3.1 Data source
Landsat TM and ETM satellite images from China Remote
Sensing Satellite Ground Station (RSGS), Chinese Academy of
Sciences (CAS) were acquired. These digital images covering
the entire studied area included 18 cloud-free or low cloudy
cover scenes of images from June to September in the years of
1986, 2000 and 2006 respectively. Land use types are easy to
identify in this period when plant grows abundantly in
Northeast China.
Land cover types for the Yanbian prefecture were summarized
as follows: (1) cropland, (2) forestland, (3) grassland, (4) water
area, (5) built-up land and (6) unused land. Forestland is
classified as closed forest, shrub, open forest, and other forest.
Closed forest is natural or man-made forest with canopy cover
of more than 30%.Shrub is land covered by tree less than 2
meters high and with the canopy cover greater than 40%. Open
forest refers to land covered by trees with canopy cover
between 10% and 30%.Other forest is land covered by orchid
and/or non- grownup forest.
Landsat images were georectified to 1:100 000 topographic
maps using ground control points (GCPs) collected by global
positioning system (GPS). The images acquired in 1986 were
used as reference to correct the other scenes in 2000 and
2006.The maximum likelihood classifier was utilized to carry
out classification of images. To revise the misclassification
errors such as boundary and spectral confusion in preliminary
classification, visual interpretation and ground survey were
performed (Liu et al. 2005b). Reference data for 1986 Landsat
images were acquired from the topographic maps (scale
1:100000) drawn in the same periods. The accuracy of
classification for land cover maps extracted from Landsat
images was assessed in 2001 and 2008 based on a reference
data of randomly selected patches. The accuracy assessment
was based on an evaluation of 400 patches. The overall
accuracies of the four land cover maps interpreted from remote
sensing images for 1986, 2000 and 2006 were 91%, 92% and
91%, respectively.
3.2 Dynamic Degree Model
The dynamic degree model was applied to reveal the change
rate of forest during different periods (Wang and Bao 1999).
The calculation is given by
U, zu,
K= <= <100% (0)
where K is the dynamic degree of forest measured as the
relative annual change rate, U, and U; are the area of forest in
the start and final year, respectively, and T is the interval of the
calculating period (in years). A positive dynamic degree value
generally indicates an increasing trend of forest coverage for a
specific period, vice versa.
3.3 Fragmentation Analysis
The spatial composition and configuration of forest landscape
pattern were described with following landscape indices,
namely the number of patches (NP), mean patch size (MPS),
patch density (PD) and area-weighted mean patch fractal
dimension (AWMPFD). NP is a simple index of the degree of
subdivision of a land cover type (McGarigal and Marks 1995),
and MPS is a commonly used metric revealing the
fragmentation in the spatial pattern analysis (Baldi et al. 2006).
PD describes the density of patches for each land use,
representing an aspect of fragmentation, which is the dissection
of patches. AWMPFD reflects shape complexity weighted by
the area of patches (Wang et al.2009, Huang et al.2012).
3.4 Temporal Trajectory Establishment
The temporal trajectory method is not only on what has
changed between dates, but also on the progress of the change
over the period (Mertens and Lambin 2000; Petit et al. 2001;
Liu and Zhou, 2004; Zhou et al. 2004; Zhou et 31.2008). In this
study, the classified land cover maps were integrated using
raster format within ArcGIS, thereafter the pixel-based change
trajectories were constructed. The result is a temporal trajectory
map, in which every pixel has a code corresponding to its
trajectory. The possible change trajectories from 1986 to 2006
in the study area are shown in Figure.l. For example, the
highlighted trajectory is specified as F-F-C, meaning that the
land was found to be forestland in 1986, forest in 2000 and
cultivated as cropland in 2006.
1986 2000 2006
C-Cropland F-Forestland G-Grassland
W-Water area B-Built-up land U-Unused land
Figure 1. All possible trajectories of land use change
4. RESULTS AND DISCUSSION
4.1 Area Change and Change Rate of Forest
The temporal changes in forestland during 1986-2006 were
shown in Table 1. The process of spatial change of the forest
land in the studied area was shown in Figure. 2. It was found
that area of forest shrank continuously in the past two decades.
In 1986, the forest covered about 82.2% of the total land area.
Owing to the degradation of closed forest and open forest,
forest area decreased significantly during 1986-2006.The net
loss of forestland was 1414.61 km?. Closed forest decreased by
1249.20 km? during 2000-2006, accounting for 92.1 per cent of
the total reduced forest area in this period. In this period, loss of
open forest and shrub were more remarkable than that from
1986 to 2000.In 2006, forest declined to 78.9 % of the total area.
The quality of forest has severely worsened in the past 20 years,
and considerable decrease in forest was mainly attributed to
reduction of closed forest.
Peri
198
-20
20(
-20
19¢
-20
Fi;