Abstract:Due to that the vegetation has been influenced by human activity for a long period, such as logging, grazing and trampling, giant pandas' habitat has been fragmented gradually, which has threatened the survival of the giant panda. Therefore, it would be an essential issue to inspect the spatial pattern of the giant panda's habitat and analyze the factors which drive these habitat changes. The aim of this study is to detect the giant panda's habitat changes by using decision tree classification approach. The study area is Mabian-Dafengding Nature Reserve (MDNR) in Liangshan Mountains of Sichuan Province. Our data include field survey samples, GIS data and Landsat TM images. Decision tree classification approach was applied to detect the giant pandas' habitat change. Compared with traditional classification method, decision tree is more capable of using both spectrum and non-spectrum information as to improve classification accuracy. According to the local climate and vegetation type, 9 habitat types were defined. Our results firstly show that the decision tree method is an effective approach on detecting the habitat change. Our results also show that the area of the mixed evergreen and deciduous broadleaf forest increases the most, while that of the mixed deciduous broadleaf and conifer forest decrease the most. In spatial, most changes were detected along the boundary and the river of MDNR, which reflects substantially that human being has strong impacts on the surrounding habitat type. All these changes should be paid attention to by the local managers for a better conservation.

DETECT HABITAT CHANGES IN MABIAN GIANT PANDA NATURE RESERVE, CHINA