Generated Burn Severity Offset Correction Values for a Composite Burn Index Field Plot Dataset

Phenological correction of pre- and post-fire imagery is used to improve remotely-sensed burn severity predictions by standardizing vegetation greenness between image pairs. However, the efficacy of existing methods across diverse fire conditions and scenarios remains underexplored. In this study, we evaluated the impact of phenological offset corrections across various methodological decision points and fire-prone environments to support analyst decision making. We generated burn severity spectral values for a dataset of Composite Burn Index (CBI) field plots across the continental US. The effectiveness of offset correction was tested across image selection techniques, spectral indices, offset generation methods, and burn perimeter sources. We compared the influence of offset corrections on the modeled relationship with CBI, agreement between burn severity thresholds, and potential bias. We found that applying offset correction consistently improved the modeled relationship with CBI by addressing extreme outlier severity values. However, offset corrections had the potential to introduce bias, systematically lowering severity values and reducing correspondence to burn severity thresholds. Our analysis suggests that offset corrections may offer benefits but also pose tradeoffs to accurately representing remotely-sensed burn severity. The utility of offset corrections may depend on the environment, methods, and scale of analysis considered. We propose a decision-tree framework for analysts to consider when employing offset correction, given their study scope.