I was hoping someone could help me out with a problem I've been battling with for a while - I need to somehow model "hotspots" of areas with high variation in pixel values across a categorical raster surface - the result is required as an input to a spatial multi-criteria analysis of tourism potential, the idea being that areas with high variation in the type of tourism attraction within a specified radius across the landscape are more favourable from a tourism perspective.
We have a categorical raster surface of tourism sites in our project area, where each pixel is coded according to the class of tourism attraction. Originally I used the focal statistics tool in spatial analyst in ArcGIS to measure variation in pixel values within a given radius around each site, but the output is a layer of discrete overlapping circles that doesn't really achieve what we need. Really, we are looking for a continuous output, that represents increasing value over space as one approaches higher densities of high variation - something like the output for kernel density, that uses a guassian kernel to model clusters I think, but obviously taking into account the variation of pixel values within these clusters. Does anyone have any ideas on that?