碩士 / 國立臺灣海洋大學 / 環境生物與漁業科學學系 / 106 / The physical environment directly influences the distribution, abundance, physiology and phenology of marine species. We used satellite-based oceanographic data of sea surface temperature (SST), sea surface chlorophyll-a concentration(SSC), and Sea surface elevation(SSE), Sea surface temperature front (JSD), Chlorophyll-a level 3(CHL) and Ocean mixed layer thickness (MLD) together with 1 degree resolution catch data of yellowfin tuna were collected during the period of 2009-2014. The study used two species distribution model, Generalized Additive Models (GAMs) and Maximum entropy model (MaxEnt) to investigate the relationship between yellowfin tuna fishing ground and oceanographic conditions and also to predict potential habitats for yellowfin tuna in the Pacific Ocean. We further explored the distribution and accuracy rate of three habitat types, suitable zone, buffer zone and core zone. The results revealed that the cumulative deviances obtained using the selected GAMs were 34%. The results suggest that areas with a higher SST approximately 25~30(°C), a SSE of approximately 0.4~0.8(m), a CHL of approximately 0 ~0.3 (mg/m3),a MLD of approximately<40 (m) and JSD of approximately 0.4~0.6 yield higher catch rates of yellowfin tuna in GAM model. The AUC were higher than 0.7 through the year in MaxEnt model, and the optimal range of hydrological variables were 25~29(°C) of SST, 0.5~0.83(m) of SSE, 0.1 ~0.19( mg/m3) of CHL, 5~32(m) of MLD, and 0.38~0.53 of JSD. Both models show that the sea surface water temperature and sea surface elevation are the main factors affecting the catch rate. In addition, we explored the accuracy of different models in predicting the distribution of catch rates. The results showed that GAM predict more than 40% accuracy in the three habitat types in second to fourth quarter, and the prediction accuracy of the buffer zone in third quarter approximately 60%. However, the three HSI models established using the MaxEnt model still differ in their prediction capabilities. The prediction of the optimal habitats in the second to third quarters with the highest AUC values predicted the highest accuracy (64 to 100%). In the fourth quarter, the accuracy of the most suitable habitat for the prediction of the lowest AUC value reached 92%. The prediction accuracy is between approximately 16.67% and 65.22% for the model of superimposing GAM and MaxEnt . In addition, the optimum habitat for the suitable habitat in all modes had seasonal changes and were mainly located in the Western and Central Pacific Ocean and in the equatorial Pacific Ocean, with seasonal changes.
| Identifer | oai:union.ndltd.org:TW/106NTOU5451009 |
| Date | January 2018 |
| Creators | Lai, Shi-Han, 賴詩涵 |
| Contributors | Lan, Kuo-Wei, 藍國瑋 |
| Source Sets | National Digital Library of Theses and Dissertations in Taiwan |
| Language | zh-TW |
| Detected Language | English |
| Type | 學位論文 ; thesis |
| Format | 74 |
Page generated in 0.0076 seconds