Integrating remotely sensed images and areal census data for building new models across scales

From a perspective of multidisciplinary studies, this paper introduces a framework for integrating remotely sensed images and areal census data for building new models across scales. The understanding of spatial scales of both data sources lays a foundation for scaling in attributes. Two specific tasks are reported in the paper. First, a range of statistics based on the sub-images after multiresolution wavelet transforms are calculated. It is found that the change rate of standard deviation over resolutions can indicate the representative scale of salient objects in an image. Second, within the valid scale range, standard deviation calculated at different decomposition levels increases almost linearly. Such a scale-independent statistic could serve a tool for scaling attributes of the ground objects for the area of an entire image or its sub-zones.