Geodetic Model of the 2017 Mw 6.5 Mainling Earthquake Inferred from GPS and Insar Data

remote sensing

Letter Geodetic Model of the 2017 Mw 6.5 Mainling Earthquake Inferred from GPS and InSAR Data

Huizi Jian, Lifeng Wang *, Weijun Gan, Keliang Zhang, Yanchuan Li , Shiming Liang, Yunhua Liu, Wenyu Gong and Xinzhong Yin

State Key Laboratory of Earthquake Dynamics, Institute of Geology, China Earthquake Administration, Beijing 100029, China; [email protected] (H.J.); [email protected] (W.G.); [email protected] (K.Z.); [email protected] (Y.L.); [email protected] (S.L.); [email protected] (Y.L.); [email protected] (W.G.); [email protected] (X.Y.) * Correspondence: [email protected]; Tel.: +86-010-6200-9427

Received: 2 November 2019; Accepted: 3 December 2019; Published: 9 December 2019

Abstract: On 17 November 2017, a Mw 6.5 earthquake occurred in Mainling County, Nyingchi City, China. The epicenter was located in the Namche Barwa region of the eastern Himalayan syntaxis. Here, we have derived coseismic deformation from Global Positioning System (GPS) data and ascending Sentinel-1A Synthetic Aperture Radar (SAR) data. Based on a joint inversion of the two datasets, we obtained the coseismic slip distribution along a curved, northeast trending, and high-angle (dip angle of 75◦) thrust fault. Our results show that the seismic moment release was 7.49 1018 N m, corresponding to a moment magnitude of Mw 6.55. The maximum slip was 1.03 m × · and the main rupture zone extended to a 12 km depth. The earthquake may have been related to the release of strain accumulated during the subduction of the Indian plate beneath the Eurasian continent. We identiﬁed a high strain rate and low b-values around the epicentral area before the earthquake, indicating that the earthquake was nucleated under a high strain/stress state. The data indicate two regions, southwest and southeast to the epicenter (the eastern Main Himalaya Thrust and northern end of the Sagaing fault), which remain under high stress/strain conditions and pose a signiﬁcant seismic hazard.

Keywords: coseismic modeling; GPS; InSAR; inversion; the eastern Himalayan syntaxis

1. Introduction

At 22:34 (UTC) on November 17, 2017, a Mw 6.5 earthquake (epicenter of 29.75◦ N, 95.02◦ E) struck Mainling County, Nyingchi City, Tibet (herein referred to as the Mainling earthquake), with a focal depth of 10 km (Figure1). The maximum intensity given by the China Earthquake Administration was VIII [1]. No casualties were reported, but more than 3000 buildings were damaged. According to the focal mechanism provided by the Global Centroid Moment Tensor (GCMT) Project, the earthquake was a thrust event that occurred in the northern part of the Namche Barwa region, along the NNW-trending Xixingla fault [2]. More than 4000 aftershocks were recorded within the first week of the mainshock, including 22 M > 3 aftershocks [3], the largest being M 4.3. Owing to the remote location and difficult accessibility of the Namche Barwa region, no clear surface rupture was identified [3]. Nevertheless, owing to the dense seismic network deployed in eastern Tibet and advances in space geodesy, the earthquake was recorded with sufficient data coverage and quality to study its rupture process, and to investigate the crustal deformation and seismic activity of the Namche Barwa region. These results provide new insights into the crust dynamics of the eastern Himalayan syntaxis. Several studies have investigated the earthquake using finite fault modeling (Table1) based on relocated aftershocks, interferometric synthetic aperture radar (InSAR) measurements, and teleseismic

Remote Sens. 2019, 11, 2940; doi:10.3390/rs11242940 www.mdpi.com/journal/remotesensing Remote Sens. 2019, 11, x FOR PEER REVIEW 2 of 14

41 of the Namche Barwa region. These results provide new insights into the crust dynamics of the 42 eastern Himalayan syntaxis. Remote Sens. 2019, 11, 2940 2 of 14 43 Several studies have investigated the earthquake using finite fault modeling (Table 1) based on 44 relocated aftershocks, interferometric synthetic aperture radar (InSAR) measurements, and 45 datateleseismic [1,3–8 ]. data However, [1,3–8]. there However, are inconsistencies there are inconsistencies among the di amongfferent fault the different models, coseismicfault models, slip 46 distributions,coseismic slip anddistributions, produced and Coulomb produced stress Coulomb changes. stress For changes example,. For using example, teleseismic using teleseismic data and 47 InSAR-measureddata and InSAR-measured surface deformation, surface deformation, Liu et al. Liu [1] reportedet al. [1] reported that the earthquake that the earthquake broke one broke asperity, one 48 andasperity, that ruptures and that were ruptures concentrated were concentrated between the between surface and the at surface a 10 km and depth, at a with10 km a maximum depth, with slip a 49 ofmaximum 0.51 m. However, slip of 0.51 Li et m. al. However, [9] identified Li et two al. coseismic[9] identified slip areas, two coseismic dominantly slip distributed areas, dominantly at depths 50 ofdistributed 5–15 km, at with depths a maximum of 5–15 km, slip with of 0.83 a maximum m. Based slip on theof 0.83 aftershock m. Based data, on the Xiong aftershock et al. [6 ]data, suggested Xiong 51 thatet al. the [6]earthquake suggested that ruptured the earthquake two adjacent rup paralleltured two faults. adjacent Their parallel inversion faults. results Their based inversion on the InSARresults 52 databased provide on the comparableInSAR data maximumprovide comparable slips of 0.36 maximum and 0.43 slips m on of the0.36 two and fault 0.43 planes. m on the In two addition, fault 53 theplanes. different In addition,studies havethe different considered studies divergent have faultconsidered plane solutionsdivergent (Table fault1 ),plane which solutions has also (Table resulted 1), 54 inwhich ambiguities has also inresulted slip modeling. in ambiguities All of thesein slip discrepancies modeling. All are of duethese to discrepancies a lack of observations are due to that a lack can 55 constrainof observations the fault that model can constrain well. the fault model well.

56 57 Figure 1.1. TectonicTectonic backgroundbackground of of the the 2017 2017 Mw Mw 6.5 6.5 Mainling Mainling earthquake. earthquake. (a) ( Regionala) Regional map map showing showing the 58 earthquakethe earthquake location. location. (b) Tectonic (b) Tectonic setting. setting. Blue arrows Blue arrows indicate indicate interseismic interseismic global positioning global positioning system 59 (GPS)system velocity (GPS) velocity ﬁelds of fields the eastern of the eastern Himalayan Himalayan syntaxis syntaxis measured measured between between 1999 and 1999 2016 and [ 12016]; error [1]; 60 ellipseserror ellipses indicate indicate a 95% a conﬁdence 95% confidence level; level the red; the rectangle red rectangle marks marks the InSAR the InSAR data coverage;data coverage; the blue the 61 rectangleblue rectangle outlines outlines the areathe area shown shown in (c). in (c). (c)M (c) >M3 > earthquakes3 earthquakes (orange (orange dots) dots) in in the the NamcheNamche BarwaBarwa 62 regionregion fromfrom 19651965 toto 20172017 andand aftershocksaftershocks (blue(blue dots)dots) ofof thethe MainlingMainling earthquakeearthquake [[10]10].. TheThe beachbeach ballball 63 indicatesindicates thethe GlobalGlobal Centroid Centroid Moment Moment Tensor Tensor (GCMT) (GCMT) focal focal mechanism mechanism of theof the earthquake; earthquake; the redthe starred 64 representsstar represents the epicenter; the epicenter; the blue the blue line denotesline denotes the fault the fault that likelythat likely generated generated the earthquake; the earthquake; black black lines 65 representlines represent other other faults faults in the in study the study region region [1] (JSZ: [1] Jiali(JSZ: Shear Jiali Shear Zone; Zone; DMF: DMF Dongjiu-Mainling: Dongjiu-Mainling fault; fault; NBS: 66 NamcheNBS: Namche Barwa Barwa region; region; ZSF: ZayuZSF: Zayu fault; fault; MAF: MAF Motuo-Aniqiao: Motuo-Aniqiao fault; fault; BLF: BLF Baqing-Leiniaoqi: Baqing-Leiniaoqi fault; fault; NJF: 67 NujiangNJF: Nujiang fault; fault; XXLF: XXLF Xixingla: Xixingla fault; fault; IYSZ: IYSZ Yarlung: Yarlung Zangbo Zangbo suture suture zone; zone; AS: Assam AS: Assam syntaxis). syntaxis).

