Geography Reference
In-Depth Information
greatest area in the cell is used). The other was geocoded with area percentage grid
method, in which each cell can be divided into 25 layers to record the area of each
type (Liu et al.
2003
). Besides, the vegetation map can provide the reference
information of vegetation since the change of forest categories is slight in the short
term. The vegetation map of China reflects detailed information on the distribution
of vegetation and includes horizontal and vertical zones of 11 vegetation groups,
54 vegetation types, 135 biome units and 796 sub-biome units (DeFries and
Townshend
1994
).
The mapping of land cover data in 2000 based on the data mining is a benchmark
of the long term land cover dataset. It is necessary to collect the ancillary data due to
the absence of other data series. Data of physical geography include information on
terrain slope, information on vegetation property variability and so on. Information
on the terrain slope and the plain area proportion are derived from DEM data
covering the entire China at the scale of 1:250,000. These data are provided by the
Data Center for Resources and Environmental Sciences, CAS. The meteorological
data, including the annual temperature and annual precipitation, are acquired from
China Meteorological Bureau. The NDVI dataset come from the Pathfinder dataset
of Earth Resources Observation System (EROS), it is extracted from the NOAA/
AVHRR-NDVI images. The spatial resolution of images is 1 km 9 1 km, and their
temporal resolution is 15 days. In order to guarantee the data quality, all the data
have all been preprocessed with the internationally accepted reliable approach
(Chen et al.
2004
). Besides, in order to eliminate the noise caused by the cloud
pollution and the atmospheric influence, we also smooth the time-series NDVI data
with the Savitzky-Golay smoothing filtering method (Ahlqvist
2008
). The NPP data
during 1985-1999 come from the remote sensing data of NOAA/AVHRR, and that
during 2000-2010 come from the NPP product of MODIS.
3.3.2 Methodology
The working procedure of the classification is as follows. First, based on the
definition of mosaics type, we produce the Cropland/Natural vegetation mosaics
data by using the grid area percentage dataset in CAS land use system. Then other
types of land use except for forest and woods are achieved by utilizing grid
maximum area mapping with two sub-classification definition between the CAS
and IGBP. Thereafter, we check out and determined the grids whose type were
consistent with the forest and woods among the WESTDC, UMD, GLC and
IGBPDIS land cover data, at the same time we identify the boundary of the forest
and woods, which are consistent with CAS land use, generate them into the sample
data. Finally, we realize the conversion of forest types of IGBP scheme with the
C4.5 classifier (Fig.
3.10
).
The land use types are first transformed into the land cover types. It is easy to
transforming some land use types, e.g., 3 classes of developed and mosaic lands,
