most buit-up areas of Colombo are in the region of 160 
persons to the hectare. 
2. REMOTE SENSING DATA USED. 
Digital data from a Landsat TM scene aquired on the 
12th of February 1993 was used in the study. 
2.1 The Proposed Index UI 
The proposed index UI was computed as shown below 
using Landsat TM band7 (B7) and band 4 (B4), 
exploiting an observed inverse relationship between the 
brightness of urban areas in the near infra-red (0.76pm - 
0.90pm) and mid infra-red (2.08pm - 2.35pm) portions 
of the spectrum. 
on E221 1.0) x 100 d) 
This index was verified by examining its relation with the 
Normalised Difference Vegetation Index (NDVI) land 
cover and building cover data of the Colombo City area. 
2.1.1 Relation of UI with NDVI. In this study NDVI 
was defined as follows using Landsat TM band 4 (B4) 
and band 3 (B3). 
  
ww (18 + 1.0) x 100 Q) 
The UI - NDVI relation was examined in two ways. 
One was by seperating pixels into those of central and 
suburban areas and the other was by seperating them into 
those of different categories of land cover by using land 
cover information. 
2.1.1.1 UI - NDVI Relation Based on Central and 
Suburban Areas. Average UI and NDVI values for 
pixels of 20X20 image pixels were computed, eliminating 
pixels from water areas by overlaying with a classified 
image. These pixels were picked up from an area of 
1024X800 image pixels covering Colombo City and its 
suburbs. The central area of the city was seperated by 
selecting an area of 400X640 image pixels covering the 
central part of the city. The remaining pixels in the 
1024X800 pixels image were considered to be of the 
suburban area. The scatter diagram of UI and NDVI for 
Colombo City is seen in Fig:2. From this figure it is seen 
that UI is high when NDVI is low in the central area of 
Colombo City. 
  
120 € CENTRAL COLOMBO 
100 L o SUBURBAN COLOMBO 
  
  
  
  
Figure 2. UI - NDVI Relation 
2.1.1.2. UI - NDVI Relation Based on Land Cover 
Type. Digital land cover data of the Colombo City ares 
was obtained by scanning a 1:50000 land use map of 191. 
The UI and NDVI images were registered with images of 
digital land cover data by removing geometric distortions 
using ground control points and Afine Transformation 
Technique. Resamplng was done using Nearest 
Neighbour Method. Each pixel was assigned to a category 
if more than 80% of the original 20X20 image pixels 
within the considered pixel belonged to that category. The 
UI - NDVI relation based on land cover category is shown 
in figure 3. This figure shows that the UI value increases 
and NDVI value decreases with increased urbanisation. 
100 
  
   
  
> 60 | oBuilt-up Land iE 
  
  
  
C)Homesteads 
40 F oPlantations 
e Paddy 
20 1 1 | 
80 100 120 140 160 
NDVI 
Figure 3 UI - NDVI Relation Based on Land Cover 
Category 
322 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996