68 InIn this this study, study, we we have have included included GPS GPS displacement displacement data data for for the the first first time. time. These data were 69 acquiredacquired inin campaign-modecampaign-mode before andand afterafter thethe earthquake.earthquake. By referring to fieldfield geological survey 70 datadata [11 [11,12],,12], we we also also determined determined a curved a curved fault plane, fault according plane, according to the relocated to the aftershock relocated distribution, aftershock 71 anddistribution, optimized andthe optimized dip-angle the according dip-angle to according the data to fitting the data quality fitting of quality the slip of models.the slip models. This paper This 72 ispaper organized is organized as follows. as follows We first. We introduce first introduce the tectonic the tectonic background background of the Namche of the Namche Barwa region Barwa 73 (Sectionregion (Section2), then 2), describe then describethe data the processing data processing of the of coseismic the coseismic displacement displacement field, field, including including the processing schemes for the GPS and SAR data (Section3). Section4 describes how the fault parameters were determined and the slip modeling results were obtained. In Section5, we discuss the two fault Remote Sens. 2019, 11, 2940 3 of 14 planes of the 2017 Mainling earthquake and the regional seismic hazard before and after the earthquake. Our conclusions are presented in Section6.

Table 1. Source parameters of the 2017 Mw 6.5 Mainling earthquake.

South Dip ( ) North Dip ( ) Source Dataset ◦ ◦ Strike Dip Rake Strike Dip Rake USGS Teleseismic 132 55 95 303 36 83 GCMT Teleseismic 109 29 56 328 66 108 CENC [10] Regional seismic data 130 27 90 310 63 90 Bai et al. [13] Regional seismic data 328 66 108 Liu et al. [1] Teleseismic and InSAR 308 76 82.2 Wei et al. [5] Aftershock 319.9 61.7 101 328 63 104 Xiong et al. [6] InSAR and aftershock 328 72 97 USGS: United States Geological Survey; GCMT: Global Centroid Moment Tensor; CENC: China Earthquake Networks Center; InSAR: Interferometric synthetic aperture radar.

2. Seismological Background The 2017 Mw 6.5 Mainling earthquake occurred in the Namche Barwa region of the eastern Himalayan syntaxis. This area is highly deformed, owing to collision between the Indian and Eurasian plates which has occurred since 40 Ma ago [14]. Modern geodetic measurements (e.g., GPS) have revealed significant crustal deformation in the Tibetan Plateau. The Mani-Yushu-Xianshuihe fault marks the northern boundary and the upper crust rotates clockwise around the eastern Himalayan syntaxis [15–18]. Tectonic movements are distributed across many large active faults, including the Jiali fault, Nujiang fault, Aparon fault, NE-trending Motuo fault, Mainling fault, Aniqiao fault, and Yarlung Zangbo fault [19,20] (Figure1). This region also su ffers from severe surface erosion (5–17 mm/yr), accompanied by powerful fluvial activity (200 kW/m), heavy river incision, and intense rock metamorphism [21,22]. The eastern Himalayan syntaxis accommodates strong seismic activity. Aside from the 2017 Mainling earthquake, at least seven M > 7 large earthquakes have occurred here in the last century [1], including the 1947 M 7.7 Nang country earthquake, located 200 km southwest of the 2017 Mainling earthquake [22], and the 1950 M 8.6 Zayu earthquake, located approximately 220 km southeast of the Mainling event [23]. The Zayu earthquake was also followed by a series of M > 6 aftershocks. The 2017 Mw 6.5 Mainling earthquake, a high-angle thrust event, caused by the NE-dipping thrust of the Namche Barwa region subducting beneath the Lhasa terrane [1], broke a regional seismic quiescence that had lasted for over 50 years.

3. Geodetic Data

3.1. GPS Data In order to study interseismic crustal deformation in the eastern Himalayan syntaxis, we deployed 13 GPS stations on the two sides of the Yarlung Zangbo River, near the Mainling epicenter, and conducted three repeated observations between 2015 and 2017. The day after the Mainling earthquake (18 November 2017), we conducted measurements at eight GPS sites located within 100 km of the epicenter and obtained position data from at least 24-h of continuous observations for each station. The GPS Receiver Independent Exchange (RINEX) data, with a 30-sec sampling rate, were processed using the GAMIT/GLOBK 10.6 software [24]. The GPS carrier phase data were ﬁrst processed to obtain loosely constrained daily solutions, along with the associated variance-covariance matrices. In this processing step, we adopted the latest absolute receiver antenna models of the International Global Navigation Satellite System (GNSS) Service (IGS), atmospheric zenith delay Remote Sens. 2019, 11, 2940 4 of 14

models [25], and the global mapping functions for the neutral atmosphere. Then, the loosely constrained daily solutions were transformed to a regional reference frame through combination with 20 IGS stations near the epicenter, after translation, rotation, and scaling using seven parameter transformations. Finally, coseismic displacements were estimated from these daily time series using the least squares method. The average measurement errors in the eastern and northern components were 2.85 and 2.69 mm, respectively. Table2 provides the coseismic displacements of the eight GPS stations (Figure2a). It is apparent that the deformation in the near field was larger than that in the far field, showing a rapid decay with distance from the fault, which suggests a rupture along a high-angle thrust fault. Although the postseismic effects in the first few days following the mainshock may be included in the estimated coseismic displacements, their magnitudes should be very small in comparison with the coseismic component and are thus negligible.

Table 2. Coseismic global positioning system (GPS) displacements of the 2017 Mainling earthquake.

a Site Longitude (◦E) Latitude (◦N) East (mm) North (mm) Cen 0KSM 94.48 29.56 2.68 2.10 5.20 1.96 0.0742 ± ± 973E 94.72 29.70 21.22 2.12 4.29 1.70 0.0438 ± ± 0YIG 94.97 30.16 1.96 2.55 2.32 2.73 0.0987 ± − ± − 0BAG 94.46 29.29 2.80 3.08 0.39 2.77 0.0447 ± ± 0MIR 94.65 29.47 4.44 4.38 6.80 4.38 0.1927 ± ± − Remote PAIZSens. 2019, 11, x FOR 94.89PEER REVIEW 29.50 3.34 2.50 15.26 2.35 0.01135 of 14 ± ± − 973F 94.79 29.94 5.37 3.18 1.67 2.89 0.0786 ± − ± − 150 pixels,0XWK there is signifi 95.38cant redundancy 29.93 in constructing8.06 a2.91 source model9.53 2.70 from all observations. 0.0610 − ± − ± 151 Therefore, we down-sampledNotes: thea CorrelationInSAR data coe ﬃbasedcient between on the north uniform and east. sampling rule, using a 150-m 152 resolution, while retaining deformation information. In total, we obtained 1365 observations.

