MOMENT TENSOR AND SOURCE PROCESS OF EARTHQUAKES IN FIJI REGION OBTAINED BY WAVEFORM INVERSION Seru Sefanaia* Supervisor: Yui YAGI** MEE0767 ABSTRACT We evaluated the uality of the loal seismi waveform reorded at Yasawa station established by the Mineral Resoures Department (MRD, Fii), and tried to estimate moment tensor solution by using this waveform data. We also estimated moment tensor solution and seismi soure proess by using teleseismi body wave so as to investigate the faulting system in Fii region. As for the evaluation of the uality of the waveform reorded at the Yasawa station, we found that EW and NS omponents were down freuently, and moment tensor solution obtained by 3 omponents of the Yasawa station is totally opposite to the Harvard CMT solution. This result may suggest that the polarity of the loal seismometer omponents was in a reverse diretion and magnifiation of the seismometer response information is not orret at this stage. We obtained a onsistant result when we divided by -0 to observed waveform. Moment tensor inversion analysis was also arried out for 30 events by using teleseismi body-wave downloaded from the IRIS-DMC. Final results are well onsistent to Harvard CMT solution. Rupture soure proess analysis was onduted for the two deep intraslab events that ourred respetively along the Tonga Kermede subdution zone and two shallow depth events whih ourred respetively along the transform fault zones. The rupture proesses for the two deep earthuakes are haraterized by the rupture propagating mainly along the strike of the fault plane. The two shallow events ourred along the transition zone also haraterized by the rupture propagating along the strike of the Fii Frature Zone but are ontrolled by the geometry of the fault. Keywords: Moment tensor inversion, Rupture proess inversion, teleseismi data, near-soure data. INTRODUCTION Earthuake foal mehanisms are basi and important information in seismology and have been utilized for understanding regional tetoni stress fields, soure mehanisms of large earthuakes, simulation of strong motion, faulting systems, tsunami generation and so on. It is important to estimate foal mehanisms of earthuakes routinely as well as to make a homogenous atalogue of moment tensors (or foal mehanisms) from small to large earthuakes for future referenes. Moment tensor and rupture proess inversion analysis has not so far being arried out in Fii. Therefore in this study, we evaluated the uality of the loal seismi waveform data obtained from the Yasawa broadband station established by Mineral Resoures Department (MRD, Fii), then we tried to estimate moment tensor solution by using this waveform data and as well as to estimate moment tensor solution and seismi soure proess using teleseismi body wave so as to investigate the faulting system in the Fii region. *Mineral Resoures Department (Seismology Setion) of Fii. **Ass. Professor, Institute of Physial Sienes, University of Tsukuba, Japan.
DATA We used both near-soure data reorded at Yasawa broadband station established by Mineral Resoures Department, Fii with range from Marh 2003 to Deember 2005 as well as teleseismi body-wave (P-wave) data reorded at FDSN and GSN network stations olleted by the Data Management Center of the Inorporated Researh Institution for Seismology (IRIS-DMC). Table : Shows the uality of Yasawa station omponents o Waveforms well reorded. x Component broken. The raw ontinuous Yasawa waveform data was onverted from nanometris to seed format by makeseed software provided by Nanometris In. Further, we obtained the waveform data in SAC format by using rdseed software whih was available from the IRIS- DMC homepage. The broadband data were onverted and numerially integrated from volts into the ground veloity m/s and then band-passed between 0.0 and 0. Hz to redue the effet of fine 3D struture and detailed soure proess. Table shows the uality of waveforms reorded at the Yasawa station in eah period. For teleseismi bodywaves we retrieved and seleted the IRIS-DMC stations from the viewpoint of good azimuthal overage with the distane range from 30 to 90. We referred to earthuake event list from the Harvard CMT atalogue from year 990 to present. Thirty events with Mw > 6.5 ourred within Fii region were analyzed for moment tensor inversion and 4 of these events were analyzed for their soure rupture proesses. Moment Tensor Inversion THEORY AND METHODOLOGY To obtain moment tensor omponents of overall earthuake, we represented the seismi soure proess as point soure model. 5 u ( t) = G ξ )* M ξ ) + eo + e = 5 = ( G t, )* T ( t) ) + eo em m = M ( ξ + () " where M and M are moment tensor at entroid of soureξ, T(t) is soure time funtion, and e m is modeling error. For simpliity, the observation euation of () an be rewritten in vetor form: d = H( T ( t), ξ ) a + e (2) 2
