利用階層線性模式探討魚類與棲地環境尺度相依關係之研究 – 以大屯溪為例

Abstract

在自然界中，尺度可視為人們藉以透視、觀察生態系統的視窗。科學家可藉由生物於生態系統中所的活動模式，邏輯性的推理生物與其生存棲地環境間之關係。在自然界中，物理性與生物性過程兩者於某一特定之時空尺度進行，兩者之交互作用可對於生態格局有決定性之影響。於Frissell et al. (1986)將河川棲地建立層級分類系統之後，尺度於生態系統相關研究中的重要性便日趨漸增。本研究的主要研究魚類及環境的跨尺度層級關聯性，更進一步探討這些跨尺度關係在不同魚種的差異性。研究中主要採用大屯溪流域常見的兩種洄游性魚類–日本禿頭鯊及台灣鏟頷魚，其中日本禿頭鯊在大屯溪為全流域分布，台灣鏟頷魚（又稱台灣鯝魚）主要分布為中上游河段。除空間分布上有明顯差異外，此兩種魚類皆佔總調查魚類的大多數，因此將其選為研究物種。魚類密度及環境變數多數由現地採樣取得。採樣時間始自2007年秋季至2008年夏季，共四季為期一年。研究中採用流況、水質以及地理相關等共16個跨尺度環境變數，對魚類進行層級迴歸分析。接續採用廣義階層線性模式（HGLM）及階層線性模式（HLM）分別對兩種魚類的出現機率及密度之相關跨尺度環境變數進行季節性和年平均模式建立以及分析。廣義階層模式之結果顯示，台灣鏟頷魚的出現機率與大尺度環境因子(高程、鹽度、水溫)的作用關係較為明顯。由階層線性模型之結果可知，日本禿頭鯊的魚類密度與小尺度（流速、水深）及大尺度（河寬、水溫、坡降、水文距離、土壤）的環境變數均有相關性，且於冬季模型中可見跨尺度交互作用關係（水深與河寬、水深與土壤）。結果顯示，魚類與環境變數的關係會隨著物種不同而改變，兩者之關聯性所在之尺度亦會隨著魚種而有差異。根據階層模式結果，日本禿頭鯊對於小尺度（section尺度）之棲地流況變數關聯性較強；反之，台灣鏟頷魚對於較大尺度（reach尺度）之地理形貌相關變數的連結性較強。為表現魚類與環境變數之間的跨時空尺度交互作用關係，本研究於最後提出一概念性示意圖，展示日本禿頭鯊在四季與全年的時間尺度下與環境之關係，更對於在保育或施行生態工法時須重視的環境因子。總論之，本研究提供一個概念性的方法，以得知魚類與其生存之環境在不同時空尺度下之交互作用關係。提醒工程師在設計河道進行魚類保育時，除考量微棲地之流速水深外，須注意到大尺度因子如高程、坡降等對於魚類的影響。除此之外，同時也證實在魚類與環境交互作用相關研究中，考量時空尺度之跨層級關聯性及季節性變化差異的重要性。

Scales could be regarded as windows for viewing through an ecosystem. Patterns enable scientists to view through systems with logical perspectives and thus are considered as key elements for illustrating natural conjunctions in an ecosystem. Yet, physical and biological patterns operating over a rage of spatial and temporal scales are able to determine the ecological patterns. To emphasize the importance of scaling in ecological fields, this study elucidates how fish and environment are associated through spatial and temporal scales under a hierarchical structure and also examined how the relationships differ among species. Sicyopterus japonicus and Varicorhinus barbatulus are fishes chosen as survey species on account of their characteristics and distributions. Both species are migratory fishes with identical movement patterns among seasons in a year. Density of the fishes and environment variables were investigated and quantified at 70 stream sections distributed among 14 reaches in the Datun stream catchment of northern Taiwan during the autumn and winter of 2007, as well as the spring and summer of 2008. Hierarchical linear generalized models (HGLMs) and hierarchical linear models (HLMs) were applied to describe the relationships between 16 environment variables and the fish density of each species. The HGLM result suggests that V. barbatulus are dependent to larger-scale, migratory-related variables – elevation, salinity and water temperature. HLM result shows that S. japonicus relates to the habitat variables at both section (current velocity and water depth) and reach (stream width, water temperature, slope, river mile and soil erosion index) levels. Cross-level interactions (depth and river width, depth and soil) could be found in winter HLM. In general, the results indicate that each species responds to the environment majorly under a specific or a range of scale according to their characteristics and life cycles. S. japonicus respond to a spatially smaller-scale habitat environment; in contrast, V. barbatulus respond to relatively greater spatial scale environment features. Eventually, a diagram explicitly illustrating the cross-level relationships between fish and environment was then provided for conceptualizing the biotic-abiotic interaction in Datun stream. In conclusion, this research has provided a measurement for discovery the potential environment variables that might have caused influence on fish distribution in streams under different spatial and temporal scales; meanwhile, yielded the essentiality of considering seasonal changes while observing patterns of certain species.