Modeling Post-Fire Vegetation Recovery in the Okefenokee Wetlands Using Remote Sensing and Generalized Additive Mixed Models

The Okefenokee is a fire-adapted swamp located in the Southeastern United States. The Okefenokee experiences varying degrees of wildfire severity, which can alter long-term vegetation dynamics. Despite frequent fires in the wildlife refuge, no study has modeled long-term vegetation recovery across multiple events using satellite time series. Our goal was to assess the long-term impacts of repeated fire on vegetation dynamics using satellite-derived indices and use a generalized additive mixed model. Landsat-derived dNBR was used to assess fire severity across 99 sites. Vegetation indices were retrieved from peak growing seasons, and recovery trajectories were modeled by severity class and site characteristics. Pre-fire conditions, cumulative fire history, and fire severity were all found significant in the NDVI model, which was the most accurate predictor out of the models tested–NDMI, EVI, and NDVI. High severity sites also recovered slower and less complete than moderate and low-severity sites. Vegetation recovery plateaued in the model after 10-12 years. This study can be used as a scalable framework for monitoring wetland fire recovery in the Okefenokee and can support adaptive fire management with additional research. Additional research may focus on integrating hydrological data and field validation to enhance understanding of post-fire wetland recovery.