umm" a 
  
a 
Ea ang a AE SA 
  
OVERCOMING URBAN MONITORING PROBLEMS 
WITH THE NEW GENERATION SATELLITE SENSORS 
B. C. FORSTER 
School of Surveying and 
Centre for Remote Sensing 
University of New South Wales 
P.0. Box 1, Kensington, 2033, 
SYDNEY, AUSTRALIA. 
ABSTRACT 
The launching of the first Landsat satellite stimulated much research into 
urban and regional applications. Studies that have been completed show that 
broad land use classes can be mapped with better than 80% reliability; but 
only Timited success has been achieved in breaking the general classes, parti- 
cularly residential, into sub-classes. In addition, it has been found that the 
monitoring of urban periphery change is difficult due to the complex nature of 
the urban Tandscape. This paper will address the spatial and spectral problems 
of the currently available satellite data (Landsat I to III) for use in more 
detailed urban studies and indicate how the new generation of satellite sensors 
will overcome many of these problems. 
Spatial problems addressed will include those due to: 
(a) the loss of contextural clues of site and association so essential in 
manual interpretation; 
(b) the heterogeneous nature of urban areas causing mixed pixel response; 
(c) the integrating effect of the sensor point spread function which signifi- 
cantly affects a pixel size single cover class; 
(d) the existence in urban areas of a continuum of cover classes not amenable 
to current classification procedures, and 
(e) the spatially varying nature of additive atmospheric effects due to the 
variation in urban bàckground reflectance. 
In addition, spectral problems due to the overlapping and parallel nature of 
urban spectral signatures will be examined and it will be shown that an increase 
in the number of spectral bands, as for example with the Landsat-D TM, will 
increase the potential for percentage surface cover prediction. 
INTRODUCTION 
The magnitude of worldwide change due to urbanization can be gauged from United 
States' figures where 295,000 ha p.a., on average, were converted from non- 
urban to urban use (NASA 1979) during the period 1960-1970. Many developing 
countries are experiencing much faster rates of urbanization, for example, in 
the two major cities of Papua New Guinea it is anticipated there will be a near 
threefold population increase between 1976 and 1986 (NORWOOD, 1978). It is 
therefore important that the most timely and detailed information is available 
to aid forward planning in urban areas. 
Throughout the world many studies have used remotely sensed data from the 
Landsat satellite to classify urban areas. One result of this work is a 
classification system for urban areas suitable for use with remote sensing 
889 
a 
- ST A 
m MS HEUS = 
mea A 
EN N =