Reconciling Precision and Scale in Monitoring Urban Land Cover Changes with Satellite Imagery

Ever increasing advances in the resolution of satellite imagery are matched by commensurate increases in the expectations of the precision of the results emanating from the interpretation of such imagery. This is particularly apparent in the case of urban planning applications of remote sensing techniques to the monitoring of land use changes at a local territorial scale. Clearly it is unreasonable to expect that remote sensing techniques will ever replace the precision of monitoring changes in patterns of land use activity on a site by site manual basis. Nevertheless urban planners and public decision makers alike share the implicit hope that enhanced resolutions in satellite imagery will lead to ever more accurate descriptions of land use dynamics. The methodological dilemma of having to reconcile precision and scale of remote sensing applications addressing urban land cover changes was faced by researchers in a recent study aimed at monitoring urban change in and around the metropolitan area of Barcelona, in Spain. The study formed part of the CNES and Spot Image jointly sponsored SPOT5 Application and Validation Programme (SAVP) (2002-2004) and drew upon the comparison of 1995 Spot 3 (20m resolution) and 2002 Spot 5 (2.5 m resolution) satellite images. The supervised classification methodology led to the identification of some eight land cover classes. In terms of comparing the resulting change in urban development, the said ‘change’ was found to be overly exaggerated. Given that the 2002 images were of a considerably higher resolution than those of 1995 led to distort or over-exaggerate the magnitude of change experienced between 1995 and 2002. Uses clearly not identified by virtue of the 20m x 20m pixel size of the 1995 20m image were in turn detected by the smaller 2.5m x 2.5m pixels of the 2002 image. Had the 1995 images been of a similar or closer resolution, then the degree of change observed would have been more credible. This was particularly relevant in fringe areas demonstrating disperse urban development and where urban development is in the process of consolidation. This ‘limitation’ does not undermine the benefits from incorporating remote sensing and satellite image analysis in urban planning. Nevertheless it is suggested that in order to achieve more accurate results, overcoming this distortion effect requires baseline imagery to be of a much closer resolution to the later imagery used for comparative purposes. Indeed it is anticipated that this limitation is something of the past, in the sense that as of this point in time, the increases in the resolution will be increasingly smaller, thereby reducing the distortion effect when comparing areas over time. Therefore it is hoped that urban planners and public decision makers with very local territorial interests, as opposed to those with more regional or strategic territorial interests, will not have to compromise the issue of scale and precision, and will be able to place more weight on the reliability of results coming out of monitoring exercises of this nature.