Application of Model Output Statistics to Tropical Cyclone Forecasting in the Western Pacific.

摘要

Analog tropical cyclone forecasting techniques inherently include some mean interaction of the cyclone with its synoptic environment. This investigation seeks to relate current synotic features to storm displacement. The objective forecasting technique, Model Output Statistics (MOS), along with the so called ridged regression biased estimator are used to predict recurvature or non-recurvature of tropical cyclones. Corrections to an analog forecast position are also predicted. It is shown that a larger data set is required before any conclusive results in predicting recurvature or non-recurvature can be made. However, the most fruitful area for application of MOS does appear to be in the recurvature / non-recurvature problem and not in the area of corrections to an analog forecast. The ridge regression estimator generally proves superior to the ordinary least squares estimator as a method of determining predictive equations. (Author)