Characterization of Tornado Spectral Signatures Using Higher-Order Spectra

Abstract Distinct tornado spectral signatures (TSSs), which are similar to white noise spectra or have bimodal features, have been observed in both simulations and real data from Doppler radars. The shape of the tornado spectrum depends on several parameters such as the range of the tornado, wind field within the storm, and the reflectivity structure. In this work, one of the higher-order spectra (HOS), termed bispectrum, is implemented to characterize TSS, in which the Doppler spectrum is considered a 1D pattern. Bispectrum has been successfully applied to pattern recognition in other fields owing to the fact that bispectrum can retain the shape information of the signal. Another parameter, termed spectral flatness, is proposed to quantify the spectrum variations. It is shown in simulation that both parameters can characterize TSS and provide information in addition to the three spectral moments. The performance of the two parameters and the spectrum width for characterizing TSS are statistically analyzed and compared for various conditions. The potential of the three parameters for improving tornado detection is further demonstrated by tornadic time series data collected by a research Weather Surveillance Radar-1988 Doppler, KOUN, operated by the National Severe Storms Laboratory.