Statistical Data Assimilation for Coastal Ocean Prediction

Kristen Foley

North Carolina State University

The South-Eastern United States has had a tremendous residential and commercial investment in coastal areas during the past 10 years. Rapid development of the coast produces a stressed ecosystem and an exponential increase in human and property exposure to storm hazards. Our goal is to improve the prediction of coastal ocean processes that are associated with tropical storms and hurricanes. Specifically we are interested in predicting storm surge, the onshore rush of seawater caused by the high winds of a land-falling cyclone. Statistical methods are used to improve the inputs needed to initiate a numerical ocean model for the coastal Carolinas and Georgia. A new class of nonstationary statistical models is proposed to model the bivariate components of the wind field inputs based on a spatially varying coregionalization model. Data assimilation methods are used to combine output of the numerical model with observed data from water level stations along the coast. Ensemble and Bayesian particle filtering methods are investigated and compared, with a focus on computational issues.

Last Update: March 25, 2005