Bayesian Parameter Estimation for Arctic Coastal Erosion under the Effects of Climate Change

Abstract

Arctic coastal erosion due to decreasing ice protection and increasing temperatures is a threat to coastal communities and infrastructure as well as a driver of long-term habitat changes. In order to respond to this threat, decadal predictive models are required that incorporate the effects of climate change under current emission trajectories. This work presents an Arctic erosion one-line model capable of estimating unknown coastal parameters through historic coastline measurements. Parameter estimation is carried out using both the extended and unscented Kalman filters, and the results compared. The model and parameter estimation are evaluated using two sections of Arctic coastline, one near Oliktok Point, AK, and the other along the coast of Barter Island, AK. Historic wave fields are modeled for both locations using downsampled historic GCM data for boundary conditions and estimating fetch distance. Future wave and temperature conditions are found using GCM projections under the RCP 8.5 pathway. Parameter estimation is performed on all coastal measurements except the most recent coastline available; this hold out measurement is then used to test the predictive power of the model. Coastlines at both locations are simulated from 1980 to 2070. It is found that root-mean-square error values for both locations are lower than purely empirical techniques and future predictions show increasing rates of erosion under the RCP 8.5 pathway.