Modelling post-seismic displacements in Thai geodetic network due to the SumatraAndaman and Nias earthquakes using GPS observations

E. Panumastrakul1, W. J. F. Simons2 and C. Satirapod*1

It is evident that the 2004 Sumatra-Andaman, 2005 Nias and 2007 Bengkulu earthquakes caused significantly large co-seismic and post-seismic displacements all over the South-East Asian region. The Thai geodetic network has been severely affected by the 2004 and 2005 earthquakes. Largest post-seismic horizontal displacements were observed in the southern part of Thailand, while moderate and small displacements were seen in the central and northern parts of Thailand. This paper will analyse the GPS observations obtained by the Royal Thai Survey Department GPS campaigns up to the end of 2009. The post-seismic displacements have been shown to follow a main direction which is towards the rupture area. A simple filtering technique is proposed to remove noise from the post-seismic displacements prior to the fitting of the post-seismic displacements with the logarithmic decay function. As a result, the t log values in the logarithmic decay function at each point were found to be more consistent in both north and east directions. The new fitting results can therefore be used to estimate the coordinate of the zero-order Thai geodetic network to any epoch with millimetre accuracy.
Keywords: GPS, Sumatra-Andaman earthquake, Nias earthquake, Thai geodetic network, Post-seismic displacement, GEO2TECDI

Introduction
The occurrence of the 9?2 Mw Sumatra-Andaman earthquake on 26 December 2004 has resulted in large co- and post-seismic displacements. In fact, this giant earthquake has been shown to generate signicant (coseismic) surface displacements up to two thousand kilometres away from its epicentre [13], [14]. It has been shown in previous studies that Thailand has been coseismically displaced and subsequently undergoing a decreasing (post-seismic) deformation at the centimetre to decimetre level and this will continue for many years [9], [8]. Large co-seismic horizontal displacements were observed in the southern part of Thailand, while moderate and small displacements were seen in the central and northern parts of Thailand. Todays deformation in Thailand is still dominated by a Southwest-Northeast trending extension [5]. Although other large earthquakes as the so-called Nias (8?6 Mw) and Bengkulu (8?5 Mw) earthquakes occurred (on respectively 28 March 2005 and 12 September 2007), the internal deformation of Thailand almost entirely
1 2

Department of Survey Engineering, Faculty of Engineering, Chulalongkorn University, Thailand DEOS, Faculty of Aerospace Engineering, Delft University of Technology, the Netherlands *Corresponding author, email chalermchon.s@chula.ac.th

results from the Sumatra-Andaman earthquake due to its much larger scale of impact in the Southeast Asia Region. The zero-order Thai geodetic network composes of seven stations (BANH, CHON, OTRI, PATT, PHUK, SRIS and UTHA), but only six stations (except for PATT) have been regularly observed with GPS since 1994. The Royal Thai Survey Department (RTSD) which is responsible for maintaining the Thai coordinate reference frame has been carrying out multiple GPS eld campaigns to monitor the post-seismic displacements. Up to the end of 2008 the total magnitude of co- and post-seismic displacements ranged from 66?4 cm in the south (PHUK), 20?7 cm in the centre (CHON), 10?5 cm in the north-east (SRIS), to 8?1 cm in the north (OTRI) of Thailand [7]. Consequently, the accuracy of the Thai geodetic network is unavoidably degraded. Satirapod et al. [8] proposed an approach for updating the Thai geodetic network to any specic time by tting a logarithmic decay function to the post-seismic time series up to the end of 2006 (approximately 2 years after the earthquake). After the end of 2006, the RTSD has been carrying out several GPS measurement campaigns and the latest was in December 2009. This paper is therefore considered as a continuing work on rening the logarithmic decay function. A simple ltering technique is proposed in this paper in an attempt to lter out noise in the GPS-derived displacements prior

72

Survey Review

2012

VOL

44

NO

324

2012 Survey Review Ltd Received 30 November 2011; accepted 30 November 2011 DOI 10.1179/1752270611Y.0000000017

Panumastrakul et al.

Modelling post-seismic displacements in Thai geodetic network using GPS observations

1 Location of epicentre of the Sumatra-Andaman earthquake and overview of Thai geodetic network sites

to the tting. The new logarithmic decay function will be used to calculate a proper correction due to an effect of the earthquakes at each geodetic point so that the Thai geodetic network can be accurately computed at any epoch. This paper is organised as follows. The GPS observations and other used data are briey described. Then, the GPS data processing strategy is explained. Next, the ltering and retting of the post-seismic displacements is detailed followed with the discussion section. Finally, some concluding remarks are made.

