Classification of thunderstorm and non-thunderstorm days in Calcutta (India) on the basis of linear discriminant analysis

S. GHOSH, P. K. SEN, U. K. DE

Abstract

The two sophisticated applied multivariate techniques, ‘Principal Component Analysis’ and ‘Two-group Linear Discriminant Analysis’ have been applied in the present work to analyze the pre-monsoon weather in Calcutta (India) and hence to forecast the pre-monsoon thunderstorms there. The work has been performed in the following two stages: i) Analysis with 20 thermodynamic and dynamic parameters derived from daily radiosonde data in Calcutta; ii) Analysis with 10 newly formed parameters which are actually the first 10 principal components formed with the 20 original parameters. The study indicates that an index known as ‘Linear Discriminant Function’ (LDF) may be constructed to predict the pre-monsoon weather in Calcutta. Not only that, the study also reveals that if the dimensionalities of the data matrices are reduced, then the accuracy of the results may improve.

Keywords

Equivalent potential temperature; saturated eqivalent potential temperature; convective instability conditional instability; principal component analysis; linear discriminant analysis

Full Text:

pdf