Geography Reference
In-Depth Information
Land Use Data from CAS
Land cver data
geocoded with area
percentage grid method
geocoded with the greatest-area method
WESTDC
UMD
GLC
IGBPDIS
Cropland/Natural
vegetation mosaics
Other Classification
Forest/woods
Overlay analysis
The agreement grids of classification
Mapping
Land cover data using IGBP scheme
C4.5 classifier
Export the spatial data to ASCII file
Fig. 3.10 The work flow of mapping based on multi-source spatial data mining approach
Sect. 3.3.2.1
Mapping the land use types to determinate the land cover classification
2 classes of artificial lands, 1 class of water among the IGBP land cover
classification.
It only needs to transform from many to one or one to one (Table
3.8
). For
example, the Paddy land and Dry land in the land use map of CAS are explicit and
correspond to the cropland class definition in the IGBP, so it only needs to
aggregate them into cropland with the binary grid. It is more feasible to judge the
land cover classification of Cropland/Natural Vegetation Mosaic with the area
percentage grid data of Paddy lands, Dry lands, Forest, Shrub lands, among which
no single type comprises more than 60 % of the landscape. The land cover of
Cropland/Natural Vegetation Mosaic is mainly located in the Inner Mongolia,
Liaoning, Hebei, Shaanxi, Shanxi, provinces, with a total area of about
730,00 km
2
in 2000 (Fig.
3.11
). The 8 classes of land cover types including the
IGBP10-IGBP17 were transformed, which account for nearly half of the total land
area. In addition, there is a little savanna in China, which is convenient to judge
based on the temperature and land use type. However, the 8 classes of vegetation
(forest, shrubs, and herbaceous vegetation) and the leaf attributes (evergreen and
deciduous), and the leaf types (broadleaved
and coniferous)
are difficult to
determine because we lack of vegetation information.
3.3.2.1 Selecting the Spatial Agreement Samples of Vegetation for Data
Mining
The closed forest and other forest classes are arbor forest classes in land use
classifications of CAS. They do not concretely specify the forest type information.
However, this provides an accurate boundary for the forest; therefore, we need an
inference rule to transform between forest in land use classification system and
IGBP
forest
categories:
evergreen
needle-leaf forest,
evergreen
Broadleaved
