PUTTING THE PIECES TOGETHER : 
COMPOSING A NATIONWIDE DEM-COVER OF BELGIUM 
S. Roovers, J. Beyen 
National Geographical Institute, 1000 Brussels, Belgium — (str, jbe)@ngi.be 
Commision IV, Working Group IV/6 
KEY WORDS: Multisource, DEM, Combination, Modeling, Production, Inter-operability 
ABSTRACT: 
Ideally a country (region, city, etc.) is covered with homogeneous data sets. Unfortunately, this ideal situation does not often present 
itself. Frequently, we have a number of different existing data sets with varying characteristics at our disposal. One then has the choice 
between starting all over from scratch in order to get a new homogeneous coverage or trying to combine the existing data and filling in 
the gaps. The decision between the two options is a matter of a number of “pushing” and “pulling” factors as for example available 
production means, urgency, required precision, market situation, etc. At the Belgian National Geographical Institute a project was started 
in 2001, aiming at the maximum use of the existing data sets in order to quickly set up a DEM with medium-scale accuracies for the 
whole country. Overall, we used more than five different sources and methods for the collection of digital elevation data, resulting in a 
nationwide DEM-patchwork. In one of these workflows we combined different powerful commercial softwares for the automatic 
generation and filtering of a DSM. When different sources of data are available for the same area, it can be very useful to combine the 
different characteristics and exploit the diversity of the data in order to improve the quality. An example of this is the combination of 
automatically generated DSMs with DTMs derived from digitized contour lines. In a densely populated area like Belgium, with highly 
irregular man-made landscapes, it is not always feasible (in a cost-effective way) to filter the DSM down to a DTM. A second DTM based 
on digitized contour lines serves as an additional information source for solving this problem. The overall accuracy of the latter DTM is 
lower in places where the DSM describes the earth surface, but on the other hand it helps to detect where the DSM does not coincide with 
the ground surface. Tests were performed for the automated and semi-automated exploitation of this diversity. 
1. INTRODUCTION 2. COMPOSING A DEM 
The Belgian National Geographical Institute (NGI) is the 2.1. The DEM situation at the NGI up to the year 2001 
National Mapping Agency for Belgium. Besides its traditional 
production of paper topographical maps (1:10.000 1:20.000 A full DTM coverage of Belgium (30 000 km?) is available in 
1:50.000 1:100.000 1:250.000 as standard products) and raster DTED (STANAG 3809) level 2 format. It has a height 
image equivalents, the NGI produces several digital vector resolution of 1 m and a grid interval of 40m in the northern and 
products such as 2D geographical databases and 3D CAD files central part of the country and 20m in the southern part. The 
as well as orthophotos. A merging of the 2D databases and 3D poor height resolution gives rise to artefacts in several 
CAD files into one single 3D database is planned for the near applications. For two thirds of Belgium there are DSMs derived 
future. The assembly of a multisource DEM fully covering the by correlation of aerial images on a scale of 1:43 000 that were 
country as a multi-purpose reference layer was started in 2001. originally used for the production of orthophotos. For 17% of 
the country a number of DTMs were constructed by manual 
This paper gives an overview of the DEM situation at the NGI, stereoplotting of points and structurelines on aerial b/w 
the present and future DEM requirements and the general photographs on 1:20 500. These were only produced for areas 
methodologies that are being used to meet these requirements. where the terrain is not too flat and for areas which are not too 
The last part of the paper deals with a test which we performed forested. Untill 2001 a lot of information on terrain height was 
to produce a DTM combining three different data sets with still in analogue form, consisting mainly of contour lines on a 
complementary properties and which are readily available at the map scale of 1:10.000. 
NGI. We discuss some preliminary results and the possibilities 
of introducing a higher degree of automation in the tested 2.2. DEM requirements 
production process. 
Within the NGI there is a demand for a better DEM for the 
The following terminology is used. A Digital Elevation Model production of orthophotos. Since 2002 we have switched from 
(DEM) can be a Digital Terrain Model (DTM) or a Digital using 1:43 000 imagery to 1:29 000 and occasionally going up 
Surface Model (DSM) or a mixture of both. The surface in a to 1:20 500 imagery. The available DSM data, complemented 
DSM corresponds to the surface that is visible on aerial images by DTED level 2, were of sufficient quality for processing of 
or satellite images. A DTM models the ground surface. the 1:43 000 imagery. However, for the larger scale imagery we 
need a better DEM. 
uA 
O3 
N 
Inte 
The 
ref 
acc 
0.5 
nee 
con 
The 
obs 
WIC 
tert 
stu 
The 
cus 
fro: 
det: 
stu 
cus 
a lc 
off- 
ver 
nee 
2.3. 
Av 
con 
ortl 
The 
als 
ope 
isst 
acti 
has 
phi 
2.4. 
Ab 
pre 
1-1 
fiel 
stil 
Tec 
thre 
fast 
the 
pos 
The 
0.51 
ach 
thai 
Buc 
ofc 
to 
imp 
deri