where d and e are N-dimensional data and error vetors, respetively, a is a 5-dimensional model parameter vetor, H is a N x 5 oeffiient matrix. The solution of the above matrix euation is obtained by using least suare approah, if we assumed soure time funtion and the entroid loation. We determined the depth of entroid and the duration and shape of soure time funtion by grid searh method sine these are needed for moment tensor inversion for the analysis of teleseismi body wave. For analyses of near-soure data, we applied low pass filter to mitigate soure time funtion effet and then estimated the depth of entroid by grid searh method. Rupture Proess Inversion To estimate rupture proess, we basially followed the formulation by Yagi and Fukahata (2008). A vertial omponent of P-wave at a station due to a shear disloation soure on a fault plane S is given by u ( t) G ξ )* D& ξ) ds + e, (3) 2 = = where G is a Green s funtion, D & is a spatio-temporal slip-rate distribution, and * denotes the onvolution operator in time domain. To formulate the inverse problem, we represent the spatiotemporal slip-rate distribution D & by linear ombination of a finite number of basis funtions: o D& ξ ) K L = k= l= a kl X k ()() ξ T t + δd& ξ), (4) where a kl are model parameters to be determined from observed data. X k ( ξ) and T l ()are t the basis funtions for spae and time, respetively. We an rewrite in vetor form: d = H a + e m, (5) where d and e m are N -dimensional data and error vetors, respetively, a is a M-dimensional model parameter vetor, and H is a N M oeffiient matrix. RESULTS AND DISCUSSIONS The results of this study are divided into three (3) main parts: Part I: Moment Tensor Inversion using near soure data obtained by Loal Seismi Station The loal data set obtained for this study was from the period of its establishment in Marh, 2003 to Deember, 2005. We estimated the entroid moment tensor solution by using loal seismi data whih ontains three omponents waveform. Green s funtion for near-field was alulated using program by Kohketsu (985), and performed waveform inversion using program by Yagi (2006). To obtain the loation of entroid, we applied the grid searh method. 3
a) b) ) We found the inonsistenies between this study solution (Figure a) and the Harvard CMT solution even if waveform fitting is good (Figure b). The result for this study is totally opposite to the Harvard CMT solution. One of the possibilities for this differene may be the polarity set up (EW and NS) omponents is in a reverse diretion. If we divide by - to the waveform in our SAC program, the foal mehanism is well onsistent with the result of Harvard CMT (Figure ). We should point out that our result (Figure ) is ten times larger in seismi moment than the Harvard result. This result implies that the magnifiation of the Yasawa seismometer information is not orret at this present time. Figure. Moment tensor solutions. a) Result obtained by this study using loal seismi data reorded at Yasawa station. b) Harvard CMT solution. ) Result for the reverse polarity waveform. The red urve is the theoretial waveform and the blak urve is the observed waveform. Part II: Moment Tensor Inversion by using Teleseismi Body Waveforms. In Part II of this study, 30 events (Mw >6.5) that ourred within the Fii region were analyzed by using teleseismi body waveforms downloaded from the Data Management Center of the Inorporated Researh Institutions for Seismology (IRIS-DMC). We performed moment tensor inversion analysis for these loal events by using moment tensor inversion programming oded by Yagi (2008). The depth of hypoenter, the seismi moment (Mo) and the final foal mehanism beahball solutions of this study was omputed with that of Harvard GMT solutions. Figure 2. Graph of Depths (Harvard Solution)-km vs. Depths (this_study) in km. Figure 3. Graph of LogMo (Harvard Solution) vs. LogMo (this_study) Our results reveal the onsistenies with the depths of hypoenters (Figure 2) and logmo (Figure 3) in the two solutions. Suh a differene observed is due to that the Harvard University routinely arried 4
out the overall waveform piking in their moment tensor analysis, whereas in this study utilizes only the P-wave for suh analysis of moment tensor. Figure 4 exhibits the final moment tensor solution of this_study plot as beahballs on the Fii region map. Shallow depth events with strike-slip faulting type ourred mainly along the two Frature zones, the Fii Frature Zone and the Hunter Frature Zone. The rest are intermediate to deep intraslab earthuakes along the two subdution zones with mostly reverse and normal faulting type. Figure 4. Foal mehanism plot of this study. Part III: Analysis of rupture soure proess Four earthuakes were analyzed for their rupture soure proesses during their ourrenes. Two deep intraslab events whih ourred along the Tonga subdution zone and two events were of shallow depths ourred along the Fii Frature Zone. We applied inversion ode originally given by Yagi and Fukahata (2008) so that the onstraints of smoothness and positivity were imposed on the solution used in this analysis. To detet atual fault plane, we found the fault plane and average slip diretion whih show best fitting using grid searh method. Figure 5. Indiates the loation of the two events plotted as beahballs on Fii map In general, for deep earthuakes rupture propagates mainly along the strike diretion due to the thinness of the seismogeni zones at deeper depths as proved by our results shown on Figure 6. As for shallow earthuakes along the Fii Frature Zone, the rupture propagates along the strike (NE and SW) diretion and breaks some part of the fault, however the propagation of the rupture seems to be ontrolled by the geometry of the fault as proved by this study solution shown on Figure 7. Figure 6. Shows the final results of inversion for one deep event 9940309. The star indiates the loation of the initial break. 5 Figure 7. Shows the final results of inversion for one of the shallow depth earthuake 99802. The star indiates the loation of the initial break.
