Postseismic Surface Deformation Associated with the Mw 6.4, 24 February 2004 Al Hoceima (Morocco) Earthquake using Time Series Analysis of SAR Images

Get Complete Project Material File(s) Now! »

Characteristics of SAR images:

As the aircraft / spacecraft platform carrying the SAR instrument moves, the beam footprint scanned along the cross-track direction, and pulses or radar waveforms are transmitted at the pulse repetition frequency in the cross-track and azimuth dimension (Figure 1). As a consequence, objects that are close to the antenna (i.e. near range) return an echo before those that are farther away. Pulse repetition frequency (PRF) used to acquire a series of backscatter each of the beam footprints along the cross track scan length. For each scan, a two-dimensional image is constructed with pixels in the along-track direction resolved by range gating and pixels in the cross-track direction resolved by the aperture size; the PRF rate and raw data recorded by SAR system is focused to form an image. Using 2-D matched pulse compression in range and azimuth directions after merging echoes in the flight direction synthesizes a large antenna aperture.Bandwidth in flight direction contains a Doppler effect associated with the variation in the sensor movement relative to Earth (Hanssen, 2001).
In SAR image, each pixel has both amplitude and phase values; the amplitude value record the radar brightness “speckle effect” in range direction, the backscaterred energy of the pulse informs about the radar contribution of that pixel (figure 2). Radar mapping using SAR signal processing enables us to remotely image a target respecting a high resolution reaching meters. The first processing step for SAR is to produce the single look complex (SLC) image of the raw image represented in 2-D array containing complex numbers recording the brightness and the phase each scatterers on the ground respecting the azimuth and the range direction of the SAR-platform.

Conventional InSAR technique limits

The differential InSAR technique are the more adopted remote sensing method used by the geophysical community to study earthquake defofrmation. This is mainly due to their applicability to different geophysical measurements going from small-scale to large-scale displacement with sub-centimeter accuracy. D-InSAR has been applied for measuring ground deformation due to volcanic (Hooper et al., 2012), co-seismic (Cakir et al., 2006; Akoglu et al., 2006), post-seismic slip (Ryder et al., 2007; Cakir et al., 2003), movement of glaciers (Gourmelen et al., 2011), fault creeping (Cakir et al., 2005). The measured deformation using D-InSAR is related to the quality of the signal correlation that can be limited by changes in scatterers property with time. Temporal decorrelation (Zebker et al., 1992) due to the dielectrically changes in the scatterers property (vegetation, water, or other phenomena) limits the use of conventional InSAR to regions having a waterless condition as in desertic areas (see also the Haoud Berkaoui case-study in Chapter 3,). In addition, interferometric phase decorrelation has another source due to the geometry of the SAR acquisitions known as the difference in incidence angles that have a big impact with large baseline separating the two SAR images (Zebker et al., 1992); this is also true with changes in the squint angle of the spacecraft that cause change in the SAR Doppler frequencies. The first step to remove or avoid this deccorelations factors, we use filtering methods and condition to imitate and help in the choice of the suitable SAR pairs to be processed respecting a small baseline and also short time∅ in order to reduce the temporal and geometric deccorelation effects. The atmospheric delay ௧ ǡο௧ is one of the most difficult signal contribution to be estimated. The signal signature can be seen as a variation in the atmospheric conditions involving a variation in the signal propagation after and before cloud penetration. This variation is compared and very soon correlated with the topographic contribution. For this reason, as the time separating the two (or more) SAR image acquisitions is important (more that one month), the time uncorrelation due to the atmospheric delay is more important and difficult to be integrally removed using the conventional D-InSAR processing system. In order to enhance the interferogram results and reduce the temporal uncorrelation, new techniques are applied based on the use of interferograms stacking and their combined information. This method is commonly called Multi-Temporal InSAR methods (MT-InSAR).

Multi-temporal InSAR

Conventional InSAR processing involve decorrelation of the interferometric signal caused by the difference in orbital position between the two SAR acquisitions, topography changes, atmospheric delay (ionosphere), atmospheric condition changes. In addition, the changes in dielectrical characteristics of scatterers favor the presence of temporal and geometric (baseline) decorrelations presented in conventional D-InSAR interferograms. Using DELFT ∅ orbital precision data (Sharoo and Visser, 1998) we can estimate and remove the orbital error ǡο . The atmospheric and topographic errors are correlated and to attenuate these decorrelations phenomena the Multi-temporal processing also known as time series analysis aims to reduce the signal decorrelation and residuals with the processing of multiple SAR images over the same track in order to get the best signal-to-noise ratio and then obtain the best phase signal correlation matrix. Nowadays, there are two MT-InSAR classes: Persistent Scatterer (PS) and Small BASline (SBAS) methods. Each of these two methods gives temporal solution to the uncorrelated phenomena based on the principle of dominant reflectivity of centers of permanent scatterers. MT-InSAR use one set of images (for phase values) and the reflected signal from each element in the ground (depending on it is resolution) is obtained with the sum of all individual wavelets reflected by the scattering centers of each elements (See Figure 5).

