A Region-based Approach for Describing Urban Morphology Based on Sub-pixel Estimation of Sealed Surface Cover

Earth observation provides regular information on urban development and could provide an important contribution to map and monitor structural characteristics of expanding cities. While most contemporary remote sensing research in urban environments tends to focus on high resolution satellite imagery, urban monitoring and modelling applications usually rely on extensive historic archives of landuse maps with a large geographic extent. Medium resolution images therefore seem better suited for these applications. Their lower resolution, however, inhibits studying urban morphology and change processes at an intra-urban level. In this research, we circumvent that problem by developing spatial metrics for use on continuous sealed surface data produced by sub-pixel classification of Landsat ETM+ imagery covering the city of Dublin, Ireland. Metrics we developed are based on the shape of the cumulative frequency distribution of estimated sub-pixel fractions at block level, as well as on spatial variation of sub-pixel fractions within each urban block. A MLP classifier is then used to relate the metric variables to urban landuse classes selected from the MOLAND topology. In combination with density information derived from the sealed surface maps, our approach allows producing maps describing urban morphology and intra-urban dynamics. These maps are a valuable source of information to aid the calibration of land-use change models. * Corresponding author.