Maxent Modeling for Extrapolating the Spatial Distribution Pattern of Forest Tree Species —Schima superba

详细信息    查看全文 | 下载全文 | 推荐本文 |

• 作者：Hou-Chang ChenNan-Jang LoWei-I ChangKai-Yi Huang
• 会议时间：2011-08-01
• 关键词：Predictive Model ; Spatial Distribution ; Maximum Entropy (Maxent) ; Discriminant Analysis (DA) ; Spatial Extrapolation ; Remote Sensing ; Geographic Information System (GIS)
• 作者单位：Hou-Chang Chen(Dept. Of Forestry)Nan-Jang Lo(Experimental Forest Management Office)Wei-I Chang(Forest Planning Section,Forest Bureau, Council of Agriculture, Taipei 100, Taiwan.)Kai-Yi Huang(Dept. Of Forestry, Chung-Hsing University, 250 Kuo-Kuang Road, Taichung 402, Taiwan)
• 母体文献：第五届海峡两岸遥感遥测会议论文集
• 会议名称：第五届海峡两岸遥感遥测会议
• 会议地点：哈尔滨
• 主办单位：中国地理学会
• 语种：chi
• 分类号：S75;X17

摘要

Explanation and analysis of species of spatial distribution pattern is an important issue that ecologists need to tackle. Predictive distribution model is the core of ecological research in late 20th century because of the development of statistical techniques, remote sensing, and GIS tools, which not only enhance the ability of predictive habitat distribution models, but also applied to biodiversity conservation and management. However, physiological behavior of species and ecological interaction are so complicated that it may often bring about highly uncertain results in species distribution. Therefore, the sample points of Chinese guger-tree (Schima superba) were obtained by GPS, and GIS technique was used to overlay five environmental factors (including topographic factors and vegetation index derived from SPOT-5 satellite images). The study applied maximum entropy (Maxent) and discriminant analysis (DA) models to predict the suitable habitat for the species in Huisun study area and determined the optimal one in terms of accuracy and efficiency. Three sampling designs with different combinations of samples taken from Tong-Feng and Guan-Dau watersheds were used for model development and validation. The results showed that the accuracy of Maxent was far better than that of DA, but both with nearly the same efficiency in model development and validation. More importantly, Maxent greatly reduced the area of field survey more than did DA, thus saving much more human and financial resources. However, the models based upon only topographic variables could not be applied at larger spatial scale due to the characteristics of topographic variables. Future studies will attempt to use variables involving spectral information associated with species extracted from high spatial, spectral resolution remotely sensed data, especially hyperspectral image data, for building a model so that it can be applied over a larger area.