followed by fuelbed depth (R2 = 0.81), dead 1 h fuel load (R2 = 0.76), and litter load (R2 = 0.67) based on remotely sensed predictors. Our spatially predicted grassland fuel maps showed seasonal variability in fuel loads in different places, reflecting the varying timing of curing by grassland types. In particular, we observed more accumulation of dead 1 h load between August and October in early senescing mixedgrass prairies than later senescing tallgrass prairies. We found an overall faster modeled rate of spread in smooth brome-dominated, invasive grasslands and native tallgrass prairies compared to mixedgrass and prairie dog-dominated grasslands. From the sensitivity analysis, we found that modeled fire behavior model is most sensitive to dead 1 h fuel load and least sensitive to litter load in this grassland system. At a time when grassland fires are increasing and resources for fuel management are limited, remote sensing approach to grassland fuel modeling may help land managers make decisions about where and when to prioritize fuel treatments to mitigate the risk of catastrophic wildfires.