Figure 1. SPOT panchromatic image, © CNES 1990, over part (4.4 x 2.6 km) of the study area in southern Metro Manila, Philippines. 
A = Lake, B = Green open space, C = Industries, D = Highway, E = Major arterial road; 
Residential areas: F = Low density, G = Medium density, H = High density, I = Under construction. 
RESULTS AND DISCUSSION 
Land-Cover Map 
The final land-cover map contains 13 categories after the tentative 
interpretation classes were grouped based on the field work 
experience (Figure 2). The categories are listed in Table 1 and 
an accuracy matrix given in Table 2. 
  
Urban Land-Cover Types 
10 Transportation 
20 Industrial/Commercial 
Residential Areas 
31 Under Construction 
32 Empty 
33 In-Filling 
34 Low Density 
35 Medium Density 
36 High Density 
Non-Urban Land-Cover Types 
40 Lake 
51 Seasonal Swamp 
52 Wooded Areas 
53 Green Open Space 
54 Golf Course 
  
Table 1. Land-cover classes. 
354 
The three classes describing residential areas under development 
may require some explanation. ’In-Filling’ means that the road 
network is finished and plots are started to be built upon and 
occupied. The ’Empty’ class contains areas where the road 
network is completed, often before 1986, without signs that the 
plots are being developed. Large parts of the area belong to this 
type of idle land: 9% of the total area with residential land-cover. 
The criteria for separating these three classes may seem 
subjective, but the differences seen in the image were confirmed 
by reference data, although some errors of commission occurred. 
The classes are also an important clue to the dynamics of urban 
growth in the area; 23% of the urbanized area belonged to these 
three classes. 
In the interpretation, it was also attempted to separate 
construction in Industrial/ Commercial areas, but the field work 
revealed a low accuracy. These areas were therefore not 
separated in the final map. The difficulties are due to that high 
reflection from existing industrial buildings was misinterpreted as 
recent construction, and that new buildings were added to already 
developed industrial estates. 
The class Industrial/ Commercial has the lowest accuracy; 77% 
regarding errors of commission and 72% for errors of omission. 
These occurred where smaller industries were located in 
residential areas, and similarly where small residential areas lay 
in industrial/commercial zones. It should be noted that separation 
of small-scale commerce and industry that is commonly 
intermixed in the residential areas of medium and high density 
was not attempted. These cannot be separated.