T43A-1308 1340h
Sensitivity Kernels of PKP Waves in the Lower Mantle and Inner Core
Traveltimes of PKP phases are often used to constrain the anisotropy of the inner core and the D'' layer structure. PKP waves on broadband records have rather short dominant periods (between 1 and 4 s), but because we need to consider the sensitivity of these waves far away from the source or the receiver, their first Fresnel zone is quite large. In order to provide a better interpretation of PKP(DF) residuals in terms of variations of anisotropy with depth in the inner core, we investigate the effects of transversely isotropic perturbations of the elastic tensor on the propagation of PKP(DF) waves. We show that three sensitivity kernels for traveltime and amplitude corresponding to three anisotropic perturbation parameters $\epsilon$, $\delta$ and $\gamma$ are required. To reduce the contributions of mantle heterogeneities, we can use PKP(AB)-PKP(DF) or PKP(BC)-PKP(DF) differential traveltimes to study the inner core anisotropy. However, heterogeneities in the lower mantle and D'' may still contribute strongly to the observed anomalies. Thus, obtaining good model of the D'' structure is a first step necessary to obtain reliable constraints on the inner core anisotropy. For this purpose, we derive and analyse the single phase and differential traveltime kernels for PKP in the mantle. The geometry of the differential kernels is complex because coupling occurs between the different PKP branches. As a result, the geometry of PKP differential kernels in D'' shows significant qualitative differences with simple ray tracing. A number of recent PKP studies have found strong variations of PKP(AB) traveltimes for very close paths. These observations suggest sharp lateral variations of seismic velocities in D'' whose origin and geometry are still poorly understood. Characterizing these anomalies will be possible using accurate 3-D sensitivity kernels.
T43A-1309 1340h
Could the Observed PKP Wave Anomalies Originate From the D'' Region?
Much work is undertaken to investigate the structure of the deep Earth, i.e. the core and lower mantle. Researching these regions is important, as they are thought to influence the generation of the Earth's magnetic field, as well as mantle convection. The interaction zone between the core and mantle is potentially the origin of plumes, and the graveyard for slabs. Much controversy exists regarding the nature of the inner core, in particular its anisotropy and rotation. Until now, body wave studies of the Earth's core have focused primarily on travel time deviations. These locate anomalous regions, but cannot determine the nature of the anomaly. The rays may have been deviated along an anomalous path (backazimuth deviation), or travelled at an anomalous phase velocity (slowness deviation). The work presented here utilises array methods both to further examine anomalies previously located by travel time studies, and to study new regions. The slowness and backazimuth of core seismic phases are calculated, and compared with global velocity models. Deviations of the different phases are analysed to resolve the location of the anomalous regions. The results are then compared with travel time deviations. Currently data has been used from the German Regional Seismic Network (GRSN), the Tien Shan Networks (KNET and GHENGIS), and the Alaska Seismic Network (ASN). Earthquakes in Tonga-Fiji are received by the GRSN, the Tien Shan Networks receive events from the Andes, and data from the ASN are used to further investigate the anomalous South Sandwich Islands to Alaska path. This path has been shown in recent inner core studies to have strong anisotropy and is also used to study inner core rotation. Results show that deviations are present in all three phases (PKPab, PKPbc and PKPdf); deviations vary for the different paths. GRSN and GHENGIS data both show different deviation patterns for PKPbc and PKPdf, possibly indicative of small-scale lower mantle structure. The South Sandwich Island to Alaska path shows large deviations, particularly so for the PKPdf branch (the branch that travels through the inner core). In order to investigate the possibility that all of the anomalies seen in the PKP phases originate from D'' region, modelling has been carried out. A 3D ray tracing method is being utilised to test this, and other scenarios.