153

154 FigureFigure 2. CoseismicCoseismic displacement displacement data. data. (a) ( aGlobal) Global positioning positioning system system (GPS) (GPS)-based-based displacements displacements (red 155 (redarrows); arrows); the black the black line marks line marks the fault the faulttrace; trace; error error ellipses ellipses indicate indicate a 95% a 95%confidence confidence level; level; the blue the 156 bluerectangle rectangle indicates indicates the aftershock the aftershock region, region, used toused demonstrate to demonstrate the along the-depth along-depth distribution distribution shown in 157 shownFigure in3a; Figure the green3a; the line green displays line displays the profile the for profile plotting for plotting the line the-of- line-of-sightsight (LOS) deformation (LOS) deformation shown 158 shownin panel in ( panelc); (b) (cRe);w (brapped) Rewrapped coseismic coseismic interferogram interferogram of the of2017 the Mw 2017 6.5 Mw Mainling 6.5 Mainling earthquake. earthquake. The 159 Thedetermi determinedned LOSLOS deformation deformation field field outlined outlined by the by red the rectangle red rectangle is color is color-coded-coded in panel in panel (a); (c (a) );LOS (c) 160 LOSdeformation deformation of profile of profile AA’, AA’, as marked as marked in panel in panel (a). (a).

161 4. Finite Fault Modeling

162 4.1. Fault Geometry 163 To obtain the coseismic slip of the 2017 Mainling earthquake, we first determined a fault model. 164 Based on field geological investigations [11,12] and the relocated aftershock catalog [3], we outlined 165 a curved fault trace comprised of three segments, namely, a northern segment with an average strike 166 of 323°, a middle segment with a strike of 307°, and a southern segment with a strike of 319°. The 167 fault has a length of 48 km, and we assumed that the fault reaches a depth of 30 km. During coseismic 168 slip modeling, we discretized the fault plane into small patches of 3 × 3 km.

169 According to dislocation theory [29,30], a slip on a fault can be linearly projected to the ground 170 surface by means of Green’s functions [31]. Therefore, with known coseismic displacements, we can 171 invert for fault slip using the constrained least-squares algorithm on a fault plane with different dip 172 angles. The optimization procedure was realized using a Steepest Descent Method (SDM) Program 173 [32]. We evaluated the slip model according to the data-fitting quality measured by the root mean 174 square (RMS).

175 The along-depth distribution of the relocated aftershocks (Figure 3a) were concentrated on the 176 eastern side of the fault trace, suggesting that the fault plane likely dips to the east in a range marked 177 by ‘OA’ and ‘OB’ in Figure 3a, corresponding to a dip angle between 60° and 85°. Therefore, we tested

Remote Sens. 2019, 11, x FOR PEER REVIEW 6 of 14

178 Remotedifferent Sens. dip2019 angles, 11, 2940 in the range outlined by the aftershock distribution and identified an optimal5 ofdip 14 179 angle of 75° to explain the surface displacements (Figure 3b). This value was used for slip modeling.

180 181 Figure 3. DipDip angle angle of ofthe the fault. fault. (a) Aftershock (a) Aftershock (green (green dots) dots) profile profile along alongprofile profile BB’, marked BB’, marked in Figure in 182 Figure2a. O denotes2a. O denotes the location the location of the fault of the trace; fault OA trace; and OA OB and outline OB outline the dominant the dominant aftershocks aftershocks and imply and 183 implythe possible the possible location location of the of fault the faultplane. plane. (b) Root (b) Root mean mean square square (RMS) (RMS) values values of the of the slip slip model model as asa 184 afunction function of of the the fault fault plane plane dip dip angle, angle, where where the the red red dot dot marks marks the the optimal optimal dip dip angle. angle. 3.2. InSAR Data 185 In the joint inversion of GPS and InSAR data, we also considered their relative weights, which 186 wereWe tested selected in the ascending range of 1:0.1 frames to 1:20. from An Sentinel-1A optimal ratio that wereof 1:3.5 acquired was found 6 days to best before explain the earthquake the entire 187 (geodetic11 November dataset. 2017 ) and 6 days after the earthquake (23 November 2017) to reconstruct coseismic surface displacements. The SAR data were processed using the GAMMA software, following a standard 188 single-pair4.2. Checkerboard differential Test interferogram generation strategy [26]. Topography effects were removed from the interferogram using the 90-m resolution Shuttle Radar Topography Mission (SRTM) digital elevation 189 To examine the geodetic data resolution for the slip distribution on the fault plane, we conducted model (DEM). We used adaptive filtering to reduce the phase noise in the differential interferogram, and 190 three checkerboard tests, using GPS, InSAR, and both datasets. According to Wang et al. [33], we then used the minimum cost flow algorithm for phase unwrapping [27]. After geocoding, we obtained 191 assigned synthetic slips of 0 and 1 m on different fault patches (Figure 4a), whose produced surface the line-of-sight (LOS) coseismic deformation field of the Mainling earthquake (Figure2). The 192 displacements were used for retrieving the slip on the fault (Figure 4b–d). We generated synthetic unwrapped coseismic deformation map (Figure2a) shows negative LOS deformation (away from 193 displacements for both the GPS stations and InSAR observation sites. These synthetic displacements the satellite), with a maximum value of approximately 0.17 m occurring on the southwest side of the 194 include errors that follow a Gaussian distribution. fault, and a positive LOS deformation (towards the satellite) with a maximum value of 0.05 m on the northeast side. However, the near-field interferogram (ascending) is decorrelated, owing to snow and vegetation coverage over the Namche Barwa region [11]. Satellite data (from Sentinel-2 and Landsat 8) after the earthquake also indicate widely distributed landslides, covering a large area near the fault, which further strengthens the decorrelation effect [1]. We extracted a displacement profile that passes through the earthquake epicenter and is normal to the fault (Figure2c). This profile shows maximum LOS displacements of 0.14–0.15 m on the southwest side of the fault and 0.05 m on the northeast side, indicating a reverse faulting process. We acquired the SAR image 6 days after the earthquake, and thus, these displacements inevitably involve postseismic relaxation processes. However, we expect that early postseismic deformation, which is mainly attributed to afterslip and/or poroelastic relaxation, is concentrated in the near-field, where the interferogram loses coherence. Thus, postseismic relaxation does not significantly contribute to our determined coseismic displacements [28]. Since the complete InSAR data comprise millions of pixels, there is significant redundancy in constructing a source model from all observations. Therefore, we down-sampled the InSAR data based on the uniform sampling rule, using a 150-m resolution, while retaining deformation information. In total, we obtained 1365 observations.

