Urban land use pattern identification using variogram on image

Abstract

The images of high spatial resolution leveraged remote sensing to the forefront of urban environments studies, as they better distinguish the elements that make up this very heterogeneous environment. Geostatistical techniques are increasingly being used in studies of remote sensing. The variogram is an important geostatistical analysis tool, because it allows understanding of the spatial behaviour of a regionalised variable, in this case, the grey levels of a satellite image. This study aims to identify urban residential patterns of three classes of use and occupation of land by the analysis of the parameters, graphics and results of variogram analysis. The hypothesis is that the values corresponding to these parameters represent the standard of each class spectral behaviour, and indicate that there is a pattern in the spatial organisation of each class. IKONOS 2002 images and a previous classification of land use and land cover of sub-basin in the Cabuçu river in São Paulo were used. Samples were taken from each class and the levels of grey in each pixel were used to calculate the variogram. After analysing the results, only the parameter range was considered, as it was observed that it was related with the degree of homogeneity of each sample. The range of values obtained in the calculation of variograms identified with better accuracy ‘multiple dwelling unit’ class rather than ‘regulated dense occupation’ and ‘irregular dense occupation’, which did not yield a good result.