GPS observations and other used data

This research focuses on the zero-order Thai geodetic network, using GPS data from the GPS stations of the RTSD, which has been regularly measured since 1994. This was also done in October 2004 as part of the EUASEAN SEAMERGES project, at a relatively short time period before the Sumatra-Andaman earthquake occurred. The complete zero-order Thai geodetic network was repeatedly re-measured in February, July and October of 2005, in April, July and November of 2006, in May and November of 2007, in November 2008 and again in December 2009. The measurement rounds in November 2008 and December 2009 were organised to redene the coordinates of the rst-order Thai geodetic network. Therefore, in addition to the seven zero-order network points also 11 rst-order network points were measured by the RTSD. Figure 1 shows the location of the Sumatra-Andaman earthquake and overview of the geodetic network points in Thailand. Table 1 presents an overview of the available GPS data observed at Thai geodetic network sites. It should be noted that from

2005 on the GPS point PAT2 replaces the original PATT point that was lost. To enable an accurate mapping of the GPS coordinate results into the International Terrestrial Reference Frame (ITRF), the widely-distributed IGS data of up to 30 carefully selected global IGS stations were added to the GPS campaign data sets. The inclusion of a substantial IGS sub-network in the data analyses allows for an independent and one-step transformation of Thai GPS coordinates into any (future) ITRF solution. The additional information required for processing are the JPL precise orbits, information of time, polar motion and earth orientation as well as satellite eclipse information.

Table 1 Overview of GPS data observed at the Thai geodetic network sites Observed station GPS campaign PHUK BANH CHON UTHA SRIS OTRI October-04 February-05 July-05 October-05 April-06 July-06 November-06 May-07 November-07 November-08 November-09 X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X X

Survey Review

2012

VOL

44

NO

324

73

Panumastrakul et al.

Modelling post-seismic displacements in Thai geodetic network using GPS observations

GPS data processing

To precisely determine the magnitude of displacement of each individual GPS point, the absolute GPS positioning technique should be employed rather than the relative GPS positioning technique. This is due to the fact that the relative positioning technique will not give a real picture of the displacement if the positions of both reference and roving stations are moved. Therefore, the absolute GPS positioning technique or so-called the GPS Precise Point Positioning (PPP) technique is selected as a processing strategy in this study. The scientic GIPSY-OASIS II software [15] is selected as the main GPS data processing software to derive precise absolute coordinate results. The GPS data were processed in daily batches with the PPP strategy. Each point position was based on the ionosphere-free combination of the undifferenced GPS carrier phase and pseudorange observations at 5 min intervals, with a cut-off elevation angle of 15 degrees. To ensure the highest precision solutions, observations from GPS satellites that were undergoing maintenance during part of the processed day (ftp://tycho.usno.navy.mil/pub/gps/) were removed. Tropospheric delays and gradients were stochastically estimated at each time interval. Ocean loading effects were modelled with the GOT00?2 model. To account for different GPS antennae, relative antenna phase centre corrections from the US National Geodetic Survey were applied. The daily solutions were then combined into campaign-averaged solutions. The median algorithm as proposed in [11] was applied to detect and remove outliers prior to the computation of nal coordinate repeatabilities. Results obtained from the processing of the December 2009 campaign show that all the GPS points performed very well, with daily coordinate repeatabilities typically ranging from 0?8 to 1?4, 1?3 to 2?0 and 2?9 to 7?3 mm for respectively the north, east and height components. These values are direct indicators for the internal precision of each station. It should be noted that the campaign-averaged coordinate solution is still in an unknown local reference frame. The campaign-averaged solutions were subsequently transformed into the ITRF2005 using the coordinates and velocities of a subset of well-determined regional and global IGS stations to estimate seven-parameter Helmert transformations. It should be noted that the computation of accurate ITRF2005 coordinates in SE Asia is not straightforward. Throughout the last decade new IGS stations in and around SE Asia became operational but unfortunately their relatively short time series became from 26 December 2004 onwards signicantly distorted by co- and post-seismic motions. As a result, any co- and post-seismic motions of IGS stations from 2005 onwards are not accounted for. Therefore, since 2005 the stations NTUS, IISC, HYDE, BAKO, COCO, KUNM, WUHN, SHAO can no longer be used in ITRF realisations of GPS coordinates in SE Asia. Although most of these stations have clear and increasing mists with their predicted ITRF2005 positions, these offsets adversely can be partly absorbed by transformation parameters when position errors in predominantly the same direction exist (e.g. the ongoing post-seismic motions as a result of the SumatranAndaman earthquake). This could more easily occur in double-differences based GPS data analysis software

