Multiscale Analysis of Landscape Spatial Heterogeneity Using Vegetation Indexes

Abstract—Remote sensing provides multiscale image data to monitoring the earth surface. The spatial heterogeneity of the surface is a function of image scales. It is also affected by variousremote sensing variables. This work uses variogram to assess the abilities of NDVI (normalized difference vegetation index) and DVI (difference vegetation index) to exploit the surface spatial heterogeneity. The decay of spatial heterogeneity as pixel size increases is measured and the spatial variability within coarse spatial resolution pixel is calculated. The results show that: 1) NDVI and DVI display a similar ability in detecting the spatial structure. NDVI variogram modeling outperforms DVI modeling in characterizing the spatial variability of the surface; 2)the spatial variability and the spatial structure both follow logarithmic relationships with stronger fits as the spatial resolution decreases; 3) the loss of spatial variability within pixel increases as spatial resolution decreases. Simple aggregation of fine resolution pixels to a coarse resolution engenders loss of image variability.