Advanced Visible and Near Infrared Radiometer type 2 (AVNIR2) onboard the Advanced Land Observing Satellite (ALOS) provides three visible and one near infrared bands with 10m spatial resolution. The spatial resolution of AVNIR2 was improved from 16 m spatial resolution of the AVNIR sensor onboard the past Japanese satellite Advanced Earth Observing Satellite (ADEOS), but mixels (mixed pixels) have been still problematic in creating classification maps from the ALOS/AVNIR2 data, especially in urban areas. Because of their heterogeneous land-cover types in urban areas, mixels can be seen in urban areas more often than in rural or other homogenous natural areas. Although higher spatial resolution images provide a better interpretability, more detailed ground features create much more diversities in each pixel, which may lead to problems for automated classification algorithms. To prevent these problems in extracting urban built-up area from ALOS/AVNIR2 data, this paper describes the results of analyzing the band reflectance characteristics of buildings in different land use zones of Nagoya, Japan. Land use zones constitute the basis for improving the city environmental aspects and preventing from the indiscriminate land use and design in the city. Nagoya city has currently twelve different land use zones. Quantitative analyses in these land use zones showed distinguishable differences from industrial/exclusively industrial use zones of Nagoya city. These differences in industrial/exclusively industrial use zones may bring lower accuracies when automated supervised-classification was applied with training data mixed with none-industrial use zones. We could clarify these problems which makes automated supervised-classification lower accuracy in urban areas, and our future proposed research will bring important keys to reduce these mixel problems by examining characteristics of physical properties of buildings in urban areas.
