Variability in quantifying leaf area index in humid, fire-affected ecosystems

Abstract Despite the prevalence of fire activity in the Southeastern United States, a limited understanding of how vegetation phenology varies after fire across different ecosystems hinders our ability to make informed land management decisions and to model vegetation responses at regional scales. Quantifying variability in post-fire observations of canopy structure is essential for land managers assessing damage and recovery, and for ecosystem modelers evaluating changes in carbon uptake, plant water use, and energetic exchanges between the land surface and the atmosphere. Land managers and scientists frequently rely on satellite remote sensing to evaluate vegetation changes in response to fire disturbances. However, satellite remote sensing data are limited by their spatial and temporal resolutions, possibly missing small fires or not capturing heterogeneous impacts to vegetation within a single pixel. To assess this gap, we quantify differences in remote sensing and ground observations of vegetation phenology, evaluating and attributing sources of uncertainty for gridded datasets typically used at ecosystem scales. We quantify phenologic changes across different ecoregions in North Carolina, USA using time series of plant density from ground observations and satellite remote sensing. Using leaf area index (LAI) as a measure of canopy structure, we evaluate changes in vegetation across different fire-affected ecosystems. Based on nearly 400 ground-based measurements collected over a 4-year span before and after prescribed fires and one wildfire, we quantify discrepancies in LAI among satellite and ground products and show how uncertainty varies with ecoregion, plant functional type, spatial resolution, and sub-pixel vegetation heterogeneity. We find that while the magnitudes of LAI estimates differ across data products, seasonal cycles before and after fire events tend to agree. Our results underscore the importance of targeted ground-validation efforts and offer practical insights for improving remote sensing based assessments of post-disturbance vegetation dynamics, carbon and water cycling, and land management.