(e.g. Bernese and GAMIT) if only regional IGS stations are used to perform the mapping in ITRF and becomes even more complicated if subsequently longer baselines need to be solved to IGS stations outside the seismically deformed region. Therefore most users here opt to connect their regional coordinate solution in an additional step to a global IGS station solution that is separately computed. For SE Asia, the zero-difference based GIPSY software appears more robust as both additional regional and global IGS data can be conveniently added and simultaneously solved for. The RMS values obtained from our mapping process of the December 2009 campaign were found to be of the order of 3 and 7 mm respectively for the horizontal and vertical components. Since most GPS points of the zero-order Thai geodetic network have long coordinate time series (1994present), their absolute coordinates and velocities were well known prior to the Sumatra-Andaman earthquake. By removing these continuous tectonic motions as being part of the tectonic Sundaland block [10], [12], their actual total co- and post-seismic displacements due to the Sumatra-Andaman and the Nias earthquakes could be accurately monitored. Because also several permanent GPS data in Thailand were available during the two major earthquakes, it was also possible to differentiate between the co-seismic and post-seismic displacements at each of the Thai sites that were observed in campaign-style [14]. By comparing the coordinate results obtained from the latest measurement campaign with the (estimated) coordinates on the 25th December 2004 (before the Sumatra-Andaman earthquake), the total horizontal displacements due to the earthquakes at each zero-order Thai geodetic network site up to December 2009 are illustrated in Fig. 2.

Filtering and refitting the post-seismic displacements

Before performing the tting of horizontal displacements with the logarithmic decay function, we rst plotted the displacement at each point in a two-dimensional plot. For the sake of brevity, only selected displacements at the CHON and BANH stations are presented in Figs. 3 and 4. Generally, the direction of displacement tends to follow the main direction which is towards the rupture area. The ltering of noise can be simply carried out by rstly tting the data to a linear equation (represented as a red line in Figs. 3 and 4). Then, each unltered position (represented as a solid triangle) is projected onto the best-t linear trend with a shortest distance to obtain the ltered position (represented as a solid rectangle). It is important to note that the difference between ltered and unltered data is generally less than 3 mm which is consistent with the noise level of displacement data. After applying the lter to all stations, the ltered horizontal displacements are plotted in Figs. 5 and 6 for north and east directions, respectively. The Y-axis represents the magnitude of displacement in cm units while the X-axis shows the period in year units since 26 December 2004. Then we t the ltered displacements with the below logarithmic decay function [2] in north and east directions for each Thai geodetic network site. It should be noted that these ttings are mainly for geodetic work

74

Survey Review

2012

VOL

44

NO

324

Panumastrakul et al.

Modelling post-seismic displacements in Thai geodetic network using GPS observations

4 Filtering the horizontal displacement trend due to the earthquakes at the BANH station

into one post-seismic time series. The estimated logarithmic model parameters are summarised in Table 2. It can be seen from Table 2 that the t log values at each point are found to be consistent in both north and east directions. In addition, the R-squared values demonstrate a very good t of logarithmic functions to the ltered post-seismic time series. Hence, the approach as proposed by Satirapod et al. [8] can still be applied to compute for the zero-order GPS station coordinates at any specic epoch.
2 Total horizontal (co- and post-seismic) displacements due to the Sumatra-Andaman and Nias earthquakes at each zero-order Thai geodetic network site up to December 2009

Discussion
Approximately 5 years after the Sumatra-Andaman earthquake, it is very exciting to nd out the postseismic displacement magnitude at PHUK station is about double that amount of the initial co-seismic displacement from the 2004 Mw 9?2 Sumatra-Andaman

u(t)~czaln (1zt=tlog )

(1)

where, t is time since the earthquake, u(t) is the displacement (east and north), c is the co-seismic offset, a is the amplitude associated with the decay, tlog is the constant logarithmic decay time. Satirapod et al. [6] have shown that the Nias earthquake did not generate signicant displacements at most of Thai geodetic network sites, so the co-seismic displacement can be ignored for all Thai geodetic network sites except for the PHUK station. The postseismic displacements due to the two earthquakes for all stations (except for PHUK station) can be combined

3 Filtering the horizontal displacement trend due to the earthquakes at the CHON station

5 Post-seismic time series in north direction due to the Sumatra-Andaman and Nias earthquakes for Thai geodetic network sites. Solid lines are best-t logarithmic decay functions

Survey Review

2012

VOL

44

NO

324

75

Panumastrakul et al.

