Using Radar Remote Sensing to Map Wetlands in the Chesapeake Bay

Brian Thomas Lamb1, Maria Tzortziou1 and Kyle C McDonald2, (1)CUNY City College of New York, Earth and Atmospheric Sciences, New York, NY, United States, (2)CUNY City College of New York, New York, NY, United States
Wetlands play a key role in Earth’s carbon cycle. However, wetland carbon cycling exhibits a high level of spatiotemporal dynamism, and thus, is not as well understood as carbon cycling in other ecosystems. Accurate characterization of wetland ecosystem parameters (vegetation communities, inundation, salinity and tidal regimes) is of particular importance for understanding these carbon dynamics. Here, we use radar remote sensing to map wetland properties in the Chesapeake Bay, the largest estuary in the US with more than 1,500 square miles of tidal wetlands, across a range of tidal amplitudes, salinity regimes, and soil organic matter content levels.

We have been using Sentinel-1 and ALOS PALSAR measurements to characterize vegetation and inundation dynamics with the future goal of characterizing salinity gradients and tidal regimes. Differences in radar backscatter from various surface targets has been shown to effectively discriminate between dry soil, wet soil, vegetated areas, and open water. Radar polarization differences and ratios are particularly effective at distinguishing between vegetated and non-vegetated areas. Utilizing these principles, we have been characterizing various wetland types using supervised classification techniques including: Random Forest, Maximum Likelihood, and Minimum Distance. In addition to using radar backscatter, we have also been using positional parameters to perform these classifications. The National Wetlands Inventory has been used as training and validation data. Ideally, the techniques we outline in this research will be applicable to the characterization of wetlands in coastal areas outside of Chesapeake Bay.