應用物種分布模型預測臺灣穗花杉之潛在生育地

人類活動和生育地的破壞，會影響物種族群之數量，甚至造成物種瀕臨絕種，當自然生育地無法保障瀕危物種生存時，則需要找出潛在生育地，並利用移地保育(Ex situ conservation)等人為方法、技術，將目標物種搬離它原本的生育環境，移至另一個生育環境中管理。因氣候變化對於林木生長、分布和種類具有一定的關聯性和重要性，在建立物種的生育環境資料時，必須先瞭解影響林木生長之主要環境因素，以此找出潛在生育地。以物種分布模型(Species distribution models, SDMs)進行生育地模擬，為近十年來被廣泛應用於全球氣候變化應對及預測物種的潛在分布，其可將目標物種的樣本資料和環境變量建立其關聯性，進行瀕危物種潛在生育地之推估，以供為區外保育及復育策略擬定之用。本研究蒐集19種環境變數和4種地形變量，配合986個臺灣穗花杉之分布樣點，藉由SDMs中最大熵(Maximum entropy, MaxEnt)模型，推估臺灣穗花杉於臺灣南部之潛在生育地。研究成果顯示，利用MaxEnt推估臺灣穗花杉潛在生育地，其擬合程度達到0.9945，其潛在生育地範圍分布，在臺灣南部地區北至北大武山、南至老佛山。本研究所建立之臺灣穗花杉潛在生育地，藉由現地調查，預測機率在0.6以上之高預測區，確實發現有臺灣穗花杉植株之零星分布，綜合各項結果，顯示MaxEent模型所推估之臺灣穗花杉潛在生育地具有其價值，可供為臺灣穗花杉生育地保育策略擬定之參考。

關鍵字

臺灣穗花杉；潛在生育地；物種分布模型； MaxEnt ；移地保育

並列摘要

Habitat destruction caused by human activity had brought effects on the population of species and might threaten endangered species to extinction. By ex situ conservation, which aims to spot new potential sites for targeted species and transplant it from its original site, a solution for endangered species conservation had provided. In order to establish the species site environmental information, comprehension on main environmental indicators is required for simulating new potential sites. Climate change is relevant and important to vegetation growth, distribution and species structure. Recently, Species Distribution Models (SDMs) are widely applied to the simulation of new potential sites for targeted species. The SDMs has been widely used, associating with environmental variables and the targeted species data for forecasting the species potential distribution and its adaptive strategy under climate change. In this study, Maximum entropy (MaxEnt) was applied to estimate new potential sites for Amentotaxus formosana, in which 986 sample sites of Amentotaxus formosana, 19 environmental variables, and 4 terrain variables were collected and taken into consideration. The results showed that the accuracy of simulation for potential sites by MaxEnt is 0.9945, where potential distributions of Amentotaxus formosana lies in North Dawu Mountain, and stretches southward to Laufoshan in southern Taiwan. By investigating areas where possesses higher accuracy of probability of 0.6 or higher, sporadic distributions of the Amentotaxus formosana were discovered. The comprehensive evaluation results of the MaxEnt model are reliable and indicates potential sites, which provide useful informations on conservation of Amentotaxus formosana.