Characterization of coastal wetlands using Synthetic Aperture Radar and Optical/IR satellite imagery

Situated at the terrestrial-estuarine interface, tidal wetlands influence and are influenced by numerous geophysical and biological processes in adjacent regions, both terrestrial and aquatic. Tidal wetlands have profound impacts on carbon and nutrient cycling, hydrology, and geomorphology in coastal regions. Accurate characterization of ecosystem parameters in these tidal wetland systems (e.g. vegetation community distribution, inundation dynamics, salinity and tidal regimes) is key for understanding how tidal wetlands are influenced by the upstream watershed and how they interact with connected estuaries. In this research effort, synthetic aperture radar (SAR) and optical satellite imagery are used to map the extent of tidal wetlands in the Mid-Atlantic United States. We employ Random Forest and other supervised classification approaches, evaluating results using datasets from the National Wetlands Inventory. Results show better than 82% classification accuracy for emergent wetlands and 80% overall classification accuracy. We map vegetation communities and inundation extents within these identified wetland systems. Observations from L-band PALSAR-1/2 (HH, HV polarization) and C-band Sentinel-1 (VV, VH Polarization) SAR satellites are used to elucidate temporal and spatial variability in wetland inundation corresponding to different tidal phases. We discuss the importance of SAR polarization and frequency and of optical spectral bands in terms of respective contributions to classification accuracy. Our findings indicate pronounced differences between C-band and L-band sensors in the ability to unambiguously classify inundation in tidal wetlands, particularly under densely vegetated canopies. In addition, an optical-SAR fusion approach was used to map vegetation in wetland systems, with a particular focus on mapping indicator species. Certain wetland species such as Phragmites australis are indicators of disturbance and high soil nutrient levels while other species such as Spartina spp. and Nuphar spp. are indicators of salinity regimes. This dual characterization of coastal wetland vegetation-inundation regimes is non-site specific and can readily be applied to regions beyond our study areas in the Mid-Atlantic United States.