Spatial Distribution, Protected Area Coverage and Hotspot Analysis of Breeding Birds in Taiwan

臺灣為特有性鳥類的重要熱點，然而應用分布資訊於保育的相關研究至今仍十分缺乏。本研究首度應用新整合之臺灣繁殖鳥類大型資料庫，以近年發展之統計模式（1）推估臺灣繁殖鳥類之物種分布，（2）重新評估各繁殖鳥種之保育現況，（3）整合各鳥種分布結果，判定臺灣的鳥類生物多樣性熱點。 本研究首先建構臺灣（145種繁殖鳥類中）116種繁殖鳥類的五種分布預測模式：判別分析（multiple discriminant analysis，MDA），邏輯迴歸（logistic regression，LR），遺傳演算法（genetic algorithm for rule-set production，GARP），生態棲位因子分析（ecological niche factor analysis，ENFA）以及最大熵函數演化法（maximum entropy，maxent）。疊合此五種模式之分布預測結果，可得各物種之「最終分布圖」。藉由計算各物種「最終分布圖」涵蓋臺灣面積的比例，可將各物種分別歸類於四分位數的四個等級中，由此得「四分位數稀有度」；22種鳥類被列為分布範圍面積小於臺灣的四分之一之「稀有」物種。 其次，本研究計算現行保護區系統涵蓋各鳥種分布範圍面積比例，各鳥種分布範圍平均有29.8% 位於保護區中。比較「四分位數稀有度」、「保護區涵蓋物種分布範圍之面積比例」與現行保育等級，本研究認為71種繁殖鳥類的保育等級可能需要重新評估。 藉由整合116種繁殖鳥類的分布圖，累加得到四種不同準則下的鳥類種豐富度分布型態：所有種豐富度、特有種豐富度、保育類種豐富度、以及稀有種豐富度。以各準則下豐富度最高的5%網格作為熱點標準，分別得到四種準則下之兩種熱點分布圖：(1)臺灣全島分布熱點，(2)現存保護區系統以外之熱點。各準則之熱點分析一致指出，臺灣山區為保存鳥類生物多樣性的重要區域；臺灣東北方的宜蘭南澳山區則為現行保護區系統外重要的鳥類分布熱點，應列為未來野外調查工作與保育策略規劃的重點區域。其餘仍未納入保護區系統內之重要鳥類分布熱點如中央山脈中段與南段，則多為現存保護區系統範圍的延伸。綜合分析與比較四種準則之熱點分布，南澳山區仍為臺灣重要的鳥類熱點，且四種不同準則之熱點間重疊性極低。因此，若想僅以單一熱點準則保育臺灣之鳥類生物多樣性，將無法涵蓋其餘準則之鳥類熱點分布區域。 由於部分物種紀錄資料不足，「以五種分布預測模式疊合以獲取最終分布圖」的分析方法，僅能得到145種繁殖鳥中的116種鳥類分布資訊。然而，建構這些未曾被研究過之稀有或隱蔽性物種的分布預測模式亦為當務之急。因此本研究續以在樣本數偏低時仍被認為能產生可信賴預測結果之最大熵函數演化法（maximum entropy，maxent）建構在第一階段研究中未曾被分析的27種稀有或隱蔽性繁殖鳥類之分布預測模式，逐一探討各物種的分布預測結果，如棲地需求、棲地完整性或破碎化程度、物種偵測度，以及資料可信度，並提出應用建議。 綜觀而論，本研究首度提供近乎完整的臺灣繁殖鳥類物種分布預測模式成果。這些模式在巨觀生態學與保育相關議題研究方面相當重要，例如評估稀有度與物種保育等級，判定生物多樣性熱點等等，亦為本論文對臺灣鳥類生物多樣性研究提出之重要貢獻。未來評估臺灣鳥類保育等級時，應納入四分位數稀有度與保護區涵蓋範圍的概念；稀有或隱蔽性鳥種則需要更多的監測努力量，以及針對各物種更適合的調查偵測技術。欲維持臺灣的生物多樣性，短期目標應著重於保護本研究指出之尚未納入現行保護區系統中的鳥類生物多樣性熱點；未來則應增加其他分類群的研究，並在選取保護區時考慮納入互補性分析的概念，亦應考量未來的土地利用情形，以及氣候變遷的影響。Although Taiwan is an important hotspot of avian endemism, efforts to use the available distributional information for conservation analyses have so far been incomplete. For the first time, a newly-assembled, large database of Taiwan’s breeding bird species has been used to apply recently developed statistical program to (1) model the distribution of Taiwan’s breeding bird species, (2) use these distribution to reassess the conservation status of these bird species, and (3) combine these models to determine hotspots of avian species richness. I modeled the distribution of 116 out of 145 breeding bird species with five GIS-based distribution modeling techniques, namely multiple discriminant analysis (MDA), logistic regression (LR), genetic algorithm for rule-set production (GARP), ecological niche factor analysis (ENFA) and maximum-entropy (maxent). The resulting five distribution maps were then combined using an ensemble approach resulting in one “final map” for each species. Each species’ coverage of the entire study area by these “final” modeled distributions was calculated. Each species was then placed into one of four quartiles, thus establishing a new measure of rarity called “range quartile rarity.” This measure thus distinguished 22 species with a limited distribution on mainland Taiwan. I then also calculated how much of each species’ distribution was covered by Taiwan’s protected areas. Overall, 29.8% of the distributional ranges of the 116 modeled species are covered by protected areas. Using this newly available information on range quartile rarity and protected area coverage, I found that the conservation status of 71 species may need to be reassessed because possible conflicts exist between previously established and these newly established conservation-relevant measures. Furthermore, these 116 distribution