Towards automation of impervious surface mapping using high resolution orthophoto

Information on the amount and pattern of impervious surface is important for hydrological modelling of urban areas. As cities expand and/or develop, hydrologic models will become outdated unless information on impervious surfaces is kept up to date. At the moment, the mapping teams are faced with choosing from among a range of alternative approaches/tools/software products to achieve this. We report here, the results of experiments conducted for a range of mapping approaches applied to high resolution orthophoto imagery covering part of the residential zone of Monash City, a local government area in Melbourne, Victoria, Australia. The application of the Expert Classification (EC) or Feature Analyst (FA) approaches requires initial human involvement to set the knowledge/learner function, which, then can be applied to any areas of similar spectral patterns. The Feature Analyst (FA) approach yielded superior results compared these 'pixel-by-pixel' methods. All of these approaches refer to mapping in aid of distributed and connectivity modelling. In the absence of access to EC or FA tools, application of the (sampling-based) Precision Method (PM), (after careful consideration of sampling stratification) will offer total impervious surface data-input estimates for lumped hydrological modelling.