聚合分析於全球鯨豚分布預測模式之初探

• Main Authors: 陳怡秀, Chen, Yi-Hsiu
• Other Authors: 李培芬, 臺灣大學：生態學與演化生物學研究所
• Format: Text
• Language: English
• Published: 2006
• Subjects: 分布; 鯨豚; 地理資訊系統模式; 預測; 聚合性分析; cetacean; distribution; GIS modeling; prediction; meta-analysis; Arctic; Killer Whale; minke whale; Polar Biology; Sperm whale; Killer whale
• Online Access: http://ntur.lib.ntu.edu.tw/handle/246246/55111; http://ntur.lib.ntu.edu.tw/bitstream/246246/55111/1/ntu-95-R93b44013-1.pdf

瞭解生物分布狀態，為延續生態研究之重要基石，亦為保育及經營管理之必須資訊。鯨豚分布研究在近數十年來蓬勃發展，其中部分區域已多次進行進行穿越線調查與分布研究，但大部分研究均具有小研究區域及密集穿越線之特徵。其研究結果可能僅代表片段資訊，難以呈現真實概況。應用地理資訊系統（geographic information system，簡稱GIS）及統計方式建立物種分布預測模式，在陸域生物之分布研究上，已經發展多年。而在近幾年中，開始應用在鯨豚分布研究中，但大部分鯨豚分布預測研究仍侷限於傳統分布研究之方法中。 本研究中，初步藉由聚合性分析方法建立鯨豚分布圖層，且進行全球分布預測。鯨豚分布資料擷取自20篇已發表鯨豚分布研究文獻與Ocean Biogeographic Information System (OBIS)資料庫，並利用地理資訊系統將鯨豚分布資訊數位化，並與海洋表層水溫、葉綠素濃度、與深度等海洋環境因子併入1° X 1°全球網格中，並利用廣義線性模式與ENFA分別進行分布預測模式建立。資料蒐集結果囊括4科8種鯨豚，各鯨種中不同方法準確度比較結果，均以廣義線性模式預測結果呈現較高之準確率。另由本研究結果可知，藉由聚合性分析方法，全球鯨豚分布預測不但可行，且深具潛力。 Distribution researches are one of central issues in cetacean studies. Over the past few decades, there are several places which have been fully investigated. All these methods share largely similar procedures: small survey area and highly concentrated transect lines. The fractional information might lead to a misunderstanding of the entirety and generate inappropriate conclusion. In recent years, cetacean-habitat modeling has been given increased attention. However, these studies were still restricted in small study areas and hardly presented the entirety. In this research, I collected the distribution information from 20 published studies and Ocean Biogeographic Information System (OBIS) database to establish a global distribution of cetacean. I used a geographic information system (GIS) to compile all available data of cetacean distributions. I also compiled environmental data, e.g., sea surface temperature (SST), chlorophyll concentration and depth. The data were in 1° X 1° square of latitude and longitude. My initial results indicate that there are 8 species in the dataset. Generalized linear model and Ecological-Niche Factor Analysis were used for model building. In this study, logistic regression was able to show the effect of different variables on species and generate better results. In conclusion, cetacean–habitat modeling represented the potentiality for predicting cetacean distributions and understanding the mechanisms determining these distributions by meta-analysis on a global scale. 1. Introduction 1 2. Methods 4 2.1 Cetacean sighting data source 4 2.2 Environmental variables 5 2.2.1 Sea surface temperature (SST) 5 2.2.2 Chlorophyll concentration 6 2.2.3 Bottom depth 6 2.3 Prediction Models 7 2.3.1 Logistic regression 7 2.3.2 Ecological niche factor analysis 8 2.3.3 Mixed model 9 2.3.4 Model evaluation 9 3. Results 11 3.1 Prediction Models 12 3.1.1 Minke whale 13 3.1.2 Sperm whale 14 3.1.3 Killer whale 14 3.1.4 Risso’s dolphin 14 3.1.5 Common dolphin 15 3.1.6 Bottlenose dolphin 15 3.1.7 Pantropical spotted dolphin 16 3.1.8 Striped dolphin 16 4. Discussions 16 4.1 Minke whale 17 4.2 Sperm whale 18 4.3 Killer whale 19 4.4 Risso’s dolphin 19 4.5 Common dolphin 20 4.6 Bottlenose dolphin 20 4.7 Pantropical spotted dolphin 21 4.8 Striped dolphin 21 4.9 Model Evaluation 22 5. Conclusions 24 6. References 25 Figure 32 Table 51 Appendix 56