Methods for Evaluating Spatial Seasonal Fire Potential Indices

Several spatial products are available for tracking changes in fire potential (fire danger) and more are under development. Such indices are developed from satellite data and gridded weather models. Evaluating a spatial fire potential product is the process of comparing fire occurrence and size data with a map of fire potential values. Methods of evaluation developed to date have included both model and non-model-based techniques, but few have attempted to explicitly account for the spatial and temporal correlation present among the fire potential readings and the fire history observations. We present the use of a randomization method, a percentile method, and the Partial Mantel Test. For each of these methods, we also address the issue of spatial scale by considering multiple spatial resolutions of the data. We demonstrate the methods for AVHRR NDVI-derived Relative Greenness (RG) and Departure from Average (DA) for Washington and Oregon.