極小尺度(hyper-local)邊坡之崩塌災害之風險評估為一相當重要之課題，由一些過去案例發現，儘管小尺度邊坡之崩塌面積很小，但仍可能致災性極高，特別是崩塌的影響範圍與地方居民高頻度活動的地區(如道路、民居等)重疊時。過去研究使物理型之崩塌模式進行災害預測或評估時，通常需要使用相當高空間解析度的地形參數。過去已有相關研究試驗了使用不同的空間解析度的數值高程模型(Digital Elevation Model, DEM)對崩塌模式產生的影響，並建議使用高解析度的DEM(5-m至10-m)可針對崩塌的發生位置有更準確的預測結果。但是，對於崩塌崩落物(如土石流)於邊坡運動的模擬應用上，DEM適用之解析度範圍仍值得探討。本研究針對印尼東爪哇省，瑪琅區本多薩里村的一個由降雨誘發之極小型之崩塌進行研究，方法為應用一物理型的崩塌-土石流整合模式進行試驗。此試驗為使用不同來源、不同解析度的DEM進行崩塌位置預測以及土石流模擬並進行比較，以了解DEM空間解析度對模式預測所造成之影響。 在本研究中使用了三種不同來源的DEM，包含(1)8公尺之印尼國家標準數值高程模型(8-m ArcDEM)，(2)使用即時動態全球定位系統測量生成的5公尺解析度之數值高程模型(5-m RtkDEM)，以及和(3)由無人飛航照產製之2公尺數值高程模型(2-m UavDEM)。在模式的應用上，本研究於現地採集土壤樣本並進行實驗室實驗以獲得模式所需之土壤物理參數，同時在當地雨量站收集歷史之日雨量記錄。研究結果發現，使用2-m UavDEM可以更準確地預測崩塌位置，而其他DEM則有較明顯的預測錯誤情形。另外，使用2-m UavDEM在模擬研究地點的土石流(崩塌影響範圍)也有較不錯的結果。本研究成果並顯示針對及小尺度邊坡進行邊坡穩定評估的重要性，對於提高當地社區對崩塌災害之認知、防治有其助益，使之能減少山坡社區之崩塌災害風險。 ;The risk assessment of landslide hazard over small (hyper-local) slopes is crucial because several cases have revealed that, a small landslide can be fatal despite its small size, especially when the affected area overlaps with the residential area. The prediction of the hyper-local landslide generally requires topographic parameters with high spatial resolution when the physically-based modeling approach is applied. Previous studies have examined different digital elevation models (DEMs) and suggested that finer resolution DEMs (5-m to 10-m) produced more accurate results for predicting the landslide initiation points. However, the suitable DEM resolution used for simulating the damaged area affected by landslide runout (or debris flow) is still questionable. Thus, the comparison of different resolutions of topographic datasets will be examined in this study, for simulating a rainfall-induced landslide and its affected area at a hyper-local slope in Bendosari village, Malang district, East-java, Indonesia. In this study, three different DEMs were used in an integrated landslides and debris flows modeling; they are (1) the 8-m Indonesian National Standard Digital Elevation Model (ArcDEM), (2) the 5-m DEM generated by real-time kinematic-global positioning system survey (RtkDEM) and (3) the 2-m DEM constructed by the unmanned aerial vehicle mapping (UAV). To perform the model, soil parameters were obtained from the laboratory test, while the daily rainfall records were collected at a local station. The result shows that the shallow landslide is more accurately predicted by the 2-m UavDEM, and it has less overestimation than the 5-m RtkDTM. For 8-m ArcDEM, it also predicts false landslides in the study site. The result indicates the 2-m UavDEM gives a more effective response for simulating the landslide and debris flow in the study site and also shows its significance to promote local community awareness on slope stability for reducing landslide disaster risks.
