FROM THE JOURNAL

TIU Transactions on Inteligent Computing

Wavelet power spectrum analysis of a nonstationary nonlinear time series data

Manaswita Das, Mourani Sinha
Department of Mathematics, Techno India University, West Bengal, Kolkata, India

 

Abstract

The significant wave height (SWH) data variability is studied for the Bay of Bengal region for the period 1996-2000 using continuous wavelet power spectrum. The averaged SWH time series were normalized by their standard deviation and then decomposed using the Morlet wavelet function. The normalized wavelet power spectrum are generated with the cone of influence, where edge effects become important. The 95% confidence level for the SWH data is shown by the black contours. For a red-noise process the significance levels were computed with a lag-1 coefficient of 0.99. For white-noise process similar contours were generated with a lag-1 coefficient of 0.00 since they are uncorrelated in time. For the year 1996 the red-noise wavelet power spectrum shows two bands of oscillations, one in the 5-20 days period and the other one in the 32-64 days period. Except for the year 2000 the maximum power is concentrated in the June-August months for the 32-64 day period. Inspite of the above fact significant region is noted only in the year 1996 for the 32-64 day period. Hence the red-noise wavelet power spectra effectively captures the oscillations in the SWH data which corresponds to seasonal variations.

Keywords: Time series, fourier transform, wavelet power spectrum,significant wave height,Bay of Bengal.