Tidal flooding can significantly impact vegetation pixel reflectance of coastal salt marshes, presenting a problem for remote sensing studies of these highly productive ecosystems. The current study aimed to spatially and temporally expand our previously developed Flooding in Landsat Across Tidal Systems (FLATS) model to detect and analyze the long-term changes in flooded marsh pixels in Landsat 5-9 imagery. As the FLATS index is only calibrated for Landsat 8, our goal was to expand the use of FLATS to a greater range of Landsat imagery and facilitate the masking of flooded pixels in long-term time series of vegetation indices. Using areas of salt marsh in the Georgia Coastal Ecosystems (GCE) Long Term Ecological Research (LTER) site, images from Landsat 5 through 9 were paired with near-coincident Landsat 7 images for a novel cross-calibration. Indices in the FLATS algorithm (Normalized Difference Water Index, NDWI and the Enhanced Vegetation Index, EVI) were calibrated for each image pair using linear regression models, and an adjusted FLATS index (FLATS+) was created to be used on a substantially expanded Landsat dataset from 1984 to 2023. The R2 scores for the vegetation index calibrations ranged from 0.78 to 0.87 for EVI, and 0.73-0.82 for NDWI. Additionally, this study sought to monitor changes in flooding patterns at the study site, utilizing the expanded temporal range of FLATS+. The trend in FLATS+ value exhibited significant spatial autocorrelation at three LTER sites, with areas of marsh experiencing significant changes in inundation over the 39-year period (Moran's I, p<0.01 at all sites). The FLATS+ index is a tool that is able to identify flooded pixels in Landsat 5-9. The index can be used to study salt marsh productivity, carbon uptake, flooding, and resiliency in response to sea level rise.
Read full abstract