maps were used to generate hotspot maps using the following criteria: total species richness, endemic species richness, threatened species richness, and rare species richness. These hotspot maps were then used to determine the 5% most species rich grid cells (1) within the entire island of Taiwan and (2) within the entire island of Taiwan but outside of protected areas. Almost all the species richness and hotspot analyses reveal that the mountainous regions of Taiwan hold most of Taiwan’s avian biodiversity. The one substantial unprotected region which was consistently highlighted as an important avian hotspot is a large area of unprotected mountains in Taiwan’s north-east (Nan’ao mountainous region) which should become a high priority for future field work and conservation efforts. In contrast, other unprotected areas of high conservation value are just spatial extensions to areas already protected in the central and southern mountains. After combining the results of the four hotspot criteria, the Nan’ao mountainous region was found to be important again. Significant for conservation efforts is also the result that different hotspot criteria overlap only partially and sometimes hardly at all. Therefore, to protect areas based on only one hotspot criterion would not protected areas of other hotspot criteria in Taiwan. The ensemble modeling approach was used on 116 out of the 145 breeding species because sufficient presence and absence data was available for these species. Because there is also an urgent need to build distribution models of the remaining bird species, which are mostly rare and cryptic species, a further data collection effort was made to assemble presence (but not absence) data for the remaining 29 species. This effort allowed the distribution modeling of another 27 breeding bird species of Taiwan using a statistical approach which has been proven to be especially reliable for modeling species with a low sample size of presence localities, namely the maximum entropy (Maxent) modeling technique. The newly available presence records were then used to model each species’ distribution. Detailed discussion of the interpretation and applicability of these distributions are provided, whereby close attention was paid to habitat requirements, the intactness and fragmentation of their habitat, the general detectability of the species and data reliability. In summary, this study for the first time provides distributional maps using modern statistical distribution approaches for almost all of Taiwan’s breeding bird species. Such models are invaluable for pursuing macroecological and conservation related topics, such as conservation assessments of rarity and species richness hotspots. I used these models to investigate a few of the many possible topics which can be explored with these novel techniques. Many of my results are new and relevant for the conservation of Taiwan’s avifauna. Therefore, range quartile rarity and protected area coverage should be incorporated into future assessments of the conservation status of Taiwanese birds. Rare and cryptic species need increased monitoring efforts and improved censusing techniques. Short-term conservation efforts should focus on protecting those species-rich hotspots which are still unprotected as highlighted in this study. Future studies should add other taxa to the analyses as well as consider complementarity analyses and future land-use and climate change effects