以整合生態模型方法評估淡水魚類分布之時空變化

Abstract

淡水生態系統多樣而複雜，位於新店溪中上游之北勢溪與南勢溪集水區具有豐富之生物多樣性，數十種淡水魚類棲居其中，各自選擇適宜之棲息環境，佔據特定河段、展現局部遷移行為或是降海洄游等特性，皆為本區常見之生態行為。然而，魚類族群生存受到棲地損失、棲地破碎化、外來入侵種、環境劣化等威脅，因此釐清淡水魚類族群在現有環境條件下的生存狀況是落實生態保育不可或缺的工作。 本研究整合穩態與動態環境因子，考量地文、氣象、水文三類因子，利用SWAT（Soil and Water Assessment Tool）模式模擬河川流量，並透過生態棲位因素分析（Ecological Niche Factor Analysis, ENFA）建立魚類分布模型，以棲地適宜性指數、分布範圍、分布樣態等面向探討魚類物種分布的時空動態變化，並建立視覺化地圖，呈現魚類於北勢溪與南勢溪集水區之適宜河段，最後利用上述分布特徵指標，以K-means進行分群，針對不同類群的魚種提供具體的保育建議。 研究結果顯示，ENFA模式模擬19個目標物種，多數模式表現達到0.6以上，而不同魚種在棲地適宜性指數、分布範圍、分布樣態等面向上多展現不同的動態類型，K-means分群結果則將19種魚類以其分布特徵分為三種類型。第一種類型包含纓口臺鰍、臺灣白甲魚、臺灣石𩼧、臺灣鬚鱲、唇䱻、粗首馬口鱲、中華花鰍、長脂瘋鱨、明潭吻鰕虎、香魚等，棲地適宜性指數高，分布範圍較廣，分布樣態相似性高，適合以維持現況與監測作為保育策略；第二種類型包含大眼華鯿、短吻小鰾鮈、圓吻鯝、日本鰻鱺，棲地適宜性指數低，分布範圍居中，分布樣態相似性亦中等，但分布範圍週期性低，適合以棲地維護為保育策略；第三種類型則包含臺灣間爬岩鰍、䱗、臺灣吻鰕虎、極樂吻鰕虎、吉利非鯽，棲地適宜性指數中等，分布範圍小，分布樣態相似性低，而分布範圍週期性高，適合以促進河川連結性作為保育策略。 本研究顯示此模型架構可應用於淡水魚類之分布動態，根據其棲地喜好與生態行為以及多魚種分布之重要河段位置，依據需求進行天然河道狀態維護、增加河川連通性、封溪護魚、族群監測等措施之配置與調整，融合適應性經營管理等治理方式，提供棲地保育策略之參考。

The freshwater ecosystem is diverse and complex. Located at the upper reaches of Xindian River, the watershed of Beishi and Nanshi River has rich biodiversity, with dozens of freshwater fish residing there. They each select suitable habitat environments, occupy specific river segments, exhibit localized migration behaviors, or engage in downstream migrations, all of which are common ecological behaviors in this area. However, the survival of fish population is threatened by habitat loss, habitat fragmentation, invasive species, environmental degradation, and etc. Therefore, clarifying the survival status of freshwater fish under existing environmental conditions is indispensable for the implementation of ecological conservation. This study integrates static and dynamic environmental factors across three key domains: geographical, meteorological, and hydrologic aspects. It applies the SWAT (Soil and Water Assessment Tool) model to simulate river flow, and establishes a comprehensive fish distribution model through Ecological Niche Factor Analysis (ENFA), delineating different spatiotemporal dynamics of fish species distribution based on habitat suitability index, distribution extent, and pattern. Through the development of visualization maps, this study identifies suitable river segments for freshwater fish in the Beishi and Nanshi River. At the end, utilizing the aforementioned distribution characteristic indicators, K-means clustering is applied to categorize different fish species groups and provide specific conservation recommendations. The research findings reveal that the species-specific ENFA model successfully simulated 19 target fish species, with a majority of the models achieving a model performance score in the Continuous Boyce Index of 0.6 or higher. These models demonstrate diverse dynamic characteristics of fish species, encompassing habitat suitability, distribution extent, and distribution pattern. The application of K-means clustering categorized the 19 fish species into three types with different distribution characteristics. The first type comprises species such as river loach (Formosania lacustre), Taiwan shovel-jaw carp (Onychostoma barbatulum), Taiwan striped barb (Acrossocheilus paradoxus), Formosan stripe dace (Candidia barbata), barbel steed (Hemibarbus labeo), thickhead chub (Opsariichthys pachycephalus), Siberian spiny loach (Cobitis sinensis), bagrid catfish (Tachysurus adiposalis), goby (Rhinogobius candidianus), and ayu sweetfish (Plecoglossus altivelis). These species exhibit a high suitability index, wide distribution extent, and a notable similarity in distribution pattern. Consequently, they are deemed suitable for conservation strategies that emphasize maintaining the current state and implementing monitoring measure. In contrast, the second type encompasses species such as large-eye Chinese bream (Sinibrama macrops), short-nose gudgeon (Microphysogobio brevirostris), round snout (Distoechodon tumirostris), and Japanese eel (Anguilla japonica). These species possess a low suitability index, a moderate distribution extent, and a moderate similarity in distribution pattern. However, their distribution extent demonstrates low periodicity, rendering them suitable candidates for conservation strategies focused on habitat maintenance. The third type consists Formosan river loach (Hemimyzon formosanus), sharpbelly (Hemiculter leucisculus), goby (Rhinogobius formosanus), barcheek goby (Rhinogobius similis), and redbelly tilapia (Coptodon zillii). These species have a moderate habitat suitability index, a small distribution extent, and a low similarity in distribution pattern. Nevertheless, their distribution extent displays high periodicity, making them well-suited for conservation strategies centered around promoting river connectivity. This study demonstrates the applicability of the model framework to the distribution dynamics of freshwater fish. Depending on specific needs of habitat preferences, ecological behaviors, and the importance of river segments, conservation strategy can be developed or adjusted the allocation of measures, including maintaining natural river environment, enhancing river connectivity, protecting fish species through stream closure and monitoring, and providing references for integrate adaptive management.