lineation
sensing,
ing work
ipdating,
H
! Object-
y sensed
99].
data i
Partial-
| satellite
mmetrie,
ummetry,
'Osjektet,
|littscene
Norway
ata with
Forestry
15795;
Opptak -
1e 1996
] image
enlarged
Chapter
4. Using
ta. 7th
ceedings
st - An
er report
ging av
satellitt.
sion 3.3.
LAND DEGRADATION RISK MAPPING USING NOAA NDVI DATA
Shiro OCHI(*) and Shunji MURAI(**)
* Assistant Prof, Dept. of Forestry, Utsunomiya Univ.
350 Mine, Utsunomiya 321 Japan
Tel:+81-28-649-5537/Fax:+81-28-649-5545
E-mail:ochi@cc.utsunomiya-u.ac.jp
** Professor, Institute ofIndustrial Science, Univ. of Tokyo
7-22, Roppongi, Manatoku, Tokyo 106 Japan
Tel:+81-3-3402-6231(ext.2560)/Fax:+81-3-3479-2762
E-mail:murai@shunji.iis.u-tokyo.ac.jp
Commission VII, Working Group 6
KEY WORDS: Land Degradation, NDVI, GVI, Land Cover, Vegetation, Global Data set
ABSTRACT :
Land degradation risk map was tried to be generated using NOAA 8k NDVI 10 days composite data set of 10 years from
1983 to 1992, and Global data set such as climate data, elevation data, and thematic maps. Before proceeding the time series
analysis, the data quality was examined using the technique of moving average for 12 months(36 scenes). The quality ofthe
NOAA 8km NDVI data was estimated as good enough to use fortime series analysis with little data modification.
The fluctuation pattern of NDVI forseveral land cover type were defined fromexisting eco-regions map, and the map was
resampled to low resolution to 8km resolution. In combination with the eco-region map with digital elevation data and
precipitation data, soil erodability is checked. And eco-region classification map and vegetation degradation map shows the
risk against the desertification, land degradation. The accuracy verification is not undertaken, however, the methodology
can be applied to the LAC or GAC data set.
1. INTRODUCTION
Due to the increase of population pressure, climate change,
etc, Land Degradation can be seen in many places on the
earth. The importance forthe assessment ofland degradation
and desertification by means of Remote Sensing technology
is pointed out in Agenda-21, however, the methodology has
not been established yet. In order to detect the land
degradation using satellite imageries, an analysis is
necessary with data which has high spatial resolution, and
long observing period as possible. Data analysis, using the
accumulated NOAA LAC(1km resolution) and GAC(4km
resolution), are expected for global land cover change
analysis. But we have to wait more some years until the
compiled LAC and GAC data sets are distributed to us, and
an advanced methodology and facilities should be developed
for the management of such huge data scts volume.
At this moment, we can obtain NOAA complied data with
8km spatial resolution which includes daily data set as well
as 10 days composite data set for more than 10 years
observing period from1981. The methodology by using
rather long time series data set must be developed for global
land cover change analysis. And it can be followed by a
aludy using more high resolution data set such as GAC and
LAC. As the basic analysis forlong time series data set of
NOAA data, this study aims following goal:
-to develop a methodology, which can be applied to LAC or
GAC data set in the future, forthe time series analysis on land
degradation and vegetation degradation in continental scale
as well as global scale, and
-10 estimate the land degradation risk by monitoring long
term land cover change using technology of GIS.
2. DATA
In this paper, 8km NDVI data of 10 days composite for 10
years from1983 to 1992 which are distributed by NOAA are
used to monitor the global vegetation degradation, and some
GIS data such as average monthly temperature and
precipitation data fromLeemans and CramerIIASA Climate
data(30 minutes pixel resolution), Baily Ecoregions of
Continents(10 minutes pixel resolution) and elevation data
fromGLOBE data set.
Before using 8km NDVI data for time series analysis, the data
quality must be examined because the intensity ofthe NDVI
is not sameon the same vegetation conditions due to the
differences ofsatellites(sensors), positions ofsensor and so
on. 10 points fromthe land of Japan were selected in order to
examine the quality. Those are:
(a) Deciduous Conifer Forest Dominant areca
(b) Deciduous Broad Leaf Forest area
(c) Rice Paddy Dominant area
(d) Urban Area
(e) Monsoon Evergreen Forest Dominant area
Fluctuation pattern of Figure-1 shows the NDVI fluctuation
pattern for deciduous conifer forest in Hokkaido, Japan. Each
curve forselected 5 points clearly demonstrate the
characteristics ofeach land cover type, however there are
some scattering points due to the climate variety and data
error. To avoid such scasonal and annual data variety, moving
average method are used to evaluate the tendency ofthe
pattern as well as to examine the data quality. The moving
average were calculated using one year data(36 scenes).
Figure-2 shows the moving average using 12 months for
547
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
