Abstract: A scheme is proposed for detecting multi-scale climate accidents by means of the combination of Morlet wavelet transform and statistical validation with which to examine multiscale spectrum structure of Meiyu intensities in the timeseries for 1885-2000 over middle-lower reaches of the Changjiang River, and make significance tests of the abrupt change points at different scales. Also the time-dependent evolution and variance contribution of Meiyu intensity series are investigated using wavelet energy density technique. Results show that on a large-scale basis, 1941 can be viewed as the division between high and low intensities and multiple abrupt change points appears at diverse time scales, for instance, during the 1885-1941 for stronger rainfall stage two phases of relatively weak rainfall occur and in 1942-2000 weak rainfall stage the 1991-2000 relatively high rainfall occur; the 2-3 and 6-7 years oscillations have higher variance in interannual variation over the study basin, the 2-3 years oscillations have pronounced variance contribution in 1978-1987 and 1996-2000, so are the 6-7 years oscillations in 1920-1932; the 23-24 and 36-37 years oscillations have higher variance contribution after the 1940s on an interdecadal basis.