應用多時域雷達干涉技術與C及L波段雷達影像分析彰化地區地表變形

Abstract

彰化地區過去因過度抽取地下水，導致地層下陷情形嚴重，而高速鐵路正行經彰化縣溪洲鄉之沉陷中心，地層下陷將可能對行車安全造成影響，因此如何快速與精確監測下陷範圍與下陷速率為重要議題。多時域雷達干涉技術(Multi Temporal InSAR, MT-InSAR)技術為涉及多時期影像之雷達干涉處理方法，不僅可獲得大範圍地表變形資訊，同時測量精度與傳統水準測量方法相當。本研究採用三種不同之MT-InSAR技術：包括PSInSAR、StaMPS/MTI及TCPInSAR技術；兩種不同波段（C波段與L波段）的雷達影像，包括：2006年8月至2008年9月共20幅Envisat ASAR影像與2006年12月至2011年2月共19幅ALOS PALASR影像，綜合分析彰化地區共六種MT-InSAR案例之地表變形速率成果，並與水準資料進行比對。經分析成果顯示上述三種技術皆能有效應用於地層下陷監測，在Envisat影像方面，三種方法以TCPInSAR獲得之時域相關點之點位密度最高（約214點/平方公里），StaMPS/MTI與水準測量比對之均方根誤差最小（約0.68公分/年）。在ALOS影像方面，三種方法中以StaMPS/MTI方法獲得之永久散射體（Persistent Scatterer）點位密度最高(約371點/平方公里)，PSInSAR與水準測量比對之均方根誤差最小（約0.73公分/年）。

Extracting groundwater has caused large-scale land subsidence in Changhua County. The Taiwan High Speed Railway (THSR) passes through a serious subsidence area in Sijhou Township, where subsidence poses a serious threat to its operation. A major issue is to effectively monitor land subsidence. Multi-temporal InSAR (MT-InSAR) techniques, involving the processing of multiple acquisitions in time, can measure ground displacements to an accuracy comparable to that of precision leveling. This study uses three MT-InSAR methods, including PSInSAR, StaMPS/MTI and TCPInSAR, with 19 ALOS PALSAR acquisitions (L-band) from 2006 to 2011, and 20 Envisat ASAR acquisitions (C-band) from 2006 to 2011 to determine subsidence. We analyze subsidence rates in six cases (three methods multipled by two kinds of SAR images). We validate the InSAR results by leveling data in the Changhua region. This study demonstrates the effectiveness of PSInSAR, StaMPS/MTI and TCPInSAR in monitoring land subsidence. With the Envisat images, the density of the temporarily coherent points determined by TCPInSAR is the highest, and is about 241 pixel/km2. StaMPS/MTI yields vertical displacements matching the leveling result to 0.68 cm/year (RMS), which is the smallest discrepancy when Envisat images are used. With the ALOS images, StaMPS/MTI results in the highest point density in the persistent scatterers and is about 371 pixel/km2. The vertical rates from PSInSAR and the ALOS images agree with the rates from leveling to 0.73 cm/year (RMS).