T43A-1310 1340h
Structure of the Inner Core-Outer Core Boundary Inferred From $PKP_{BC}$ Diffracted Waves
Examining the structure of the inner core-outer core boundary (ICB) is important for understanding the evolution and dynamics of the Earth's core. Seismic waves that diffract past the C cusp ($PKP_{Cdiff}$) preferentially sample the ICB and so contain important clues about this region. Just as $P_{diff}$ has been studied to infer core-mantle boundary structure, we expect that analysis of $PKP_{Cdiff}$ will reveal ICB structure. However, there have been few systematic studies of $PKP_{Cdiff}$. We examined all of the IRIS temporary networks for which data are currently available and found three that provide enough high-quality $PKP_{Cdiff}$ data for array processing techniques to be used effectively: INDEPTH-II, INDEPTH-III, and BANJO. We apply the cross-correlation and adaptive stacking techniques to the broadband waveforms to extract the differential travel times of $PKP_{Cdiff}$ across the arrays, and use a weighted least-squares technique to invert theses times for the three-dimensional $PKP_{Cdiff}$ slowness vector. We generate standard errors for the ray parameter and backazimuth using a bootstrap-type resampling algorithm. The earthquakes we analyze at the three networks sample different regions of ICB. For the earthquakes recorded at INDEPTH-II and INDEPTH-III arrays, $PKP_{Cdiff}$ phases sample the ICB beneath north Africa and southwest Europe, and for the earthquakes at BANJO, they sample the ICB beneath Antarctica and west North America. We find $PKP_{Cdiff}$ ray parameters vary from $1.79 \pm 0.11$ to $1.95 \pm 0.05$ s/deg beneath north Africa and southwest Europe,from $1.60 \pm 0.08$ to $1.74 \pm 0.07$ beneath Antarctica and have values around $1.82 \pm 0.09$ beneath west North America. In order to lessen the effect of shallow structure on the travel times, we also invert the differential travel times ($PKP_{Cdiff}-PKP_{DF}$) for differential slowness vectors. We find less variation in them, however the differences appear to be significant. Therefore, our initial results support the idea of modest lateral heterogeneity at the ICB. We intend to examine the robustness of this result by refining our observational technique, quantifying the relationship between $PKP_{Cdiff}$ ray parameters and ICB structure, and identifying more $PKP_{Cdiff}$ phases from regional networks to make a global study of ICB heterogeneity.
T43A-1311 1340h
Regional variation of PKP C-diff slowness observed by high-dense network -implication for lateral velocity variation of the lowermost outer core
Clear core phases are observed at a recently deployed Japanese seismic network (Hi-net) for two intermediate depth events which occurred beneath S. Bolivia on 2001/06/29 and beneath Argentina on 2002/09/24. The epicentral distances are between 144 and 161 degrees, and between 150 and 164 degrees, respectively. The differential travel times of PKP(BC) or PKP(Cdiff) minus PKP(DF) for the S. Bolivia event suit calculated results using by the previous model of Kaneshima et al. (1994). The differential travel time residuals for the Argentina event show positive value, and become larger as the epicentral distance becomes large. Although the above results suggest that there is anomalous structure in the inner or outer core along the ray paths for the Argentina event, it is unclear which T(DF) or T(BC or Cdiff) is anomalous. Hi-net is a dense short-period seismic network which consists of about 700 stations in almost whole extent of Japan. Close look at core phases can be carried out using data from Hi-net. We checked the slowness of PKP(DF) and PKP(Cdiff) for the events. We select data whose epicentral distance is greater than 153 degree. The slowness of PKP(DF) and PKP(Cdiff) for the S. Bolivia event, and that of PKP(DF) for the Argentina event are almost same as those of PREM. However, the slowness of PKP(Cdiff) for the Argentina event is 7 percent larger than that of PREM. Thus we can conclude that the large slowness of PKP(Cdiff) is a main cause of the large differential residuals. This result suggests that a low velocity anomaly exists in the lowermost outer core which the PKP(Cdiff) for the Argentina event samples, and that the velocity structure in the inner core along the rays for the two events and in the outer core along the rays for the Argentina event sample is close to the normal, although the influence of the heterogeneous structure beneath stations and D_h cannot be ruled out yet. The central points of the diffracted rays on the inner core boundary are located in between 7 and 21N, S. of those for the S. Bolivia event. These results may imply that the velocity of the lowermost outer core is slower in the equatorial area than in the mid-latitude.
