探讨一种处理多年气象观测数据的新方法,该方法以地理信息系统为技术支撑,以数字图像处理和标准主成分分析为核心,适用于处理空间分布广、时间序列长的多类型气象观测数据。在实例研究中,首先对中国160个站10年气温、降水观测数据进行空间双线性插值,得到一系列数值图像;并对这些数值图像进行标准主成分分析,得到反映多年气候状况且相互独立的各主成分;然后选取其中前6个主成分进行空间迭代聚类分析,得到中国气候分类图;通过对所得类别进行多年气候指标统计分析,得到中国气候区划图;最后将所得气候区划图与传统气候区划图进行比较分析,并指出该方法在处理多年气象观测数据工作中的可行性与局限性。

The objective of this study is to develop a new approach for long term climate monitoring data mining by using geo-information system (GIS), digital image processing, principal components analysis (PCA) and spatial iterative self-organizing unsupervised classification technologies. Monthly mean air-tempurature and precipitation data collected from 160 cities in China from 1987 to 1996 were interpolated spatially into digital images, then principal components analysis and spatial iterative self-organizing unsupervised classification were performed. In comparison with the previous classification results by using traditional schemes, climate zoning result from this method is similar to traditional climate zoning schemes.But in some districts zoning result is bad due to sparsity or scarsity of monitoring data. But GIS and SPCA
showed advantage in analyzing long term climate monitoring data.