中文摘要：

以盐城地区为例,通过滑动t突变检验、EEMD（集合经验模态分解,Ensemble Empirical Mode Decompositon）以及Markov Chain（马尔科夫链）统计模型,对盐城地区54 a年雷暴日数据进行动态与趋势预测研究,并对周期研究结果进行显著性和误差检验,结果表明：盐城地区54 a年雷暴日呈现减少的趋势,突变年为1966年,其54 a年雷暴日序列可以分解成4个IMF（Intrinsic Mode Function）分量和1个趋势分量。主要存在0~0.05、0.45~0.50 Hz两个频率区间,相应的其周期为2.22 a年际周期和27 a年代际周期（可信度均超过95%）。EEMD重构雷暴日数据与原始数据的误差百分比处于±0.8%范围内,其马尔科夫链长周期预测显示未来年雷暴日处于25~35 d之间的概率为45%,大于35 d的概率为32%。

英文摘要：

Through adopting the moving t-mutation test, EEMD （Ensemble Empirical Mode Decomposition） and Markov chain statistic model, the author studied the dynamic and trend prediction of annual thunderstorm day number in the past 54 years in Yancheng area, and conducted the significance and error test for the periodicity research results. The results indicated that the annual thunderstorm day number in the past 54 years in Yancheng area revealed a decreasing trend, and the mutation year was 1966. The sequence of annual thunderstorm day number in 54 years could be decomposed into four IMF （ Intrinsic Mode Func- tion） components and a trend component. There existed two main frequency scopes： 0-0.05 Hz and 0.45-0.50 Hz, and their corresponding period was 2.22 years and 27 years （their reliability all exceeded 95%）. The error percentage between the recon- structed thunderstorm day data by EEMD and the original thunderstorm day data ranged from -0.8% to 0.8%. The long-period prediction by Markov chain shows that the probability which the number of annual thunderstorm days in the future is 25- 35 d will be about 45%, and the probability which the number of annual thunderstorm days in the future is over 35 d will be about 32%.