多时相NOAA-AVHRR数据主成分分析的生物学意义

Bio-implication of Principal Component Analysis to LandCover Using Multitemporal AVHRR Data

利用多时相 NOAA- AVHRR的中国归一化植被指数 NDVI数据进行主成分分析 ,并与从NDVI派生的 4个生物学参数作相关分析 ,结果表明 :主成分变换既压缩了信息 ,将 1 2个月的信息主要压缩到前 4个主分量 ,又提取了关键的变化信息。第一主分量反映基本植被覆盖信息 ,第二、第三和第四主分量反映植被季相变化信息。正是由于一年 1 2个月的 NDVI曲线反映了植被季相变化特征 ,使得主成分变换得到的各主分量具有一定的生物学意义 ,而且 1 7种中国典型植被在这 4个主分量图像上存在一定的差异性 ,使其具有进行较高精度土地覆盖分类的潜力。

By using 12 months' 1km AVHRR data in China to make principal component analysis (PCA), and correlating with the NDVI-derived four bio-parameters, this paper shows that the PCA-transformed first four principal components (PCA1,PCA2,PCA3,PCA4) contribute the 95.24% cumulation of variance, they can represent the main information of 12 months'data. And PCA1 has good correlation with NDVI cumulation of whole year, PCA2 has good correlation with NDVI difference of winter and summer,PCA3 has good correlation with NDVI difference of spring and summer,PCA4 has good correlation with NDVI difference of spring and autumn.\;Moreover, this paper shows that different vegetation type has different feature in the four bio-parameter images, that is (NDVI, NDVIws, NDVIss and NDVIsa, and these four bio-parameters (similar to the first four principal components) have the great potential to make land cover classification in China. \;This paper indicates that as for multitemporal NDVI data of one year, PCA not only compresses the information to the first four principal components, but also extracts the key seasonal change information. The PCA1 represents the basic land cover information of vegetation, the other three main principal components (PCA2,PCA3,PCA4) extract the seasonal change information of vegetation. In all, they should be included to aid the land cover classification to increase the accuracy.