T43A-1312 1340h
Seismic Evidence for a Complicated Inner Core Boundary
We present new seismic evidence for strong lateral heterogeneity near the inner core boundary (ICB). As a probe of the ICB we examined the amplitudes of waves (PKiKP) reflected from the ICB at precritical distances. Specifically, we selected a distance range of $50^{\circ}-90^{\circ}$, in which the PKiKP amplitude is especially sensitive to the shear velocity at the top of the inner core. At these distances the theoretical reflection coefficient at the ICB is nearly zero, and so observations of significant PKiKP energy are inconsistent with the standard model (PREM) for the density and velocity jumps at the ICB. We searched for PKiKP waves using data recorded by the short-period, small-aperture arrays of the International Monitoring System (IMS), for which coherence based stacking procedures can be used to high frequencies. We selected about 180 source-array combinations that had the highest potential for generating PKiKP waves and used a sliding window, time-domain beamforming process for phase identification. In 23 cases PKiKP was positively identified, in 32 cases the data were inconclusive, and in the remainder PKiKP was definitely not present. For data in the positive category we measured PKiKP/P amplitude ratios from optimally tuned array beams, and for data in the inconclusive and negative categories we measured upper bounds on PKiKP/P amplitude ratios from appropriately formed beams. In all cases we corrected the ratios for the effect of the source mechanism using Harvard CMT solutions. The positive PKiKP data are clustered geographically and most have PKiKP/P amplitude ratios 1-2 orders of magnitude larger than predicted by PREM; however, many of the negative data have upper bound PKiKP/P ratios that are significantly lower. We carried out a series of robustness tests to affirm the anomalous PKiKP/P amplitude ratios, and also found evidence that the relevant P waves are not unusually small. Although mantle effects cannot be completely ruled out, the most straightforward explanation for our observations is that the ICB has significant lateral variations in structure. This implies that the sedimentation and compaction processes responsible for growing the inner core are not uniformly simple, and that some of the heterogeneity present deeper within the inner core may reflect this complexity.
T43A-1313 1340h
Characteristics of PKP-Cdiff coda observed with small-aperture seismic arrays
PKP-Cdiff is a compressional wave diffracted around the inner core boundary (ICB), which may contain much information on the structure of the ICB. Here I have determined slowness of PKP-Cdiff and its coda with small-aperture arrays as large as several kilometers in size to approach the unsettled issues on the origin of the PKP-Cdiff coda (Morita, 1987; Umetsu, 1989; Nakanishi, 1990). The short-period seismic array data is collected from the Prototype International Data Center (pIDC) of the International Monitoring System (IMS) as Koper et al. (2004) did. The criteria of data selection (focal depths greater than 100 km, mb greater than 5.5, epicentral distances from 153 to 160 degrees, the period since Jan. 1995 through Feb. 2000) yielded twenty-nine event-array pairs, the band-pass filters comprising sixteen frequency bands from 0.25 to 4 Hz with 0.5 Hz bandwidths were operated to the array seismograms. Then the beam-forming as functions of slowness and incident azimuth were applied to each band-pass filtered array record. The difference of array response functions were adjusted by using the phase weighted stack with variation of the power index of the phase stack. The stacked powers for the PKP-Cdiff coda were investigated on the vespagram for all the frequency bands and the slownesses were distributed around two peaks of 1-3 s/deg and 4-5 s/deg, corresponding to the slowness of PKP-Cdiff and PKP-AB, respectively. The incident azimuths concentrated around the great-circle directions connecting the events and arrays. The specific stacked powers with slowness of 1-3 s/deg observed at distance 153-155 degrees tended to occur just after the PKP-Cdiff first arrivals whereas those at 156-160 degrees scattered uniformly in the time window between PKP-Cdiff and PKP-AB and are a few around the PKP-Cdiff first arrivals. Such characteristics of the slowness filtered PKP-Cdiff coda imply that the PKP-Cdiff coda is affected with the ICB structure. Because the small first arrival and large coda observed around larger distances are similar to the characteristics of PKiKP coda observed in 50-90 degrees by Vidale and Earle (2000) and Koper et al (2004), the nature of the PKP-Cdiff and its coda may reflect the scattering property of the ICB.
