臺灣陸域保育類哺乳動物的空間分布預測、保護區涵蓋及熱點分析

Abstract

生物的空間分布是進行研究與保育的重要資訊，當面對稀有、調查困難、出現記錄少或分布範圍廣大無法全面調查的物種時，可以利用物種分布預測模式（species distribution modeling）推估出物種的潛在棲地分布，以利進一步分析。物種分布預測模式能夠藉由連結物種出現地點資料與研究區域的環境因子資料推估出物種的潛在棲地分布，目前已被廣泛運用於各種生物與地理的相關研究。 臺灣的保育類哺乳動物多為食物鏈高級消費者，生存常面臨人類活動、棲地開發的威脅，然而，至今有關臺灣保育類哺乳動物的空間分布研究仍不完整，也缺乏應用分布資訊的進一步分析。 本研究整合1988年至2013年間11種臺灣保育類哺乳動物的全臺出現記錄，將物種出現資料與環境因子輸入maximum entropy、genetic algorithm for rule-set production和ecological niche factor analysis三種分布預測模式分析，結合三種模式的預測結果，計算各物種在全臺的棲地適合度分布圖。運用兩種不同閾值做為棲地適合度的切分點，分別呈現各物種大膽及保守估計的潛在分布範圍，以克服資料分布不均的缺陷。最後，利用地理資訊系統整合物種潛在分布與臺灣現行保護區的含括範圍，分析個別物種的潛在棲地受到保護區涵蓋的程度；並進一步將11種物種的潛在分布套疊，得出臺灣保育類哺乳動物的分布熱點，與現行的保護區系統做比較，以評估臺灣現行保護區系統對保育類哺乳動物的保護功效。 研究結果顯示，分布受保護區涵蓋的比例在物種間差異頗大，其中臺灣水鹿、黃喉貂及臺灣野山羊的比例高於50%；但石虎及白鼻心的比例卻低於20%。考量所有保育類哺乳動物分布受保護區涵蓋的平均比例時，大膽估計的比例為38.9%；保守估計的比例則為43.7%。另外，熱點分析的結果顯示，大膽估計的保育類哺乳動物的熱點只有41.7%受保護區涵蓋；保守估計的保育類哺乳動物的熱點則有53.7%在保護區內。從海拔分布來看，保育類哺乳動物分布熱點的海拔略低於現行保護區。 本研究建議加強針對中低海拔物種的保育，例如將此類物種的重要棲地劃設為保護區，或限制部分地點（如淺山地區）的開發。

Spatial distributions of flora and fauna are critical information for research and conservation. When studying distribution of rare, enigmatic, sparsely recorded or broadly distributed species, researchers can model their potential habitats with species distribution modeling techniques. Species distribution modeling can depict the potential habitat distribution of species by analyzing their occurrence records and environmental variables of study area. To date, these techniques have been applied to various biological and geographic research. Most protected mammals in Taiwan are of high-level consumers in the food chain, threatened by human activity and habitat destruction. However, research on distribution of protected mammals in Taiwan has so far been incomplete. There is also a lack of further application of the distribution information. In this study, I collected observations of 11 protected mammal species between 1988 and 2013, and modeled the potential distribution of each species with three distribution modeling techniques namely maximum entropy, genetic algorithm for rule-set production and ecological niche factor analysis. The resulted habitat suitability map of each species was obtained by ensembling the outputs of three species distribution models. In order to conquer the data defects caused by uneven sampling, two thresholds were selected, translating the habitat suitability maps into boldly predicted and conservatively predicted presence–absence maps respectively. For further analysis, I calculated protected area coverage of each species and compared the distribution of protected mammal hotspots with protected areas, evaluating the effectiveness of protected areas in Taiwan. The results showed that the protected area coverage varies between species, with Cervus unicolor swinhoei, Martes flavigula chrysospila and Naemorhedus swinhoei higher than 50%; and Prionailurus bengalensis chinensis and Paguma larvata taivana lower than 20%. Considering the average protected area coverage of all 11 protected mammals, is 38.9% when boldly predicted; and 43.7% when conservatively predicted. Moreover, 41.7% of boldly predicted protected mammal hotspots are covered by protected areas; 53.7% of conservatively predicted protected mammal hotspots are covered by protected areas. The elevation of protected mammal hotspots is slightly lower than protected areas. I recommend enhancing the protection of low to medium altitude species, such as designating habitat of low to medium altitude species as protected areas or restricting lowland forest exploitation.