Coupled soil and vegetation properties toward remotely sensed coastal terrain characterization

There is a need to quantify relationships among vegetation structure and soil properties in coastal areas to better understand resilience, erosion, land use impacts, eutrophication, and mobility in this terrain. Remote sensing observations have been shown to effectively capture both vegetation and soil properties. However, identifying linked vegetation-soil properties and inferring soil properties from remote sensing when vegetation obscures the pixels remains challenging. Leveraging multiscale and multisensor remote sensing data and fusion techniques, we investigated the capability to identify linked vegetation-soil properties and infer the strength and stability of a barrier island along a swath where the soil surface is bare to partially to fully obscured by vegetation. To this end, we asked (1) Which remote sensing and field measured vegetation properties relate most strongly to soil properties, especially strength and moisture? (2) Do observations of soil conditions, geotechnical descriptors, vegetation species and health, and sensible hyperspectral signatures (HSI) allow accurate characterization of the combined soil-vegetation complex? We used multilevel and multi-sensor ground-, uncrewed aerial system (UAS)-, and satellite-based observations, namely HSI, lidar, ground optical, and geotechnical measurements, to test the variability and relationships among remote sensing-based measurements and geotechnical measurements, such as soil bearing strength. Firstly, we found high capability of HSI data to discriminate soil and soil moisture when any soil was exposed (beach, foredune, back dune, and marsh) and discrimination of vegetation at UAS and satellite scale. We found strong relationships among relative vegetation structure and soil properties, namely biomass estimates and soil strength, when using combined ground-based observations and UAS-based HSI observations and relative high accuracy upscaling to high-resolution satellite level maps of soil and vegetation properties. In addition, we found that adding slope and aspect data moderately enhanced the assessment of soil strength parameters in vegetated areas, although improvement of lidar data collection protocols in subsequent data collections promise further improvements in upcoming studies. In areas with tallest vegetation or soils that were highly saturated (inundation), results were mixed, likely due to poorer inference of soil background from remote sensing and soil strength from field measurements approaching zero, respectively. However, using a combination of shortwave infrared data, full spectra for analyses (spectral unmixing, dimensionality analyses, and supervised classification techniques), water-specific indices, and vegetation type information, wetland soil delineation was improved. Differences in soil and vegetation properties detected using field optical measurements were used to test upscaling techniques, i.e. training, for UAS-based HSI. With the help from ground-based data, a framework of mapping vegetation and soil specific properties was developed which enabled finer spatial analyses to be carried out with respect to the interdependence of vegetation and soil properties from remote sensing observations on a coastal barrier island.