Modelling post-seismic displacements in Thai geodetic network using GPS observations

become slightly distorted since the displacement appears to be slowly decreasing over a large area in SWNE direction. For large scale mapping (i.e. 1 : 1000, 1 : 4000), there is no need to remap since relative distances within Phuket can be considered to remain unchanged. This implies that most of the existing maps produced before the earthquakes are still valid and no obvious effect would be noticed by large scale map users. On the other hand, the displacement will cause a distortion on the Thai geodetic network whose required accuracy is on a centimetre level. Therefore RTSD has announced the use of November 2008 campaign solutions as a new national reference frame for Thailand. This is done to mitigate the effect of the coand post-seismic displacements due to the earthquakes. In addition, the readjusting of the rst-order and second-order geodetic networks is currently being carried out by RTSD to maintain the accuracy of such control points over Thai region. The new coordinates of the control points will subsequently be used to serve large scale mapping, engineering and cadastral surveying works in Thailand.

Concluding remarks
6 Post-seismic time series in east direction due to the Sumatra-Andaman and Nias earthquakes for Thai geodetic network sites. Solid lines are best-t logarithmic decay functions

earthquake. These displacements are much larger than the expected displacements estimated from elastic geophysical post-seismic models [3], [4]. Based on an extrapolation of the new logarithmic decay function for PHUK this indicates an additional 10 cm of postseismic deformation could be accumulated by the end of 2015. Fleitout [1] pointed out that up to 4 years after the earthquake, the mean-eld and near-eld post-seismic deformations derived from GPS observations are approximately 17 and 40% of the co-seismic displacements. On the other hand, the deformation results in the region of Thailand, which are considered as a far-eld data, were shown to be more than double (200%) of the amount of the initial co-seismic displacement. The deformations in the far-eld region are therefore extremely valuable for geophysical research. In addition to the fact that the displacement in Phuket already reaches 70 cm the whole island of Phuket has

In this paper, the new GPS observations obtained from RTSD GPS campaigns up to the end of 2009 have been processed using the PPP strategy of the GIPSY-OASIS II software. Displacements obtained from the latest campaign (December 2009) conrmed that the earthquakes have resulted in total horizontal (co- and postseismic) displacements, ranging from 70?2 cm in the south to about 8?4 cm in the north of Thailand. The simple ltering technique was applied to remove the post-seismic displacements noise. The ltered postseismic motions were subsequently retted with a logarithmic decay function using longer time series up to approximately 5 years after the Sumatra-Andaman earthquake. The logarithmic model parameters were computed in north and east directions for each Thai geodetic network site. The new tting is shown to be much improved. Our future work will focus on the modelling of corrections due to the post-seismic displacements to the rst-order and second-order Thai geodetic network points. In addition, the vertical motion after the Sumatra-Andaman earthquake will be investigated by using also the existing permanent GPS stations in and around Thailand.

Table 2 Co-seisimic and post-seismic best-t logarithmic model parameters for the two earthquakes StationDirection PHUKNorth1 PHUKEast1 PHUKNorth2 PHUKEast2 BANHNorth BANHEast CHONNorth CHONEast UTHANorth UTHAEast SRISNorth SRISEast OTRINorth OTRIEast Validity 261204 to 28305 261204 to 28305 28305 to 11209 28305 to 11209 261204 to 11209 261204 to 11209 261204 to 11209 261204 to 11209 261204 to 11209 261204 to 11209 261204 to 11209 261204 to 11209 261204 to 11209 261204 to 11209 a/mm 64.03 130.30 64.03 130.30 253.37 297.47 26.6 42.78 227.09 231.43 213.19 219.40 220.03 212.56 tlog/year 0.304 0.294 0.304 0.294 0.453 0.452 0.356 0.356 0.388 0.391 0.324 0.320 0.561 0.578 c/mm 2101.9 2252.5 2116.4 2267.0 276.8 2154.0 239.7 267.5 239.3 247.4 219.7 231.5 223.7 222.1 R2 0.995 0.994 0.995 0.994 0.995 0.995 0.993 0.993 0.987 0.987 0.974 0.974 0.990 0.990

76

Survey Review

2012

VOL

44

NO

324

Panumastrakul et al.

