AbstractsGeography &GIS

Landscape level wildfire prediction can be used to allocate wildfire resources and guide land management practices. Wildfire prediction in arid habitats in the Southwestern United States is of specific concern because of the negative ecological impacts of fire on desert habitats and the current lack of accurate fire prediction tools for such areas. This study examines the ability to predict previous fire occurrences and estimate future fire potential using satellite imagery and on the ground field survey techniques along with ignition potential data (lightning strikes and distance to roads), topographical data (elevation and aspect), and climate information (maximum and minimum temperatures). The satellite data was used to create a suite of potential fuel load models that were then evaluated for the best fit models using AIC model selection. The best fit fuel load model (Fuel Load Model 1) was then used in conjunction with 2005 remote sensing and fire occurrence data to predict fire potential for that year. Fuel load Model 1 along with spring Fuel Moisture Content (FMC), lightning strikes, distance to roads, and perennial vegetation type were modeled and a Receiver Operating Characteristic (ROC) curve was used to evaluate the agreement between model predictions and actual fire occurrence. The ROC evaluation rendered an Area Under the Curve (AUC) value of 0.90 indicating accurate prediction of fire occurrence for 2005. This study provides evidence that remote sensing techniques can be used in combination with field surveys to accurately predict wildfire potential in Mojave Desert habitats.