Reproducing a subjective classification scheme for atmospheric circulation patterns over the United Kingdom using a neural network

Abstract

Atmospheric circulation patterns are currently classified manually according to subjective schemes, such as the Lamb catalogue of circulation patterns centred on the United Kingdom. However, the sheer volume of data produced by General Circulation Models, used to investigate the effects of climatic change, makes this approach impractical for classifying predictions of the future climate. Furthermore, classification extending over long periods of time may require numerous authors, possibly introducing unwelcome discontinuities in the classification. This paper describes a neural classifier designed to reproduce the Lamb catalogue. Initial results indicate the neural classifier is able to out-perform the currently used rule-based system by a modest, but significant amount.