From publications in this field (Table 2), it becomes apparent that
even in the USA, where the morphology of cities is generally not very
complex, results obtained will not impress real users very much.
The data lack a lot in both thematic resolution and accuracy.
For certain specific applications, however, delineation of urban areas
by digital classification may be competitive with other methods. Such
applications include the indication of urban areas on small-scale maps
(1:100,000 and 1:250,000 map series, DANTHINE et al, 1978), delineation
of census tracts at the urban periphery (CORNER, 1976; CHRISTENSON &
LACHOWSKI, 1976), and information at the national scale on the amount
of urban land (CARTER et al, 1977). Depending on accuracy and consis-
tency, such delineations may also be used to estimate urban populations
in developing countries, and to define the area for airphoto coverage,
but the problems of such an approach are many and they should not be
underestimated.
Considering these different aspects, a limited research project was
defined at ITC, mainly for obtaining case-study material for the
Urban Survey courses, but with an eye on the potential applications
mentioned above. Four objectives were defined:
1) to become acquainted with techniques and procedures for digital
classification of multispectral images;
2) to form an idea of the possible potential of future satellite
data with higher spatial resolution;
3) to test claims by others on accuracy and number of different
urban classes obtainable;
4) to assess the potential of the present Landsat data for developing
countries to monitor their urban development.