Effects of sensor spatial resolutions and classification themes on urban landscape analysis: a case study in Shanghai, China

Abstract

Urbanization is a significant cause of change in the urban landscape. Landscape analysis and remote sensing provide effective means for researching the process of urbanization, but the problems of scale and accuracy in landscape studies are yet to be resolved. In particular, the impacts of sensor spatial resolutions and classification themes on the urban landscape with changing scale size have rarely been reported. Roads are an important land-use type in urban areas. Unfortunately, this is also one of the land-use types that are relatively easy to neglect because of sensor resolutions and classification themes. To better understand the characteristics of the urban landscape, the landscape pattern of Shanghai was analyzed using land-use maps produced from Landsat Thematic Mapper (TM) and Indian Remote Sensing satellite PAN (IRS-PAN) images with changing grain size (i.e., resolution) and extent size (i.e., size of study area). Landscape metrics were computed along a 51 km2 transect cutting across Shanghai with a moving window. The results showed that both sensor spatial resolutions and classification themes provide incomplete information, such as missing linearity corridors, from urban landscape analysis. This could greatly affect urban landscape analysis and planning. The missing information may be detected by changing both the grain and extent sizes in the analysis parameters. However, varying the sensor resolutions did not produce the same effects as varying the classification themes. Varying the classification themes led to within-class variance, and thus was more significant in fine-resolution imagery than in coarse-resolution imagery. Linearity corridors such as urban roads should be identified and classified in research focusing on the details of urban fragmentation, urban ecologies related to road corridors, and urbanization issues.