Classification and machine recognition of severe weather patterns

Abstract

Forecasting and warning of severe weather conditions are treated from the vantage point of pattern recognition by machine. Pictorial patterns and waveform patterns are distinguished. Time series data on sferics are dealt with by considering waveform patterns. A severe storm patterns recognition machine is described, along with schemes for detection via cross-correlation of time series (same channel or different channels). Syntactic and decision-theoretic approaches to feature extraction are discussed. Active and decayed tornados and thunderstorms, lightning discharges, and funnels and their related time series data are studied.