IRR 1-km
oject was
pril 1992
and cover
reference.
re used in
| and
| in this
inute grid
1 in Table
n existing
n Central
und truth
2.3 Classification
The feature extracted for land cover mapping in this study
is phenological information which is monthly ratio of Ts
over NDVI. Here, Ts is land surface temperature
calculated by split window algorithm (Price 1984) from
channel 4 and channel 5 of AVHRR. The Ts is affected
by rainfall and the maximum Ts corresponds no or less
rain condition. Therefore monthly Ts/NDVI is derived
from the maximum Ts in a month and the maximum
NDVI in a month.
Basic classification method is the same as the case of four
arc minute except the use of the ratio Ts/NDVI instead of
NDVI.
3. PREPROCESSING OF TEMPORAL
AVHRR DATA FOR LAND COVER
CHANGE DETECTION
Land cover change cannot be detected by a simple
subtraction of two AVHRR images at different times
because AVHRR data is affected by unrelated factors to
land cover; for example, cloud, noise, difference of solar
zenith angle, difference of scan angle, and change of
sensor sensitivity. Within these factors, change of solar
zenith angle and change of sensor sensitivity give
AVHRR data low frequency temporal variations, and
other factors do high frequency temporal variations.
Authors are pursuing methods to remove these variations.
Figure 2 shows the concept to remove high frequency
Tropical forest
——
NDVI
variation. The basic idea of this method is to make
monotonous increase( or decrease) of NDVI in the pre-
determined time range(for example, four 10-days units) by
changing valley(hill) value up(down) in temporal series
data(Park 1998).
To remove low frequency variation, ground truth data
from stable land cover have to be used as a reference.
4. FURTHER NECESSARY ACTIONS/
STUDIES FOR SUCCESSFUL
LAND COVER MONITORING
4.1 Accumulation of Ground Truth Data
Key action for better global/continental land cover
monitoring is ground truth collection. There are many
studies or projects to produce local land cover data sets as
a final product or a by-product. However most of them
are not open to other researchers. The Data and
Information System(DIS) sub-committee and the Land
Use and Cover Change(LUCC) sub-committee under
Japan national committee for International Geosphere and
Biosphere(IGBP) initiated a new action to solve this
. problem. This is Asian Wide Land Use and Cover Meta-
Database (AWLC meta-database). ^ An AWLC meta-
database is an internet-accessible database of meta-data of
land use, land cover, and other land surface variables in
Asia. The main reason of the development of an AWLC
meta-database is to remove barrier which avoid to use the
pre-acquired knowledge of land use and land cover of Asia.
—
Deciduous forest
NDVI
Jan Dec
ee Spun NIE)
Original Data
Corrected Data
Figure 2 Removal of high frequency variations in time series NDVI
International Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998
493
