Probabilistic models of fire occurrence across National Park Service units within the Mojave Desert Network, USA

摘要

The frequency and size of wildfires within the Mojave Desert are increasing, possibly due to climate and land cover changes and associated increases in non-native invasive plant biomass, as measured by normalized difference vegetation index (NDVI). These patterns are of particular concern to resource managers in regions where native plant communities are not well adapted to fire. We used an information-theoretic and mixed-model approach to quantify the importance of multiple environmental variables in predicting, separately, the probabilities of occurrence of all fires and the occurrence large (>20 ha) fires in five management units administered by the National Park Service in the Mojave Desert Network and based on fire ignition data obtained for the period 1992-2011. Fire occurrence was strongly associated with areas close to roads, high maximum NDVI values in the year preceding ignition, the desert montane ecological zone, and high topographic roughness. Large fire probability was strongly associated with lightning-caused ignition events, high maximum NDVI values in the spring preceding ignition, high topographic roughness, the middle-elevation shrubland ecological zone, and areas further from roads. Our probabilistic models and maps can be used to explore patterns of fire occurrence based upon variability in NDVI values and to assess the vulnerability of Mojave Desert protected areas to undesirable fire events