Indian rainfall data over the period 1870 to 1990 have been analysed using continuous wavelet transforms. Results obtained are found to be particularly revealing with regard to the temporal structure of the time series if local scaling is used for the transform coefficients, normalizing their values at any wavelet scale with the maximum at that particular scale. Tracking the maxima in such normalized values leads to a picture of seven prominent modes, which are modulated in wavelet scale (and hence also frequency) and in amplitude. Because the modes wander over the parameters of the wavelet map, they cannot always he easily detected by classical spectral analysis. The quasi-cyclic modes so identified exhibit appreciable jitter, and have average periods varying from a little less than 2 years to nearly 80 years. [References: 19]
展开▼