台灣留學生出席國際會議補助

2009年6月18日 星期四

Multi-scale analysis of InSAR time series to estimate variations in topographically correlated propagation delays with application to the Makran Subduction Zone

論文發表人: 林玉儂(加州理工學院)

 

Many InSAR observations are plagued by propagation delays that correlate with topographic variations within a given scene.  These delays are frequently termed tropostatic delays and are assumed to result from temporal variations in horizontal stratification of the troposphere.  We present a robust approach to estimating tropostatic compensation coefficients (K values) that is relatively insensitive to confounding processes (e.g., earthquake deformation, phase ramps from orbit errors, etc). Our approach takes advantage of a multiscale perspective by adopting wavelet decomposition of both topography and observed phase.  By decomposing topography and observed phase in a given interferogram into several spatial scales, we determine the bands spanning different characteristic length scales wherein correlation between topography and phase is significant and stable.  Our approach also uses the inherent redundancy provided by multiple interferograms constructed with common scenes. We define a unique set of component time intervals, Tint, using a suit of interferometric pairs. The pair-based K values (Kpair) are then combined to estimate temporally consistent values for each time interval (Kint).   The Kint values are then recombined to make final values of Kscene in order to correct each interferogram. 

We are testing our approach in the region of the Makran subduction zone, located in western Pakistan and eastern Iran, within the influence zone of South Asian monsoon. We use twenty-nine ENVISAT images to develop the time series. Preliminary results find large variations in estimates of Kscene and Kint.  Generally, the tropostatic correction accounts for a relatively small portion for the phase observed, although significant effects are found for selected pairs.  The typically small impact of the tropostatic correction implies that in the future we must consider more complex dynamic atmospheric models.

 

差分合成孔徑雷達影像是用來做同震地表變形判釋的重要工具,然而用在震間低幅度的地表變形上,卻有很大的限制,因為低幅與低頻的震間變形行為所產生的訊號容易受到傳輸延遲的干擾,這種延遲與地形起伏呈共變現象,又被稱作對流層延遲,起因於對流層在垂直向分層的時間變化。本研究使用地形與相位間的線性相關模型,建立一個方法以估算其轉移函數K,此方法對於其他地質現象(構造變形、軌道誤差造成的大尺度相位傾斜)相對不敏感,因而能提供較可靠的K值。此多尺度分析法利用小波轉換將地形與差分雷達影像分成數個不同的空間尺度,其中某些特定尺度顯示出較高的地形相位共變關係,而且其所得到的K值比起其他尺度相對穩定,表示不受到其他因素的干擾。本研究同時利用多幅影像所提供的內在冗餘性來建立非重覆性的時間序列,首先我們定義一組非重覆的時間段落Tint,與其對應的轉移函數Kint,則以各差分合成孔徑雷達影像為基礎的轉移函數Kpair便是Kint的線性組合。解此線性系統便可得到Kint的時間序列,並可由其推測所以其他差分合成孔徑雷達影像的Kpair

我們將此一方法應用在巴基斯坦與伊朗邊界的莫克蘭隱沒帶,該地區曾於1945年發生規模8的強震,並在周圍印度洋海岸造成海嘯災害。我們希望透過傳輸延遲校正可以得到較為精確的震間變形量,進而了解目前在隱沒帶上的應力狀態。測試的結果發現,我們的方法可以在山區得到不錯的校正量,但是校正後影像仍舊呈現出不小的對流層訊號,推測應該與動態的對流層變化諸如局部對流或是背風波有關。這表示我們下一步需要考慮動態對流層變化的校正。