利用歷史航照分析斷層地形與活動之可行性： 以縱谷斷層關山段為例

Abstract

縱谷斷層系統為臺灣最活躍的斷層系統之一，其中自玉里春日向南延伸至鹿野瑞源的池上斷層段，已發現是一條具高速率（約20–30 mm/yr）震間潛移特性的活動斷層。然而，這條活躍的潛移斷層仍有一定的孕震風險，根據歷史地震紀錄，1951年11月的地震序列僅在玉里至池上一帶觀察到地表破裂現象，而在池上以南至鹿野的關山地區（本文稱為關山段）則無地表破裂紀錄。目前針對關山段的斷層特性研究十分有限，主因是該地區受人類活動及河流作用顯著影響，斷層跡及相關地形特徵多已被抹去，導致調查困難。 隨著攝影測量技術的發展，目前已能利用歷史航照重建已消失的地形特徵。儘管許多研究成功運用「運動回復結構」（Structure from Motion, SfM）與「多視角立體匹配」（Multi-view Stereo, MVS）技術於分析大尺度地形，但其在小尺度斷層變形的地形研究中應用仍然有限。本研究利用1947年至1974年間的歷史航照，藉由SfM-MVS方法重建關山段的歷史數值地表模型與正射影像，以評估此方法應用於小尺度的活動斷層地形與活動性分析之可行性。 透過檢視成功與失敗的重建結果，本研究歸納出以歷史航照進行斷層地形分析時所面臨的主要限制，並提出具體的應用建議。本研究也提出一套可供解釋歷史地形的地表指標，經由分析電光階地於1947年至1974年間的地形變化，本研究認為此地區的斷層跡應位於現代河蝕崖地形更西側的位置。此外，根據研究結果，該階地地形在13年間（1961至1974年）大致未見顯著變化，顯示此段期間未有明顯的地表變形跡象。結合實地調查結果推測，縱谷斷層關山段可能並非潛移型斷層，而更傾向具震間鎖定特性的斷層行為。

The Longitudinal Valley fault system (LVFs) is one of the most active fault systems in Taiwan. The Chihshang fault segment, which extends from Yuli to Luyeh, has been identified as an active creeping fault with a high slip rate of approximately 20–30 mm/yr. However, historical earthquake records indicate that surface ruptures of the November 1951 earthquake series were only observed between Yuli and Chihshang, with no evidence of rupture between Chihshang and Luyeh in the Guanshan area (referred to as the Guanshan section). To date, limited research has focused on this section, partly due to significant modification of fault traces and topographic features by human activity and fluvial processes. Advances in photogrammetry now enable the reconstruction of historical landforms using historical aerial photographs. While numerous previous studies have successfully applied structure from motion (SfM) and multi-view stereo (MVS) techniques for large-scale geomorphic analyses, their application to small-scale fault deformation has remains limited. In this study, we utilized historical aerial photographs from 1947 to 1974 to reconstruct historical digital surface models and orthomosaics along the Guanshan section using the SfM-MVS method, with the goal of assessing the feasibility of this approach for analyzing fault-related landforms and activity at a smaller scale. Through the evaluation of both successful and unsuccessful modeling results, we identify key limitations and offer recommendations for the effective use of historical aerial photographs in active fault topographic analysis. We also propose a surface indicator useful for historical topographic interpretation using orthomosaics. An analysis of topographic changes along the Dianguang terrace between 1947 and 1974 suggests that the actual fault trace potentially located farther west. Furthermore, our results indicate that the topography of the Dianguang terrace remained largely unchanged during the 13-year period from 1961 to 1974, suggesting the absence of surface deformation. Combined with field observations, these findings imply that the Guanshan section of the LVFs may not exhibit creeping behavior and is more likely characterized by interseismic locking and strain accumulation.