vation
3RS-1
radar
)ptical
, tidal
)conut
lasses
d use
) was
hange
The
nainly
forest
h and
3.4 ha
op) and
reased
. Ob of
while
d as it
nifting
essure
ılation
imited
'h will
> land.
1lation
vation,
n the
forest
ositive
hifting
jlation
en the
gative,
ılation
id data
1e first
yt data
'nized,
Multi-
, while
JERS-
cover
use the
ication
clearly
om the
.] and
ofthe
rugged terrain, high moisture content of the whole
scene, short wavelengths and low incidence angle,
single polarization of both images, and the quality
of the images.
- Visual interpretation of radar data gave good
results. It was possible to recognize four more
classes on radar images than on the optical image.
These classes were: swamp forest, tidal forest,
rubber plantation and coastal coconut plantations.
- It was possible to detect more classes from a
combined image of optical and microwave, which
would have been difficult using TM or radar data
alone.
- A comparison of the results obtained by
supervised classification of 1973 Landsat MSS and
1990 TM images shows that the latter gave more
land cover classes and better overall accuracy
because of its improved spectral and spatial
resolution. However, it was difficult to distinguish
between different forest types such as swamp
forest, tidal forest and rubber plantations. It was
also difficult to differentiate between wetland rice
fields and rivers.
- For detecting settlements, ERS-1 was found to
be better than JERS-1 and optical data (TM and
MSS). In the present study, the settlement
information was thus incorporated in the
classification result obtained using TM data. This
was a good example of improving the results of
the multispectral classification of optical data by
tr the settlement information from radar
ata.
- Using remotely sensed data, it was possible to
detect changes in land use. The major change
took place in the lowland forest area. The
reduction in lowland forest was replaced mainly
by shifting cultivation, with secondary forest
followed by bush and clearcut and settlement.
; The main factors which effect land use changes
In the study area are the increasing population
(both natural and immigration), social and
economic aspects and cultivation management
practices.
The following conclusions can be drawn from the
results of the second test site:
- Spot image was able to recognize more classes
(ll to 9) and in higher accuracy (95.4190 to
21,5%) comparing to Landsat TM image.
- ERS-1 multi-temporal images was able to detect
dynamic object (e.g. three cycle of rice field).
- ERS-1 multi-temporal images detect more land
cover type classes than single JERS-1 image (e.g.
8 to 5).
- JERS-1 image was better in classifying forest
cover types comparing to JERS-1.
- Combining optical and radar images can enhance
the ability of each one of them alone to detect
more land and forest cover types in the tropics.
S. REFERENCES
Verstappen, H. Th, 1973. "A Geomorphological
Reconnaissance of Sumatra and Adjacent Islands
(Indonesia), Groningen: Wolters-Noorhoff
Publishing. 182 p.
35
307
251
2071
Area (ha)
(Thousands)
Beach Wet Bush Pri-for Shift Settle
Land use/cover classes
[77] MSS 1973 classes M TM 1990 classes
Figure 1. Comparison between Landsat MSS and
TM Classes period (1973-1990).
g
p
©
2X
22
5
=
Q
=
A
o
14 T T^ T T T T T
0.5 0.6 0.7 0.8 0.9 1 1.1 1.2 1.3
Population (peron)
(Millions)
-"- Shifting culti. —— Lowland forest
Figure 2. Relationship between population, shifting
cultivation and primary (lowland) forest.
319
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