4. Finite Fault Modeling

4.1. Fault Geometry

195 To obtain the coseismic slip of the 2017 Mainling earthquake, we ﬁrst determined a fault model. 196 BasedFigure on ﬁeld 4. Checkerboard geological investigations test. (a) Input model, [11,12 where] and theblack relocated = 1 m and aftershock white = 0 m. catalog Retrieved [3], fault we outlinedslip 197 a curvedusingfault (b) synthetic trace comprised global positioning of three segments, system (GPS) namely, data a only; northern (c) using segment synthetic with interferometric an average strike 198 of 323synthetic◦, a middle aperture segment radar with (InSAR) a strike data of only; 307 ◦(d,) and using a southernboth datasets. segment with a strike of 319◦. The fault

Remote Sens. 2019, 11, x FOR PEER REVIEW 6 of 14

178 different dip angles in the range outlined by the aftershock distribution and identified an optimal dip 179 angle of 75° to explain the surface displacements (Figure 3b). This value was used for slip modeling. Remote Sens. 2019, 11, 2940 6 of 14

has a length of 48 km, and we assumed that the fault reaches a depth of 30 km. During coseismic slip modeling, we discretized the fault plane into small patches of 3 3 km. × According to dislocation theory [29,30], a slip on a fault can be linearly projected to the ground surface by means of Green’s functions [31]. Therefore, with known coseismic displacements, we can invert for fault slip using the constrained least-squares algorithm on a fault plane with different dip angles. The optimization procedure was realized using a Steepest Descent Method (SDM) Program [32]. We evaluated the slip model according to the data-fitting quality measured by the root 180 mean square (RMS). 181 TheFigure along-depth 3. Dip angle distributionof the fault. (a of) Aftershock the relocated (green aftershocks dots) profile (Figure along profile3a) were BB’, concentratedmarked in Figure on the 182 eastern2a. sideO denotes of the the fault location trace, of suggesting the fault trace; that OA the and fault OB plane outline likely the dominant dips to the aftershocks east in a and range imply marked 183 by ‘OA’the andpossible ‘OB’ location in Figure of3 thea, corresponding fault plane. (b) toRoot a dip mean angle square between (RMS) 60 values◦ and of 85 ◦the. Therefore, slip model we as testeda 184 differentfunction dip anglesof the fault in the plane range dip outlinedangle, where by thethe red aftershock dot marks distribution the optimal and dip identifiedangle. an optimal dip angle of 75◦ to explain the surface displacements (Figure3b). This value was used for slip modeling. 185 InIn thethe jointjoint inversion inversion of of GPS GPS and and InSAR InSAR data, data, we we also also considered considered their their relative relative weights, weights,which which were 186 testedwere tested in the in range the range of 1:0.1 of 1:0.1 to 1:20. to 1:20. An An optimal optimal ratio ratio of of 1:3.5 1:3.5 was was found found to to best best explain explain thethe entire 187 geodetic dataset.

188 4.2. Checkerboard Test 189 To examineexamine thethe geodeticgeodetic datadata resolutionresolution forfor thethe slip distribution on the fault plane, we conducted 190 threethree checkerboard tests, tests, using using GPS, GPS, InSAR, InSAR, and and both both datasets. datasets. According According to Wang to Wang et al. et [33], al. [ 33we], 191 weassigned assigned syntheticsynthetic slips slips of 0 of and 0 and 1 m 1 mon on different diﬀerent fault fault patches patches (Figure (Figure 4a),4a), whose produced surfacesurface 192 displacements were used for retrieving the slip onon thethe faultfault (Figure(Figure4 4bb–d).–d). WeWe generated generated synthetic synthetic 193 displacements for both the GPS stations and InSAR observationobservation sites.sites. These synthetic displacements 194 includeinclude errorserrors thatthat followfollow aa GaussianGaussian distribution.distribution.

195 196 Figure 4. Checkerboard test. test. (a (a) )Input Input model, model, where where black black = 1= m1 mand and white white = 0= m.0 Retrieved m. Retrieved fault fault slip 197 slipusing using (b) ( syntheticb) synthetic global global positioning positioning system system (GPS) (GPS) data data only; only; (c (c) ) using using synthetic interferometric interferometric 198 syntheticsynthetic apertureaperture radarradar (InSAR)(InSAR) datadata only;only; ((dd)) usingusing bothboth datasets.datasets.

The checkerboard test results show that the GPS network has a relatively poorer fault slip resolution compared with InSAR, owing to the sparse GPS site coverage (Figure4b,c). The joint Remote Sens. 2019, 11, x FOR PEER REVIEW 7 of 14

Remote Sens. 2019, 11, 2940 7 of 14 199 The checkerboard test results show that the GPS network has a relatively poorer fault slip 200 resolution compared with InSAR, owing to the sparse GPS site coverage (Figure 4b,c). The joint 201 inversion based based on on both both the the GPS GPS and and InSAR InSAR data better data better resolves resolves the fault the slip, fault decreasing slip, decreasing the modeling the 202 errorsmodeling by 10%errors and by 10% 30% and when 30% compared when compared with the with results the basedresults onlybased on only the on InSAR the InSAR or GPS or data,GPS 203 respectively.data, respectively.The modeling The modelingerrors were errors evaluated were evaluated according according to the di toﬀerences the differences between the between synthetic the 204 andsynthetic inverted and slips.invertedTherefore, slips. Therefore, the GPS data the GPS play data an important play an important role in constructing role in constructing the slip model the slip of 205 themodel 2017 of Mainling the 2017 earthquake.Mainling earthquake. Overall, the Overall, available the geodeticavailable data geodetic have data a relatively have a betterrelatively resolution better 206 aboveresolution the 12above km depth.the 12 km depth.

207 4.3. Modeling Results for the 2017 Mw 6.5 Mainling Earthquake 208 Figure5 5 shows shows the the modeling modeling results results for for the the 2017 2017 Mainling Mainling earthquake. earthquake. The The inverted inverted coseismic coseismic slip 209 distributionslip distribution (Figure (Figure5a; fault 5a; fault slip data slip aredata provided are provided in the in supplementary the supplementary materials) materials) reveals reveals one large one 210 sliplarge patch, slip patch, extending extending 30 km 30 along km thealong strike. the Thestrike. maximum The maximum slip (1.03 slip m) (1.03 was locatedm) was on located the northern on the 211 segmentnorthern ofsegment the fault, of the near fault, the surface.near the Thesurface. slip The model slip also model indicates also indicates that the earthquakethat the earthquake rupture 212 reachedrupture the reached surface the and surface was concentrated and was concentrated above a 12 abovekm depth. a 12 The km slip depth. was The mainly slip characterized was mainly 213 bycharacterized reverse faulting, by reverse consistent faulting, with the consistent GCMT solutionwith the determined GCMT solution from the determined teleseismic fro datam the [1]. 214 Theteleseismic determined data coseismic [1]. The determinedslip is dominantly coseismic located slip in is the dominantly zone where located the geodetic in the observations zone where have the 215 goodgeodetic resolution observations (according have good to the resolution checkerboard (according test), suggesting to the checkerboard that the slip test), is well suggesting constrained that the by 216 theslip data.is well Assuming constrained a shearby the modulus data. Assuming of 30 GPa, a shea ther determinedmodulus of 30 slip GPa, model the determined has a total moment slip model of 18 7.49 10 N m, corresponding18 to a moment magnitude of 6.55, which is close to the oﬃcially released 217 has a× total moment· of 7.49 × 10 N∙m, corresponding to a moment magnitude of 6.55, which is close 218 valueto the ofofficially Mw 6.5 released [9]. value of Mw 6.5 [9].

219 220 Figure 5. Modeling results. ( a) Coseismic slip distribution determined from both globalglobal positioningpositioning 221 system (GPS) and interferometric synthetic aperture radar (InSAR) data; green arrows indicate slip 222 directions. ( b) Comparison between between the the observed (blue (blue arrows) arrows) and modeled (red (red arrows) GPS data data;; 223 lowerlower panels show show the the (c ()c )down down-sampled-sampled InSAR InSAR data data,, (d) ( dmodeling) modeling result, result, and and (e) their (e) their residuals. residuals. The 224 Thered rectangles red rectangles in panels in panels b–e b–eindicate indicate the fault the fault plane plane and andthe solid the solid red redline linemarks marks the fault the fault trace. trace.

