(3) 8 monthly NDVI data from March to October, 1990 
for global land area 
The following three features derived from monthly NDVI 
were analyzed for each land cover class of ground truth 
data in order to decide rules of decision tree classification 
method. 
(1) Clustering results 
(2) Maximum NDVI: the maximum monthly NDVI 
value in twelve months 
(3) Minimum NDVI: the minimum monthly NDVI 
value in twelve months 
By the method describing above, 
(1) global 4-minute land cover data set and 
(2) 4-minute ground truth data set in Asia and Oceania 
were produced and are being distributed. 
2. LAND COVER DATA SET OF THE 
WHOLE ASIA(30 ARC SECOND GRID) 
Thirty arc second grid land cover data set of the whole 
Asia is the final product by the LCWG/AARS. The 
developed data set will be released by CD-ROM early 
1999. 
2.1 Used data and information 
- AVHRR data 
NDVI, channel 4, and channel 5 of NOAA AVHRR 1-km 
10-days composite produced by IGBP-DIS project was 
used in this study. The used data ranges from April 1992 
to March 1993. 
- Thematic maps 
Maps of ecoregions, vegetation, land use, and land cover 
were used as the source of ground truth and as a reference. 
Twenty-five different types of thematic maps are used in 
this study. 
2.2 Land Cover Classification System and 
Ground Truth Collection 
The land cover classification system applied in this 
project was the same one as the four arc minute grid 
global land cover data set, that is the one shown in Table 
1. Ground truth data were collected mainly from existing 
thematic maps and partly from field survey in Central 
Asia. 
Figure 1 shows the regions of the collected ground truth 
data. 
  
Figure 1 Ground truth regions for 30 arc-second grid land cover data set of Asia 
492 International Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998 
t2 
Th 
is | 
ove 
cha 
by 
rair 
froi 
ND 
Bas 
ND