Morpho-spatial extraction of urban nuclei in diffusely urbanized metropolitan areas

Remote sensing has permitted estimates of the built-up surface as a regional measure of urban sprawl. However, the accuracy of the estimates and the urban forms associated with this measure are still scarcely documented, especially on the periphery of metropolitan areas where urban land use is scattered in the landscape. An operational framework is presented for the structural description of metropolitan urbanization at regional scale, with a view to enhancing the urban cartography in Mexico. This framework is composed of (1) the extraction of the regional urban sprawl pattern, (2) the accuracy estimate per population density zone, and (3) the automated extraction of urban nuclei within this pattern. The method was applied to the Toluca-Atlacomulco valley, a significant component of the diffuse urbanization surrounding Mexico City. In the first step, a Landsat TM image is classified using a decision tree based on prior knowledge and a combination of unsupervised and supervised algorithms. In the second step, the results of the classification strategy are compared with alternative classification strategies. In the third step, compact objects are extracted from the built-up diffuse pattern using operators of mathematical morphology. The distribution of sizes and shapes of the extracted urban nuclei is compared with those appearing in existing fine scale national land use maps, generated via aerial photograph delineation. Among the assets of the method are the relatively objective criteria in the extraction of the urban sprawl pattern, the low cost of its generation, and the description of urban nuclei at a user-defined scale. (C) 2011 Elsevier B.V. All rights reserved.