and observed on N-th day. It is found out from these
figures that some areas in the composite image of OCTS
or GLI will catch only backward reflectance radiance and
the others only forward ones. This property is very
eminent for GLI. On the other hand, any point in the
composite image of AVHRR will have the high probability
to catch the radiance from various scan angles. In
general, the backward and forward radiance from the
same objects are different. The backward reflectance
radiance from a dense forest is bigger than the forward
one, whilethe case of a desert is vice versa.
When the period for image composite is set longer, for
example almost one month (27 days), the composite
image of AVHRR is usually produced using three
sequential nine-days composite images. The method is
adaptable for GLI. For OCTS, however, the image should
be produced using fifteen scenes acquired on N-1, N-12,
N+3, N-8, N+7, N-4, N+11, N, N-11, N+4, N-7, N+8, N-
3, N+12 and N+1st days. In this case, the pixels on the
composite image will be probably observed with various
scan angles.
The solar zenith angle at observation affects NDVI.
Figure 3 shows the coverage patterns and solar zenith
angle of AVHRR and GLI at the vernal equinox. The case
of OCTS is very similar to that of GLI. The gaps of GLI
coverages can be identified in the regions of about 60
degrees in latitude owingto the tilt function. The patterns
of solar zenith angle for AVHRR is changed even on the
same season because of the delay of the local time at
descending node. The characteristics of the solar zenith
angle of composite image is as same as that of scan
angle. Any pixel of AVHRR composite image has the
probability to be observed with various solar zenith
angles. While the pixels of OCTS or GLI are observed with
some limited solar positions. If the monthly composite
image is prepared by the method as mentioned before,
various solar zenith angles will be identified on OCTS
composite images.
5. METHOD OF IMAGE COMPOSITE
The method to make a composite image of OCTS and GLI
as same as that of AVHRR proposed before is as follows;
1) The area of images with scan angle up to 30 degrees
are primarily utilized.
2) The period for composite is set to 9 days. The numbers
of images acquired during nine days are three for OCTs
and six for GLI. Cloud free composite image wl
sometimes not be able to be produced, especially for
OCTS because only three images are available during
nine days. When the period is set to one month, three
nine-day composite images of GLI are used to produce
one month composite and fifteen images of OCTS
indicated in chapter 4 (not nine day composite images)
are used for one month.
3) The scan angle and the solar zenith angle of any pixel
in the composite images of AVHRR, OCTS and GLI are
different from each other. When those composite
images are used integratedly for vegetation monitoring,
cross calibration have to be examined with the
considerations of these differences.
6. CONCLUSION
In this paper, the geometric properties of making the
composite image of AVHRR, OCTS and GLI for
vegetation monitoring are examined and the method of
image composite is proposed. The OCTS and GLI data
are not available. After the launch of ADEOS in this
summer, | will examine to check the method using
AVHRR and OCTS data.
If the method is adaptable, it willbe applied to GLI data in
the future.
REFERENCES
Hashimoto, T., Murai, S., 1992. Generation of cloud free
vegetation index map, Proc. of the 13th ACRS.
Kajiwala, K., Ryutaro, T., 1992, Aaalyses of problems on
the utilization of NOAA GVI data (in Japanese), Journal of
JSPRS, No.31, No.3, pp.16-24.
Singh, S. M., 1988, Simulation of solar zenith angle effect
on global vegetation index (GVI), Int. J. Remote Sensing,
Vol.9, No.2, pp.237-248.
Matsumoto, M., et al, 1991, A study on the dependance
of NOAA GVI on solar zenith angle and its correction (in
Japanese), Journal of JSPRS, Vol.30, No.3, pp.34-41.
278
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996