Using satellite data and data assimilation to improve analysis of historical shoreline trends from 1984 to the present and future projections along the U.S. Atlantic coast

Abstract:

Multi-decadal shoreline change rates are historically used to quantify the vulnerability of coastal communities and guide management decisions. Shoreline positions are typically derived from temporally sparse data that do not resolve the constant evolution driven by anthropogenic and oceanographic forces. Additionally, it is unclear whether linear shoreline change rates can be used to skillfully predict shoreline positions, particularly in locations where processes like sea level rise may have caused modern accelerations in shoreline erosion. Recently, tools have been developed that allow users to download and extract shorelines from publicly available satellite images. In this study, CoastSat, a python-based Google Earth imagery extraction tool, was used to acquire roughly 500 satellite-imagery-derived shorelines at various sites spanning the United States Eastern coast over 30 years in order to derive shoreline change rates from this temporally dense data. We evaluate the accuracy of the satellite-derived shoreline and compare shoreline change rates to determine the impact of temporal resolution. The predictive capacities were explored for both static shoreline change rates (satellite and traditional) by withholding the last 5 years of satellite-derived shorelines from the analysis. A Kalman Filter data assimilation method is also applied to each dataset to assess dynamic changes in the shoreline change rate and assess potential improvement to the prediction of shoreline migration over decadal timescales. This data assimilation approach will also be used to objectively identify linkages between temporal changes in shoreline change rates and forcing mechanisms to aid in developing appropriate coastal resilience plans.