Capability of SPOT XS data in producing detailed land cover maps at the urban-rural periphery

Abstract

The capability of SPOT multi-spectral (XS) data in generating detailed land cover maps at the urban-rural fringe is critically evaluated using two images, one recorded in summer and another in winter. The factors affecting the mapping accuracy are also identified and assessed in this study. Covering an area of 90km2 in South Auckland, New Zealand, two subscenes of SPOT XS images were used to map 10 categories of land cover at level II of the Anderson scheme with an overall accuracy of 76.2 and 81.4 per cent from the winter and summer data, respectively. The higher accuracy achieved using the summer image is due to the higher distinctiveness of vegetative covers in summer. The main limiting factors are identified as the high heterogeneity of land-use patterns commonly found at the urban periphery, poor representativeness of training samples caused by the presence of the same land cover element across a diverse range of land cover classes, and varying conditions of vegetative covers.