> between
ing and in
0 uncover
; indicated
used large
spectral
se of the
proposed
; Hodgson
jo method
crease the
g samples
ed regions
ation.
distinguish
vegetation
- R) were
classified
cause the
| was Very
Vegetation
supervised
1d met the
veral trail
lerived for
is drought
banks was
cloud
Table 1 Land Use Changes
Landuse 1993
hectare
dense forest 533,314 50
scattered forest 2,057 80
orchard 1,353 70
vegetable farm 2222
tea : 2
grass 5825
buildup area 22.6
barren land and landslide 1.106. 10
water 864.2
cloud 629.8
total 60,155.30
exposed. These dry river beds were very easily confused
with landslides along the river bank. A method was
devised to handle this problem. The area occupied by
normal river flow was digitized from 1:10000
topographic maps and used to mask the 1993 and the
1995 images.
Possible training sites for supervised classification were
first identified on Arial photos and topographic maps.
Field survey was taken to check the ground truths and to
select final training sites. Training sites are chosen based
on their representative and accessibility. Differential GPS
were used to record the exact location of these training
sites.
The results of supervised classifications were checked by
field investigations.
Land Use Change Analysis
The results of the classification are shown in Table 1. The
vegetable farm increased from 222.2 hectare to 542.9
hectare. The orchard increased from 1353.7 hectare to
1755.4 hectare. Forest reduced 555.2 hectare. These
figures indicated that farming practice was still increasing
during the study period. The vegetable farm used large
quantify of fertilizer and pesticide that could pull the
river. The dramatic increase in vegetable farm should be
closely monitored.
The increase in barren land and grass was due to two
forest fires happened on December1994 and January
1995.
The analysis results have shown that satellite images
could be very effective in monitoring the land use
changes in Der-Chi watershed. The procedures developed
133
1995 change
% hectare % hectare %
88.6 52.7559 .30 37.7 2855 2 -1.0
3.4 2 2530 00 3.7 192.2 9 9
2.2 1,755 40 2.9 401.7 29.7
‚4 542.9 0.9 320.7. 144.3
0 8.1 0 6.143.050
1 624.7 1 42.2 7 d
0] 25.8 0.1 3.2 14.2
1.8 1 250.90 2 ci 144.8 I3.
1.4 822.3 1.4 -41.9 -4.8
1.1 114.1 0.2 z8158.755/-81.9
100 60,155.30 100
in this research could be used for processing satellite
images for land use monitoring.
Environmental Database Construction
The satellite images and other data were used to setup an
environmental database for the Der-Chi area to support
future researches and environment management tasks.
Although, many data had been collected by forest
services, bureau of soil conservation, Taiwan power
company, bureau of aborigine management, national park
services, environmental protection agency and local
government These data were stored at each organization
and had not been organized into a coherent database. In
this research, these data were converted into a standard
digital format and formed an integrated database.
The database is consists of three data sets: the physical
environmental data set, the cadastral data set and image
set (figure 3).
environematal data base
for Der-chi Watershed
|
physcial cadastral image
environment data set data set
data set
satellite land use
images photos
Figure 3 environmental database structure
Physical Environment Data Set
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
