以環境DNA及視覺調查探究亞熱帶近岸魚類群聚中不同層級的多樣性

Abstract

與其他方法相比，環境DNA複合分子條碼（eDNA metabarcoding） 的問世大大增加了魚類相調查的力度。然而概括地以eDNA取代傳統方法（如視覺調查或實體採樣）尚有爭議，因此有必要事先了解eDNA的調查空間尺度。我的研究旨在使用三種調查方法── eDNA、水下視覺辨識（underwater visual census, UVC）、潛水員錄製影片（diver-operated videos, DOV），比較北台灣21個樣點（包含6個離島樣點）中三種空間尺度下魚類多樣性的差異。具體而言，本研究調查區域中的物種庫（γ多樣性）、分析魚類群聚的空間分布情形（β多樣性）、釐清魚類群聚及底棲組成之間的交互作用（α多樣性）。 三種方法一共記錄到438種魚類，當中有5種可能為台灣新紀錄種。在γ多樣性中，eDNA記錄到最高的物種豐富度（383種），而UVC及DOV分別記錄到111種及83種。雖然eDNA額外找到312種魚，但其未能偵測約43%在UVC及DOV下記錄到的魚種（55/127）。在UVC及DOV下記錄到的皆為礁區相關魚種，其中大於90%為大洋底棲性魚種，其他皆屬於底棲性魚種。相較之下，eDNA除了調查到礁區相關魚種（包含62%的大洋底棲性及26%的底棲性魚種），亦記錄到12%的大洋性及4%的深海魚種。eDNA所調查到的魚類群聚與UVC及DOV所調查到的有顯著性的差異，但三種方法皆觀察到顯著的區域內變異，亦即離島與北部海岸樣點之間的魚類群聚差異。β多樣性的空間分布情形在三種方法中甚為相似，皆由替代（turnover）造成（~98%）。曼特爾檢驗（Mantel test）更進一步顯示成對替代（pairwise turnover）與地理距離之間有顯著的相關性，而最大相關性出現在UVC中，其次為DOV及eDNA。在α多樣性中，eDNA的樣點平均物種豐富度（69 ± 19.5）顯著性地高於UVC（20 ± 4.8）及DOV（15 ± 4.8）。然而從基於距離的冗餘分析（distance-based redundancy analysis, dbRDA）的結果來看，UVC調查之魚類群聚中的變異量最能夠被底棲組成所解釋，且此解釋變異量高於DOV及eDNA。此外，即使將UVC及DOV的豐度（abundance）或生物量（biomass）資料加入dbRDA分析，所解釋的變異量並未增加。 在γ多樣性中，，若有充足的樣本重複，eDNA在調查區域中的物種庫方面有極佳的表現，且為UVC及DOV所望塵莫及的。在β多樣性的層級下，所有方法皆成功地揭示相同的β多樣性空間分布，以及觀察到魚類群聚相似度隨著距離衰減的情形。在α多樣性的層級下，沒有任何一種方法能完整調查棲息在當地的魚類相，但UVC及DOV適合用來釐清魚類群聚及底棲組成之間的交互作用。我建議要根據欲探究的生態問題來選擇相應的調查方法，而非一味地綜合eDNA與其他傳統調查方法。綜上所述，採用同樣的一採樣單位使本研究能夠實際比較各種方法在探究不同層級魚類多樣性的差異。

The environmental DNA (eDNA) metabarcoding emerges as a powerful method, allowing a more exhaustive investigation of fish fauna than any other methods. Yet, a generalization of eDNA’s use in replacing traditional methods such as visual surveys or physical sampling remains debatable. Therefore, a prior understanding of eDNA’s spatial resolution is necessary. My study aimed to compare the assessments of fish diversity at three spatial scales by eDNA, underwater visual census (UVC), and diver-operated videos (DOV) surveys across 21 reef sites in northern Taiwan of which six were located in offshore islands. The specific objectives were to explore the regional species pool (γ-diversity), reveal spatial patterns of fish assemblages (β-diversity), and disentangle the relationships between fish assemblages and benthic composition (α-diversity). A total of 438 marine fish species were detected across methods (γ-diversity), among which five are new to Taiwan. The eDNA detected the highest number of species (n = 383), followed by UVC (n = 111) and DOV (n = 83). Although eDNA uncovered 312 unique species, it failed to detect ~43% of the species observed in UVC and DOV surveys (55/127). All fishes identified in UVC and DOV surveys were reef-associated, of which >94% were bentho-pelagic species, and the remainings were benthic species. The eDNA detected not only reef-associated fishes including bentho-pelagic (62%) and benthic species (26%), but also some pelagic (12%) and deep-sea fishes (4%). Fish assemblages delineated by eDNA significantly differed from those assessed by UVC and DOV. Still, all methods captured an intra-regional variation isolating offshore islands from the sites along the northern coast of Taiwan. The spatial patterns of β-diversity were similar among survey methods and mainly driven by turnover (~98%). Mantel tests further revealed significant correlations between pairwise turnover and geographic distance among sites with the highest correlation in UVC, followed by DOV and eDNA. At the α level, species richness per site was the highest in eDNA (69 ± 19.5 species), followed by UVC (20 ± 4.8 species) and DOV (15 ± 4.8 species). However, distance-based redundancy analysis (dbRDA) demonstrated that variation among fish assemblages constrained by benthic composition was the greatest in UVC, followed by DOV and eDNA. Interestingly, the use of information on abundance or biomass in dbRDA did not raise the constrained variations in DOV and UVC. At the γ-diversity level, eDNA exhibits an extraordinary power to explore the regional species pool given sufficient replication, which is unachievable by DOV and UVC. At the β-diversity level, all the methods successfully apprehend the spatial patterns of β-diversity and the distance decay of similarity in fish assemblages. At the α-diversity level, none of the methods is capable of investigating the entire resident fish fauna, but visual surveys are more suitable for scrutinizing interactions between fish and benthos. Instead of undiscriminatingly recommending a combination of eDNA with traditional survey methods, I suggest implementing specific surveys according to the ecological questions of interest. Overall, the adoption of the same one sampling unit in each method allows us to realistically compare the different scales of fish diversity.