f Patna and 
ment study 
mbai using 
nt study of 
ote sensing 
t along the 
INPT, New 
/s. National 
nt study of 
ing remote 
Limited. 
1, Director, 
|" Dr. 7A) 
plications), 
agement in 
cientists of 
port during 
IAPRS & SIS, Vol.34, Part 7, “Resource and Environmental Monitoring”, Hyderabad, India, 2002 
IMPROVING THE QUALITY OF DIGITAL ELEVATION MODELS DERIVED 
FROM OPTICAL SATELLITE DATA FOR TERRAIN EVOLUTION STUDIES 
Ian Dowman 
Department of Geomatic Engineering, University College London, Gower Street, 
London WCIE 6BT, United Kingdom 
Tel: +44 20 7679 7226, Fax: +44 20 7380 0453, email: idowman @ ge.ucl.ac.uk 
KEYWORDS: Earthquakes, DEM/DTM, Correlation, Fusion, Accuracy 
ABSTRACT : 
Digital elevation models (DEMs) have been generated from satellite data for many years now, and techniques for automatic 
generation of DEMs have evolved as scientists have gained a better understanding of the techniques, and the new demands placed by 
high resolution data. The establishment of a Centre for the Observation and Modelling of Earthquakes and Tectonics (COMET) in 
UK has led to an assessment of whether current techniques are producing the best DEMS for specific applications, and in particular 
to analyse the detail of the terrain to study landscape evolution. Digital elevation models are vital to this study. This paper analyses 
current methods of matching optical images for DEM generation and asks the question: how can the quality of DEMs be improved? 
It can be concluded that there are three main directions which research could take: 
e Improved stereomatching, including use of new techniques such as wavelet transformations, the use of breaklines and the 
use of new strategies that recognise terrain slope and land cover. 
e Improved representation of the terrain through selection of grid size and method of interpolation. 
e Data fusion techniques that exploit synergy and remove blunders. 
This paper will review current techniques and propose a number of methods that will produce DEMs of the quality required for the 
study of terrain evolution. 
1.0 INTRODUCTION 
As more and more sources of digital elevation models 
(DEMs) are available, and the range of applications 
increases, so there is an inevitable demand for the quality of 
the DEMs to improve. What is meant by quality, however, 
is very much subject to the perspective of the user. The 
requirements of a DEM for producing orthoimages are very 
different from those for terrain analysis, or for extracting 
buildings in urban areas. The concept of accuracy is 
similarly subjective, and might refer to absolute elevation 
accuracy of a point, or to the definition of micro relief or 
building edges. Accuracy is also linked to 'scale' and to the 
spacing between points. 
The representation of the data is also a crucial issue, as is 
the transformation that has taken place between the original 
measurements and the representation. A surface could be 
exactly represented by a set of accurate measurements with 
infinitely small spacing between those. As the spacing is 
increased so the accuracy of the representation will 
deteriorate. 
Stereomatching is now generally used to generate DEMs. 
There are many techniques used in stereomatching. The 
basic categories of area based matching and feature based 
matching can themselves be broken down to include the 
method of correlation, the strategies used, use of break lines 
eic. All of these have particular advantages and 
disadvantages and the user should be aware of these and be 
able to choose the best techniques for his or her application. 
For the study of terrain evolution, the use of break lines are 
particularly important, and the definition of micro relief. 
The detail available from the matching may be lost if 
unsuitable storage schemes and interpolation techniques are 
used. It is necessary to select the most suitable. Finally, DEMs 
can be improved by synergistic use of two or more DEMs with 
different characteristics. Interferometric SAR and optical data can 
both produce good DEMs, but are complementary, and the best 
characteristics of each can be used to improve the final result. 
An understanding of these factors may, in itself, contribute to 
improve the quality of the DEM. Improvement of the accuracy of 
the measurements will also be a factor: 
This paper will limit itself to discussing DEMs derived from 
optical images obtained from Earth observation satellites for the 
purpose of terrain analysis. The paper will first present the 
COMET project and then set out the factors which could improve 
the DEM. The more promising of these will then be analysed and 
conclusions reached on which research directions to follow. 
2.0 THE COMET PROJECT 
A Centre for the Observation and Modelling of Earthquakes and 
Tectonics (COMET) has been established between the Universities 
of Oxford, Cambridge and University College London in UK. The 
overall objective of COMET is to use the complete range of Earth 
Observation techniques that pertain to the measurement of crustal 
deformation, and to use these to determine the accumulation and 
release of crustal strain, and to reveal the effects of past 
earthquakes and faulting. Together with theoretical analysis, these 
measurements will be used to study the earthquake cycle, the 
response of the Earth’s surface to faulting and uplift, and the 
physics of continental deformation. All of these areas of research 
have a direct bearing on the quantification of seismic hazard and 
interpretation of past environments controlling the distribution of 
natural resources. The primary Earth Observation techniques that 
will be used to start with are InSAR, GPS and Satellite 
Photogrammetry. 
617