Remote sensing of vegetation in coastal dune systems
Despite vegetation’s key role in barrier islands stability, we lack an adequate characterization of the spatial distribution of barrier island vegetation in relation to topographic determinants. The relatively recent wide availability of remotely sensed data at suitable resolutions makes it now possible to observe barrier island eco-topographic patterns, but Remote Sensing (RS) techniques, widely used in inland ecosystems, have received little attention to monitor vegetation in coastal dunes. In this study, we apply RS techniques to extend local field observations to larger scales and explore the relationship between vegetation density and topography. To this end, we address the following questions: 1) Can we map dune vegetation species distribution by hyperspectral RS? 2) Can we use LiDAR data to estimate canopy properties (Leaf Area Index, LAI)? 3) Can we establish a relationship between LAI and Normalized Difference Vegetation Index (NDVI) for dune vegetation? 4) What are the relationships, over large scales, between LAI/ NDVI and soil elevation, distance from the shoreline, curvature, and other topographic features? A linear unmixing algorithm was applied to the hyperspectral data to retrieve the vegetation species abundance of coastal dunes on Hog Island in Virginia. We also used LiDAR data to retrieve the Digital Terrain Model (DTM) and to estimate LAI with a conceptually based approach.
The results show that mapping dune vegetation species can be challenging because of low vegetation density and similarity between species. The results also suggest that LiDAR data can be successfully used to retrieve dune vegetation LAI. The comparison of LAI estimates and NDVI shows a coherent relationship. Moreover the spatial distribution of NDVI and LAI show that plants mainly develop on the back side of the front dune system, while the front side of the dune is characterized by low presence of vegetation. These results will contribute to inform dune system dynamical models.