Thesis (Ph.D.)--Massachusetts Institute of Technology, Dept. of Architecture, 1982. / MICROFICHE COPY AVAILABLE IN ARCHIVES AND ROTCH. / Bibliography: leaves 200-204. / by Robert Mohl. / Ph.D.

M.Phil. (Geography) / Please refer to full text to view abstract

M.Sc. (Geography) / Landsat TM and SPOTMSS data were analysed and classified using two different procedures and the resulting maps were evaluated with respect to land use in the Sandton urban-rural fringe. The Landsat TM data consisted of 6 spectral bands (0,45-0,52, 0,52-0,60, 0,63-0,69, 0,760,90, 1,55-1,75, 2,08-2,35 IJm). The SPOT MSS data (one image taken in summer and one in winter) consisted of 3 spectral bands (0,50-0,59, 0,61-0,68, 0,79-0,89 IJm). The data from the two systems were stretched statistically so that all bands showed similar spread on both sides of the median. A ground truth map was obtained from the Sandton Town Council against which the final land use maps derived from Landsat and SPOT were compared for accuracy. The satellite data were analysed in two steps to compile the land use maps: The first step was a cluster analysis based on ISODATA of Ball and Hall (Ball, m..al, 1965). The result were 3 maps with 34, 30 and 35 spectral classes for Landsat TM and the SPOT seasonal images. The next step was a combination of cluster analysis and nearest-neighbour analysis. Examples of the land uses required for the final maps were chosen and for each a histogram of spectral classes was compiled. A nearest-neighbour analysis was done to determine how many pixels of the same class lie next to each other. All the pixels in the spectrally classified image were viewed in conjunction with the surrounding pixels; a histogram and nearest-neighbour analysis was done for each. The results were then compared to that of the land use examples and each pixel was allocated to the land use class which it most resembled. The evaluation involved a computerised comparison of the land use maps with the ground truth map obtained from the Sandton Town Council. The final results were three different land use maps, each created with one image (Landsat TM, SPOTsummer, or winter images). The land use classes identified on each map were: agricultural holdings; high density residential areas; low density residential areas; townhouses (only from the SPOTimages); a combination of commercial and industrial areas; parks; unused land; recreational areas. Comparative use of the two satellite based data acquisition systems leads the author to conclude that: Landsat TM was best for mapping agricultural holdings and high density residential areas; the SPOT summer image was best for mapping townhouses, parks, unused land and recreational areas, the SPOT winter image was best for mapping low density residential areas and commercial/industrial areas. Both systems may be regarded as data sources for urban research, for the mapping of land use in urban-rural fringes. The result of this study is the provision of an easily updated land use map of the Sandton urban-rural fringe to aid effective planning and control where future development will take place.

This thesis identifies a problem in the current practice for storage of locational data of entities in the cadastral layer of a land information system (LIS), and presents as a solution an information model that uses an object-oriented paradigm.