Applications: Earthquakes and Tectonics (2)

Geodetic inversions for slip versus depth for several major strike-slip ruptures (Mw > 7, e.g. 1992 Landers, 1999 Hector Mine, 2010 El_Mayor-Cucapah) show a 10% to 40% reduction in the shallow slip (< 2 km) with respect to the slip at depths of 5 to 8 km.  This has been called the shallow slip deficit (SSD).  The large magnitude of this deficit has been an enigma since it cannot be explained by shallow creep during the interseismic period or by surrounding earthquakes’ triggered slip.  One potential explanation for the SSD is that the previous geodetic inversions did not use displacement data close to the fault so the shallow portions of the slip models were poorly resolved and generally underestimated. In this study we improve the geodetic inversion, especially at shallow depth by:  1) refining the InSAR processing with non-boxcar phase filtering, model-dependent range corrections, more complete phase unwrapping by SNAPHU using a correlation mask and allowing a phase discontinuity along the rupture; 2) including near-fault offset data from optical imagery and InSAR azimuth offsets; 3) using an more detailed fault trace; 4) and using additional GPS survey data.  With these refined data, the slip inversion has much better resolution at shallow depth.  For the Landers rupture the SSD is reduced from 45% to just 18%.  Similarly for the Hector Mine rupture the SSD is reduced from 15% to 3% and for El Mayor Cucapah rupture it is from 60% to 11%. Recently we have acquired and processed InSAR data from two new sources - Sentinel-1a and ALOS-2.  At the meeting we will report on their data quality and processing methods using GMTSAR.

While pixel-offset measurements are less precise than InSAR observations, they can provide more reliable and more robust results in areas prone to decorrelation and have been successfully applied to several cases where InSAR failed. Moreover, pixel offsets can provide unambiguous ground displacements in both the line-of-sight (LOS) and the along-track (azimuth) directions, allowing us to derive the three-dimensional displacement field. Most SAR-offset applications have used the conventional offset-measurement method, which is based on cross-correlating regularly-spaced sub-images (windows) throughout the SAR images. This method does not take the variety of scattering characteristics on the ground into consideration.

Local pixel offsets are determined from the similarity between sub-images, similarity that comes from the coherent speckle patterns and the ground reflectivity. Even though a pair of SAR images is completely incoherent, similar reflectivity of ground targets may still provide useful information for estimating offsets. Here we show how SAR-offset measurements can be improved by considering the reflectivity and speckle features in SAR images, and by utilizing time-series of SAR images. We further develop the conventional pixel-offset method mainly in two ways: 1) we detect and enhance correlated amplitude targets, such as strong reflectors and coherent speckles, and 2), if applicable, we can also systematically suppress un-correlated features originating from incoherent speckles and low-coherence pixels. We demonstrate our novel offset-tracking techniques by applying them to four earthquakes:

1) The 2011 Van (Turkey) earthquake. In this case we show that by focusing the offset estimations on sub-images containing strong reflectors, detected with a 2D sinc-function template, the reliability and accuracy of SAR offsets can be significantly improved.

2) The 2013 Balochistan (Pakistan)earthquake. In this case we demonstrate that decorrelated ScanSAR images, even in non-interferometry modes, can still be useful for deriving 3D coseismic displacement fields.

3) The 2001 Bhuj (India) earthquake. In this case we show that by mapping the pre- and post-seismic reflectivity using time-series of SAR images, SAR offsets between ERS and ENVISAT datasets can be used to reveal, for the first time, the near-field coseismic deformation.

4) The 2005 Fukuoka (Japan) earthquake. In this case we show that by properly weighting the amplitude images using the averaged coherence map from time-series SAR images, we are able to retrieve small (smaller than 40 cm) displacements in the azimuth direction from a mid-resolution ENVISAT dataset.

These four examples demonstrate that the improved pixel-offset technique can be applied to measure ground deformation in many cases where we lack coherent SAR images. The proposed pixel-offset processing framework also allows for cross-sensor offset estimation, e.g. between ERS and Envisat images, which may provide new information for a number earthquakes for which we have limited data. 

Averaging long time series of InSAR images have allowed us to observe and measure small deformations of the Earth's surface associated with processes occurring at a very slow rate. Steady deformation processes such as the motion of tectonic plates across a plate boundary can be observed by combining image series spanning long time intervals. The study of transient processes however, requires dense temporal sampling of the deformation to allow mitigating errors in the signal over shorter periods of time. Here we compare two studies of the Asal Rift in the Republic of Djibouti using two different data sets: The first study used a 10-year long series of 24-day spaced, C-band data from the Canadian Space Agency Radarsat-1 mission. The second study uses a 1.6-year long series of X-band data acquired by the COSMO-Skymed (CSK) constellation of satellites and provided by the Italian Space Agency. The return periods of the CSK spacecrafts go in sequences of 1, 3, 4, and 8 days, making a dense temporal series of 120 acquisitions over the 2012-2013 time periods.
Over the 1997-2008 decade, the first study combining data from ascending and descending passes of the spacecraft revealed that the Asal rift was accommodating 15 mm/yr of horizontal opening associated with uplift over a broad region (~40 km) and localized subsidence of the rift inner floor at a rate of 6 mm/yr. Most of the deformation appears to be accommodated by 2 of the rift-bounding faults along the North (Gamma) and South (E) shoulders of the rift. A detailed observation of the data revealed that the movement on the faults was characterized by periods of slow creep interrupted by events of accelerated movement resulting in millimeter-scale increments of slip on faults, coeval with bursts of microseisms under the Fieale volcano in the middle of the rift.
The analysis of the CSK data is in progress. Preliminary results of the descending pass time series indicate that the Northern bounding fault (Gamma) is active during the 2012-2013 period but that movement seems to have died on the Southern fault (E). The rate of subsidence of the inner rift with respect to the northern shoulder is estimated to be unchanged compared to the previous decade covered with Radarsat-1 data. The high spatial resolution of the CSK data shows the complex slip transfer between individual segments composing the northern fault system. However, the shorter duration of the CSK series prevents us from identifying any clear event of accelerated slip on the bounding faults during the 1.5-year observation period. This is mostly due to a residual seasonal signal that could not be effectively filtered from the short data series spanning only 1.5 season cycles. CSK high temporal sampling rate potentially offers the means to detect transient processes of short duration provided that atmospheric signal can be effectively mitigated. Further results will be presented at the meeting.

