Heterogeneous interseismic coupling along the Xianshuihe-Xiaojiang fault system, eastern Tibet

Get Complete Project Material File(s) Now! ».

GPS measurements and data processing

In Chapter 1, I have introduced the network architecture of the CMONOC and the GPS data collection. Here, I describe the GPS measurements along the Altyn Tagh fault and the Haiyuan fault carried out by our group in the Institute of Geology, China Earthquake Administration. The locations of GPS stations are shown in Figure 2.1.
Figure 2.1. GPS sites distribution. (a) Distribution of GPS stations in northern Tibet. (b) and (c) show GPS stations along the Altyn Tagh fault and the Haiyuan fault system. White and red triangles show the location of campaign-mode and continuous GPS stations of CMONOC respectively. Blue triangles represent the campaign-mode stations that we re-measured in 2017 across the Altyn Tagh fault. Green and Cyan triangles indicate the location of campaign-mode and continuous GPS stations that we deployed across the Haiyuan fault.
Our team revisited the same 17 benchmarks used by He et al. (2013) in October 2017. These benchmarks were measured 2–3 times in the period 2009–2011, across the western segment of the Altyn Tagh fault (86°E). These GPS stations were deployed roughly perpendicular to the Altyn Tagh fault, extending about a 400 km long profile across the fault. These GPS benchmarks were installed with forced centering system (Figure 2.2a). Of the 17 stations, 16 are well preserved; only one station near Qiemo County (north of the Altyn Tagh fault) was damaged. We used Trimble NetR9 receiver with Trimble GNSS Choke Ring Geodetic Antenna and collected continuous observations for 72 to 240 hours at the GPS sites at a 30s sampling rate. The collected GPS data, together with CMONOC GPS data in 1999-2017, constitute the dataset that I am going to use in Chapter 3. Along the Haiyuan fault, our team started to deploy campaign-mode GPS sites since 2013. In 2015, 21 new stations were established in order to compensate the low spatial resolution of CMONOC in the near field of Haiyuan fault (Figure 2.1). The construction of GPS station follows the standard of CMONOC. The base of GPS stations has a depth of at least 70 cm on the bedrock and a depth of at least 2 m in soil layer. Each station was equipped with forced centering system. We measured these GPS stations annually during the summer in order to reduce the effect of annual and semiannual components when estimating velocities. We used Trimble NetR9 and Topcon NetG5 receivers with choke ring antennas to perform continuous observations at 30s sampling rate, with at least 72 hours per site during each campaign. As of September 2019, we accumulated 5–7 campaigns of GPS data. In my work, I only used the GPS data spanning the 2013–2017 period. Figure 2.2. Pictures of GPS stations. (a) The typical campaign-mode GPS station across the Altyn Tagh fault. (b) The typical campaign-mode GPS station across the Haiyuan fault. (c) The continuous GPS station across the Haiyuan fault.
Cavalié et al. (2008) first identified a shallow creeping segment along the Laohushan segment (~103.8°E), Haiyuan fault, using the ERS InSAR spanning 1993–1998, which was later confirmed by Jolivet et al. (2012) using the Envisat InSAR spanning the 2003–2009 period. Furthermore, Jolivet et al. (2012, 2013) found that the shallow creep extend a 35 km-long segment, with transient slipping behaviors. However, there have been no GPS measurements across the segment. Consequently, our group deployed four continuous GPS stations in the near-field of the fault and normal to the fault in September 2017 (Figure 2.1, 2.3). These GPS stations were constructed on bedrock, equipped with Trimble NetR9 receiver and Trimble GNSS Choke Ring Geodetic Antenna (Figure 2.2c). The GPS sites started to operate around October 2017, and at the time of my PhD, since only a year of data were available, I did not include these GPS stations to the work presented in Chapter 4. Now, we have accumulated around two years of data, and I show the position time series of the four GPS stations in Figure 2.3. It is obvious that across the Laohushan fault, GPS sites velocities show variation, both in horizontal and vertical. I leave the interpretations of those four GPS sites for future work.