T43A-1314 1340h
Inner Core Anisotropy in Attenuation
It is now well established that the compressional velocity in the Earth's inner core varies in both direction and geographic location. The compressional waves travel faster along the polar directions than along the equatorial directions. Such polar-equatorial difference is interpreted as a result of inner core anisotropy in velocity (with a magnitude of about 3%) and such anisotropy appears to be stronger in the ``western hemisphere" (180$^{o}$W -40$^{o}$E) than in the ``eastern hemisphere" (40$^{o}$E-180$^{o}$E). Along the equatorial paths, the compressional velocity also exhibits a hemispheric pattern with the eastern hemisphere being about 1% higher than the western hemisphere. Possible explanations for the causes of the velocity in anisotropy and the hemispheric difference in velocity along the equatorial paths include different geometric inclusions of melt or different alignments of iron crystals which are known to be anisotropic in velocities. Here, we report an observation of ubiquitous correlation between small (large) amplitude and fast (slow) travel time of the PKIKP waves sampling the top 300 km of the inner core. We study this correlation by jointly analyzing the differential travel times and amplitude ratios of the PKiKP-PKIKP and the PKPbc-PKIKP phases recorded by the Global Seismographic Network (1990-2001), various regional seismic networks (BANJO, BLSP, FREESIA, GEOFON, GEOSCOPE, Kazakhstan, Kyrgyz, MEDNET, and OHP), and several PASSCAL Networks deployed in Alaska and Antarctica (XE: 1999-2001, XF: 1995-1996, and YI: 1998-1999). Our dataset consists of 310 PKiKP-PKIKP and 240 PKPbc-PKIKP phases, selected from a total of more than 16,000 observations. PKIKP waves exhibit relatively smaller amplitudes for those sampling the eastern hemisphere along the equatorial paths and even smaller amplitudes for those sampling the polar paths in the western hemisphere. One simple explanation for the velocity-attenuation relation is that the inner core is anisotropic in attenuation and the direction of high attenuation correlates with that of high P velocity. Different anisotropic behaviors in velocity and attenuation can be best explained by different alignments of iron crystals under the hypothesis that iron crystals are anisotropic in both velocity and attenuation and their axes of high P velocity correspond to those of high attenuation.
T43A-1315 1340h
Quantifying Inner Core Scattering from PKiKP Coda Waves
It is thought that the core is mainly composed of iron, with a few percent of lighter elements. Despite its rather uniform composition, several authors have reported scattered energy coming from the inner core. This implies that the inner core has lateral variations in structure or composition with a scale length of tens of kilometers. These lateral variations represent seismic scatterers and could be caused by crystal grain size, variations in the orientation of anisotropy, or the presence of partial melt and/or impurities. However, it is still debated whether the scatterers are distributed throughout the volume of the inner core, or simply along the inner core boundary. This work pursues the characterization of inner core scattering (ICS) using a data set of high frequency PKiKP waveforms, recorded at short-period, small-aperture seismic arrays. We apply a methodology used in the study of heterogeneities in the crust and upper mantle, which consists of curve fitting of the observed scattered energy with a standard model, along with a stripping technique to isolate the PKiKP coda from the rest (eg. P, PcP, and PP codas). The coda energy is usually obtained from the mean square of the velocity, but the envelope seems to be more suitable. The selected model consists of an exponential decay in time, which depends upon the quality factor of coda waves (Qc), the geometrical spreading of the scattering medium, and a constant that depends on the general properties of the medium. This model can be easily linearized allowing the use of well-developed least squares techniques. We also apply the Coda Normalization method to remove the source and site effects. Following this procedure, we have been able to isolate and study in detail the PKiKP coda. Our preliminary results show a strong correlation between the different parameters (Qc, the geometrical spreading, and the medium constant), not allowing a definitive interpretation of their numerical values. Nevertheless, the resulting decay rate appears to be clearly distinguishable in each case, permitting a visual comparison. We believe that the detailed study of more events might let us constrain spatial and/or temporal variations in the ICS. Our findings will help in the understanding of the inner core 3D structure and possibly help resolve the issue of the inner core differential rotation.
