ANALYSIS OF THE LAND COVER AND ITS CHANGE
OVER YELLOW RIVER BASIN USING SATELLITE DATA
M. Matsuoka *, T. Hayasaka *, Y. Fukushima *, Y. Honda b
? Research Institute for Humanity and Nature, 335 Takashima-cho, Kyoto, Japan
- (matsuoka, hayasaka, yoshi)@chikyu.ac.jp
b Center for Environmental Remote Sensing, Chiba University, 1-33 Yayoi-cho, Chiba, Japan
- yhonda@cr.chiba-u.ac.jp
KEY WORDS: Land cover, Classification, Yellow River basin, Terra/MODIS
ABSTRACT:
Land cover classification over Yellow River basin, China, by means of simple decision tree classification method using multi-
temporal metrics derived from MODIS data is shown in this paper. Since the land cover is one of the input parameters for the
hydrological model, the critical feature of classification method is to control the classification result using threshold values.
The tentative result shows good correlations with existing digital land use map, though the small overestimation or
underestimation are recognized in several categories and the improvement is necessary in discriminating the single cropped
agricultural field from natural grassland. The adjustments of the threshold values established in the decision tree would be
done by the aid of higher resolution satellite data.
1. INTRODUCTION is low as once per 16 days (Sheffner, 1999) and the
scheduled acquisition of the data is restricted by the cloud
1.1 Background cover. TM series are mainly utilized for the validation and
: : tuning of the land cover classification and change detection,
The development of water resource management model in the and the detailed analysis intended for the drastically
Yellow River basin, China, is important because the stream changed area. Land cover classification using MODIS data
ofthe river had not reached to the Bohai Sea for many days (number 1 in the figure) is the subject of this paper.
in a year since 1970s as a result of climate change and 3
human activities. The land cover and its change reflect the
interactions between human activities, such as agriculture
and urbanization, and nature as dynamics of water and
energy, therefore, it is one of the essential parameters for the
development of the model. The purpose of this study is to
analyze the land cover and its change for 20 years since
1980s over Yellow River basin using satellite data.
'02 '82
. Land cover classification using MODIS
. Comparison of land cover with AVHRR
. Change detection using AVHRR
. Validation and detailed analysis by TM
1.2 Outline of this study
Three kinds of sensor, Terra/MODIS, NOAA/AVHRR, and
Landsat/TM or ETM+, are utilized in this study according to
its spatial and temporal resolutions. The outline of this
study, comprised of four frames, is shown in Figure 1.
MODIS onboard Terra, which is new sensor operated since
1999, has two 250 meters bands in visible and near infrared.
MODIS are viewing the entire Earth's surface every 1 to 2
days (Conboy). The land cover classification map is
generated from the time series MODIS data in 2002 by means Figure 1. Outline of this study
ofthe supervised decision tree method. AVHRR is operated
over twenty years from the end of 1970s (Kidwell, 1998),
and precious and long term seamless data are archived. The 2. METHODOLOGY
spectral and temporal characteristics of AVHRR data in 2002
are compared by each land cover categories of classification 2.1 Overview
map, and thereafter the changes ofthe land cover since 1982
are detected using time series of AVHRR data. TM and ETM+ Land cover classification on the scale from regional to
onboard Landsat series have much higher spatial resolution global has been implemented with the use of AVHRR data by
than AVHRR and MODIS, though the observation frequency means of a variety of classification method as clustering
836
Intei
———
(Par
Cen
al,
199
net\
sim
the
199
inpi
feat
clas
22
MO
of
atm
by «
pro
(abt
250
fron
648
esti
WOU
atm
Dist
of
to !
mea
Dist
cov
lati!
lon;
foll
seri
Anr
5)
Sec
Apr
the
deri
min
ma
eS
sele
con
clas
cov
per!
