在介绍重构相空间技术的主要定量指标（关联维数D2和柯尔莫奇诺夫（Kolmogorov）熵）的基础上，针对长江上游川中地区降水时间序列，探讨了不同嵌入维m下其关联维数的变化规律。计算结果为该时间序列的饱和关联维D2=5.11，最低嵌入维m=10，Kolmogorov熵k=0.338；采用主分量分析（PCA分布）方法进一步验证了该序列具有混沌特性，并且得到该序列的预测年限不应超过2.96 a，为降水预测提供了较为科学的依据。

Based on introducing the main quantitative indexes ofcorrelation dimension D2 and Kolmogorov entropy in rebuilding time series imbedding space, the relationship between built-in dimension m and correlation dimension D2 is discussed with the precipitation time serial in Sichuan middle region of Yangtze River upstream. The saturation correlation dimension, minimum built-in dimension and Kolmogorov entropy are calculated and given, that is D2=5.11,m=10 and k=0.338. Primary component analysis method is applied to validate its chaotic character, and obtained that the forecasting length for this precipitation time serial would not exceed 2.96 year. This time serial chaotic analysis will provide a scientific gist for precipitation forecasting.