The Spectral Dimension in Urban Land Cover Mapping from High- Resolution Optical Remote Sensing Data

Recent sensor development including high spatial resolution spaceborne systems like IKONOS as well as hyperspectral sensors like AVIRIS have the potential for more detailed and accurate mapping of urban land cover and land use. However, the new spaceborne sensors are limited in their spectral resolution with just four multispectral bands and few studies have focused on the issues of spectral properties of urban materials and their representation and mapping from remote sensing data. The research in this study analyzes the spectral dimension of urban materials from comprehensive field spectra measurements and hyperspectral AVIRIS as well as simulated IKONOS and LANDSAT TM data in 4 m spatial resolution. The results indicate highly complex and diverse spectral properties of urban land cover types. We explore and identify the most suitable wavelengths in mapping urban land cover and show that the spectral separability of urban land cover types is strongly dependent on spectral sensor characteristics. Particularly, IKONOS and LANDSAT TM have specific limitations in accurate separation of several urban land cover categories compared to mapping with AVIRIS that provides an overall accuracy of ~78% for a detailed map of urban land cover.