WAVELET ANALYSIS OF INTERANNUAL VARIATION OF TROPICAL CYCLONES IN GUANGDONG

Abstract:

Climatological laws are studied for the annual frequency of tropical cyclone occurrence and the date of the yearly first landfall, which take place in the Guangdong province or pose serious threats on it from 1951 to 1999, using the data in the Yearly Book on Typhoons. A new method that has developed over recent years for the study of temporal sequences, the wavelet analysis, is used, in addition to more common statistical approaches. By analyzing two wavelet functions, MHAT and MORLET, we have compared the results of transformation of the wavelets provided that other conditions remain unchanged. It is discovered that the variance of MORLET wavelet has better indication of primary periods; period-time sequence charts can reflect major affecting periods for individual sections of time; when compared with the original sequence, the chart shows a little shift. On the other hand, such shift is absent in the MHAT wavelet, but its higher frequency part of variance covers up the primary periods to make its variance less predominant as compared to the MORLET wavelet. Besides, the work compares two different assumptions of an amplifying factor a. It is found that primary periods can be shown more clearly in the variance when a takes the exponential of 2 than it takes values continuously. Studying the annual frequency of tropical cyclones and the date of first appearance for periodic patterns, we have found that the primary periods extracted by this approach are similar to those obtained by wavelet transformation.