This slip model ﬁts well with the observed coseismic displacements, providing RMS values of 4.3 mm for the GPS data and 15.2 mm for the InSAR data. Comparisons between the observed and

Remote Sens. 2019, 11, 2940 8 of 14 modeled displacement fields (Figure5b–e) demonstrate small misfits overall, without systemic bias. Some large localized misfits could be related to heterogeneity that our elastic, half-space model could not account for. As seen in Figure5a, aftershocks were dominantly located around the coseismic rupture, where coseismic stress perturbation is expected to be high, owing to the strong slip gradient. The occurrence of aftershocks demonstrates one type of postseismic relaxation (beside afterslip, viscoelastic relaxation, and poroelastic rebound) in response to the coseismic stress perturbation. We also tested the modeled coseismic slip from the GPS and InSAR data independently (Figure6), and the results indicate that although the modeled slip based on each of these two data types is different from that of the joint inversion (Figure5a), they share common features. This suggests that both InSAR and GPS data play important roles in constraining the coseismic slip distribution, with good consistency, despite the data being collected during slightly different time periods.

Figure 6. Slip models. Models using (a) interferometric synthetic aperture radar (InSAR) data only and using (b) global positioning system (GPS) data only.

5. Discussion Incorporating both the GPS and InSAR measured coseismic displacement data, including the relocated aftershocks, we constructed a slip model for the 2017 Mw 6.5 Mainling earthquake. The model can explain the geodetic data well, and, in general, the slip displays a disjointed distribution with the aftershocks. Using our results, we addressed two questions: (1) Was the secondary parallel fault involved in the coseismic rupture of the 2017 Mainling earthquake, as suggested by Xiong et al. [6]?; (2) What were the stress/strain conditions before and after the 2017 Mainling earthquake and what are the implications for regional seismic hazard?

5.1. Single-Plane Rupture or Double-Plane Rupture Most studies have suggested that the Mainling earthquake ruptured a single fault [1,5,8], and this is consistent with our results. However, Xiong et al. [6] suggested that the Mainling earthquake ruptured two parallel faults [6], which are outlined by the aftershock distribution (Figures1c and7). To investigate whether both fault planes were involved in the coseismic rupture of the 2017 Mainling earthquake, we investigated the spatiotemporal evolution of the relocated aftershocks (Figure7). The results indicate that the aftershocks ﬁrst occurred on the fault constructed in our coseismic slip model (Fault A). The ﬁrst aftershock on the secondary parallel fault (Fault B) occurred 15 min after the mainshock. Both Fault A and Fault B belong to the Xixingla fault system [11,12]. The seismic moment release level along Fault B was higher in the later time period than that along Fault A. This temporal-spatial evolution of relocated aftershocks suggests that the aftershocks along Fault B were likely triggered by the coseismic rupture along Fault A.

Remote Sens. 2019, 11, 2940 9 of 14 Remote Sens. 2019, 11, x FOR PEER REVIEW 9 of 14 Remote Sens. 2019, 11, x FOR PEER REVIEW 9 of 14

262262 263263 FigureFigure 7. 7. 7.Early EarlyEarly aftershocks. aftershocks. aftershocks. (a) Aftershocks(a) Aftershocks within within within the first the the 15 first firstmin 15 after 15 min min the after mainshock. after the the mainshock. mainshock. (b) Aftershocks (b ) (b) 264264 withinAftershocksAftershocks the first within within hour the the after firstfirst the hourhour mainshock. after the (mainshock.mainshock.c) Aftershocks ( c()c )Aftershocks withinAftershocks the firstwithin within 7 days the the first after first 7 days the7 days mainshock. after after the the 265265 mainshock.mainshock. To further investigate the triggering effect, we calculated the coseismic Coulomb stress change, 266266 basedTo onTo further further our slip investigate investigate model (Figure thethe triggering5a), using effect,effect, the method we we calcul calcul ofated Scholzated the the [ 34coseismic coseismic]: Coulomb Coulomb stress stress change, change, 267267 basedbased on on our our slip slip model model (Figure(Figure 5a), using thethe methodmethod of of Scholz Scholz [34]: [34]: ∆σc = ∆τs µ(∆σn), (1) − Δσc = Δτs − μ(Δσn), (1) Δσc = Δτs − μ(Δσn), (1) where ∆σc is the Coulomb stress change on the receiver fault, ∆τs and ∆σn represent changes in shear 268 where Δσc is the Coulomb stress change on the receiver fault, Δτs and Δσn represent changes in shear 268 andwhere normal Δσc is stressthe Coulomb (positive stress along change the slip on directionthe receiver of thefault, fault), Δτs and respectively, Δσn represent and µchangesis the einffective shear 269 and normal stress (positive along the slip direction of the fault), respectively, and μ is the effective 269 frictionand normal coeffi stresscient (ranging(positive from along 0 tothe 1) slip [34 ].direction Using our of slipthe modelfault), (Figurerespectively,5a), we and considered μ is the aeffective typical 270 friction coefficient (ranging from 0 to 1) [34]. Using our slip model (Figure 5a), we considered a typical 270 friction coefficient coefficient, (rangingµ, of 0.4 from [35,36 0] to and 1) [34]. calculated Using Coulombour slip model stress (Figure changes 5a), at we 4, 8,considered and 12 km a depthstypical 271 friction coefficient, μ, of 0.4 [35,36] and calculated Coulomb stress changes at 4, 8, and 12 km depths (Figure8). The results indicate that Fault B is located in the zone of positive Coulomb stress change, 271272 friction(Figure coeffic 8). Theient, results μ, of indicate 0.4 [35,36] that andFault calculated B is located Coulomb in the zone stress of positivechanges Coulomb at 4, 8, and stress 12 kmchange, depths and that the magnitude could be high enough to trigger earthquakes along Fault B. 272273 (Figureand that 8). the The magnitude results indicate could bethat high Fault enough B is locatedto trigger in earthquak the zone esof alongpositive Fault Coulomb B. stress change, 273 and that the magnitude could be high enough to trigger earthquakes along Fault B.

274 274275 FigureFigure 8. 8.Coulomb Coulomb failurefailure stressstress change based based on on the the slip slip model model in in Figure Figure 5a.5a. Results Results for for depths depths of ( ofa) ( a) 276 4 km, (b) 8 km, and (c) 12 km. Yellow points denote aftershocks. 275 4Figure km, ( b8.) Coulomb 8 km, and failure (c) 12 km.stress Yellow change points based denote on the aftershocks. slip model in Figure 5a. Results for depths of (a) 276 4 km, (b) 8 km, and (c) 12 km. Yellow points denote aftershocks. 277 WeWe also also modeled modeled the the coseismiccoseismic slip using two two parallel parallel fault fault planes planes (Figure (Figure 9).9 ).The The double double-plane-plane 278 modelmodel shows shows strong strong rupturing rupturing along along Fault Fault A, A, similar similar to to that that shown shown in in Figure Figure5 a,5a, but but a a much much weaker weaker slip 277 We also modeled the coseismic slip using two parallel fault planes (Figure 9). The double-plane 279 alongslip along Fault B.Fault The B. results The results show show that thisthatmodel this model does does not improvenot improve with with the geodeticthe geodetic data dat ﬁtting,a fitting, where 278 model shows strong rupturing along Fault A, similar to that shown in Figure 5a, but a much weaker 280 thewhere residuals the residuals increased increased by 12.24% by when12.24% compared when compared with the with one-fault the one model.-fault model. On this On basis, thiswe basis, conclude we 279 slip along Fault B. The results show that this model does not improve with the geodetic data fitting, 281 thatconclude Fault Athat played Fault aA dominant played a dominant role during role the during 2017 the Mw 2017 6.5 Mw Mainling 6.5 Mainling earthquake, earthquake,and thatand Faultthat B 280282 whereFault theB was, residuals at most, increased marginally by involved12.24% when in the compared dynamic rupture. with the one-fault model. On this basis, we was, at most, marginally involved in the dynamic rupture. 281 conclude that Fault A played a dominant role during the 2017 Mw 6.5 Mainling earthquake, and that 282 Fault B was, at most, marginally involved in the dynamic rupture.

