| Satellite remote sensing has not been widely studied and applied to urban environments. This research integrated spatial, spectral and temporal information from optical satellite data of various resolutions for the analysis of urban landscape dynamics. Medium-resolution Landsat TM/ETM+ and SPOT panchromatic data were collected over Beijing, China, during the 1980s and 1990s. In addition, high-resolution IKONOS data collected over London, Ontario, Canada, were studied as a new promising source of data. Conventional multispectral classification is insufficient for the processing of medium-and-high-resolution data because of the heterogeneous nature of urban areas. New methods for the analysis of these data for the dynamic urban environment were proposed and evaluated.; Landsat TM images were used to detect urban expansion of Beijing between 1984 and 1997. In order to separate different types of built-up land from various non-built-up lands, derived structural information (i.e., road density) was calculated and integrated with spectral information for urban change detection. Change-detection accuracy of 85% was achieved from the structural-spectral approach. Detailed urban spatial patterns in Beijing were mapped using SPOT panchromatic data of improved spatial resolution. Texture features from the SPOT panchromatic image were used to classify various urban spatial patterns (i.e., low-rise, old and new multi-storey, high-rise, construction, etc.) It was found that these patterns could be mapped with about 80% accuracy using four to five texture features. Furthermore, multi-storey buildings were identified and newly constructed multi-storey buildings in Beijing were detected using merged Landsat TM/ETM+ and SPOT panchromatic data. This study proposes a rule-based inferring method based on land cover compositions in the neighbourhood for detailed urban land use mapping from IKONOS data. Urban land uses, including single-family and multi-family residential, commercial, etc., were mapped from the IKONOS data for London, ON, with about 90% accuracy. The final GIS-compatible per-parcel land use map was produced from the detailed land use recognition result.; This is the most intensive study using medium-resolution satellite data for the changing environment of a large complex city; it is also the first study to digitally apply a rule-based approach for detailed urban land use mapping using IKONOS data. |
