Scales of spatial autocorrelation and self-similarity of urban landscape pattern

Choosing the center area of Shanghai as regional background, this paper analyzed the urban landscape pattern characters with different grains by the technology of RS and GIS. The results showed that landscape indices of Moran I and fractal dimension were all related to the grains. The spatial autocorrelation and self-similarity of urban landscape pattern depended on different scales within certain range of scale, with the depending degree differed greatly. The spatial autocorrelation highly depended on the scaling, while the spatial self-similarity did not follow the scaling regularly. The spatial autocorrelation and self-similarity of urban landscape pattern had different sensitive points to the scaling. For the landscape of residence, Moran I, which characterized the spatial autocorrelation of the urban landscape structure, had the sensitive point to the scaling at the level of 50 m. However, sensitive points to fractal dimension differed greatly among different types of landscape patches.