Full text: Proceedings of the Symposium on Global and Environmental Monitoring (Pt. 1)

Robert DE WULF, Roland GOOSSENS and France GERARD, 
Laboratory of Remote Sensing and Forest Management, Gent State University 
Coupure 653, B-9000 Gent, Belgium 
Boudewijn DE ROOVER 
Satellite Reimte Sensing Research Programme (TELSAT-II/04) 
Coupure 653, B-9000 Gent, Belgium 
WWF International, World Conservation Centre 
CH-1196 Gland, Switzerland 
Laboratory of Remote Sensing in Forestry, Southwest Forestry College 
Hd: Springs, Kunming, P.R. of China 
The study area is located in the Xishuangbanna Dai autonomous prefecture 
(Yunnan province) in the southwest of the People's Republic of China. The 
Mengyang Nature Reserve forms part of the Xishuangbanna Nature Reserve 
totalling some 200000 ha. 
The variety in relief and soil types, combined with its location in the 
subtropical region, yields an exceptional richness of vegetation types, 
offering a challenge for a digital image classification exercise. In addition, 
human influence profoundly changed the landscape in the past 30 years, 
resulting in shifting cultivation and various forms of permanent agriculture 
and cash crop plantations. 
A Landsat 5 Thematic May.per image forms the basic material for this forest 
cover classification case study. Extensive field data collection conducted in 
the summer of 1989 served as reference for supervised classification 
procedures. The classification exercise involves following experimental 
variables: class level, classifier type (layered and single step), structure 
of layered classifier and channel selection. The main objective of this study 
was to find out if relevant forest classes could be reliably classified. The 
distinction between primary forest and secondary forest was examined in detail 
in order to evaluate the usefulness of satellite remote sensing for nature 
conservation management In an endangered tropical environment. 
It was concluded that, combining several classification methods, an accuracy 
of 80% for Level III vegetation classes is possible. Secondary forest can be 
separated from primary forest (Level II classes) with an accuracy of about 85 
%, and Level I classification is possible with an accuracy of 95 %. The best 
results are obtained using the mid-infrared and the thermal infrared TM 
channels in conjunction with the near-infrared and the red bands. 
Key Words : classification, Landsat 
nature reserve, China 
Mengyang, as well as the jther nature 
reserves of Xishuangbanna, features an 
exceptional floral and faunal richness, 
and is currently the subject of joint 
efforts by the Chinese Ministry of 
Forestry and the World Wide Fund for 
Nature (WWF) to set up a conservation 
management plan. 
Thematic mapping of the natural and 
man-made vegetation types is a primary 
requirement for management planning. 
Spaceborne remote sensing represents a 
convenient and cost-effective technique 
to map large areas which a> e difficult 
to survey by traditional ground 
The main questions to be answered by 
Thematic Mapper, forest degradation, 
the exercise described in this paper 
1. How well can natural vegetation be 
distinguished from agricultural land 
use classes. 
2. In view of a reliable assessment of 
the forest degradation in 
Xishuangbanna, is it possible to 
distinguish between primary forest 
and secondary forest? 
3. Which classification strategy yields 
the best results. In particular, 
what is the contribution of the 
Thematic Mapper channels 5, 6 and 7 
to the classification accuracy, in 
addition to channels 2, 3 and 4.

Note to user

Dear user,

In response to current developments in the web technology used by the Goobi viewer, the software no longer supports your browser.

Please use one of the following browsers to display this page correctly.

Thank you.