A NEW GLOBAL LAND COVER MAP, GLCNMO
R. Tateishi 2 3,
*, Bayaer 3 , M. A. Ghar 3 , H. Al-Bilbisi 3 , J. Tsendayush 3 , A. Shalaby 3 , Alimujiang Kasimu 3 , N. T. Hoan 3 , T. Kobayashi 3 ,
B. Alsaaideh 3 , Md. M. Rahman 3 , E. Tsevengee 3 , Y. Yamada b , and S. Kajikawa b
3 Chiba Univeristy, Center for Environmental Remote Sensing, 1-33 Yayoi-cho, Inage-ku Chiba, 263-8522 Japan -
tateishi@faculty.chiba-u.jp
b Geogaphical Survey Institute of Japan, 1 Kitazato Tsukuba 305-0811 Japan - kajikawa@gsi.go.jp
KEY WORDS: Land Cover, Global, Mapping, MODIS
ABSTRACT:
The Global Mapping Project organized by International Steering Committee for Global Mapping (ISCGM) has produced a new 1km
global land cover data called Global Land Cover by National Mapping Organizations (GLCNMO) as one of eight layers of Global
Map data. It has twenty land cover classes defined by Land Cover Classification System (LCCS). Fourteen classes out of twenty
were derived by supervised classification, and the rest of six classes such as urban, tree open, mangrove, wetland, snow/ice, and
inland water were classified independently using MODIS and/or other source data. This land cover mapping has the following
features, (a) Eight periods of 16-day composite 7-band 1-km MODIS data of 2003 year were used as primary source data, (b)
Training data for supervised classification were collected using Landsat image, MODIS NDVI seasonal change pattern, and
comments from National Mapping Organizations with the reference of Google Earth, Virtual Earth and existing regional maps, (c)
Urban areas were extracted using Gridded Population density, DMSP data, and MODIS NDVI. (d) Tree open areas were mapped
from percent tree cover data derived from the same MODIS data, (e) Inland water and ocean were mapped by clustering of MODIS
data 1 .
1. INTRODUCTION
Land cover is one of key environmental parameters. After the
availability of global satellite data such as NOAA/AVHRR,
global land cover map was produced using satellite data. The
first one was IGBP-DISCover(GLCC website) using AVHRR
data of 1992. University of Maryland and Boston University
have produced global land cover data using MODIS data
independently (UMDLCC website)(BULC website), and Joint
Research Centre (JRC) of European Commission produced
GLC2000 using SPOT/VEGETATION data of 2000(GLC2000
website). Though several global land cover data are available,
their accuracy is not very high. Better global land cover data is
necessary for reliable scientific use of the data.
Global Mapping project (ISCGM website) is an international
project to develop eight basic global datasets by the cooperation
of National Mapping Organizations. One of the products is
global land cover data which is called Global Land Cover by
National Mapping Organizations (GLCNMO). GLCNMO was
produced using mainly MODIS data observed in 2003. This
paper described how GLCNMO was produced.
2. POLICY OF LAND COVER MAPPING
Basic policy of the production of GLCNMO is as follows.
(1) use of global 1km satellite data
(2) cooperation of National Mapping Organizations for
intermediate check of the classification trial
(3) publication of training data
There have been several global land cover mapping projects so
far. However no project releases or publishes training data or
reference data they used. Accumulation and common use of
training data (or ground truth data) by different projects
increases reliable knowledge of global land cover, and it
follows better land cover mapping in the future.
(4) independent mapping for the classes which are difficult to
be classified by automatic digital classification method.
In this mapping project, six out of twenty land cover classes
were found difficult to be classified by supervised method.
Therefore these six land cover classes were classified
individually by different methods.
3. DATA USED
3.1 MODIS data
The main data used in this mapping project is MODIS data
observed in 2003. The source MODIS data is MODIS/TERRA
Nadir BRDF- Adjusted Reflectance 16-DAY L3 Global 1 KM
SIN Grid Product (MOD43B4 NBAR). This is 16-day
composite, 7-band, 1-km, 10-degree tile data. The source data
were processed to remove cloud contamination, and reprojected
into latitude/longitude coordinate system, and mosaicked to
make five continental data(Eurasia, North America, South
America, Africa, and Oceania) and two island regions in the
Ocean.
3.2 Other data
3.2.1 Satellite data: Other satellite data used are Landsat
+ETM data(GLCF website) and DMSP/OLS data(DMSP/OLS
website). Landsat data were used to collect training data, and
DMSP/OLS data were used for urban mapping.
3.2.2 Global land cover data: Existing global land cover data
such as GLC2000(GLC2000 website), land cover by Boston
Corresponding author.