多重模式評估方法模擬虎杖在台灣之生態棲位

Abstract

生態棲位模擬 (Ecological Niche Modeling) 結合地理資訊系統 (Geographic Information Systems) 為近幾年來生態學、生物地理學和保育生物學所廣泛使用之分析方法。動植物的空間分布型態為生物和所在棲地生態環境因子間綜合交互作用的結果，由於物種生理習性與生態系統交互作用的複雜問題，使得生態棲位分佈模擬的結果往往具有高度的不確定性，在使用不同模式分析方法時，常會出現不一致的模擬結果。因此，本研究提出以多重模式綜合評估方法 (Multi-Modeling Assessment)，使用多個模式綜合推估出物種之潛在空間分佈。近二十年來，虎杖 (Polygonum Cuspidatum Sieb. & Zucc.) 經人為傳播而成為歐美各國嚴重的外來入侵植物，為能了解其空間分佈特性，本研究分析台灣本島原生虎仗之棲地特性，以 GARP (Genetic Algorithm for Ruleset Prediction)、MAXENT (Maximum Entropy Method) 以及邏輯迴歸 (Logistic Regression) 等三種最為廣泛使用之生態棲位模擬分析方法，結合 20 個生態環境因子，預測全台灣虎杖之潛在棲地分布並歸納其原生棲地之環境特徵。研究使用現有虎杖調查樣點共 75 筆，以及現地野外調查 71 筆，共 146 筆觀測資料進行分析，比較三種模式建立之環境因子變數並評估預測準確度，以多重模式評估方式綜合預測出虎杖潛在棲地空間分布。

Ecological niche modeling combined with a geographic information system can model and display the spatial distributions of species efficiently. Species’ spatial distributions are the result of dynamic interactions between species and environmental factors. However, the complexity and uncertainty due to the interplay of ecosystem and species can cause difficulties when an individual model is used for the assessment purpose. Therefore, the multi-modeling assessment approach is proposed in this study. We combined various modeling techniques and records of known species occurrences with digital layers of environmental variables to predict species’ probability distributions over broad areas. This study aims to explore the native habitat characteristics of Polygonum Cuspidatum Sieb. & Zucc. (Japanese knotweed) in Taiwan, which has seriously invaded North America and Europe. This study applies three widely-used ecological niche models, GARP (Genetic Algorithm for Ruleset Prediction), MAXENT (Maximum Entropy Method) and logistic regression, with 20 environmental variables to predict the potential distribution of Japanese knotweed in Taiwan. Further, the results are compared between the models, and an optimum combination of habitat characteristics of the species is derived from intersecting the models.