GPS data processing

Here, I take the GPS data surveyed across the western Altyn Tagh fault (86°E) as example to illustrate data processing. In the following Chapters, I will also introduce briefly the GPS data used in each study, the data sources, and corresponding data processing or integration.
I processed the campaign-mode GPS data measured in 2017 together with those data collected between 2009 and 2011, and with seven regional CMONOC continuous sites using the GAMIT/GLOBK 10.6 software (Herring et al., 2016), which consist of three steps:
(1) Loosely constrained daily solutions from GAMIT. Raw phase observations are first processed together with 40 globally distributed sites from the International Global Navigation Satellite System Service (IGS, http://www.igs.org/) to obtain loosely constrained daily station coordinates and satellite orbit parameters. The processing includes IGS precise orbits and Earth orientation parameters, absolute antenna phase center calibrations, and the Finite Element Solution 2004 ocean tide loading model during the daily data processing (Lyard et al., 2006).
(2) Position time series. Loosely constrained daily solutions are then expressed in the International Terrestrial Reference System 2008 (ITRF2008) by estimating seven parameters transformation (orientation, translation, and scale) using the 40 globally distributed IGS stations. The time series for each station is carefully examined to exclude outliers. I set 3-sigma from the mean as the threshold for outlier selection. I show one example of position time series in Figure 2.3.
(3) GPS site velocity and its uncertainty. Currently, there are two methods to estimate the GPS site velocity and its uncertainty. The first one uses position time series to simultaneously estimate linear terms (including offsets and slopes), nonlinear terms (annual and semiannual), and the amplitude and spectral index of a time-correlated (or power law) noise model, such as the Maximum Likelihood Estimation implemented in the CATS software (Williams, 2008). Such method fits flicker and white noises to all time series to estimate the slope (velocities) and their realistic uncertainties. This method works well with continuous and clean time series but has drawbacks (Saria et al., 2013). First, it requires at least 4 years of continuous data collection to provide reliable spectral index estimates. Second, episodic time series or sites with abundant observation gaps lead to unrealistic velocity and uncertainties estimates.

InSAR dataset

The interseismic InSAR displacement data that I used in the thesis were provided by Simon Daout from the publication of Daout et al. (2018) (covering the western Altyn Tagh fault, Chapter 3); provided by Xiaogang Song from the publication of Song et al. (2019) (covering the Haiyuan fault, Chapter 4); and provided by Lei Zhang from the publication of Zhang et al. (2018a) (covering the Xianshuihe fault, Chapter 5). The readers are recommended these papers for details on the InSAR data processing. Here, I introduce briefly the SAR data time span, spatial coverage of data, and some basic technical points.

The Altyn Tagh fault

Daout et al. (2018) processed the complete Envisat descending archive along four 500 km long and 100 km wide Envisat overlapping tracks (162, 391, 119, and 348) between 2003 and 2011 (Figure 2.6). The InSAR track encompasses our GPS profile. They constructed 484 small baseline differential interferograms with the New Small Baselines Subset chain (NSBAS) based on the ROI_PAC software. In their data processing, they applied a series of corrections for atmospheric delays, separation of the effects of hydrology and permafrost seasonal changes in the high plateau and corrections for the orbital residuals. The resulting LOS rate map is shown in Figure 2.4.

How to consider vertical deformation

InSAR displacements are typically most sensitive to vertical deformation than horizontal deformation (Wright et al., 2004). To highlight the effects of vertical deformation in data interpretation, I use the Envisat descending satellite as an example to illustrate the sensitivity of LOS to east, north and vertical displacements. Here, I take the average incidence angle to be 23°, azimuth to be 193°, then the LOS could be decomposed: VLOS=[VN, VE, VU]∙[-0.094, 0.391, 0.916]T (2.1) in this case, LOS is more sensitive to vertical than horizontal by a factor of 2.5. Here, I take the creeping fault as an example to illustrate how vertical deformation might bias our interpretation. Assuming there is 1 mm/a vertical deformation across one E-W orientation strike-slip fault, such as the case for the Haiyuan fault, which has a shallow creep rate of 5 mm/a; then, theoretically, the observed LOS rate difference should be 2.871 mm/a across the fault. Should we neglect the vertical deformation across the fault and use the observed LOS displacement to invert for the fault creep rate, we could obtain a creep rate of 7.34 mm/a, which is overestimated almost 50%. The simple calculation highlights the importance of accounting for the vertical deformation in the InSAR LOS when horizontal deformation only to modeled. In the following, I explain how to consider vertical deformation. Let’s write the observed horizontal GPS velocity as: = ℎ− + + (2-1).
where ℎ− represents the ground true horizontal deformation, is the reference frame of GPS (for instance, the Eurasian plate reference frame), and is the random errors. Note that in the following, unless otherwise specified, GPS corresponds to horizontal crustal deformation. The observed InSAR displacement is written as: =ℎ∙ℎ+ ∙ + + (2-2).
where ℎ and represent the ground true horizontal and vertical deformation respectively, ℎ and are the projection vectors for horizontal and vertical respectively, is the reference frame of InSAR (for instance, referenced to the unwrapping point), and is the errors, which includes random error and long wavelength error.

