Geoscience Reference
In-Depth Information
vegetation cover on this area, and much rocky land exposed to the air, so the RD here is
intense; if the NDVI value is between 0.2 and 0.4, the extent of RD is moderate; if NDVI is
between 0.4 and 0.6, the extent of RD is mild; when the NDVI is above 0.6, it means these
areas are good protected.
3.2.2 Land cover classification
After the detection of NDVI change, it is still needed to know the specific changes of land
cover types in the study area. Generally, there are two methods to distinguish and interpret
the remote sensing image: supervised classification and unsupervised classification.
Supervised - image analyst "supervises" the selection of spectral classes that represent
patterns or land cover features that the analyst can recognize. Unsupervised - statistical
"clustering" algorithms used to select spectral classes inherent to the data, more computer-
automated.
Supervised classification was used in this study. It is much more accurate for mapping
classes, but depends heavily on the cognition and skills of the image specialist. The strategy
is simple: the specialist must recognize conventional classes (real and familiar) or
meaningful (but somewhat artificial) classes in a scene from prior knowledge, such as
personal experience with what is present in the scene, or more generally, the region it is
located in, by experience with thematic maps, or by on-site visits. This familiarity allows the
individual(s) making the classification to choose and set up discrete classes (thus
supervising the selection) and then, assign them category names.
Training ground and training sample selection is very important in supervised
classification, the classification result will have a big different in supervised classification if
the training sample is different. So it should be careful to select the training ground and
choose the represented training sample correctly. These are the key points to produce a
good classification result. In this study, supporting data and local land cover maps were
used to help distinguish the land cover types. Due to the coarse resolution of MODIS data, it
is not credible to classify many detailed land cover types. Thus based on information from
supporting data and local land cover maps, six types of land cover were to be classified:
water, wood, grassland, residence, farmland and unused land. After the classification, post-
processing of image classification should be performed to get more reliable land cover
maps, whilst accuracy assessment would be done.
4. Results
4.1 NDVI distribution and change of RD from 2000 to 2010
NDVI values were calculated by equation (1) and their distribution maps of 2000 and 2010
were obtained using ENVI software. Difference can be seen in the NDVI maps of the two
years. In north part of the study area, NDVI values decreased in November 2008 compared
with that in November 2000. However, NDVI maps can not show these changes distinctly.
To distinguish the extent of the RD from 2000 to 2010, a decision tree upon Table 1 was
produced as shown in Fig. 3. The Decision Tree classifier performs multistage classifications
by using a series of binary decisions to place pixels into classes. Each decision divides the
pixels in a set of images into two classes based on an expression. Based on these rules, the
extent distribution of RD from 2000 to 2010 can be mapped clearly (see Fig. 4).
Search WWH ::
Custom Search