Philippines the only operational way of carrying out the project
was by using SPOT, any other considerations about data
sources could be disregarded. The mapping project for the
National Population Census in Nigeria used SPOT multispectral
instead of the technically more feasible panchromatic images, as
this was the only available data and the time was limited. The
point behind this is that all these things have to be considered
for each project, a general answer does not exist. It also means
that these issues have to be considered concisely and can not be
neglected.
4.6 Production of DEMS and Contours
It is possible and feasible to produce Digital Elevation Models
and contours from stereo SPOT images. It is, however, for many
reasons problematic in all areas except the most arid. As the
stereo is not obtained on the same track the same day, both the
images used have to be observed under cloud-free conditions.
This limitation allows only a few choices of images. The
consequences are; (i) the base-to-height ratios vary, giving
inconsistent and unpredictable results, (ii) the time differences
between the scenes vary, not only giving variations in precision
but, even worse, also resulting in low reliability, (iii) any time
planning is very difficult if data does not exist in an archive and
programming is required - systematic coverage of an area by
archived stereo data is rare; (iv) to cover partly cloudy stereo
pairs the number of images has to be significantly increased.
When the image material is sufficiently good (only a few days
between the image observations and a base-to-height ratio better
than 0,5) the reliability and overall accuracy is sufficient for
. many purposes. An r.m.s. accuracy of 7-10 metres is routinely
obtainable from SPOT stereo. This is sufficient for many
purposes. It is sufficient for production of Digital Elevation
Models for certain applications, such as planning of cellular
networks. It is sufficient for production of 20 metre height
contours at 1:50,000 (and corresponding elevation data bases)
for areas where maps do not exist and the budget is limited.
However, in general the limitations of today's stereo satellite
data are too large for most general mapping purposes.
4.7 Topographic Map Revision
Revision of topographic maps can be, and has already been,
carried out to a large extent using high resolution satellite
images such as SPOT. However the feasibility varies for
different types of objects. For revision of detectable objects it is
a fully operational technology, as was the case in the mapping
projects in Nigeria. When the requirements are tuned to the
possibilities, as in the Baltic projects, even completely new map
series can be established very successfully. If very detailed
information is needed, as was the case in Malaysia, only aerial
photographs, or orthophotomaps based on aerial photographs
have sufficient resolution.
The resolution of SPOT has been sufficient for identification of
area and most line objects in topographic maps at 1:50,000.
Infrastructural elements - roads, tracks, waterways - are usually
well detectable. Powerlines in open terrain and old railroads
might not be detectable, but that is more a problem of contrast
than resolution. The area classes of ordinary topographic maps
are usually very well detectable both with multispectral and
panchromatic SPOT images. Most point objects, such as single
houses, are not detectable in SPOT or Landsat images. A
significantly higher resolution is needed for detection of point
objects.
714
4.8 A Typical Land Cover Mapping Project
How shall a typical land cover mapping project be carried out.
À description of such a project is made in Fig 5. This
description puts very much emphasis on the production of
image maps. À more general description of how a work flow
could look is:
Step 1. Acquisition of low cost information for planning of the
project. The information is typically found in existing maps,
NOAA-images and quick-looks of satellite images. It will
give an increased understanding of the main features of the
study area.
Step 2. A visit to the project area by the leading staff to get an
overview of the distribution of the different classes. If
possible photographs and/or video films should be obtained,
from ground and air. Inventory of available information
sources is essential. A preliminary classification schedule
should be made.
Step 3. Test and selection of image data, product types, spectral
channels, etc. Ordering and/or production of image material.
Step 4. Preliminary interpretation. Test of classification system.
Identification of areas for field investigation.
Step 5. Field investigation in areas selected as typical and in
ambiguous areas. The field trips should give the interpreters
detailed knowledge about the area and the classes.
Step 6. Final interpretation based on the preliminary
interpretation and refined by the knowledge acquired
through the field work.
Step 7. Digitising, check and editing of the interpretation result.
Step 8. Cartographic processing and production of printing
originals. Printing of maps. Production of final report.
Formal training should be performed early in the project,
making on-the-job training possible to carry out throughout the
entire project.
5 CONCLUSIONS
The following conclusions can be made from the projects:
Satellite remote sensing is a reliable and fully operational tool
for thematic mapping of vegetation and land cover at scales
between 1:50,000 - 1:250,000. In applications covering large
areas it is actually the only realistic alternative.
Satellite images are fully operational for topographic map
revision of detectable objects. When the requirements are tuned
to the possibilities even completely new map series can be
established very successfully. However today’s satellites do not
fulfill the same requirements as aerial photographs.
Only geometrically corrected images should be used in mapping
projects. Normally Orthophotomaps should be used unless
accuracy requirement and data allows the use of non-
orthocorrected image maps. Satellite data is an excellent
material for production of image maps. For digital mapping
geometrically corrected scenes can be considered an alternative
to image maps.
Image availability is a major problem for data acquired in the
visible and infrared bands. The major factor limiting availability
is cloudiness. Operationality of the data acquisition and
distribution are major factors to consider for data choice.
Interpretation on photographic hard copies is preferable to
direct interpretation on computer screen in most applications
where new maps are produced. For revision purposes, however,
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996