外來入侵物種銀合歡族群擴散模擬之研究

本研究以遙測為研究基礎，探討南台灣恆春半島，外來入侵植物銀合歡(Leucaena leucocephala (Lam.)de Wit)之分佈與擴散，研究方法係結合遙測、邏輯斯迴歸與細胞自動化(cellular automaton, CA)等，建構一物種擴散模型，評估銀合歡族群入侵情形。首先，利用影像分類法獲取空間分佈，分析銀合歡族群1988~1997年之動態過程，結果顯示，該年間銀合歡入侵面積約3,865 ha，占全區的6.5%，爾後據此空間分佈資訊推估邏輯斯迴歸機率，並建構CA模擬所需之細胞演化規則，研究結果發現，CA模擬與2007年銀合歡實際範圍作比較後，總體準確度可達95%，Kappa統計值為0.8，由此可知，本研究所發展的CA模型可準確地模擬銀合歡之族群擴散。總體而言，本研究成果可作為銀合歡防治策略擬定時之參考，另一方面，亦能作為探討其他外來入侵種之參考範本。

關鍵字

遙測；恆春半島；邏輯斯迴歸；細胞自動化；生物入侵

並列摘要

The objective of this study is to assess the distribution and expanding process of an alien-invasive plant: Leucaena leucocephala (LELE) based on remote sensing in the Hengchun Peninsula of Taiwan. This study developed a spatially explicit logistic regression imbedded cellular automaton (CA)-based species expansion model to simulate the invasion of LELE. The first, among the dynamics of LELE for 1988-1997, the invasion sites (from non-LELE to LELE) were 3865 ha, ccupying 6.5% of the study area. Second, 1988-1997 LELE spatial distributions were ingested into a logistic model to build transition rules for CA simulation. The results demonstrated that the enhanced CA model can precisely (an overall accuracy of 95%, with a kappa statistic of 0.8 by comparing with the 2007 LELE distribution) model the population expansion of LELE. The findings may provide insights into the control of LELE, which may be very important for other alien-invasive species research.