Remote Sens. 2019, 11, 2940 10 of 14 Remote Sens. 2019, 11, x FOR PEER REVIEW 10 of 14

283 284 Figure 9. Slip distribution along Fault A (a) andand FaultFault BB ((bb).). (c) Overall view ofof thethe double-planedouble-plane 285 modeling results.

5.2. Stress/Strain State before and after the Mainling Earthquake 286 5.2. Stress/Strain State before and after the Mainling Earthquake To understand the stress/strain state before and after the 2017 Mw 6.5 Mainling earthquake, 287 To understand the stress/strain state before and after the 2017 Mw 6.5 Mainling earthquake, we we calculated the b-value from the regional seismicity and the strain rate from the long-term geodetic 288 calculated the b-value from the regional seismicity and the strain rate from the long-term geodetic data. Then, we analyzed the impact of the earthquake on regional seismic hazard from the perspective 289 data. Then, we analyzed the impact of the earthquake on regional seismic hazard from the of Coulomb stress change as shown in Figure8. 290 perspective of Coulomb stress change as shown in Figure 8. The frequency-magnitude distribution of an earthquake usually follows the Gutenberg-Richter law: 291 The frequency-magnitude distribution of an earthquake usually follows the Gutenberg-Richter 292 law: log10N = a bM, (2) − where N is the number of earthquakes, equal to or greater than magnitude M, a reﬂects the seismic log10N = a − bM, (2) level of the target region, and b represents relative frequency of small and large earthquakes and 293 haswhere been N thoughtis the number to measure of earthquakes, the variation equal of theto or stress greater in the than crust magnitude [37]. Therefore, M, a reflects spatial the/temporal seismic 294 variationslevel of the of target the b region,-value provideand b represents important relative insights freque intoncy stress of small states and relating large toearthquakes large earthquakes, and has 295 andbeen are thought important to measure in seismic the hazard variation assessment. of the stress Previous in the studies crust [37]. have Therefore, shown that spatial/temporalb-values in the 296 epicentralvariations areaof the are b-value small beforeprovide the important occurrence insights of a large into earthquakestress states [ 38relating–41]. For to large example, earthquakes, both the 297 2011and are Mw imp 9.0ortant Tohoku in and seismic the 2004 hazard Mw assessment. 9.1 Sumatra Previous earthquakes studies were have preceded shown by that decreasing b-valuesb-values in the 298 inepicentral the main area rupture are small zones before [42]. Shithe etoccurrence al. [43] found of a thatlargeb-values earthquake decreased [38–41]. systematically For example, before both the 299 20082011 MwMw 7.99.0 WenchuanTohoku and earthquake. the 2004 Mw 9.1 Sumatra earthquakes were preceded by decreasing b- 300 valuesFor in the the 2017 main Mw rupture 6.5 Mainling zones [42]. earthquake, Shi et al. Han[43] etfound al. [ 4that] analyzed b-values temporal decreased variations systematically in the 301 bbefore-value the before 2008 the Mw earthquake, 7.9 Wenchuan however, earthquake. they did not include spatial variations. In this study, we used a 50-year regional catalogue (from 1965 to the occurrence of the 2017 Mainling earthquake) [10] to map 302 For the 2017 Mw 6.5 Mainling earthquake, Han et al. [4] analyzed temporal variations in the b- spatial variations in b-values [44]. According to the frequency-magnitude distribution, the complete 303 value before the earthquake, however, they did not include spatial variations. In this study, we used magnitude in this region is approximately M 2.0. We used the ZMAP [45] to calculate b-values over 304 a 50-year regional catalogue (from 1965 to the occurrence of the 2017 Mainling earthquake) [10] to a grid size of 0.2 0.2 (Figure 10a). The results show low b-values in the epicentral region prior to 305 map spatial variations◦ × ◦ in b-values [44]. According to the frequency-magnitude distribution, the the Mainling earthquake, indicating a high stress state. 306 complete magnitude in this region is approximately M 2.0. We used the ZMAP [45] to calculate b- 307 values over a grid size of 0.2° × 0.2° (Figure 10a). The results show low b-values in the epicentral 308 region prior to the Mainling earthquake, indicating a high stress state.

Remote Sens. 2019, 11, 2940 11 of 14

Remote Sens. 2019, 11, x FOR PEER REVIEW 11 of 14

309 310 FigureFigure 10. 10.Map Map view view of of b-value b-value (a(a)) andand secondsecond invariantinvariant of of the the strain strain rate rate (b (b) )before before the the Mainling Mainling 311 earthquake.earthquake. Blue Blue arrows arrows denote denote global global positioningpositioning system (GPS) (GPS)-derived-derived displacement displacement rates rates and and the the 312 greengreen star star denotes denotes the the epicenter epicenter of of the the Mainling Mainling earthquake.earthquake.

313 WeWe also also calculated calculated thethe strain rate rate of of the the eastern eastern Himalayan Himalayan syntaxis, syntaxis, based based on the on long the-term long-term GPS 314 GPSdisplacement displacement rates rates [46]. [ 46Using]. Using an L- ancurve L-curve method method [47], we [47 interpolated], we interpolated the GPS thevelocity GPS field velocity by 0.4° field 315 by× 0.4 0.4° and0.4 calculatedand calculated the second the second invariant invariant of the strain of the rate strain tensor rate (effective tensor (e strainffective rate), strain which rate), ◦ × ◦ 316 whichcharacterizes characterizes the strainthe strain state state of the of theupper upper crust crust [47]. [47 The]. Theresults results indicate indicate a high a high strain strain rate rate in inthe the 317 epicentralepicentral area area (Figure (Figure 10 b),10b), which which is consistentis consistent with with the the high high stress stress conditions conditions reflected reflected in the in bthe-values. b- 318 Bothvalues. the b Both-values the andb-values strain and rate strain results rate suggest results that suggest the 2017 that Mainlingthe 2017 Mainling earthquake earthquake was nucleated was 319 nucleated under high stress/strain conditions. Furthermore, we identified two regions in the eastern under high stress/strain conditions. Furthermore, we identified two regions in the eastern Himalayan 320 Himalayan syntaxis, region A (along the eastern Himalaya Main Thrust) and region C (along the syntaxis, region A (along the eastern Himalaya Main Thrust) and region C (along the northern ends of 321 northern ends of the Sagaing fault), with high strain rates and low b-values. These regions pose a the Sagaing fault), with high strain rates and low b-values. These regions pose a high seismic hazard. 322 high seismic hazard. From the perspective of Coulomb stress change (Figure 8), we have found that From the perspective of Coulomb stress change (Figure8), we have found that the triggering e ffect of 323 the triggering effect of the Mainling earthquake was local. The earthquake imposed only very small 324 thestress Mainling perturbations earthquake on wasadjacent local. faults, The earthquakeand, as such, imposed its impact only on verythe regional small stress seismic perturbations hazard was on 325 adjacentsmall. faults, and, as such, its impact on the regional seismic hazard was small. 6. Conclusions 326 6. Conclusions 327 InIn this thi study,s study, we we modeled modeled GPS GPS andand InSARInSAR displacementdisplacement data data for for the the 2017 2017 Mw Mw 6.5 6.5 Mainling Mainling 328 earthquake.earthquake. Incorporating Incorporating the the relocated relocated aftershocks,aftershocks, we performed performed a a joint joint inversion inversion toto determine determine the the 329 seismogenicseismogenic fault fault plane plane and and coseismic coseismic slip slip distribution.distribution. Based Based on on our our resu results,lts, the the earthquake earthquake occurred occurred 330 alongalong a curved a curved fault fault plane plane outlined outlined by relocated by relocated aftershocks. aftershocks. The mainshock The mainshock likely likely triggered triggered a parallel a 331 faultparallel plane, fault along plane, which along numerous which aftershocks numerous also aftershocks occurred. also The occurred. coseismic The rupture coseismic extended rupture30 km 332 alongextended strike 30 and km 12 along km strike downdip. and 12 The km determineddowndip. The rupture determined reached rupture the surface,reached the with surface, a maximum with 333 slipa of maximum 1.03 m. The slip earthquake of 1.03 m. The released earthquake a total releasedseismic moment a total seismic of 7.49 moment1018 N ofm, 7.49 corresponding × 1018 N∙m, to × · 334 a magnitudecorresponding of Mw to a 6.55.magnitude Using of the Mw interseismic 6.55. Using GPS the interseismic velocity field GPS for velocity 1999–2016 field and for 1999 the regional–2016 335 seismicityand the betweenregional seismicity 1965 and 2017,between we calculated1965 and 2017, the secondwe calculated invariant the of second the strain invariant rate and of theb-values strain of 336 therate eastern and b Himalayan-values of the syntaxis. eastern Himalayan The results syntaxis. demonstrate The results that the demonstrate epicentral that area the of epicentral the 2017 Mw area 6.5 337 Mainlingof the 2017 earthquake, Mw 6.5 Mainling before itsearthquake, occurrence, before had its high occurrence, strain rates had and high low strainb-values, rates and indicating low b-values, a high 338 stressindicating state. This a high suggests stress state. that theThis 2017 suggests Mainling that earthquakethe 2017 Mainling was nucleated earthquake under was anucleated high stress under/strain 339 state.a high Furthermore, stress/strain thestate. spatial Furthermore, distributions the spatial of strain distributions rate and ofb -valuesstrain rate show and thatb-values zones show along that the 340 easternzones Himalaya along the Maineastern Thrust Himalaya and the Main northern Thrust end and of the the northern Sagaing faultend of remain the Sagaing under highfault stressremain and 341 under high stress and continue to pose a high seismic hazard. continue to pose a high seismic hazard.

