This paper develops a logistic regression model for predicting the probability of a sea-breeze occurring at Fort Denison in Sydney Harbour during September, the period during which the Olympic Games were held in Sydney in 2000. Starting with approximately 140 potential predictor variables, a reduced set of nine derived predictor variables was obtained from which the final prediction model was derived. These variables appear to be meteorologically consistent and the final model performs very well on the training dataset and on a set of test data and gives acceptable false prediction error rates.
Dunsmuir et al. (Sun,) studied this question.