Statistical Data Assimilation for Coastal Ocean Prediction

Montserrat Fuentes
Statistics Department, North Carolina State University

Kristen Foley
Statistics Department, North Carolina State University

Lian Xie
Marine Earth and Atmospheric Sciences Department, North Carolina State University

We present a model-uncertainty estimation procedure and a data assimilation method for a coastal ocean prediction and assessment system (COPAS), to provide timely and accurate assessments and predictions of coastal marine environment. Gridded wind fields are used to spin up and force these ocean numerical models. Currently, these wind field drivers are specified by deterministic models, that are a function of the central pressure and location of the storm center and parameters to define the shape of the pressure profile of the storm. While these equations incorporate important physical knowledge about the structure of hurrican wind fields, they cannot always capture the asymmetric and dynamic nature of a hurricane. We present a statistical framework to account for variability not captured by the wind physical model. A linear model of coregionalization is used to account for spatial variability in the horizontal and vertical wind components as well as the covariance between components at the same location. A Bayesian framework allows for estimation of the parameters of the multivariate spatial model and the physically based wind data from buoys, ships, aircraft and satellite. The proposed statistical model is used to create an ensemble of wind fields for a sequential ensemble data assimilation method to improve hurricane forecasting using COPAS.

This methodology is applied to a couple of case studies in the Eastern US coast, one in September 1989 when we had Hurricane Hugo, and the other in September 1999 when we had Hurricane Floyd.

Last Update: September 27, 2005