关键词: 戈壁, 砾石粒径, 主成分分析, 遥感, 哈密

Abstract: The size of gobi surface gravel is correlated to the factors such as multispectral remote sensing data, vegetation indexes and geological factors. However, these factors are usually strongly correlative. The size of gobi surface gravel model will become an ill-posed one if the model is built directly with the factors. The principal components (PCs) for those factors are obtained by principal components analysis (PCA). In that case, not only the main information of these factors can be reserved in the model, the multicolliearity problem of the factors can also be avoided. Moreover, the number of the variables decreases and the model is optimized. Based on the data obtained from Landsat TM images of 2010 and 30 m DEM at a alluvial-fan in Hami, Xinjiang, China, the paper analyzes the PCs by PCA for the 43 factors, which include 6 multi-spectral bands, 2 kinds of vegetation index of NDVI and GEMI, surface roughness generated from DEM, mean, variance, entropy, correlation and contrast extracted from texture analysis. The results show that the accumulative ratio of contribution of the first 5 PCs is 98.0%. Then the size of gobi surface gravel model is set up by regression analysis of SPSS 18 based on these first 5 PCs. F test examination shows that the size of gobi surface gravel is correlated significantly to these first 5 PCs. Finally, the study estimates the size of gobi surface gravel based on the model, and the precision is above 80%. We could learn about the gobi characteristics, cognize the laws of sand grain move and desert extension by studying the grain size of gobi surface gravel.

Key words: gobi, size of gravel, principal components analysis, remote sensing, Hami