Modelling post-seismic displacements in Thai geodetic network using GPS observations

Acknowledgements
This research is supported by the National Research University Project of CHE and the Ratchadaphiseksomphot Endowment Fund (CC307A), a grant from the Netherlands Organisation for Scientific Research (NWO) User Support Programme Space Research and the GEO2TECDI-SONG project. Finally the authors would like to thank to the Royal Thai Survey Department (RTSD) for providing the GPS data used in this study. This paper is based on the presentation made at the 2010 International Symposium on GPS/GNSS.

References
1. Fleitout, L., 2010. Observation and modelisation of POStseismic deformation in the SUMatra region. Presentation made at the GEO2TECDI final workshop, 27 April, TUDelft, The Netherlands. 2. Marone, C. J., Scholz, C. H. and Bilham, R., 1991. On the mechanics of earthquake afterslip. Journal of Geophysical Research, 96: 84418452. 3. Pollitz, F. F., 1992. Postseismic relaxation theory on the spherical earth. Bulletin of the Seismological Society of America, 82: 422 453. 4. Pollitz, F. F., 1997. Gravitational viscoelastic postseismic relaxation on a layered spherical Earth. Journal of Geophysical Research, 102(B8): 1792117941. 5. Satirapod, C., Bamrungwong, S., Vigny, C. and Lee, H. K., 2010. Change of strain rate in Thailand after the 26 December 2004 and 28 March 2005 earthquakes using GPS measurements. KSCE Journal of Civil Engineering, 14(2): 215220. 6. Satirapod, C., Laoniyomthai, N. and Chabangborn, A., 2007. Crustal Movement of Thailand Disc Due to 28 March 2005 Earthquake as Observed from GPS Measurements. International Journal of Geoinformatics, 3(1): 2933.

7. Satirapod, C., Simons, W. J. F., Panumastrakul, E. and Trisirisatayawong, I., 2010. Updating Thai Reference Frame to ITRF2005 Using GPS: Diversion between ITRF2000 and 2005 in Southeast Asia, Survey Review, in press. 8. Satirapod, C., Simons, W. J. F. and Promthong, C., 2008. Monitoring deformation of Thai Geodetic Network due to the 2004 Sumatra-Andaman and 2005 Nias Eauthquakes by GPS. Journal of Surveying Engineering (ASCE), 134(3): 8388. 9. Satirapod, C., Simons, W. J. F., Promthong, C., Yousamran, S. and Trisirisatayawong, I., 2007. Deformation of Thailand as detected by GPS measurements due to the December 26th, 2004 mega-thrust Earthquake, Survey Review, 39(304): 109115. 10. Simons, W. J. F., Ambrosius, B. A. C., Noomen, R., Angermann, D., Wilson, P., Becker, M., Reinhart, E., Walpersdorf, A. and Vigny, C., 1999. Observing Plate Motions in S. E. Asia: Geodetic Results of the GEODYSSEA Project. Geophysical Research Letters, 26(14): 20812084. 11. Simons, W. J. F., Panumastrakul, E. and Yousamran, S., 2009. Processing the RTSD November 2008 GPS Campaign. DEOS official report to RTSD, Delft, The Netherlands. 12. Simons, W. J. F, Socquet, A., Vigny, C., Ambrosius, B. A. C., Abu, S. H., Promthong, C, Subarya, C., Sarsito, D. A., Matheussen, S., Morgan, P. and Spakman, W., 2007. A Decade of GPS in Southeast Asia: Resolving Sundaland motion and boundaries. Journal of Geophysical Research, 112: B06420, doi:10?1029/ 2005JB003868. 13. Stein, S. and Okal, E. A., 2005. Seismology: Speed and Size of the Sumatra Earthquake. Nature, 434: 581582. 14. Vigny, C., Simons, W. J. F., Abu, S. H., Ronnachai, B., Satirapod, C., Chhoosakul, M., Subarya, C., Omar, K., Abidin, H. Z., Socquet, A. and Ambrosius, B. A. C., 2005. Insight into the 2004 SumatraAndaman earthquake from GPS measurements in Southeast Asia. Nature, 436: 201206. 15. Zumberge, J., Heflin, M., Jefferson, D., Watkins, M. and Webb, F., 1997. Precise point positioning for the efficient and robust analysis of GPS data from large networks. Journal of Geophysical Research, 102: 50055017.

Survey Review

2012

VOL

44

NO

324

77