Comparison of GPS and InSAR

Daout et al. (2018) processed Environmental Satellite (Envisat) descending tracks covering the northwestern Tibetan Plateau between 2003 and 2011. They used the ROI_PAC software (Rosen et al., 2004) and applied a series of corrections for atmospheric delays, separation of the effects of hydrology and permafrost seasonal changes in the high plateau and corrections for the orbital residuals. Here, we compare our GPS horizontal velocity field with LOS rate map from Daout et al. (2018) for tracks 348 and 119 which overlap our GPS network (Figure 3.2b).
As the GPS and InSAR data are referenced to different frames, to compare them, we align them to the same reference frame. For each GPS site, we first calculate the dispersion of the InSAR LOS values for pixels located within a circle radius of 1 to 10 km (Figure A.1). Three GPS stations (AT13, AT14 and AT15), where the dispersion of the co-located InSAR LOS values is stable and small, are chosen as reference sites to align the two datasets. Horizontal GPS velocities are projected to LOS according to the local incidence angles. We estimate a mean InSAR LOS rate inside a circle of radius 1 km. An averaged value of offsets between the LOSGPS (GPS velocity in LOS) and the LOSInSAR (InSAR LOS velocity) is calculated and added to the LOSInSAR. The comparison allows us to quantify the contribution of vertical deformation to InSAR. The procedure used assumes negligible relative vertical deformation for the 3 reference GPS sites. Consequently, all the vertical deformation discussed in the following are referenced to the 3 GPS sites.

Table of contents :

Chapter 1∙ Introduction 
1.1. Seismic hazard in the Tibetan Plateau
1.2. Earthquake cycle
1.2.1. Interseismic deformation
1.2.2. Coseismic deformation
1.2.3. Postseismic deformation
1.3. Previous geodetic observation and modelling of crustal deformation in the Tibetan Plateau
1.4. The aims of this thesis
1.5. References
Chapter 2∙ Geodetic datasets and data processing 
2.1. GPS measurements and data processing
2.1.1. GPS measurement
2.1.2. GPS data processing
2.2. InSAR dataset
2.2.1. The Altyn Tagh fault
2.2.2. The Haiyuan fault
2.2.3. The Xianshuihe fault
2.3. Geodetic data integration
2.3.1. How to consider vertical deformation
2.3.2. Methodology and data integration
2.3.3. Minor comments
2.4. References
Chapter 3∙ Measuring the crustal deformation along the Altyn Tagh fault using GPS and InSAR 
3.1. Abstract
3.2. Introduction
3.3. Geodetic data and analysis
3.3.1. GPS measurements and data processing
3.3.2. Comparison of GPS and InSAR
3.4. Results
3.4.1. GPS velocity field
3.4.2. Corrections of the geodetic data for block-rotation
3.4.3. Modeling results
3.5. Discussion
3.5.1. Slow strain accumulation rate along the western Altyn Tagh fault
3.5.2. Do asymmetric patterns of interseismic velocity exist across the western Altyn Tagh fault?
3.5.3. High spatial resolution crustal deformation in the northwestern Tibet
3.6. Conclusions
3.7. References
Chapter 4∙ Surface creep on the Haiyuan fault system, northeastern Tibet, constrained from GPS and InSAR
4.1. Abstract
4.2. Introduction
4.3. Geodetic data
4.3.1. GPS data and processing
4.3.2. InSAR data
4.3.3. Geodetic data comparison and combination
4.4. Modeling strategy
4.4.1. Interseismic horizontal GPS velocities
4.4.2. Combined GPS-InSAR inversion
4.5. Results
4.5.1. GPS velocity profile fitting
4.5.2. Fault creep from the restored InSAR
4.5.3. Block modeling of the GPS and restored InSAR
4.6. Discussions
4.6.1. Shallow creep segments along the Laohushan fault
4.6.2. Mechanism of shallow fault creep along the Laohushan fault
4.6.3. The correlation between fault creep and the M~8 1920 Haiyuan earthquake
4.6.4. Earthquake scenarios along the Haiyuan fault system
4.7. Conclusions
4.8. References
Chapter 5∙ Heterogeneous interseismic coupling along the Xianshuihe-Xiaojiang fault system, eastern Tibet
5.1. Abstract
5.2. Introduction
5.3. Data and model
5.3.1. GPS data processing
5.3.2. GPS data modeling
5.3.3. InSAR data and processing
5.4. Interseismic coupling results
5.5. Discussion
5.5.1. Fault behavior and moment budget along the XXFS
5.5.1.1. Overall characteristics
5.5.1.2. Northern Xianshuihe fault
5.5.1.3. Southern Xianshuihe fault
5.5.1.3. Anninghe-Zemuhe fault
5.5.1.4. Xiaojiang fault
5.5.1.5. Seismic potential along XXFS
5.5.2. Shallow creep along the Xianshuihe fault
5.5.3. Temporal variation in creep rate along the Xianshuihe fault (~30.2°N–30.4°N)
5.6. Conclusions
5.7. References
Chapter 6∙ Discussions and conclusions
6.1. Fault coupling and seismic potential
6.2. What we learn about the shallow fault creep in Tibet
6.2.1. Interseismic fault creep on crustal faults in Tibet
6.2.2. Very long-term creep/afterslip on crustal faults in Tibet
6.2.3. The role of shallow creeping in seismic behavior
6.3. What we learn about the geodetic data
6.4. Future work
6.5. References
Appendix A-Appendix for Chapter 3
Appendix B-Appendix for Chapter 4
Appendix C-Appendix for Chapter 5

GET THE COMPLETE PROJECT