Please use this identifier to cite or link to this item:
Environmental factors determining benthic communities in northern Taiwan
Benthic community,Environmental determinism,Abiotic factors,Unsupervised clustering,Supervised clustering,Ecosystems,
|Publication Year :||2020|
本研究目的為探討不同底棲群聚與其對應的環境條件(如:物理、化學、及地質因子)之間的關係,並探討群聚組成會受哪些決定性環境因子影響。實驗地點為台灣東北角10個樣點(5個地點X 2個深度)以形態功能群(morpho-functional group)為區分方式進行底棲群聚調查及周遭環境因子資料蒐集。藉由兩種方法,(1)非監督式k均質集群分析(Unsupervised k-means clustering) 和 (2)監督式多變項回歸樹分析 (Supervised Multivariate Regression Tree, MRT)探討決定性環境因子與底棲群聚之間的關聯,並比較兩種方法的差異性和檢驗非監督式k均質集群分析對於環境因子的預測力是否準確 。
研究結果顯示殼狀珊瑚藻 (Crustose coralline algae, CCA) 在本研究實驗地點中有約一半的覆蓋率,為東北角相對優勢的形態功能群。其他形態功能群在實驗地點間擁有較大的標準差, 表示在各個樣點有不同的覆蓋率。化學因子在深度以及地點間並無明顯的差異,波浪(wave motion)強度及底質的不同在兩種分析方法中解釋了大部分的群聚差異,可視為東北角的決定性環境因子。MRT 和 k-means 的分群結果獲得0.82的調整蘭德指數 (Adjusted Rand Index),顯示兩種分析方法所獲得的結果有高度的相似性,也間接證實了非監督式學習方法對於相對應環境因子的高度預測能力。本研究所獲得的結論除了能夠評估未來群聚潛在的動態變化之外, 也可以對海洋沿岸底棲群聚的保育及管理有所幫助。
On a global scale, human activities and their effects on climate jeopardized the future of tropical reef ecosystems. On a regional scale, sensitivity difference could exist among communities that may originate from how they are locally shaped by environmental conditions. However, despite the possible importance of these abiotic factors, relatively few studies have focused on how they could influence benthic communities on a small spatial scale.
Here, I propose to uncover the determinism of marine coastal benthic communities by a detailed investigation of their response to small-scale modification of the environmental conditions including physical, chemical, and geological factors. At ten locations (confounding site and depth) in Northern Taiwan (NT), benthic communities were delineated using a morpho-functional classification of the organisms observed on photo-transects. k-means clustering was used to identify k homogenous groups among transects assuming it can ascertain the diversity of communities present. Their environmental determinism was later examined by combining this result with 16 environmental variables of transect conditions into a regression tree (RT) framework. Biotic and abiotic information were further analyzed with a Multivariate Regression Tree (MRT) to figure out the hierarchical environmental determinism. The classifications produced by both approach were compared using the Adjusted Rand index (ARI) to assess the predictive power of unsupervised clustering (UC) on its missing explanatory components (abiotic variables).
Crustose coralline algae (CCA) was dominant in NT and represented consistently about half of the benthic cover. Other morpho-functional categories presented high deviation across NT suggesting various cover among locations. Both MRT and k-means produce 5 communities with a similarity of 0.82 in ARI. Chemical factors in this study were quite unanimous and did not stand out as determinism while wave motion followed by substrate type resolves most of the variance. The high value in ARI and comparable structure between MRT and RT concluded that the community itself may inform us of the environmental conditions it is thriving in.
A greater consideration of these different communities and their environmental context is important to determine their trajectories under global change. Expansion of our analytical work will have implications with further potential ramifications on their dynamics.
|Appears in Collections:||海洋研究所|
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.