Permanent Scatterers (Persistent Scatterers, PS)

Persistent scatterer InSAR (PS) (Ferretti et al., 2000, 2001; Kampes, 2005; Hooper et al.,2004; van der Kooij et al., 2006) marks the second generation of InSAR processing systems that belong to the family of time series analysis and MT-InSAR techniques. PS was developed to give solutions to the conventional InSAR decorrelation especially those due to temporal and geometric effects (Ferretti et al., 2001). The errors due to spatial and temporal variations cause the temporal and spatial phase errors producing disconnected areas in space and time (Hooper, 2006). This makes the interpretation of geodetic measurements from interferograms difficult and sometimes ambiguous by reducing the signal-to-noise ratio (SNR). Man-made structures can be used as corner reflectors and reference to calibrate the InSAR system due to their coherent phase center. The principle of permanent reflector method is to identify points (scatterers) from a series of InSAR images that preserve their coherence (in term of signal backscattered) in time and space. These scatterers will form a connected network from their phase measurements (Hooper, 2006). The interferometric phase distribution ranging between [π -π], to improve the phase signal, one of the existing solutions is to choose the dominant element from the interferometric phase by statistical estimations from the brightness of each reflector and the dominant signal with its corresponding dominant scatterer. This method guarantees that the obtained signal is less  affected by the decorrelations phenomena using the one signal of the dominant scatterer for eachpixel to perform the time series evolutions of the studied phenomena. The chosen pixels have invariance in their amplitude values and/or their phase values over a set of PS SAR images. Note that the variance correlates with the brightness of the non-selected scatterers of the same pixel. In fact, as their brightness raises, the variance of the pixel grows. The signal-to-clutter ratio (SCR) is obtained in this case by measuring the ratio between the echo of the dominant scatterer and to ∅ the sum of echoes from the other reflectors of the same pixel. The aim of using the PS method is to separate the noise signals and extract the signal of deformation ( ௗ ǡο௧ ) from formula 1. In order to remove from the interferograms the phase contribution due to the topography and the imaging geometry ௧ ǡο௧ , PS use the flattening correction. The principle is to suppose that all scatterers are in the WGS-84 projection system of the digital elevation model (DEM) in radar coordinate and calculate the topographic contribution and then remove it. Most common way is to use the SRTM data (Farr et al., 2007). Following Kampes (2006), the DEM error is estimated from the phase signal using: ∅௧ ǡο௧ ൌ ఒ Ʌ గ ୧୬ (5).
Where: r is the distance from the satellite to the pixel on the ground. Ʌis the look angle of the satellite. is the perpendicular baseline. Where r is the distance from the satellite to the pixel on the ground, Ʌ is the look angle of the satellite, and is the perpendicular baseline.

READ  Basic Concepts of Machine Learning

Merged Permanent Scatterers and Small Baseline

Since the appearance of MT-InSAR methods (PS and SBAS), several software solutions have been developed. The Stanford Method for Persistent Scatterers (StaMPS) (Hooper, 2008) is the only software package that incorporates both PS and SBAS algorithms. In addition, StaMPS/MTI proposes a combined method from both of PS and SBAS results. StaMPS uses i) ROI-pac (Rosen et al., 2004) in order to focus the raw SAR data, ii) Doris (Kampes., 2006) to perform interferograms from SLC images and iii) SNAPHU (Statistical-Cost, Network-Flow Algorithm for Phase Unwrapping ) (Hooper et al., 2007) for 3D unwrapping. StaMPS PS algorithm use the information from SAR amplitude dispersion (Ferreti et al., 2001) with threshold value of ~ 0.4 in order to select the candidate scatterers. According to (Hooper, 2008) the pixels of filtered phase that exhibit a small decorrelation in short time and are denoted as slowly-decorrelating filtered phase (SDFP), are selected by the SBAS method. The merged method selects pixel from the PS and SDFP groups that are in coincidence to improve the spatial sampling of the signal. The used PS algorithm from Hooper (2007) requires a minimum constraint of five SAR data. In order to have the possibility to process large areas, Hooper (2007) proposes to divide the whole area into small patches (in azimuth and in range) and process InSAR for each of them separately. The final map will be reconstituted from the processed patches.

