Abstract
Various studies have shown that alpine ecosystems in the Tibetan Plateau are very sensitive to climate change and they are facing degradation. However, such studies are regional and either use coarse resolution satellite imagery as a monitoring tool or rely upon secondary data. Traditional digital image classification methods normally can't meet specific needs of extracting land cover categories from satellite images. This paper aims to map land use/land cover (LULC) particularly the predominant rangeland vegetation in the Lhasa River Basin by applying an accurate expert system classification (ESC). Landsat ETM+ satellite imageries taken in 2000 and other digital datasets were used. To show the advantages of ESC, the results were compared with the results of maximum likelihood classification (MLC) system. The ESC resulted in a classified map with a significant higher accuracy (84.31%) than maximum likelihood classification (74.31%). This demonstrated the successful application of ESC in the Lhasa River basin and, most importantly, three rangeland vegetation categories were extracted with acceptable accuracy. Furthermore, the results showed an effectively used of the two classifications in extracting LULC categories in zone above 4200m than in zone below 4200m, due to the relatively homogeneous landscapes. It is suggested that data of improved spatial and spectral resolution and more correct, complete and relevant expert knowledge could potentially improve the accuracy of ECS. Compared with the MLC, the ESC requires more time to extract and tune the knowledge in order to create rule bases. This drawback can be reduced by developing an automatic and reproducible system to extract knowledge from the data layers to a specific landscape in the future work.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have