Deriving urban land use with metric-based signatures: Comparing Landsat ETM+ and SPOT 5 imagery

Medium resolution (MR) satellite images are ideally suited for mapping spatial patterns of the built environment and their changes through time. Because urban form and function are closely related, a time-series of MR images is also useful to infer historical land-use patterns, which is required for calibrating urban growth models. As the structural detail that can be resolved from imagery inevitably depends on spatial resolution, the research presented in this paper intends to examine the effect of image resolution on the accuracy with which land use can be derived. To this end, a supervised classification strategy using spatial metric based signatures derived from continuous impervious surface maps was applied separately on a Landsat ETM+ image (30m resolution) and a SPOT 5 image (10m resolution). Results indicate that although broad land-use classes could not be further subdivided on an image of higher resolution, the distinction between these classes clearly improved when the SPOT 5 image was used instead of Landsat.