文章以小波诊断技术为基础,对建国以来中国洪涝灾害成灾面积变化进行了多时间尺度分析。研究表明,建国以来中国洪涝灾害成灾面积变化具有明显的3年左右、9年左右和17年左右的特征时间尺度和相应的周期性变化特征;并且特征时间尺度近年来逐渐缩短,即成灾面积的变化渐趋频繁。根据这几个特征时间尺度上小波系数的演变趋势发现,在未来几年内,中国洪涝成灾面积呈现下降的总体趋势。根据小波方差分析发现中国洪涝灾害成灾面积变化具有显著的3年、9年和17年的主要周期,并以9年和17年的周期尤为显著;同时,根据功率谱的比较分析同样发现,中国洪涝灾害成灾面积同样存在着3年左右、7~9年和17年的主要周期。

The major superiorities of wavelet analysis lie in good part characteristics in time (or spatial) domain and frequency domain, as well as focusing attention on any signal details. Therefore, it is suitable for wavelet analysis to extract the implicit periodical characteristics from the time series and to predict qualitatively the trends. Based on the wavelet analysis, time series of area affected by flood disaster after 1949 have been studied at multiple time scales. The results show: (1) the area affected by flood disaster in China has 3-year, 9-year and 17-year characteristic time scales. The characteristic time scales tend to diminish in recent years, which represents that the inundated cropland area varied more frequently than ever. There is a trend that the area affected by flood disaster in China will decrease in whole in the recent years on the base of the analysis of changes of the wavelet coefficients at 3-year, 9-year and 17-year time scales. (2) The area affected by flood disaster in China has the 3-year, 9-year and 17-year main periods from the wavelet variance analysis, which is consistent with the results of the wavelet coefficients analysis on the whole. The 19-year period and 17-year period are especially outstanding. The wavelet energy of the area is mainly focused on 9-year, 17-year and 3-year characteristic time scales, which means the change of the inundated cropland area is rested on these three time scales. (3) Based on the power spectra analysis, we also find that the area affected by flood disaster in China has the 3-year, 9-year and 17-year main periods. In contrast with wavelet analysis, the distribution characteristic of different periods can not be reflected at local time-domain by power spectra analysis, that is to say, there is no resolution in time-domain and power spectra analysis can not be used to analyze locality. In a word, the application of wavelet analysis in the area affected by flood disaster and its development will deepen the knowledge of characteristic of flood disaster, and provide some reference to the prediction of the inundated cropland area. It also provides a new tool to explore spatial-temporal complexity of flood disaster, and it will draw more and more attention.