|
Activity Number:
|
56
|
|
Type:
|
Topic Contributed
|
|
Date/Time:
|
Sunday, August 6, 2006 : 4:00 PM to 5:50 PM
|
|
Sponsor:
|
Section on Bayesian Statistical Science
|
| Abstract - #306981 |
|
Title:
|
A Bayesian Framework To Combine Multivariate Spatial Data and Physical Models for Hurricane Surface Wind Prediction
|
|
Author(s):
|
Kristen M. Foley*+ and Montserrat Fuentes
|
|
Companies:
|
North Carolina State University and North Carolina State University
|
|
Address:
|
8203 Campus Box, Raleigh, NC, 27695,
|
|
Keywords:
|
Bayesian inference ; coregionalization models ; hurricane surface wind fields ; non-separable multivariate models ; spatial statistics ; storm surge forecasts
|
|
Abstract:
|
We wish to improve the prediction of the coastal ocean response to hurricanes by combining disparate sources of observational data and output from numerical forecast models. A new Bayesian modeling framework is introduced to allow for estimation of the parameters of a multivariate spatial statistical model for wind data and parameters of a physically based deterministic model while accounting for potential bias in the observed data. For real-time storm surge prediction, we use an empirical Bayesian approach to predict hurricane surface wind fields at high spatial and temporal resolution to be used as input for a 3-D coastal ocean model. We find that this spatial model improves prediction of the wind fields when compared to the original deterministic model. These methods also are shown to improve storm surge estimates for a case study of Hurricane Charley (2004).
|
