Analysis of Water Quality Pattern near the Lake Hinuma using Passive Image Data
The principal component analysis
The principal component analysis is tried on the
variable of the pelitic sediment (3). Primary main
component (Z1) and second main component (Z2) that
exceeded 80% at the accumulation contribution ratio
were adopted. It was divided into 3 groups by the
plotting of the main component score on the Z1....Z2
plane. The ratio of coarse grain sediment of the sand
and silt is big for first group Gs1, and there are features
of which all nitrogen and moisture are comparatively
small. Second group Gs2 and third group Gs3 has the
contrary property to the first group Gs1.
It is the characteristic of second group Gs2 that mud
quantity, all nitrogen and moisture, etc. are big. These
of third group Gs3 are comparatively small. These
group Gs1, Gs2 and Gs3 are shown in Figure-4.
Fig.4 Result ofthe principal component analysis
Satellite image and the classification result
The utilization image
As satellite image data from the passivity sensor,
Landsat
·TM data in the Ibaraki Prefecture area in
Japan is used. This image data are extracted on 24 Apr.
1994, and it is a scene of path 107-row35 and 0% cloud
amount.
CCT value of water area.
Training area of assumed 6 sites was selected from
water area in the image analysis district. The difference
of mean CCT value of these training area affects to the
evaluation of the classification. The almost equal class of
the mean value of CCT value is arranged, and Table-1 is made.
Table-1 Distribution of the band data in water areas
| water area | Band 1 |
Band 2 | Band 3 | Band 4 | Band 5
| Band 6 | Band 7 |
| NO.1 | 92.5 | 35.5 |
34.1 | 20.2 | 14.1 | 117.7 | 6.7 |
| NO.2 | 98.1 | 40.1 | 40.1 | 24.8 | 18.5 | 122.3 | 8.8 |
| NO.3 | 92.7 | 38.9 | 40.1 | 25.5 | 19.4 | 128.1 | 8.6 |
| NO.4 | 94.3 | 30.9 | 39.7 | 26.6 | 19.1 | 130.1 | 8.6 |
| NO.5 | 97.3 | 40.2 | 41.1 | 26.5 | 19.1 | 128.4 | 8.7 |
| NO.6 | 97.7 | 41.1 | 43.3 | 29.2 | 20.9 | 131.7 | 9.3 |
Next, there is seldom difference of CCT
value of band 1 and band 7 as the result of examining
the CCT value with band data of each class. Then, the
classification was carried out next selected bands.
1) Band 1, 2,3, 4, 5, 7 3) Band 2, 3, 4, 5
2) Band 2, 3, 4, 5, 6
Thematic map preparation by these band variables is
carried out by remote sensing analysis system of
RSIPS[4].
The image classification result
Waters in Lake Hinuma, Nakagawa River and
coastal zone have been classified into six kinds.
Clustering processing of a round robin inspection needs
a long time. So the clustering uses sampling mesh data
instead of all pixel data of study area. The sampling
mesh data number was made to increase from 250, and
test of the clustering was carried out. The sampling
mesh data number was increased till about 450 and
classification result was stabilized. The selectivity of
superscription variable is examined from the
classifi-cation result of waters. Referring to the
pattern of this classification result, the high similarity
training district selected 12 classification items from
study area (4). In making this training district to be
the supervised data, the classification was applied to the
satellite image data for the purpose of water area in the
lake Hinuma, Nakagawa River and coast. This result
is shown in Figure-5. The classification map was output
by the application of maximum likelihood method
(MLHM). Figure-6 is this classification result.
Discussion
Sea surface area is divided into 2 layers when the
relation of the classified pattern of the water system of
LANDSAT satellite image data by remote sensing
technique and water depth. The lake Hinuma water
and the Nakagawa River water were distinguished. It
is possible to distinguish the lake Hinuma water to
suicidal drowning and effluent area and lake area of
Shishido river. The complicated pattern property is
shown near the center in the lake, and the classification
result changes by selection position of the training area,
and stabilizing pattern property is not shown. It is
difficult to under-stand the complicated pattern
property near the center of the lake by satellite image
data analysis in a time.
Conclusion
The analysis becomes easy, if the part of business
work analyzed from existing material and drawing and
field study data is simplified by the new technique.
Until now, the situation of surficial extension and
distribution pattern has been easy to understand by
diagrammatizing characteristic group got from the
principal component analysis of the 3, 4 chapter. This
pattern situation may be replaced to the investigation of
the remote sensing from the sky (5) under the
expectation of the superscription. The classification
technique of remote sensing was applied to image
datum. This result is as follows. Shishido river effects
category is classified as class A of Lake Hinuma water
system. It was also possible to find its boundary. This
is correspondent to group a1 shown in the water quality
data analysis. The Hinuma watershed is interpreted as
effect category class B of the coastal water.
This is corresponded to group a3 of the water quality
data analysis. The Hinuma watershed is understood
as effect category class B of the coastal water. This is
corresponded to group a3 of the water quality data
analysis. It is not possible to show clear boundaries
near the center of the lake for the complicated pattern
property such as class C and D. This is corresponded
for group a2 of the water quality data analysis and
group of Gs2 and Gs3 of the soil analysis. The pattern
in the sea water area in the coast is classified into the 2
classes. Class F is the deep water, and class E is the
shallow water depth area in the effect category of the
running water of Nakagawa River.
Reference.
- National Astronomical Observatory (1998):
Chronological Scientific Table’98, Maruzen,pp.662-663.
- Yukio Asuma, Horie, M. Ishii, T. Miura, K. and
Ohshima, K(1998):Environment Resources of Ibaraki
Prefecture Hinuma, Region Integrated Lab. of Ibaraki
Univ., pp.1-21.
- Noboru Ohsumi (1989): Statistical Data Analysis and
Software, Education Promotion Association of
Broadcasting University, pp.116-136.
- Takashi Hoshi(1979) : Processing System “USAS” of
MSS Data from an Aircraft and It’s Application,
Published in the Proc. of JSCE ,No.285, pp.69-83.
- D J Maguire, M. F Goodchild and D. W Rhind
(1991):Geographical Information Systems: Principles
and Applications,,Longman Group UK Limited, Sec.14.
- Kazuhito Tanii and Takashi Hoshi (1999) : Useful of
unsupervised classification residual image using RSIPS,
Proc. 58th annual conference of IPSJ,1-D,pp.2-17~18.
Fig.5 The classification result of Ward's method
Fig.6 Classification result of MLH method
