Presenter: Sean Vitousek (USGS)

Description:
Satellite-derived historical shoreline data combined with dynamic shoreline models enable predictions of coastal erosion on unprecedented spatiotemporal scales.  Here, we present a satellite-data-assimilated, ensemble Kalman filter shoreline change model (CoSMoS-COAST) to predict large-scale coastal change and associated uncertainty due to waves, sea-level rise, and other natural and anthropogenic processes.  The model utilizes ensemble data assimilation to improve model reliability and uncertainty quantification using a straightforward technique based on the spread of the (individually deterministic) model trajectories with a range of model parameters and climate forcing scenarios.  Here, we apply the developed model to the entire California coastline (approximately 1,350 km), many sections of which are poorly monitored with traditional survey methods, where 5 or fewer data points is common.  The most intensive local monitoring programs in California have performed ~200 topographic beach surveys over the last two decades, yet satellite imagery can typically provide 500-1000 shoreline observations spanning the entire coast over the same period. We demonstrate (for purposes of model calibration/validation) that data assimilation of satellite observations vs. extensive in situ observations provides nearly the same level of model accuracy.  When comparing model predicted shoreline positions to satellite-derived observations, the model achieves an accuracy of <10 m RMSE for nearly half of the entire California coastline for the validation period of 2015-2020. The data-assimilation model presented here is generally applicable to a variety of coastal settings around the world owing to the global coverage of satellite imagery.