4 DISCUSSION
4.1 Geometric Corrections and Image Maps
In almost all the projects described above satellite image maps
of some kind were used. Contrary to image map prodution from
aerial photographs orthocorrection is not always necessary
using satellite images. The viewing angle is sometimes
insignificant, compared to the scale and accuracy requirements.
The advantages of image maps are many. The existing
information in maps, etc. can be interpreted simultaneously and
be directly compared with the images. The results of an
interpretation are already in a known and correct system. The
distribution, visualisation and cartographyic production of the
information can be done directly. Projects using uncorrected
satellite images usually get messy when trying to handle the
geometrical problems, combining information sources and
making cartography. When visual interpretation on hard copies
is done the possibility of working with the same image
geometry as in the end products is an immense advantage. In
digital systems geometric correction is necessary, although a
division into image maps is not always needed.
In general, excellent orthophoto maps can be produced very
efficiently from satellite data. A geometric accuracy around half
a pixel can be obtained routinely. The large size of the original
images admits homogeneous radiometry in large areas.
4.2 Availability and Clouds, Clouds and Availability
À common theme in the description of the above projects is
dependence on the availability of image data. The choice of data
type, acquisition period of the year and even project areas has
often been a function of availability. The limiting factor is
primarily cloudiness, in combination with the programming
capabilities of the system. Especially in systems with neither
side-looking capability nor tape recorder, such as Landsat,
image access has been the major limitation.
Availability is not only one of many limitations. It is even likely
that the single most important problem for the utilisation of
today's satellite image technoloy is the unreliability of the
access to data.
Availability has to be considered at the earliest possible stage in
preparing a project. In the more favourable cases data exist but
are too old. In the worst cases a complete coverage is not
possible during the project execution time. To combine
different data sources is commonly the solution. In many of the
projects in equatorial countries this has been solved by
combining a high resolution data set (SPOT) with a lower
resolution complementary data set (Landsat). Another solution
in areas where cloud-free images hardly exist is to include
acquisition of extra images in the budget and in the project plan
- to be used for cloud mosaicking.
43 Analog or digital mapping?
In all the thematic mapping projects described here the
interpretation was done on overlays of photographically
produced transparent satellite images. Afterwards the results
were digitised and transferred into a GIS system.
As an alternative method the mapping can be done directly on
the screen - with the raster image as the background for the
vector mapping. Why did we not use such methods? There are
many reasons: (i) this was not feasible technology at the time
many of these projects were carried out; (ii) there would have
been an increased risk of delays in the projects because of
technical problems, insufficient training, etc; (iii) the projects
713
might become more concentrated on technology instead of
obtaining results; (iv) the purchase of the hardware would
increase the cost of the project - connected with this is the
problem of obtaining future funding for maintenance of the
equipment; (v) all experience points towards the fact that
interpretation on photographic hard copies is faster to do than
on-screen mapping.
On the other hand, it appears that revision of data bases can be
done very efficiently digitally. It is easy to compare and detect
changes and edit them with the vector data overlaid on the
images. It is also possible to semi-automate revision - in a way
almost impossible for mapping without any previous
knowledge.
4.4 Use Auxiliary Data!
To carry out an operational project is not the same as doing
research! In research projects a basic methodology is tested on a
specific material. In fully operational projects it is necessary to
use all available data sources - it is one of the clues to
effectiveness, accuracy and success. The two most obvious
sources are existing maps and supplementary sets of satellite
images or aerial photographs. In existing maps thematic areas,
object and height information are available. Supplementary
image sets can be used to identify features and classes not
visible in the base data set. They can be used for overlapping
cloudy areas and for identifying changes. The geometric
problems of aerial photographs and of not corrected satellite
images are usually not relevant when used together with image
maps. An inventory of existing relevant maps should be made at
an early stage of a project. Usually topographic maps have the
most important information, but various thematic information
usually exists at different ministries. Any kind of existing files
from earlier inventories are also useful.
4.5 Which Image Material to Use - Stereo, Mono, B/W,
Colour, Resolution, etc?
The use of aerial photographs creates naturally, through the
acquisition schemes, availability to stereo images. However as
individual photographs they are complicated to use and handle.
Satellite images can relatively easily be corrected into
orthophotomaps, as can aerial photographs with somewhat more
time consuming methodology. What is the situation with
satellite data? To get stereo coverage with satellite is twice as
expensive as mono coverage. It can only be done with SPOT
operationally and is anyhow practically more or less impossible
except in arid areas.
The question is, which is more important for a good
interpretation result, stereo coverage or the easy and reliable
handling based on orthophoto maps?
A similar question is whether the texture in higher resolution
images (such as SPOT P) is more important than the spectral
information in, for instance, Landsat TM. When there are
differences in interpretability, are they worth trading against
other factors as operationality, price, etc.
Such questions are totally impossible to answer generally.
Stereo can be more important than, for instance, colour images -
that was part of the conclusion of the forest inventory project in
Chile. The texture of SPOT panchromatic data would have been
preferred against the spectral information of Landsat TM in the
forest and biomass mapping of Malawi. In the forestry mapping
of Northern Namibia the advantages of the middle infrared
channel were considered negligible, despite the fact that this
was a vegetation mapping. In the land cover mapping of the
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996