An Improved Approach for Selecting and Validating Burn Severity Indices in Forested Landscapes

Abstract

Burn severity maps based on remotely sensed reflectance data provide a useful way for land managers and researchers to represent and compare spatial variation in fire effects among wildfires and prescribed fires. A need exists for an objective and rigorous selection approach that ensures the best possible spatial predictions of burn severity. The aim of this study was to present and test a methodology for selecting the optimal burn severity index from a suite of calculation and validation options that can be used to produce data for more rigorously comparing ecological effects of fire that occur in contrasting phenologies. In our study, we cross-validated remote sensing data with field data and we tested the predictive ability of 12 cross-validated index calibrations that were generated using common statistical approaches, to predict field-measured burn severity indices collected at burned and unburned areas in New Jersey Pinelands National Reserve. We demonstrate the utility of our approach, provide convincing evidence for the use of CBI as a field-based index over WCBI, and provide a cross-validated method for calculating burn severity in this vegetation type that can be used by managers and researchers.