Estimating the impacts of saltwater intrusion on Coastal Farmlands of Delmarva Peninsula using Sentinel-2 Imagery and spectral unmixing

Keywords: Saltwater instrusion, spectral unmixing, Sentinel-2, sea level rise
Abstract Type: Paper Abstract

Abstract

One-fifth of the world’s irrigated farmland witnesses an annual economic loss of US$27.3 billion as a result of crop yield reduction or abandonment due to soil salinization. Saltwater intrusion (SWI), the inland encroachment of saltwater into freshwater aquifers, in coastal areas, is mainly driven by sea level rise, storm surge, and excess groundwater pumping. Collecting field-level data on soil salinity can provide an accurate measure of the extent of salinization, but it is challenging to scale up to geographically larger regions. We aim to capture the spatial distribution of salt patches on the soil surface across 14 coastal counties in the eastern USA states of Delaware, Maryland and Virginia. While high-resolution aerial images can be used to quantify the extent of these fine-scale salt patches, these images do not have high temporal resolution. Our objective is to derive annual maps of salt-impacted lands to better aid landowners' decision-making process. In order to capture the annual evolution of these fine-scale salt patches, we used Sentinel-2 Level 2A (surface reflectance) bands that have a spatial resolution of 10-20 meters and a temporal resolution of 5 days. We performed spectral unmixing on these images resulting in each pixel with a fraction value corresponding to the percentage of pixel area covered by salt patch. To train and test the spectral unmixing algorithm, we used field data as well as high resolution (0.6 to 1.0 meters) drone and aerial imagery.