A statistical framework to combine multivariate spatial data and physical models for hurricane surface wind prediction

Storm surge is the onshore rush of seawater associated with hurricane force winds and can lead to loss of property and loss of life. Currently storm surge estimates are created by numerical ocean models that use a deterministic formula for surface wind fields. A new multivariate spatial statistical framework is developed to improve the estimation of these wind field inputs while accounting for potential bias in the observations. Using Bayesian inference methods we find that this spatial model improves estimation and prediction of surface winds and storm surge based on wind data for a case study of Hurricane Charley.