Scale dependence of autocorrelation from a remote sensing perspective

摘要

Spatial autocorrelation has been proved to be a useful tool in many fields, including spatial heterogeneity research and spatial structure investigation. With the increasing of remote sensors, images of different resolutions are being acquired and put into usage. So how to select images of appropriate spatial resolution becomes to be a great challenge. Therefore, it's necessary to investigate the scale dependence of the spatial autocorrelation in remotely sensed images, as Jupp et al (1989) has declared that the spatial autocorrelation in an image is related with the spatial resolution. In this paper, panchromatic band of the QuickBird imagery is aggregated into a series of images of coarser spatial resolution and used to investigate the scaling effects. Both global and local spatial autocorrelation measures at different scales are calculated. Results show that global autocorrelation increases as the resolution becomes coarser and lag distance decreases. Local autocorrelation shows dependence on scale and the land cover type. It's necessary to combine global and local measures together to explore the intrinsic of spatial autocorrelation.