T43A-1316 1340h
A Temporal Change Of Inner Core Travel Times Confirmed By High-Quality Waveform Doublets
The first claim of temporal change of differential travel times between PKP(DF) waves (traversing the inner core) and PKP(BC) waves (avoiding the inner core) was made by Song and Richards (1996) and interpreted as evidence of inner core super-rotation. This observed/apparent time dependence of the BC-DF times was instead proposed by Souriau and Poupinet (1997; 1998; 2000) to be an artifact of systematic earthquake mislocations. Li and Richards (2003) showed how to avoid such artifacts by use of BC-DF measurements made on a waveform doublet (two earthquakes at essentially the same location, as evidenced by their highly similar waveforms) for events separated by several years, but they found only one example in the South Sandwich Islands (SSI), recorded at just one station (COL). Here we report on three additional doublets, each showing a change of differential BC-DF travel times at several stations and always with the same trend: signals of the later event travel faster on paths from SSI to Alaska and Canada. From high-quality waveform data recorded at seismographic stations of the International Monitoring System, the U.S. Atomic Energy Detection System, the Alaska Seismic Network, and station INK in Canada, we have found a change of differential travel times at 50 stations for one of these doublets; and at 5 stations for each of the other two. The observed changes in BC-DF times lie in the range 0.08 to 0.18 s over 10 years. For the best-recorded doublet (events on 1993 Dec 01 and 2003 Sep 06), waveform similarity restricts hypocenter separation to be less than 2 km. From array data at ILAR we measure the rate of change of BC-DF times with respect to epicentral distance as 0.013 s/km --- far too small a value to allow plausible mislocations to cause the observed changes in travel time, and leading us to conclude that the changes are indeed temporal.
T43A-1317 1340h
Constraining Inner Core Rotation From a Worldwide Search of Waveform Doublets
A major debate in determining inner core rotation using seismic PKP differential times has been potential systematic event mislocation. The problem can be avoided by using waveform doublets separated by several years (Li and Richards, 2003). The high similarity of the whole waveforms in a doublet ensures that the two events indeed occur at the same location and sample the same Earth structure. The waveform similarity also allows measurements of relative time shifts with high precision. Motivated by the recent successful discovery of waveform doublets of South Sandwich Islands (SSI) earthquakes that show clear temporal changes at stations in Alaska and Canada (Zhang et al., this session), we are conducting a worldwide search for waveform doublets. We have designed an efficient search scheme, which involves initial pairing of earthquakes at a targeted region using catalog locations and semiautomatic correlation and visualization using MATLAB. Our search on SSI earthquakes have yielded 12 new doublets that are separated by 0, 1, 3, 4, 6, 7, 8, 13, 15, 18, 18, 23, and 33 years. When aligned with the PKP(BC) phase (turning at the bottom of the outer core), the PKP(DF) phase (traversing the inner core) of the later event is consistently earlier for the pairs with the time separation greater than 7 years, but pairs separated by less than 4 years show little travel time shifts. The Jul 25, 1970 event forms a triplet with the 1993 and 2003 doublets discovered by Zhang et al. (this session). The DF phase is earlier by 0.35 to 0.40 s over 33 years. We will report our search results on Alaska and Kermadec earthquakes for which evidence for temporal changes in BC-DF times has been reported (along Alaska to South Pole and Kermadec to Norway paths). Waveform doublets will also allow us to examine possible changes in travel times or amplitudes of other phases that sample the inner core (such as PKiKP, PKKP, and inner core scattering waves).