The collapse of OKN32 and OKN32bis wells and geologic context

The Haoud Bekaoui is an active oil field extraction located 30 km southwest of Ouargla city, northeast of the Algerian Sahara desert (Fig. 1). Several wells were installed in the region to exploit oil natural resources, and among them the drilling of OKN32 in January 1978 (Fig. 2). The Haoud Berkaoui oil field is situated on a monoclinal structure of outcropping and gently west dipping Upper Cretacious units (Figs. 3 and 4). The upper 2000 m stratigraphic log shows from bottom (Fig. 4): 1) 1400-m-thick succession of clay and limestone units from Neocomian to Turonian, 2) 600-m-thick Senonian units made of ~200-m-thick saline deposits overlaid by ~250-m-thick of anhydrite covered by ~150-m-thick carbonate, and 3) ~40-m-thick Mio-Pliocene sandy-clay (Sonatrach, 2004). In February 1978, OKN32 oil well has accidentally exposed aquifer eruptions that belong to the well-known continental sandstone deposit (Albo-Barremian). This drilling has been monitored afterward by the Sonatrach petroleum company that tried repeatedly to stabilize the collapse and stop the spreading of the subsidence area. Indeed, soon after the drilling the two oil fields Okn32 and Okn32bis, 80 m distance from each other, were badly damaged in this incident, causing an important environmental and ecologic impact in the region.
The oil well collapse formed a cavity crater or sinkhole showing 80-m-depth and 320 m in diameter that expands from year to year (Fig. 5). It is surrounded by 0.25 to 0.30 m wide and ~50-m-deep cracks visible at the surface. According to previous studies made by Sonatrach (unpublished reports, Hocine et al., 2008),the collapse of oil wells is caused mainly by the interaction between the ~200-m-thick lower Senonian salt layersand pressurized (50°C) Albo-Barremian fossil water (Akretche et al., 1992).An important collapse episode of OKN32 oil drilling occurred in October 1986,despite many attempts since 1978to stabilize the collapse by tubing to protect against aquifer eruptions and pollution. Fossil water rises through the~200-m-thicksalt layer and to the surface at 300 m3/hr and a temperature of over 50°C at a speed of 1 to 1.5 m/sec. The hot and pressurized water that induces the gradual dissolution of salt units is at the origin of an expanding large cavity located at 450 to 650-m-depth (Fig. 4). In May 1991, a concentric and meter-wide crack system was formed at 600 m from the crater center. Hence, the underground collapse chamber may increase from 300-400 m to 600 m in diameter or larger if the dissolution process will continue in the future (Fig. 4).

Table of contents :

I. Summary 
II. Résumé 
III. Introduction 
Chapter I: Basic Principles of Conventional InSAR and New InSAR Processing Generations PS-InSAR, Small Baseline and Merged Methods
1. Introduction
2.1. Synthetic Aperture Radar Interferometry: Methods and techniques
2.1.1. Principles of SAR image acquisition:
2.1.2. Characteristics of SAR images:
2.2. Conventional InSAR technique limit
2.3. Multi-temporal InSAR
2.3.1. Permanent Scatterers (Persistent Scatterers, PS)
2.3.2. Small Baseline (SBAS)
2.3.3. Merged Permanent Scatterers and Small Baseline
Chapter II: Monitoring Ground Deformation in the Haoud Berkaoui Oil Field (Sahara, Algeria) Using Time Series Analysis of SAR Images
Summary
1. Introduction
2. The collapse of OKN32 and OKN32bis wells and geologic context
3. InSAR data and processing
3.1. Differential InSAR results
3.2. MT-InSAR processing and results
4. Modeling
5. Discusion and Conslusion
Acknowledgments
References
Chapter III: Monitoring Landslide in the Urban Area of Constantine (Northeast Algeria) using Advanced Merged PS-SB Insar Analysis
1. Introduction
2. Constantine Landslides Description
3. Constantine Geology
4. InSAR analyses
5. Conclusion
Acknowledgment
References
Chapter IV: Postseismic (Interseismic) Deformation in the El Asnam Fault Region (Algeria): Results From Merged PS-InSAR and Small Baseline Methods
1. Introduction
2. Tectonic setting
3. InSAR Analyses
3.1 Chelif basin-scale deformation using time series
4. Modeling
5. Conclusion
References
Chapter V: Postseismic Surface Deformation Associated with the Mw 6.4, 24 February 2004 Al Hoceima (Morocco) Earthquake using Time Series Analysis of SAR Images
1. Introduction
2. InSAR Analysis
3. Modeling
4. Disussion and conclusion
References
General conclusion
List of All references
Appendex  GMTSAR Processing: Application to the Surface Deformation of A Moderate Mw 5.7 Seismic Event (1999 Ain Temouchent Earthquake, Northwest Algeria)
1. Introduction
2. Seismotectonic Context
2. GMTSAR systems
3. InSAR data analysis
4.1. A new program for automatic fringe counting
4.2. A new Graphical User Interface (GUI) for GMTSAR
5. Discussion
References

GET THE COMPLETE PROJECT

Related Posts