Two basic approaches have been investigated. In the first, sea breeze and other important features of wind behaviour throughout the day are derived directly from the observed half hourly wind measurements. In the second, parsimonious parametric regression models that characterise the daily wind profiles have been developed. Methods for predicting the occurrence of sea breeze events include multiple logistic regression and classification and regression trees. Conditional on a sea breeze event being forecast the key features, such as onset time, duration and strength of winds, are forecast using multiple regression and multiple adaptive regression splines. Predictors used in these various models include those obtained from numerical prediction models and observed data prior to the forecast horizon.
Models for the spatial description of surface winds on the microscale of Sydney Harbour will also be described.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner