Modelling soil sealing in Flanders and the Brussels Capital Region: relating residential soil sealing density to spatial context and urban morphology
ISBN 978-85-88783-11-9
Authors
1Canters, F.; 2Vanderhaegen, S.
1VRIJE UNIVERSITEIT BRUSSEL - CARTOGRAPHY AND GIS RESEARCH GR Email: fcanters@vub.ac.be
2VRIJE UNIVERSITEIT BRUSSEL - CARTOGRAPHY AND GIS RESEARCH GR Email: svdhaege@vub.ac.be
Abstract
The spatial structure of the Flanders and Brussels Capital Region is often described as a scattered network of sprawled cities. Ribbon development is omnipresent along the well-developed transport infrastructure, resulting in a highly fragmented landscape. Due to the rapid and still ongoing urbanization in Flanders, monitoring urban sprawl and understanding its dynamics is of major importance to assess its environmental impact. The conversion of natural areas to impermeable surfaces has a direct connection with environmental phenomena like urban heating, increased surface runoff and decreased recharge of groundwater bodies. As such, knowledge about the degree of soil sealing, its spatial pattern, and how this pattern is linked to ongoing processes of urbanisation is essential for planners and policy makers. Combining medium-resolution remote sensing data covering large areas with high-resolution data for subareas of limited extent allows detailed mapping of soil sealing, both now and for the past, and provides valuable information for quantifying urban growth. Future estimates of soil sealing can be obtained by applying observed relationships between soil sealing and population density on population forecasts at municipality level. Alternatively, observed average levels of soil sealing per land-use category can be linked to land-use forecasts. Both these approaches, however, do not provide information with sufficient spatial detail, as required for urban environmental studies. Recent work by Vanderhaegen et al. (2015) attempts to improve the modelling of soil sealing distribution by relating the proportion of soil sealing within residential areas to the spatial context of these areas, more in particular by taking into account densities of different land uses in the areas’ surroundings, proximity to transport features (roads, road exits, railways, railway stations), and positioning with respect to urban clusters of different size. Although the use of a piecewise, regression-tree-based modelling approach in this study shows satisfactory overall results, the spatial distribution of the residual error reveals the inaptness of the model to account for significant local deviations from the global trend, mostly within low-density residential development in the immediate vicinity of major cities. In the present work the context-based approach described above is extended by including key characteristics of the morphology of the built-up area within each neighbourhood in the modelling, using a set of urban metrics, derived from large-scale reference data. This way not only spatial context, but also structural properties of the neighbourhood itself become an explanatory factor in the modelling of soil sealing distribution. The model is calibrated for the Flanders and Brussels Capital Region, which is covered by four Landsat 8-OLI scenes, for which 30m sub-pixel soil sealing proportions are estimated, using high-resolution IKONOS imagery for a subset of each scene to provide reference data for regression-based spectral unmixing. Including information about the spatial configuration of built-up areas within residential zones, like e.g. patch size, compactness and building adjacency, results in an overall increase of model performance and a decrease of the range of residual errors, and in a model capable of accounting for trends in soil sealing characteristics for different types of residential neighbourhoods. References Vanderhaegen, S., De Munter, K., and Canters, F. (2015). High resolution modelling and forecasting of soil sealing density at the regional scale. Landscape and urban Planning 133, 133-142. DOI: 10.1016/j.landurbplan.2014.09.016