- 學位論文
應用多時域雷達干涉技術分析長周期與潛在之季節性地表變形
Applying the multitemporal SAR interferometry (MTInSAR) technique for detecting long period and potential seasonal surface displacement
摘要
長時間與季節性的地表變形可以呈現一個地區的地質概況、水文地質條件和潛在的災害風險。因此，本研究選擇受地層下陷影響的沖積扇地區和有潛在滑移跡象的邊坡做為研究區域。在人口稠密的沖積扇區，地下水資源的需求和地表沉降間之平衡是很重要的議題，本研究的第一部分將結合多時域雷達干涉技術(MTInSAR)、衛星導航系統、精密水準儀、地下水監測井和分層壓密監測井的數據，分析台灣濁水溪沖積扇內地表位移與地下水位變化的關係。由多項監測資料整合的時間序列分析成果，可以得到在以年為單位的時間尺度上，當孔隙水壓的影響占主導地位時，地下水位的上升會造成地表垂直抬升的現象。反之，當水體荷重的影響大於孔隙水壓時，地下水位變化和地表垂直變形將是負相關。透過了解地下水位變化與地表位移間的相互關係，將對地層下陷和地下水資源利用之間的平衡有很大幫助。另一方面，位於台灣北部華梵大學校區內的順向坡，自1990年以來持續觀測到邊坡有潛移的現象，由於現場測量設備缺乏持續的維護或監測資料不公開的情況，2018年後較難獲得可靠的監測數據，因此，在本研究的第二部分，應用多時域雷達干涉技術監測此潛移型山崩2014-2019的活動性。此技術所偵測到的潛移區域範圍接近現地監測與調查結果所定義之兩個滑動塊體的範圍。根據永久散射點(PS點)的時間序列，可以清楚地觀測到此邊坡的長期重力變形現象，此外，亦注意到在台灣地區較少被討論的短期季節性波動訊號。為了進一步了解所偵測到的潛移型山崩季節性波動訊號，本研究的第三部分同樣利用多時域雷達干涉技術針對台灣其他潛移型山崩進行2019-2021年間的活動性分析。從時間序列的分析成果得知，只有位於華梵大學(Shiding-T001)和清境農場(Renai-D057)的潛移型山崩具有季節性波動訊號，推測來源可能是由於垂直分層大氣效應造成之相位延遲或孔隙水壓變化造成之地表變形，亦或兩種影響皆有。本研究的成果顯示，多時域雷達干涉技術可以與現地監測資料互補不足之處，並洞察地表位移變化，呈現垂直地表位移與沖積扇地下水位變化之間的關係，以建議可以利用地下水資源之處，以及監測潛移型山崩的活動性，並進一步從時間序列的分析了解這些邊坡的運動特性，以提供後續在防災應變作業上之參考與應用。而山區的季節性波動訊號則需要在未來的研究中進一步討論與了解，以利潛移型山崩監測上的實際應用與發展。面對SAR影像的大數據時代來臨，機器學習或深度學習在SAR影像處理的應用也將是未來的一個重要課題。
並列摘要
Long period and seasonal surface deformation can represent the geological condition, hydrogeological issues and hazard potential of an area. In this study, we selected an alluvial fan affected by subsidence hazard and potential slow-moving landslides as study areas. In populated alluvial fan areas, balancing the mitigation of land subsidence and the demand of groundwater resources is a very important issue. The first part of this study attempts to combine data from MTInSAR, GNSS, precise leveling, groundwater monitoring wells and multi-layer compaction monitoring wells to analyze the relationship between groundwater level change and surface displacement within the alluvial fan of the Choshui River in Taiwan. In a yearly time scale, the positive or negative relationship between groundwater level change and surface displacement is depend on the combined effect of pore water pressure and water-mass loading at different location of the alluvial fan. Thus, understanding the combined effect of pore water pressure and water-mass loading is very helpful for management of land subsidence and usage of groundwater resources in an alluvial fan. On the other hand, a slow-moving landslide on the Huafan University campus in northern Taiwan has been observed since 1990. However, the reliable monitoring data are difficult to acquire after 2018 due to the lack of continuous maintenance of the field measurement equipment or the data not in the public domain. In the second part of this study, the MTInSAR technique is applied to monitor this slow-moving landslide from 2014 to 2019. The slow-moving areas detected by the PS pixels coincide with the two sliding block areas, which are defined by in situ monitoring data and field survey. A long period surface deformation of the slow-moving landslide can be clearly observed by the time series of the PS pixels. Moreover, a short period seasonal fluctuation signals of the slow-moving landslide, which may present the vertical seasonal surface displacement, can also be detected in this study. In order to further explore the seasonal fluctuation signals of the slow-moving landslide, the third part of this study investigates the movement characteristics of other potential slow-moving landslides during 2019 to 2021 by the MTInSAR technique. The time series show that only the slow-moving landslides of the Huafan University campus (Shiding-T001) and the Qingjing Farm (Renai-D057) present the seasonal fluctuation signals, which may due to the delay of the vertical stratigraphic atmospheric effects, or the vertical surface displacement cause by the pore water pressure, or the both. From the results of this study, the MTInSAR technique can complement the lack of in situ monitoring data and provide further insight into land surface changes. The relationship between the vertical surface displacement and the groundwater level change of the alluvial fan can be presented with more detailed surface information data. The activity of the slow-moving landslides can be monitored to further understand the movement characteristics of these areas, which could provide important information for landslide hazard prevention and emergency response. However, the seasonal fluctuation signals in mountainous areas should be clarified in the next study for the widely application of monitoring slow-moving landslides by MTInSAR technique. The applications of machine learning or deep learning for dealing with the big-data era of SAR images processing would also be an important issue in the future.