Remote Sens. 2019, 11, 2940 12 of 14

Supplementary Materials: The following are available online at http://www.mdpi.com/2072-4292/11/24/2940/s1. Author Contributions: Methodology, H.J., L.W., Y.L. (Yanchuan Li) and X.Y.; software, H.J., L.W., Y.L. (Yanchuan Li); validation, W.G. (Weijun Gan), S.L., Y.L. (Yanchuan Li), K.Z., Y.L. (Yunhua Liu), X.Y.; data curation, H.J., W.G. (Weijun Gan), S.L., Y.L. (Yanchuan Li), K.Z., Y.L. (Yunhua Liu), W.G. (Wenyu Gong), X.Y.; writing—original draft preparation, H.J.; writing—review and editing, H.J., L.W., W.G. (Wenyu Gong), S.L., Y.L. (Yanchuan Li), K.Z., Y.L. (Yunhua Liu), X.Y.; funding acquisition, L.W. Funding: This work was supported by the National Natural Science Foundation of China (grant number 4167040536 and U1839211). Acknowledgments: We are grateful to all of the staff involved in GPS data collection, to Sarah Henton De Angelis for the language revision, to ESA for providing Sentinel InSAR data (https://scihub.esa.int), to Wang Rongjiang for the SDM inversion program, and to Zhou Yijian for his assistance in the b-value calculation process. Figures were drawn using the GMT software (https://gmt.soest.hawaii.edu). Conflicts of Interest: The authors declare no conflict of interest.

References

1. Liu, Y.; Shan, X.; Zhang, Y.; Zhao, D.; Qu, C. Use of seismic waveforms and InSAR data for determination of the seismotectonics of the mainling Ms 6.9 earthquake on NOV.18, 2017. Seismol. Geol. 2018, 40, 79–100. 2. Bai, L.; Li, G.; Song, B. The source parameters of the M 6.9 Mainling, Tibet earthquake and its tectonic implications. Chin. J. Geophys. 2017, 60, 4956–4963. (In Chinese) 3. Yin, X.; Zhou, B.; Chen, J.; Wei, W.; Xie, C.; Guo, Z. Spatial-temporal distribution characteristics of early aftershocks following the M6.9 Mainling earthquake in Tibet, China. Chin. J. Geophys. 2018, 61, 2322–2331. (In Chinese) 4. Han, J.; Yang, J.; Wang, W. Relocation of the aftershock sequence of Mainling Ms 6.9 earthquake in 2017 and spatio-temporal variation characteristics of b-value. Acta Seismol. Sin. 2019, 41, 169–180. 5. Wei, W.; Xie, C.; Zhou, B.; Guo, Z.; Yin, X.; Li, B.; Wang, P. Location of the mainshock and aftershock sequences of the M 6.9 Mainling earthquake, Tibet. Chin. Sci. Bull. 2018, 63, 1493–1501. (In Chinese) [CrossRef] 6. Xiong, W.; Chen, W.; Wen, Y.; Liu, G.; Nie, Z.; Qiao, X.; Xu, C. Insight into the 2017 Mainling Mw 6.5 earthquake: A complicated thrust event beneath the Namche Barwa syntaxis. Earth Planets Space 2019, 71, 71. [CrossRef] 7. Hu, S.; Yao, H. Crustal velocity structure around the eastern Himalayan syntaxis: Implications for the nucleation mechanism of the 2017 Ms 6.9 Mainling earthquake and regional tectonics. Tectonophys 2018, 744, 1–9. [CrossRef] 8. Wang, W.; Yang, J.; Wang, Y.; Jiang, X.; Zheng, Y. Imaging velocity structures and aftershock distributions in the source region of the 2017 Mainling M 6.9 earthquake. Chin. J. Geophys. 2019, 62, 2048–2058. (In Chinese) 9. Li, W.; Xu, C.; Yi, L.; Wen, Y.; Zhang, X. Source parameters and seismogenic structure of the 2017 Mw 6.5 Mainling earthquake in the Eastern Himalayan Syntaxis (Tibet, China). J. Asian Earth Sci. 2019, 169, 130–138. [CrossRef] 10. China Earthquake Datacenter. Available online: http://data.earthquake.cn/index.html (accessed on 6 October 2017). 11. Xie, C. A Study on Tectonic Geomorphology of Namche Barwa and Activity of the Faults. Ph.D. Thesis, Institute of Geology, China Earthquake Adminstration, Beijing, China, 2019. (In Chinese). 12. The 1:250000 Geological Map of Tibet Autonomous Region. Chengdu Inst. Geol. Miner. Resour. China 2003, 1. (In Chinese) 13. Bai, L.; Li, G.; Khan, N.; Zhao, J.; Ding, L. Focal depths and mechanisms of shallow earthquakes in the Himalayan–Tibetan region. Gondwana Res. 2017, 41, 390–399. [CrossRef] 14. Xu, Z.; Ji, S.; Cai, Z.; Zeng, L.; Geng, Q.; Cao, H. Kinematics and dynamics of the Namche Barwa Syntaxis, eastern Himalaya: Constraints from deformation, fabrics and geochronology. Gondwana Res. 2012, 21, 19–36. [CrossRef] 15. Wang, Q.; Zhang, P.; Freymueller, J.; Bilham, R.; Larson, K.; Lai, X.; Liu, J. Present-day crustal deformation in China constrained by Global Positioning System Measurements. Science 2001, 294, 574–577. [CrossRef] [PubMed] 16. Zhang, P.; Shen, Z.; Wang, M.; Gan, W.; Bürgmann, R.; Molnar, P.; Hanrong, S. Continuous deformation of the Tibetan Plateau from Global Positioning System data. Geology 2004, 32, 809–812. [CrossRef] Remote Sens. 2019, 11, 2940 13 of 14

17. Gan, W.; Zhang, P.; Shen, Z.; Niu, Z.; Wang, M.; Wan, Y.; Cheng, J. Present-day crustal motion within the Tibetan Plateau inferred from GPS measurements. J. Geophys. Res. 2007, 112–126. [CrossRef] 18. Liang, S.; Gan, W.; Shen, C.; Xiao, G.; Liu, J.; Chen, W.; Zhou, D. Three-dimensional velocity field of present-day crustal motion of the Tibetan Plateau derived from GPS measurements. J. Geophys. Res. 2013, 118, 5722–5732. [CrossRef] 19. Burg, J.; Nievergelt, P.; Oberli, F.; Seward, D.; Davy, P.; Maurin, J.; Diao, Z.; Meier, M. The Namche Barwa syntaxis: Evidence for exhumation related to compressional crustal folding. J. Asian Earth Sci. 1998, 16, 239–252. [CrossRef] 20. Ding, L.; Kapp, P.; Wan, X. Paleocene–Eocene record of ophiolite obduction and initial India-Asia collision, south central Tibet. Tectonics 2005, 24, TC3001. [CrossRef] 21. Tapponnier, P.; X, Z.; Roger, F.; Meyer, B.; Arnaud, N.; Wittlinger, G.; Yang, J. Oblique stepwise rise and growth of the Tibet Plateau. Science 2001, 294, 1671–1677. [CrossRef] 22. Wang, P.; Scherler, D.; Liu-Zeng, J.; Mey, J.; Avouac, J.; Zhang, Y.; Shi, D. Tectonic control of Yarlung Tsangpo Gorge revealed by a buried canyon in Southern Tibet. Science 2014, 346, 978–981. [CrossRef] 23. Chen, W.; Molnar, P. Seismic moments of major earthquakes and the average rate of slip in central Asia. J. Geophys. Res. 1977, 82, 2945–2969. [CrossRef] 24. Herring, T.; King, R.; Floyd, M.; McClusky, S. GAMIT Reference Manual. GPS Analysis at MIT GLOBK, Release 10.6. Available online: http://www-gpsg.mit.edu/~{}simon/gtgk/GAMIT_Ref.pdf (accessed on 6 October 2017). 25. Saastamoinen, J. Atmospheric Correction for the Troposphere and Stratosphere in Radio Ranging Satellites. Geophys. Monogr. Ser. 2013, 247–251. 26. Werner, C.; Wegmüller, U.; Strozzi, T.; Wiesmann, A. Gamma SAR and interferometric processing software. In Proceedings of the Ers-Envisat Symposium, Gothenburg, Sweden, 16–20 October 2000; p. 1620. 27. Werner, C.; Wegmüller, U.; Strozzi, T.; Wiesmann, A. Processing strategies for phase unwrapping for INSAR applications. In Proceedings of the European Conference on Synthetic Aperture Radar, Cologne, Germany, 4–6 June 2002; pp. 353–356. 28. Shen, Z.; Sun, J.; Zhang, P.; Wan, Y.; Wang, M.; Bürgmann, R.; Zeng, Y.; Gan, W.; Liao, H.; Wang, Q. Slip maxima at fault junctions and rupturing of barriers during the 2008 Wenchuan earthquake. Nat. Geosci. 2009, 2, 718. [CrossRef] 29. Ben-Menahem, A.; Singh, S. Multipolar elastic fields in a layered half space. Bull. Seismol. Soc. Am. 1968, 58, 1519–1572. 30. Singh, S. Static deformation of a multilayered half-space by internal sources. J. Geophys. Res. 1970, 75, 3257–3263. [CrossRef] 31. Okada, Y. Internal deformation due to shear and tensile faults in a half-space. Bull. Seismol. Soc. Am. 1992, 82, 1018–1040. 32. Wang, R.; Diao, F.; Hoechner, A. SDM—A geodetic inversion code incorporating with layered crust structure and curved fault geometry. In Proceedings of the General Assembly European Geosciences Union, Vienna, Austria, 7–12 April 2013. 33. Wang, L.; Hainzl, S.; Zöller, G.; Holschneider, M. Stress-and aftershock-constrained joint inversions for coseismic and postseismic slip applied to the 2004 M 6.0 Parkfield earthquake. J. Geophys. Res. 2012, 117, B07406. [CrossRef] 34. Scholz, C. The Mechanics of Earthquakes and Faulting; Cambridge University Press: New York, NY, USA, 1990. 35. King, G.; Stein, R.; Lin, J. Static stress changes and the triggering of earthquakes. Bull. Seismol. Soc. Am. 1994, 84. 36. Yin, F.; Han, L.; Jiang, C.; Shi, Y. Interaction between the 2017 M 6.9 Mainling earthquake and the 1950 M 8.6 Zayu earthquake and their impacts on surrounding major active faults. Chin. J. Geophys. 2018, 61, 3185–3197. (In Chinese) 37. Scholz, C. On the stress dependence of the earthquake b value. Geophys. Res. Lett. 2015, 42, 1399–1402. [CrossRef] 38. Guha, S. Premonitory crustal deformations, strains and seismotectonic features (b-values) preceding Koyna earthquakes. Tectonophysics 1979, 52, 549–559. [CrossRef] 39. Imoto, M. Changes in the magnitude—Frequency b-value prior to large (M 6.0) earthquakes in Japan. ≥ Tectonophysics 1991, 193, 311–325. [CrossRef] Remote Sens. 2019, 11, 2940 14 of 14

40. Molchan, G.; Kronrod, T.; Nekrasova, A. Immediate foreshocks: Time variation of the b-value. Phys. Earth Planet Inter. 1999, 111, 229–240. [CrossRef] 41. Molchan, G.; Dmitrieva, O. Dynamics of the magnitude—frequency relation for foreshocks. Phys. Earth Planet Inter. 1990, 61, 99–112. [CrossRef] 42. Nanjo, K.; Hirata, N.; Obara, K.; Kasahara, K. Decade-scale decrease in b value prior to the M 9-class 2011 Tohoku and 2004 Sumatra quakes. Geophys. Res. Lett. 2012, 39. [CrossRef] 43. Shi, H.; Meng, L.; Zhang, X.; Chang, Y.; Yang, Z.; Xie, W.; Katsumi, H.; Han, P. Decrease in b value prior to the Wenchuan earthquake (Ms 8.0). Chin. J. Geophys. 2018, 61, 1874–1882. (In Chinese) 44. Sreejith, K.; Sunil, P.; Agrawal, R.; Saji, A.; Rajawat, A.; Ramesh, D. Audit of stored strain energy and extent of future earthquake rupture in central Himalaya. Sci. Rep. 2018, 8, 16697. [CrossRef] 45. Swiss Seismological Service. Available online: http://www.seismo.ethz.ch/en/research-and-teaching/ products-software/software/ZMAP/ (accessed on 6 October 2017). 46. Li, Y.; Shan, X.; Qu, C.; Liu, Y.; Han, N. Crustal deformation of the Altyn Tagh fault based on GPS. J. Geophys. Res. 2018, 123, 10309–10322. [CrossRef] 47. Li, Y.; Shan, X.; Song, X.; Jiang, Y.; Gan, W.; Qu, C.; Wang, Z. Fault locking and slip rate deﬁcit on the middle and southern segment of the Tancheng-Lujiang fault inverted from GPS data. Chin. J. Geophys. 2016, 59, 4022–4034. (In Chinese)

© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).