恆春半島銀合歡擴張之動態模型

Abstract

外來物種的入侵，是當前棘手的生態議題之一，為了減緩外來種所造成的衝擊，不論經濟或生態層面，往往都得付出高額的成本。銀合歡(Leucaena leucocephala)，為一種原產於中南美洲的速生小喬木或灌木。在近幾十年間，恆春半島因為其特殊的地形與氣候特性，深受銀合歡擴張的危害。 預測外來物種分布，使管理者能聚焦在目標物種線機率較高的區域，有利於及早偵測，以達到管理上的效率。近年來，生態棲位模式已逐漸被運用在物種分布的推估上，然而因為生態棲位模式只能以環境因子推估物種與環境達到某種平衡時的分布情形，以至於單獨使用生態棲位模式往往不足以預測快速擴張的外來物種。本研究結合了生態棲位模式與行為者基礎模式，使得空間中的單元能夠被獨立操控，進而動態地模擬銀合歡在恆春半島擴張的情形。 本研究模擬銀合歡自1988年擴張到2007年的狀況，並且以2007年的分布資料進行評估，發現最佳的參數組合具有良好的表現，其AUC達0.96，且Mantel’s r達0.79。不過，模擬2007年到2013年間新被入侵的區域時，發現模式的預測的範圍過於保守，因此建議在管理上，仍可配合生態棲位模式所推估的物種分布進行決策。

The impact of invasive alien species is one of the thorniest environmental calamities. Considerable costs, both economic and ecological, are spent for the attempt to manage and mitigate the impacts. Leucaena leucocephala, a fast growing tree or shrub native to southern Mexico and northern Central America, has widely invaded the Hengchun peninsula for the past few decades. Early detection is a critical principle to accomplish the efficiency of managing invasive species. And predicting the distribution pattern could be a practical way, because it enables managers to narrow down the potential extent of invasion. Ecological niche models (ENMs) have been increasingly applied to accomplish such projections according to environmental variables; however, ENMs alone do not sufficiently meet the requirement of predicting the expansion of invasive species, because they are incapable of providing further information of how close the species is to full expansion. In this study, we built a dynamic model by integrating ENMs and agent-based techniques to predict the expansion process and the future distribution of L. leucocephala in the Hengchun peninsula. The model simulated the expansion process from 1988 to 2007, and the prediction result was evaluated with observed data in 2007. The optimist model showed excellent coherence with an AUC of 0.96 and its Mantel’s r as 0.79. However, while we simulated the newly infested areas during 2007 to 2013, the extent of infestation is relatively conservative. Consequently, the predicted extent of the static ENM would be recommended for further reference.