CONCLUSION In this study, we first investigate the uality of waveforms reorded at Yasawa station whih is operated by the Mineral Resoures Department of Fii. Using three omponents waveform reorded at Yasawa, we tried to estimate the entroid moment tensor solution so as to ompare it with the result of Harvard CMT. We found that the solution by this study is not onsistent with the Harvard CMT. If we divided by -0 to the observed waveform, we an obtain a onsistent result as the Harvard CMT. On the other hand, moment tensor inversion analysis was arried out by using teleseismi bodywaveforms for 30 past loal earthuakes (Mw>6.5) retrieved from IRIS-DMC. We found that middle earthuake with shallow depth loated along the two maor frature zones, the Fii Frature Zone (FFZ) and the Hunter Frature Zone (HFZ) and intermediate and deep earthuakes loated within the two subduting slab, the Tonga Kermede subdution zone and the Vanuatu Trenh. The foal mehanisms are well onsistent with tetoni setting. In the final part of this study, soure proess inversion was performed for four events. As for deep earthuake, sine thikness of seismogeni zones within the slab dereases with depths, the diretion of rupture propagation during a large earthuake should be mainly along the strike of the slab. For shallow earthuake ourred within the Fii Frature Zone, rupture propagated along strike diretion, NE and SW diretion or parallel to the diretion of the Fii Frature Zone. In general, for shallow earthuakes during suh rupture, it breaks a part of the fault and the rupture propagation diretion seems to be ontrolled by the orientation or the geometry of the frature zone. Our results agree well to the theory. RECOMMENDATION Based upon the purpose of this study, the following reommendations are put forward.. To investigate and hek the polarity of Yasawa broadband station as well as the magnifiation of the seismometer information and efforts should be made for our Mineral Resoures Department (MRD) to monitor and maintain this station. 2. Also to hek and maintain our CTBTO broadband station. 3. Moment tensor analysis to be arried out routinely for all medium-large size earthuakes that will our within the Fii region by using both near soure data and teleseismi body-wave. 4. Database for this analysis should be ompiled and aessed as referenes for future purposes. ACKNOWLEDGEMENT I would like to take this great opportunity to express my heartfelt thanks to my advisor Dr Hiroshi Inoue for the important advie given all the time espeially regarding our seismi network in Fii. REFERENCES Dregar, D., Ritsema, J. and Pasyanos, M., 995, Geophys. Res. Lett., Vol. 22, 8, 997-000. Kohketsu, K., 985, J. Phys. Earth, 33, 2-3. Rognvaldsson, S. T. and Slunga, R., 993, Bull Seis. So. Am. Vol. 83, 4, 232-247. Tibi, R., Estabrook, C. H. and Bok G., 999, Geophys. J. Int. (999), 38, 625-642. Yagi, Y., Kikuhi, M., 2000, Geophys. Res. Lett., Vol. 27, 3, 969-972. Yagi, Y., Mikumo, T., Paheo, J. and Reyes, G., 2004, Bull Seis. So. Am. Vol. 94, 5, 795-807. Yagi, Y., 2007-2008, IISEE/BRI. 6