CHANGE DETECTION IN URBAN AREAS USING SATELLITE IMAGES 
AND SPECTRAL MIXTURE ANALYSIS 
F. Kressler and K. Steinnocher 
Department for Environmental Planning 
Austrian Research Centre Seibersdorf 
2444 Seibersdorf, Austria 
Commission VII, Working Group 9 
KEY WORDS: Change Detection, Urban, Landsat, Multitemporal, Spectral Mixture Analysis 
ABSTRACT: 
Monitoring changes of the environment is a problem faced by many different institutions today. Especially government 
agencies have the duty to detect and record these changes which may take place in urban, forest, agricultural, desert areas, 
and so forth. The challenge is to gather the necessary information at acceptable costs and to develop suitable techniques to 
detect and record the changes in question. Due to their low costs of recording data and their spatial, spectral, and temporal 
resolution, satellite sensors are very suitable for a number of different applications. As the sensors supply large quantities of 
data the techniques used for the analysis of the recorded satellite images should not only be reliable but also very efficient. 
The strategy proposed in this paper applies a spectral mixture analysis to satellite images and uses the results of this analysis 
for change detection in an urban area. It will be shown how those areas, where construction activities have taken place, may 
be derived and highlighted to facilitate the update of existing landuse data bases. 
1. INTRODUCTION 
Cities are constantly developing, and to keep up with the 
changes is a very difficult task for urban and regional 
planners. The challenge is to gather the necessary data at 
acceptable costs in the intervals required, and to develop 
suitable techniques for the extraction of the required 
information. For urban studies this task is typically done by 
photo interpretation. Small-scale photographs (e.g. 
1:10,000 and larger) and vertical stereopairs from mapping 
cameras are used for urban studies on a general scale. Air 
photography has the disadvantage of being very expensive 
in obtaining the pictures, making regular repeat coverages 
prohibitively expensive (Richards, 1992). Also a lot of 
manpower and time is needed for the analysis of analogue 
photographs. Compared to areal photography, satellites 
offer a number of advantages which are necessary for a 
system, that will be used for change detection, despite their 
still rather coarse resolution (e.g. Landsat TM, 30 x 30 m?). 
These advantages are: regular repeat coverage; recording 
data from the same geographic area at the same time of 
day; maintaining the same scale and look-angle; recording 
reflected radiant flux in consistent and useful spectral 
region; and lower costs compared to ohter mehtods (Jensen, 
1986). As the spatial resolution of satellite systems 
Improves, it will be easier to take advantage of these 
features, but even today satellite images may be used for 
change detection in urban areas. A number of techniques 
are being used for these purposes. These include image 
differencing, image overlay, image ratioing, classification 
comparison, principal component analysis, and change 
Vector analysis (Jensen, 1986). These methods have the 
disadvantage of either supplying no information as to the 
nature of the change or being very cumbersome to 
implement. The aim of this study is to show how satellite 
Images, recorded at different dates, which were transformed 
379 
using the spectral mixture analysis (SMA), can be used to 
determine where building activities have takem place. 
The SMA has been used in a number of studies in the 
natural environment, for instance in the estimation of 
sediment concentration in the Amazon River (Mertes et al., 
1993), the analysis of rock and soil types at the Viking 
Lander 1 Site (Adams et al, 1986), the abundance of 
vegetation in deserts (Smith et al., 1990), and the analysis 
of inland tropical water (Novo and Shimabukuro, 1994). 
The SMA has also been used for classification and 
consequent change detection (Adams et al., 1995) and for 
mapping evaporite minerals on a playa surface (Bryant, 
1996). One study using the SMA in an urban environment 
deals with the analysis of data collected by an airborne 
thematic mapper (ATM) of the University College of 
Swansea, UK (Foody and Cox, 1994). 
Two Landsat TM quarter scenes covering the City of 
Vienna were available for this study. The images were 
recorded on June 5, 1986 and July 1, 1991. Of the seven 
available bands the six reflective bands covering the visible 
light and parts of the near and middle infrared were used. 
Except for geocoding the images were not pre-processed in 
any other way. 
2. IMAGE ANALYSIS 
2.1. Spectral Mixture Analysis 
The aim of SMA is to estimate how each ground pixel's 
area is divided up among different cover types. The results 
are a series of images, each the size of the original image, 
and each giving a map of the concentration of a different 
cover type across the scene (Settle and Drake, 1993). 
Before these proportions can be calculated a set of spectra 
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996