506
As a principle, always two satellites are
to be functioning. Hence, everywhere is
covered four times a day.
A vegetation index data are made from
one week data of AVHRR. Daytime data are
sampled and mosaicked by the lowest
radiance principle. Cloud covers are
thus almost eliminated from the results.
As for the band selection, only visible
and near infra-red bands are selected.
Bands 1 and 2 of vegetation index data
corresponds those of AVHRR bands. Bands 3
and 4 of the vegetation index data are
calculated from bands 1 and 2 according to
the following equations.
band 3 = band 2 - band 1 +
band 4 = band 1 - band 2 + c
band 1 + band 2
C^,C2 : constant
3. DATA PROCESSINGS
One data set was used in this study. They
are shown in Fig. 1.
A preliminary classification revealed a
fairly large shading effect mainly caused
by sun angle deviations. Those shading
effects were first eliminated by the
following equation.(See Fig. 2)
1' = I/cos[ {7r/2-2tan _1 (r/2R) }- (X ]
Here,
1' : image value at point P after the
elimination of earth curvature
effects
I : observed image value at point P
R : radius of the earth
(X : declination of the sun
and
r = 2Rtan { ( 71/2- (D)/2 }
where
(p : latitude of point P
The classification was done by a maximum
liklihood method. Training areas were
chosen according to the World Vegetation
Map made by Preston James et al. . At
the first stage, 67 categories were
selected from this map and spectral
signatures of the image. After the
classification, these 67 categories were
unified to 17 categories as shown in
Table 1.
Four kinds of band combinations, i.e.
band 1 and 2, band 3 and 4 , band 1,2 and
4, and all 4 bands, were tried. From
these results, the original two bands
combination shows a good discrimination
between forests and crops while the
vegitation index bands show a relatively
comprehensive discrimination.
From the purpose of this study, the 4
bands combination was selected for further
processings. The classified results were
converted to longitude and latitude
coordinates as shown in Fig.3 and the
acreages of each categories were
calculated as shown in Table 2.
4. CONCLUSION
As a result of this study, the following
conclusion were derived.
1. The NOAA vegetation index data were
proved to be a suitable data for
world wide vegetation monitorings.
2. A method for the vegetation
classification of NOAA vegetation
index data was proposed.
3. Though vegetation index data are
nearly cloud-free, remaining clouds
were still problems for classi
fication .
4. As the actual ground truth is almost
impossible for this kind of studies,
more reliable sources for world
vegetation distributions are
necessary.
5. REFERENCES
1) 'World Vegetation Map',
Preston E James et al,
The TIMES Atlas of the World, TIMES
Books, London, 1981.
