Study On The Characteristic Scale Of Eastern Qilian Mountainous Landscape

With three periods of remote sensing images from landsat, the characteristic scale of eastern Qilian mountainous landscape was studied by semivariance analysis and wavelet transform. The results were as follow:1)There were four scales of structure feature in eastern Qilian mountainous landscape: The main plaques type were shrubs, grass and waters between 4100～4800m; Forest, shrubs, grassland and waters between 5400～6500m; Glacier between 11000～15000m; Glacier and wasteland between 20000～22000m.2)The characteristic scales were different in the same phase and different directions. The characteristic scales in the east-west and north-south were less than that in the northwest-southeast as a whole, which related to the northwest-southeast strike of geological structure system in eastern Qilian mountainous landscape.3)The comparison of characteristic scales between the same directions and different phases showed that: Glacial and wet land reduced due to the climate warming in the late 1990s in the eastern Qilian mountain; The heterogeneity of eastern Qilian mountainous landscape was influenced obviously by artificial reasons such as overcutting of forest and overgrazing. All of which caused the correlation of plaques such as forest, shrubs, grassland and waters in the study area were weakened and characteristic scales got smaller.4)The characteristic scale of eastern Qilian mountain landscape became larger with the increase of sampling spacing, but the increasing range was significant different in time and direction. If the sampling spacing continued to increase thousands of meters, the autocorrelation of landscape pattern could not show better. Therefore, the change of sample step in semivariance function and calculating spacing could overcome the lack of multiscale analysis of semivariance.5)The method of semivariance analysis could choose calculating spacing and direction according to the requirement. It could also decide the characteristic scale in different directions by the theoretical model fitting, and explain the fact that the ecological effects diminished with the distances. But it was insensitive to the analysis of the structure characteristics in different directions, and could not reveal the multi-scale features of landscape pattern in the research areas. Whereas the wavelet transform could make up the insufficiency very well. Therefore, the paper suggested that the combination of the two methods had obvious advantages in grasping spatial characteristics of landscape and analyzing the characteristic scale of landscape.