臺灣特有鳥種現況多樣性與未來分布

Abstract

氣候變遷是21世紀最受關注的焦點議題。人為所造成的暖化現象對於生態系統造成了多面向的影響。藉由過去資料的剖析、物種分布預測模式的比較與利用，可了解生物和環境間的關係，同時幫助探討現今物種分布的熱點區域以及受保護的情形。續搭配多重的氣候假設，則可透析物種面臨氣候變遷可能的分布改變情形，進一步有利於考量可能的調適與減輕策略，達到地球永續發展與利用的長遠目標。 本研究結合1993至2004年鳥類調查觀察記錄，識別出17種特有鳥種在臺灣個別與獨特的分布形態，透過一平方公里的空間解析度，剖析臺灣特有鳥種現今與未來的巨棲地特徵、重要影響之環境因子與整體分布。 研究結果發現，依據物種出現記錄的網格數，可將此17種臺灣特有鳥種分為普遍種 (具有多於200個網格的出現記錄) 、不普遍種 (100至200個網格) 與稀有種 (少於100個網格) 共三大類型，其中稀有種之一的帝雉 (Syrmaticus mikado) 含有最少的出現記錄，而普遍種之一的五色鳥 (Megalaima nuchalis) 則有最多的發現記錄。各特有鳥種具有其特定的分布範圍以及棲地類型喜好。整體而言，17種特有鳥種棲息地乃於相異的海拔與氣候條件下。植被覆蓋度高、森林密度大以及植被指數中至高的棲地類型普遍為其普遍之偏好。比較五種物種分布預測模式，包含邏輯迴歸 (Logistic Regression, LR) 、多變量區別分析 (Multiple Discriminant Analysis, MDA) 、基因演繹法 (Genetic Algorithm for Rule-set Prediction, GARP) 、類神經網路 (Artificial Neural Network, ANN) 與最大熵函數演算法 (Maximum Entropy, MAXENT) ，可發現非線性模式 (GARP、ANN、LR與MAXENT) 預測能力較線性模式 (MDA) 佳，藉由Kappa、敏感性 (sensitivity) 、特異性 (specificity) 與正確性 (accuracy) 四種模式評估值檢測，GARP與MAXENT為其中預測能力最佳的模式，適用於臺灣特有鳥種的現況分布預測。以三種相異的生物多樣性熱點方法 (基於典型對應分析之熱點; CCA-based potential hotspots, CBPH、賦予相同比重之熱點; same-weighted hotspots, SWH與賦予相異比重之熱點; differentiated-weighted hotspots, DWH) ，比較臺灣特有鳥種可能熱點位置，發現SWH最能捕捉實際觀測之生物多樣性位置 (72%) ，DWH次之 (61.6%) ，CBPH最差 (35%) 。其中，特有鳥種熱點主要受國家公園保護，約可保護22-23%的熱點區域，野生動物保護區與自然保留區則僅保護少於6%的熱點，多數的特有鳥種熱點區域仍未受保護。而於氣候變遷的假設下，分別預測特有鳥種於A2以及B2情境中於五種大氣環流模式 (general circulation models, GCMs) 在不同階段 (2020、2050、2080與2100年) 的分布情形，可發現氣候變遷對於臺灣特有鳥種可能分別有正、負之影響，此影響和物種現今的海拔分布中位數有高度的相關，同時，預測結果顯示，呈現負影響之物種，將可能往高海拔移動以及減少其分布，呈現正影響之物種，則將不僅棲息與原有棲息地並擴張到高海拔處。 透過完整的生物分布預測，確實可反映物種受現今環境以及暖化增溫後之分布情形。長時間的監測、針對性的實際觀測以及跨學科的實驗將是了解複雜的生態系統整體關係之所必須，亦是保育與經營管理之重要基石，更可幫助人類與地球生物有效且永續的共依共存。

Climate change is the focus of a 21st- century global issue. Anthropogenic warming has caused, and will continue to cause, multi-faceted effects on ecosystems. Combining long-term observation data and species distribution models is helpful in understanding relationship between environment and species distribution and in recognizing biodiversity hotspots and currently protected situations. Analyzing various assumptions under possible climatic changes in order to predict future species distributions would provide further consideration of possible adaptation and mitigation strategies and help to achieve global sustainable development and utilization. I used two bird inventories conducted in 1993–2004, and extracted data for 17 endemic bird species, with a spatial resolution of 1 km2, to identify individual and specific features of their distributions and to predict current and future potential distributions in Taiwan. According to species’ occurrences, the 17 species were classified as common (present in >200 grids), uncommon (100–200 grids) or rare (<100 grids). The Mikado Pheasant (Syrmaticus mikado), as a rare species, had the lowest occurrence records, while the Taiwan Barbet (Megalaima nuchalis), as a common species, had the highest. Each species had a specific distribution range and habitat preference. In general, these 17 species occupied heterogeneous elevation and climatic conditions, and they favored habitats with high vegetation cover, dense forest and median-to-high Normalized Difference Vegetation Index (NDVI). In comparison with five species distribution models, including logistic regression (LR), multiple discriminant analysis (MDA), genetic algorithm for rule-set prediction (GARP), artificial neural network (ANN), and maximum entropy (MAXENT), the nonlinear models (GARP, ANN, LR, and MAXENT) provided better predictions than did the linear (MDA) model. Based on kappa, sensitivity, accuracy, and specificity values for each species and the three species categories (common, uncommon, and rare species), GARP and MAXENT were the most consistent models for predicting current distributions of the 17 endemic bird species. By overlapping species predictive distributions and defining areas with the upper 5% of species richness as biodiversity hotspots, three hotspot criteria were designated: CCA-based potential hotspots (CBPH), same-weighted hotspots (SWH), and differentiated-weighted hotspots (DWH). These potential hotspot zones of the endemic bird species can be used to assess the efficiency of protected areas. The SWH showed the most coverage (72%) of actual biodiversity hotspots where species richness is higher than 7 species, followed by the DWH and CBPH (61.6% and 35%, respectively). National Parks provided the greatest protection for the 17 endemic bird species, protecting 22-23% of hotspot areas, whereas nature reserves and wildlife refuges protected less than 6% areas. Most potential biodiversity hotspots were not protected adequately. The effects of climate change on species distributions showed that species would have both positive and negative responses (i.e. increases and decreases) that correlated highly with the median value of the species’ originally occupied elevations. The geographical patterns indicated that the negative-effect species would shift up in elevation, with decreased distribution over time while the positive-effect species would remain in the original habitats and expand to higher elevation. Species predictive distributions are proved usefully to reflect their distributions both in the face of current environments and future increasing temperature. Long-term monitors, targeted field-based observations, and interdisciplinary experiments are necessary and helpful to resolve complicated problems across the natural systems.