應用多期數值地形模型於海崖侵蝕之研究-以臺東成功海岸為例

近年來無人飛行載具（Unmanned Aerial Vehicle，簡稱無人機、UAV）與相關軟硬體發展快速，已成為環境監測與科學研究的重要工作。本研究之目的為探討簡易UAV拍攝的影像和所產製的DEM，可如何偵測地形起伏較大之海崖的侵蝕後退歷程，並評估結合其他測量方法的海崖後退量作業模式。研究樣區為臺東縣成功鎮八邊溪口的海岸，長約2公里，過去一百年來海崖持續侵蝕後退。本研究自多期圖資影像擷取崖線、濱線、海崖形態、海灘形態、海灘粒徑、人為活動等因子，包含（1）自行以簡易空拍機搭配RTK拍攝產製之3期影像與DTM（精度3公分），2018至2020年；（2）UAV產製之正射影像，2013年；（3）航空攝影測量產製之海岸像片地形圖（比例尺1:1000），1986年。此外，採用其他計畫之LiDAR-DEM（0.2x0.2公尺）與影像（2020年）為各期圖資套疊的基準。根據多期圖資的套疊比對，本研究測得八邊溪口的海崖在1986-2020年（34年）間崖頂最大後退量可達34公尺，後退量空間差異很大，整體而言崖頂後退與崖底呈現水平後退的模式，該段期間曾經歷多次大颱風事件。在2013-2020年（7年）期間，本段海崖變化很有限，只判釋出4處上邊坡崩落，該段期間幾乎沒有颱風巨浪記錄。逐步迴歸分析亦確認，類似本海岸之窄灘、弱岩特性的海崖後退，主要受控於極端事件的發生與否。本研究呈現了低成本的簡易空拍機容易在短時間內多次作業，又具有低高度、多角度拍攝的彈性，由此獲得的高解析度影像與3D模型，可判釋海崖頂的確切位置與坡面的細微地形變化，而有助於推知海崖後退歷程。若結合RTK測得之地面控制點或以既有之高成本、高空間精度LiDAR-DEM為基準，則可解決海岸研究常見之控制點分布不佳的問題，而測得海崖或海灘的量體變化。本研究發展的海崖研究可操作模式，對地形起伏大的海岸環境而言特別重要。

關鍵字

海崖後退；無人機攝影測量；數值地形模型；逐步回歸；八邊溪

並列摘要

In recent years, the use of UAV (Unmanned Aerial Vehicle) and relevant software have increased and become crucial in environmental monitoring and scientific research. The purpose of this research is to explore how the images taken by a simple UAV and the generated DEM can detect the erosion and retreat process of sea cliffs with large undulations, and evaluate the operation mode of sea cliff retreat combined with other measurement methods. The study sample area is at Babian Coast, Taitung. The coast is about 2 kilometers long. The sea cliffs in that area have been eroding in the past 100 years. The main method of this research is to extract cliff lines, coastlines, and factors such as sea cliff shape, beach shape, beach particle size, human activities, etc. from multi-phase data images, including (1) self-use of a simple aerial camera with RTK filming products Phase 3 images and DTM (accuracy 3 cm), 2018 to 2020; (2) Orthographic images produced by UAV, 2013; (3) Orthophoto Base Map in the coastal area produced by photogrammetry (scale 1: 1000), 1986. In addition, the LiDAR-DEM (0.2x0.2m) and images (2020) of other projects are used as the basis in each phase. According to the overlap comparison of multiple maps, this study has measured that the maximum retreat of the cliff top of the Babian Coast Sea Cliff from 1986 to 2020 (34 years) can reach 34 meters. This area underwent many typhoons during this period, and the amount of retreat varied greatly in space. On the whole, the retreat of the clifftop and the bottom of the cliff present a horizontal retreat pattern. From 2013 to 2020 (7 years), there was almost no record of typhoons and huge waves during this period. The sea cliff changes in this section were very limited, only 4 upper slopes were found to have been released. The stepwise regression analysis also confirmed that the retreat of sea cliffs, which are similar to the narrow beaches and weak rocks on the coast, is mainly controlled by the occurrence of extreme events and is sporadic. This research shows that the simple aerial camera is easy to operate multiple times in a short time, and has the flexibility of low-altitude and multi-angle shooting. The high-resolution images and 3D models obtained from this can be judged as the top of the sea cliff. The exact location and subtle topographical changes of the slope are valuable for inferring the retreat process of the sea cliff. If combined with control points measured by RTK or based on the existing high-cost, high-spatial accuracy LiDAR-DEM, the amount of Body change of sea cliffs or beaches can be measured, and the problem of poor distribution of control points common in coastal research can be solved. The operational model of sea cliff research developed in this research is especially noteworthy for coastal environments with wide undulations.