全色態衛星影像於灘線變遷的應用及黃金海岸灣裡段侵蝕的探討

本文擷取衛星全色態影像的水線，並修正水線至平均灘線，以分析黃金海岸的灘線變遷，並探討灣裡段侵蝕的影響因素。影像前置作業包括去除白泡沫雜訊、影像強化及形態學，再利用Canny邊緣偵測法來獲得精準的水線。透過實測潮位內插至影像拍攝時間的海水位，並配合實測地形的前灘坡度，決定平移距離來修正擷取的水線至平均灘線。由2006年及2010年實測灘線與影像灘線比較，二者的Bias分別為0.35m與0.92m，及RMSE為6.52m與5.5m，此精度評估值與蔡(2017)相近，證實本方法的可行性。分析2003年至2014年歷年影像的平均灘線變化率，發現灣裡段灘線在冬季、夏季及全年皆呈現不同程度的侵蝕。根據港灣研究中心的波浪資料及水利署的水文資料，以經驗公式估算沿岸平均年漂沙量及河川平均年輸砂量。後者約占前者約1-2%，且2003年至2014年的沿岸漂沙量逐年增加，但河川輸砂量卻逐年減少。以雙變數迴歸及F檢定，說明夏季與全年的沿岸漂沙量為灣裡段灘線變化的主要影響因素，而河川輸砂量影響灘線變化不顯著。

關鍵字

台南黃金海岸；沿岸漂沙；全色態影像

並列摘要

The waterlines in pan-satellite images are detected and shifted to the shorelines of the Gold Coast in Tainan in this thesis. These shorelines are analyzed to show the shoreline evolution of the Gold coast and to investigate the Wan Li beach erosion. The techniques of image processing include erasing wave whitecaps, image enhancement, morphological image processing and Canny methods are used to accurately detect the waterline of a pan-satellite image. Interpolated water level at the shooting time of an image from observed tidal data associated with shoreface beach slope is to determine the shoreline by a horizontal distance through trigonometrical identities. The biases between the detected shorelines and measured ones are 0.35m and 0.92m respectively for 2006 and 2010. The corresponding RMSEs are 6.52m and 5.5m. The results are close to those of Tsai(2017), indicating the application of the proposed method to the shoreline change. The rates of summer, winter and anuual shoreline from 2003 to 2014 are negative, showing that the Wan Li beach is suffering from erosion. The longshore sediment discharge and river sediment discharge are estimated by some empirical formulas based on wave data of the Harbor and Marine Technology Center and discharge as well as sand suspension data of the Water Resources Agency. The annual averaged river sediment transport is approximate 1-2% of in the longshore sediment transport from 2003 to 2014. The longshore sediment transport increases with time, but the river sediment transport decreases with time. Through two-variable regression analysis and F-value test the longshore sediment transport is the dominant factor in the shoreline changes. The river sediment transport is not significant to affect the shoreline changes.