摘要:【應分段包括計畫目標、架構(重要工作項目)及預期效益;限2000字以內】
本研究計劃探討海洋環境因子如何影響秋刀魚在北太平洋的時空分佈。提供即時海洋環境資料及海況預報,與漁業資料整合。以提高漁民的投資報酬率,並提升管理的有效性。最終目標是達到資料整合與漁場即時預報。本年度持續整合並更新環境資料,包括表水溫(Sea Surface Temperature),表水鹽度(Sea Surface salinity),水溫鋒面 (ocean temperature front),表水溫鋒面頻度(temperature front frequency),海面高度(Sea Level Height anomaly),海洋生產力(ocean color),洋流流向及流速(geostrophic flow),渦旋動量(eddy kinetic energy) 及海洋生產力(ocean color),以高解析度圖相,每日更新與發佈。並整合過去漁業資料(CPUE),0.5 X 0.5 度方格解析度,將CPUE的空間分佈,按年月份與上述之各環境參數之歷史資料套疊。提供預報系統給秋刀魚漁船實驗性試用,將提供三項資訊:1. 每日更新與即時發佈(daily)前述之環境資料,發布之網站,包含彩色與黑白兩個版本,並可以自由放大縮小,以利漁民方便使用。2. 提供最相關之歷史資訊。亦即,透過比對歷史和當下環境資料,提供最接近當下之歷史環境資料,並套疊CPUE之分布。3. 利用我們2018年建立的環境資料與歷史CPUE關聯性之Boosted regression tree,代入即時環境資料,預測當日秋刀魚的出現機率,並呈現所預測機率之空間分布圖。實驗船公司可以透過網路,進入資料庫伺服器,看到這些資訊並提供給漁船。執行後續將與漁民討論如何改善此系統。
利用2002-2018 CPUE資料,結合環境資料,以每月0.5 度方格解析度,發展最大熵學習法(Maximum Entropy)做預測模式。並整合所有已發展之方法,包含Zero-inflated GLM, Zero-inflated GAM, Boosted regression tree, Habitat suitability index及Maximum Entropy,做組合預報(ensemble forecast)。再者,計算月平均之CPUE,及漁場平均之環境因子量,使用GLM和GAM分析CPUE的月別變動與環境因子的關係。進一步以GLM和GAM分析生物分佈中心經度和緯度與環境因子的關係,以瞭解秋刀魚每年洄游路徑的變化。
Abstract: 【請配合中文之執行內容撰寫】
This project aims to develop a now-cast system for using environmental data to predict fishing ground for Pacific saury, in order to provide useful tools for fisheries management and lower the cost of fishery operations. Here, we propose to integrate environmental data, including Sea Surface Temperature, Sea Level Height anomaly, ocean color, sea surface salinity, geostrophic flow velocity and direction, and calculate ocean temperature gradient, frequency of temperature front, and eddy kinetic energy. We will display the ocean conditions with high-resolution maps and update the information daily. We then integrate CPUE data of Pacific saury with the environmental data (monthly data with 0.5 degree resolution). We will work with saury fishermen to test our system, using experimental fishing boats. We will provide three sources of information: First, the daily updated environmental data, with color and black-and-white versions and zoom-in function; second, the most relevant historical environmental data associated with CPUE; third, the distribution map of occurrence probability of saury predicted by a month-specific Boosted regression tree using the updated environmental data. Here, the specific Boosted regression tree model was parameterized by historical environmental data and CPUE. Registered fisheries companies can access the information in our data servers through internet. We can then evaluate the performance of our now-cast system.
We will develop two forecast models. First, based on the 0.5-degree grid data, we will try the Maximum Entropy method to link CPUE with environmental factors. Second, we will integrate all models that we developed (including Zero-inflated GLM, Zero-inflated GAM, Boosted regression tree, Habitat suitability index, and Maximum Entropy) to do ensemble forecast. We will also calculate the fishing ground averaged CPUE and environment variables, and employ GLM and GAM. Finally, we will estimate centroid of the CPUE distribution, in order to study the migratory trajectory of Pacific saury. We will then link the longitude and latitude with environment variables, using GLM and GAM.