numbers
ined
and a
alues
guire 1),
istent
of NDVI
TVI
geneity
ts. Table
y
) the
B.
our
hods can
it terrain,
ndition.
[ain area
‘€ also
ns.
ctance
be limited
he
OWS
-VNIR
| cover
n the river
ind near-
igh water
ween
ted to
he
|]
| and non-
ed
at in the
ow &
| off-nadir
are
tions
derived
ity of
rbid water
water
The spectral reflectances of freshwater and salt water
decreased as wavelength increased.
Additionally, the decreasing rate of freshwater reflectances
was greater than that of the salt walt reflectances.
Therefore, red and near-infrared reflectances of the river water
body were higher than those of the sea water body (see Figure
3 (B)). js | |
This may be due to high turbidity causing high reflectance and
low absorption in red and near-infrared bands.
It should be noted that the higher vegetation indexes for the
river were related to higher suspended solid concentrations
present compared to the sea water (see Figure 3 (A) ) because
yellow substances absorb mainly in blue region of the
spectrum (Tassan, 1988). As a consequence, the spectral
signatures of water bodies vary with turbidity and its
composition so that OPS VNIR data are useful for monitoring
of water quality.
4.3 Spectral Angle Mapper and Maximum Likelihood
Classification
With the limitation of flat water surfaces the supervised
classification method using a maximum likelihood decision
rule provided the best display of the location and movement of
both freshwater and salt water. Because the classified pixels
have the highest probability, the resulting classes are
sufficiently accurate for the mapping of the diffusion Nakdong
River water into the coastal sea, and also its influences on the
adjacent coastal environments (Figure 4).
As shown in Figure 4 maximum likelihood classification
shows promise in detection of water properties.
Finally, a major advantage of SAM algorithm is that it is
available for discriminating the given seven classes within
shaded polygons by using as an additional parameter the
measure of maximum acceptable angle between spectra
vectors, because the decision rule of spectral angle depends
only on the direction of the spectra, not their length (Research
Systems, 1995). The site-specific accuracy assessment that
compared and ground truth data with a GIS supporting
verification the results of SAM classification are presented in
Table 4.
The SAM classification was conducted using an additional
parameter measured in “Maximum Angle (radians)”. To
evaluate the effect of this parameter unclassified areas were
selected, and compared to the corresponding areas on shaded
relief image on the color composite scene. When unclassified
mountain areas were added to the forest class the resulting
value was very similar to the sample value.
The ground truth data collected at the end of December,
1994 were used for the validation of the proposed land cover
classes, Also during the study I found that the GIS supported
In-situ analysis could be used for detecting the land cover
changes of the classified image data in agricultural areas and
wildlife and bird sancturies related mainly to human impact.
The identification of shadow areas(mainly termed
unclassified) was accomplished through visual interpretation,
and could be exactly displayed on the processed images.
5. CONCLUSION
l. The SAM method produced a more accurate land cover
classification of areas with steep slopes, various aspects
and low solar angles than conventional classification
methods.
2. The spectral digital numbers and vegetation indexes of
mountainous forest areas were higher for aspects facing
the sun than for aspects away from sun under low sun
elevation condition.
3. The spectral digital numbers and vegetation indexes for the
flat terrains could be used as ecological and environmental
parameters, regardless of geometric illumination conditions.
4. The supervised classification of water surface was useful
for monitoring changing water properties in esturaine
and coastal areas.
ACKNOWLEDGEMENTS
This work was financially supported by the Korea Science
and Engineering Foundation under Contract KOSEF 941-
1300-008-1 with the Kookmin University.
NASDA(National Space Development Agency) of Japan
provided the JERS-1 OPS data, whose ownership belong to
MITI(Ministry of International Trade and Industry) / NASDA.
The author wishes to thank Dr. Jerry Korol at the
Department of Forest Resources, University of Idaho for his
help in reviewing the manuscript.
REFERENCES
Center for the Study of Earth from Space (CSES) and
Cooperative Institute for Research in Environmental Sciences
(CIRES), 1992. SIPS User's guide : Spectral image processing
system, version 1.2. University of Colorado, Boulder, pp.16-
29.
Deering, D. W., Middleton, E. M. , Irons, J. R., Blad, B. L.,
Walter-shea, E. A., Hays, C. J., Walthall, C. L., Eck, T. F.,
Ahmad, S. P. and B. P. Banerjee, 1992. Prairie grassland
bidirectional reflectances measured by different instruments at
the FIFE site. J. Geophys. Res., 97, pp. 18887-18904.
Hanan, N. P., Brown de Colstoun and S. D. Prince, 1995.
Estimation of canopy photosynthetically active radiation
(PAR) interaction using bidirectional aircraft reflectance
measurements of the HAPEX-sahel bush fallow subsite. In :
Proc. of IGARSS’95 Symposium, Firenze, pp. 1299-1301.
Jensen, J.R., 1996. Introductory Digital Image Processing : A
remote sensing perspective. Prentice Hall, Upper Saddle River,
NJ, pp 240-247.
Maselli, F., Conese, C., Filippis T. D. and S. Norcini, 1995.
Estimation of forest parameters through fuzzy classification of
TM data. IEEE Trans. Geosci. Remote Sensing, 33(1), pp 77-
83.
Research Systems, Inc., 1995. ENVI Version 2.0 Tutorial.
Boulder, Colorado.
Tassan, S., 1988. The use of Thematic Mapper for coastal
water analysis. In : International Archives of Photogrammetry
and Remote Sensing, Kyoto, Japan, Vol. 16.
Wang, Y. and D. L. Civco, 1992. Post-classification of
misclassified pixels by evidential reasoning : A GIS approach
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
