SPOT DATA FOR URBAN LAND-COVER MAPPING AND ROAD MAP REVISION 
Bengt Paulsson 
UNCHS (Habitat) 
P.O. Box 30030 
Nairobi, Kenya 
Commission IV 
ABSTRACT: 
Urban managers in the developing world, where cities face rapid population growth and expansion 
in area, urgently need updated maps that can be produced quickly and economically. The potential 
of SPOT data for mapping and revision of road and land-cover information was tested for a study 
area in Manila, Philippines. Large-scale maps complemented with field work were used as 
reference to evaluate the accuracy of maps produced at 1:25 000 scale. Visual interpretation of 
panchromatic data produced the best results. The revised roads had an overall accuracy of 87%, 
which is comparable to the existing base-maps produced from aerial photos. 13 land-cover types 
were separated with an overall accuracy of 92%. Areas under construction and varying degrees of 
completion covered 23% of the urbanized area. 
KEY WORDS: Urban land-cover, Map revision, Developing country, SPOT 
INTRODUCTION 
It is not uncommon for cities in the developing world to face 
annual population growth of up to 10 percent, and an expansion 
in area by 50 to 100 percent every 10 years. When having to 
respond to these development pressures, the urban managers 
often have no up-to-date base-maps and systematic information on 
the extent of settlements, land-use patterns, environmental 
problems, and infrastructure facilities. 
New solutions need to be found that meet the urgent requirements 
for updated and well maintained information bases. Traditional 
techniques, such as ground surveying and conventional aerial 
photogrammetric methods are frequently too costly for the 
municipal budget. Because of the time required for production, 
the resulting maps are repeatedly out of date and need revision on 
delivery. 
With this background, a pilot project was initiated within the 
framework of the UNDP/UNCHS(Habitat)/World Bank Urban 
Management Programme. The project seeks to advise the urban 
manager on the utility of satellite remote sensing as a source of 
information in developing countries (Paulsson, 1992). The pilot 
project is undertaken in collaboration with the GRID centre at 
UNEP in Nairobi. 
This paper presents findings from a case-study in Manila, 
Philippines, regarding the potential for urban land-cover mapping 
and road map revision. 
352 
URBAN LAND-COVER MAPPING 
The nature of urban land-use and land-cover makes computer- 
based methodologies like classification difficult to use. The 
primary obstacle is that materials such as concrete maintain their 
reflectance characteristics as measured by the satellite also when 
they are configured by man to accommodate a diverse array of 
land-uses. The proportion of vegetation can help the computer; 
e.g., to discriminate commercial zones from single-family and 
multiple family residential. But the importance of size, shape and 
texture/pattern for separating urban features cannot be 
overemphasized. 
Computer-based analysis has been used in developing country 
environments to produce fast mapping with the limited scope of 
mapping the extent of urbanized area; e.g., Brouwer et al. 
(1990). The limitations of computer-based classification based 
only on reflectance data have prompted research into methods 
that include texture and context information and knowledge-based 
processing (Gong and Howarth, 1990, Maeller-Jensen, 1990). 
These methodologies are, however, not yet operational. Often, 
a final delineation through visual interpretation is needed to 
improve the results (Gastellu-Etchegorry, 1990). In the latter 
study, standard visual interpretation of panchromatic SPOT prints 
was also done and produced significantly better results. 
When the satellite data is supplemented in the analysis by field 
verifications, air photos, old maps, and general familiarity with 
the area, visual interpretation can produce more detailed