3.3 Environmental Mapping of two Malaysian
Provinces
In co-operation with the Malaysian centre of Remote Sensing
(MACRES) SSC carried out a geological and land cover
mapping of the two states Selangor and Kedah in West
Malaysia (Andersson C. et al, 1993). Selangor is the most
urbanised state in the country, with the capital Kuala Lumpur in
the central part, and with large plantations of oil palm and
rubber outside the city. Kedah, in the north, is known as the rice
bowl of Malaysia with extensive paddy fields along the coast.
Altogether 31 Land Cover Maps at 1:50,000 and 12
Photogeological Maps at 1:100,000 were produced during the
project, which was finalised in 1992. Satellite Image Maps at
1U Urban and associated
areas
E 1U; Industrial area
Estate building
Recreational area
1T Mining area
CULTIVATED LAND
| 2H Mixed horticulture
Rubber
Oil palm
Coconut
Forest plantation
Figure 2. Map legend of Land Cover Map, Malaysia.
3.4 Forest and Biomass Mapping of Malawi
During 1992 and 1993 a land use, forestry and biomass
mapping of Malawi was carried out in co-operation between the
Ministry of Forestry and Natural Resources of Malawi and SSC
(Rasch et al, 1994). The mapping was based on Satellite Image
Maps from Landsat TM at a scale of 1:250,000 of the entire
country. The interpretation was performed at 1:150,000.
Landsat TM images were selected due to cost limitations. Two
main problems due to the image material occurred, both of
which had to be compensated through more extensive field
work. Firstly many of the available cloud-free images were from
the time of year when the deciduous forests were defoliated.
This occurs only a few months after the beginning of the dry
period and resulted in difficulties in identifying forested areas
compared to shrubs. Secondly the lack of texture in Landsat as
compared to SPOT increased the difficulties of the vegetation
interpretation further, something not compensated by the middle
infrared channel.
The unique feature of this project was the use of the satellite
image maps, combined with rather extensive field surveys, to
carry out a forest volume estimation. A number of field survey
areas, consisting of clusters of test areas, were selected at
random within the forest areas (Fig 4). The areas were also used
for checking the land use mapping. One tree in each sample plot
was selected for measurement of tree height, in total 544 trees.
Altogether 43 trees were also felled for measurements of tree
volumes and estimation of a volume function for individual
706
4P
4Y
4C
NON-CULTIVATED LAND
NATURAL/SEMI-NATURAL VEGETATION
i eR
6B
7E
7TFc Coastal swamp forest
7F,
7L.
the scales of 1:50,000 and 1:100,000 were produced from
existing multispectral SPOT images. Landsat TM images were
used as a supplement to the SPOT images. Two separate teams
carried out the project, one for the land cover mapping and one
for the geological mapping. The use of other available maps and
information sources was a main feature of the project. Pre-
interpretation was carried out before the field work and was
followed by a final interpretation. As in most other projects the
final maps were produced digitally after manual digitising of the
interpretation overlays. In total 44 land cover classes based on
24 main classes were mapped (Fig 2 and 3). A stochastic
accuracy test of 100 check points gave an accuracy of better
than 95%.
NON-CULTIVATED LAND,
NON-VEGETATED
Paddy field
i 9C Cleared land
and
Sugar cane
|i 9N Natural bare land
Diversified crops
10W Large water body
Grassland, scrub-
grassland and other
erbaceous vegetation 10K Large fishpond
Bush MIXED CLASSES are shown as
combinations of the above symbols and
Dry-land forest OTHER MAP ELEMENTS
Selected road
Railway
National boundary
State boundary
po ^m River
Peat and freshwater
swamp forest
Logged forest
Secondary forest
trees. The final result provided estimates on forest volumes per
class, per area and nationwide.
A forest change detection comparing old Landsat MSS images
from 1972/73 with the Satellite Image Maps from 1990/91 was
also done. The low resolution of Landsat MSS limited the
possibilities of the change detection. Certain assumptions, for
instance that the bare rocks and marshes inside forest areas had
not been changed, and that deciduous forests and evergreen
forests covered the same areas as before, had to be made. For
different classes the change in area size was estimated for three
regions and for the entire country.
Eee
800 m
Figure 4. Layout of Cluster with Sub-clusters. One sample plot
was located at each corner of the sub- clusters. The
sample plots were circular with three fixed radii
depending on diameter class; <10 cm: 5 m radius, | 0
cm - 20 cm: 10 m radius, >20 cm: 15 m radius.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B4. Vienna 1996