Extension in the Red Sea gradually increases from about 7 mm/year in the northwest to 16 mm/year in the southeast. To the west the Arabian plate is bounded by the Gulf of Aqaba transtensional basin and the Dead Sea transform fault, whereas in the southeast by the Gulf of Aden and the Owen’s Fracture Zone. Earthquakes in the region occur primarily on the main plate boundaries. However, some damaging earthquakes have taken place historically and in recent past in western Arabia, indicating that active faults exist within the Arabian plate and that the extension is not entirely focused at the middle of the Red Sea. The 2004 Tabuk earthquake (magnitude 5.1) is an example of such an intra-plate earthquake, occurring on the Red Sea flank in northwestern Saudi Arabia, over 200 km away from the center of the Red Sea rift. The earthquake could not be well located using seismic data and reported locations range up to ~20 km from each other. By using InSAR imaging and Bayesian estimation we constrain its location and source parameters and associate it with a likely source fault, improving the seismic hazard information near the city of Tabuk. Another earthquake, the largest to occur in the Red Sea region for several decades (Mw 7.2), took place ~250 km northwest of Tabuk in 22th November 1995 within the Gulf of Aqaba. While this earthquake has been studied before by using InSAR, we have analyzed previously unused InSAR data and used newer data processing methods, as well as field observations, to improve earlier fault-model estimations. The Red Sea extension is also marked by magmatic activity, including the 2009 Harrat Lunayyir dike intrusion on the Red Sea flank in northwestern Saudi Arabia, about 300 km southeast of the Tabuk earthquake. Three volcanic eruptions have also taken place in the southern Red Sea during the past several years, on Jebel at Tair Island (2007-8) and within the Zubair archipelago (2011-12 and 2013). To study this activity, we use InSAR and optical images to constrain the feeder-dike geometry and the associated stress directions. On Jebel at Tair we find evidence for temporarily varying stress orientations that appear isolated from the regional Red Sea stress field. The two eruptions in the Zubair archipelago were surtseyan and produced two new islands. The islands were formed entirely from explosive phreatomagmatic activity, as the eruptions did not last long enough to progress to an effusive eruption. All the InSAR data of magmatic activity in the region indicate that magma is transported directly from the mantle to the surface without significant resident time in the crust. In addition, the reawakened volcanic activity in the southern Red Sea comes after more than century-long quiescence and seems to be a part of the recent increase in activity in the region near the Afar triple junction, following the onset of the Dabbahu (Afar) rifting episode in 2005.

The Sept. 24, 2013 Mw 7.7 Awaran, Pakistan earthquake ruptured a fault of ~214 km long. The surface deformation detection using imagery from the newly launched Lansat-8 satellite by sub-pixel offset measurements, reveals a rarely smooth-curved rupture in the Makaran region. The surface deformation distributes along a simplex rupture, without any branches, and significant displacements occurred at the west of the fault, indicating strong shallow slip on a west dipping fault.

Our modeling with the Landsat-8 deformation data and the fixed fault location at surface, shows a gentle-variable west-dipping fault, with relatively smooth slip-distribution and obvious rake variations, rather than the pure strike-slip motion on a shallow west-dipping fault as inferred by others. Though the left-lateral strike-slip dominates the rupture, the thrust motion is also clearly concentrating on 4 areas. This slipping feature is confirmed by calculating the ratio of horizontal displacement components, and comparing it with model prediction. Unlike the shallow dipping angles (e.g. 45° from the USGS teleseismic solution), we found they might be in a range of 50°~70°along the rupture, which is normal for a strike-slip dominated fault, and is also consistent with the tectonic environment of the accretionary wedge in the Makaran region. The slip solution can also be tested against teleseismic body waves.

We keep monitoring the post-seismic motion of the seismogenic fault of the Awaran earthquake immediately after the earthquake using the Landsat-8 imagery and the TerraSAR-X data in the middle section of the fault. A series of post-seismic deformation images are obtained using the Landsat-8 data and the sub-pixel displacement measurements. Though the resolution of the data is as large as 15meters, we still can identify some interesting signals across the fault. However, it is impossible to do geophysical modeling using these data for a quantitatively analysis. We asked DLR for some data acquisitions covering the middle section of the fault after the earthquake using its high resolution TerraSAR-X radar satellite. There are 5 temporal acquisitions across the middle part of the fault and 1 acquisition at the north end of the fault. The data show high coherence in this period and it is possible for us to get clear post-seismic motion signals in the key portion of the fault. Depending on these data, we can do some geophysical modeling of the post-seismic motion in details. We expect for further monitoring of the post-seismic motion along the Awaran earthquake fault in the following years using both TerraSAR-X data, and the new lunched Sentinel-1A data in the near future. These data will contribution to the rehology studies of the accretionary Wedge of Pakistan.