| 摘要: | 自從雷達差分干涉影像成功的監測Landers地震的同震地表變形位移量之後，雷達差分干涉技術(DInSAR)已經被廣泛的應用在許多地球科學的領域上。近20年來，許多理論與演算法的精進已突破傳統雷達差分干涉的限制，如永久散射體差分干涉技術(Persistent Scatterer InSAR, PSI)、短基線長子集演算法(Small Baseline Subset, SBAS)和多時雷達差分干涉技術(Multi-Temporal InSAR, MTI)，並且成功的運用火山監測、冰河流動、新期構造活動和自然災害等議題上。然而許多地質議題仍因為其各自的環境因素而造成此技術在應用上的限制與問題，因此本研究針對三個不同的地質危害進行不同角度的探討，並且利用地質統計的方法，結合雷達差分干涉技術的結果和其他測地資料與監測資料，提升InSAR結果的精確性與合適性。
第二章是探討位於台灣中西部的濁水溪沖積扇，該地自1970年以來即遭受到抽水引發的地層下陷問題。藉由共同克利金(cokriging)演算法整合1993年到2008年的InSAR和水準測量結果，提升空間和時間上的監測解析度，並以連續GPS站的資料作為驗證。研究結果顯示共同克利金演算法可以提升InSAR的精確度並保留空間上較好解析度的特性，同時發現濁水溪沖積扇的地層下陷中心從沿海往內陸移動的時間分界點位於1998年，而地層下陷線速率自2005至2008年有逐漸減緩的趨勢。
第三章研究是探討台南市區因為美濃地震所引起的災害問題。將地震前後的同調性(Coherence)影像對進行分析，透過統計方法處理地震前30幅升降軌道的同調性影像，再搭配地表物隨時間的去相關特性，建立兩步驟(two-step)門檻值方法，挑選出因受到地震而造成同調性差異的像素(coseismic coherence difference, CCD)，而這些像素可製作因地震造成災害的位置分佈圖，甚至包含建築物損壞與結構破裂的位置。此外，解算同震位移升降軌道的相位資訊所獲得的二維方向（垂直與東西方向）位移量，與現地調查量測的結果相互吻合，並且顯示在關廟市區的同震地表位移有受到關廟向斜的影響，造成關廟市區南北相反的地表活動狀況。
第四章研究是探討義大利Apennine山區的坡地潛移活動問題。位於Caramanico內的坡地潛移活動都曾受到場址效應的地震事件、強降雨的事件和融雪的泉水事件所誘發，因此透過PSI技術分析自2015年至2018年的48幅干涉影像對，以獲得衛星視角(line-of-sigh, LOS)方向的時間序列變化結果。當發生地震事件時，位於曾發生過坡地滑移位置上的PS點位，會顯示LOS方向伸長的變化資訊。透過MATLAB®軟體的detrend函數消除趨勢後的隨時間變異結果，可以發現地震事件對場址放大的效應在Caramanico比在Lettomanoppelo的影響還顯著。此外，副影像的大氣效應影響在Caramanico和Lettomanoppelo都非常類似，但是仍與每月降雨和消除趨勢後的隨時間變異結果無法吻合，未來仍須更詳細的大氣資訊、更精確的紀錄資料與更高頻率的監測結果，才能有效分析各個因素對消除趨勢隨時間變異結果的影響。;Since the first radar interferometry of Landers earthquake was successful to estimate the surface deformation, the interferometric synthetic aperture radar (InSAR) technique have been applied widely in earth science. In recent 20 years, the improved algorithms and theories, such as Permanent Scatterers™, Small Baseline Subset (SBAS), Persistent Scatterer InSAR (PSI), and Multi-Temporal InSAR (MTI), have overcame the InSAR limitations and prospered in many geoscience fields, such as volcanology, glaciology, neotectonic activities, and natural hazards. However, every geologic issue still existed the limitations and problems in its environment. For upgrading the accuracy and appropriateness of InSAR data in different geologic issues, we combine the InSAR results with other geodetic or monitoring data with the geostatistics method and focus on three different types of geologic hazards in different orientation analysis.
The chapter 2 is the land subsidence issue due to groundwater withdraw in the Choushui River Fluvial Plain (CRFP) of western central Taiwan. Exploiting various observational techniques this study employed a geostatistical cokriging algorithm to integrate multiple types of observations for mapping land subsidence in the CRFP between 1993 and 2008 to effectively reduce the regional effects and interpolation biases in InSAR observations. Precise leveling data and persistent scatterer InSAR results were first assessed through variogram analysis, the results of which revealed similar directional variograms and no nugget effect for the experimental variograms. The accuracy of a cokriged land subsidence map was verified using continuous GPS data at the same cell locations and was significantly improved compared to when the precise leveling data were used. The cokriging interpolations of land subsidence in the CRFP indicated that the severe subsidence areas had moved from the coastal area to the central CRFP. An obvious critical subsidence migration point for the southern CRFP occurred in 1998. Overall subsidence rate trends increased in the early periods (1993–1998 and 1998–2003) and decreased in the most recent period (2005–2008).
The chapter 3 is the Meinong earthquake-induced disasters in urban area. This study proposes a workflow that enables the accurate identification of earthquake-induced damage zones by using coherence image pairs before and after an earthquake event. The workflow uses InSAR processing to account for coherence variations between coseismic and preseismic image pairs. The coherence difference between two image pairs is useful information to detect specific disasters in a regional-scale area after an earthquake event. To remove background effects such as atmospheric effect and ordinal surface changes, this study employs the two-step threshold method to develop the coseismic coherence difference (CCD) map for our analyses. 34 Sentinel-1 images between January 2015 and February 2016 were collected to process 30 precoseismic image pairs and 2 coseismic image pairs for assessing multiple types of disasters in the Tainan City of southwestern Taiwan, where severe damages were observed after the Meinong earthquake event. The coseismic unwrapping phase results were further calculated to estimate the surface displacement in east-west and vertical directions. Results in the CCD map agree well with the observations from post-earthquake field surveys. The workflow can accurately identify earthquake-induced land subsidence and surface displacements, even for areas with insufficient geological data or for areas that had been excluded from the liquefaction potential map. In addition, the CCD details the distribution of building damages and structure failures, which might be useful information for emergency actions applied to regional-scale problems. The inversion of 2D surface displacement reveals complex behavior of geological activities during the earthquake. In the foothill area of the Tainan City, the opposite surface displacements in local areas might be influenced by the axis activities of the Kuanmiao syncline.
The chapter 4 is the creeping landslide in the Apennine mountain of central Italy. Because the amplification effect of earthquake event, strong rainfall event, and the spring of melting snow could trigger the creeping behaviors of landslide in the Caramanico, this study processes 48 interferometric images between 2015 and 2018 with PSI technique to analysis the line-of-sight (LOS) time-series displacements in the landslide area. When occurring the earthquake events, the obvious drop down behaviors in time-series displacement results could be notable, especially in the area with previous landslide event. Exploiting the variation of LOS time-series results with the detrend function of MATLAB® software, the amplification factors of the earthquake events with magnitude larger than 4 were considerably higher than reference site in the Caramanico and the movement features could be the sliding directions of some potential landslide area. The site amplification effect of earthquake event influences the differences of de-trend results between Caramanico and Lettomanippello partially. The slave atmospheric effects in Caramanico and Lettomanoppello have similar pattern beside during March to November, 2016. However, the monthly rainfall data is hardly match with the slave atmospheric effects. The relationship between de-trend results, slave atmospheric effect, and monthly rainfall needs more detail atmospheric data, more precise recording data, and higher frequent observation data to analysis the weighting of each factor. |
