Long-period Body Waves Research Articles

Joint earthquake source inversions using seismo-geodesy and 3-D earth models

A joint earthquake source inversion technique is presented that uses InSAR and long-period teleseismic data, and, for the first time, takes 3-D Earth structure into account when modelling seismic surface and body waves. Ten average source parameters (Moment, latitude, longitude, depth, strike, dip, rake, length, width and slip) are estimated; hence, the technique is potentially useful for rapid source inversions of moderate magnitude earthquakes using multiple data sets. Unwrapped interferograms and long-period seismic data are jointly inverted for the location, fault geometry and seismic moment, using a hybrid downhill Powell-Monte Carlo algorithm. While the InSAR data are modelled assuming a rectangular dislocation in a homogeneous halfspace, seismic data are modelled using the spectral element method for a 3-D earth model. The effect of noise and lateral heterogeneity on the inversions is investigated by carrying out realistic synthetic tests for various earthquakes with different faulting mechanisms and magnitude (Mw 6.0-6.6). Synthetic tests highlight the improvement in the constraint of fault geometry (strike, dip and rake) and moment when InSAR and seismic data are combined. Tests comparing the effect of using a 1-D or 3-D earth model show that long-period surface waves are more sensitive than long-period body waves to the change in earth model. Incorrect source parameters, particularly incorrect fault dip angles, can compensate for systematic errors in the assumed Earth structure, leading to an acceptable data fit despite large discrepancies in source parameters. Three real earthquakes are also investigated: Eureka Valley, California (1993 May 17, Mw 6.0), Aiquile, Bolivia (1998 February 22, Mw 6.6) and Zarand, Iran (2005 May 22, Mw 6.5). These events are located in different tectonic environments and show large discrepancies between InSAR and seismically determined source models. Despite the 40-50 km discrepancies in location between previous geodetic and seismic estimates for the Eureka Valley and Aiquile earthquakes, the seismic data are found to be compatible with the InSAR location. A 30° difference in strike between InSAR and seismic-derived source models is also resolved when taking 3-D Earth structure into account in the analysis of the Eureka Valley earthquake. The combination of both InSAR and seismic data further constrains the dip for the Zarand earthquake, and in all cases the seismic moment is more robustly constrained in the joint inversions than in the individual data set inversions. Unmodelled lateral heterogeneities in Earth and the models could partly explain some of the observed source parameter discrepancies related to the seismic data. © The Authors 2014. Published by Oxford University Press on behalf of The Royal Astronomical Society.

Synthetic seismograms for a synthetic Earth: long-period P- and S-wave traveltime variations can be explained by temperature alone

Current interpretations of seismic observations typically argue for significant chemical heterogeneity being present in the two large low shear velocity provinces under Africa and the Pacific. Recently, however, it has been suggested that large lateral temperature variations in the lowermost mantle resulting from a strong thermal gradient across D? may provide an alternative explanation. In case of a high heat flux from the core into the mantle, the magnitude of shear wave velocity variations in tomographic models can be reconciled with isochemical whole mantle flow and a pyrolite composition. So far, the hypothesis of strong core heating has been tested in a consistent manner only against tomographic S-wave velocity models, but not against P-wave velocity models. Here, we explore a new approach to assess geodynamic models and test the assumption of isochemical whole mantle flow with strong core heating directly against the statistics of observed traveltime variations of both P and S waves. Using a spectral element method, we simulate 3-D global wave propagation for periods down to 10 s in synthetic 3-D elastic structures derived from a geodynamic model. Seismic heterogeneity is predicted by converting the temperature field of a high-resolution mantle circulation model (MCM) into seismic velocities using thermodynamic models of mantle mineralogy. Being based on forward modelling only, this approach avoids the problems of limited resolution and non-uniqueness inherent in tomographic inversions while taking all possible finite-frequency effects into account. Capturing the correct physics of wave propagation allows for a consistent test of the assumption of high core heat flow against seismic data. The statistics of long-period body wave traveltime observations show a markedly different behaviour for P and S waves: the standard deviation of P-wave delay times stays almost constant with turning depth, whereas that of the S-wave delay times increases strongly throughout the mantle. Surprisingly, synthetic traveltime variations computed for the isochemical MCM reproduce these different trends. This is not expected from a ray-theoretical point of view and highlights the importance of finite-frequency effects. Most importantly, the large lateral temperature variations in the lower mantle related to strong core heating are able to explain most of the standard deviation of observed P- and S-wave delay times. This is a strong indication that seismic heterogeneity in the lower mantle is likely dominated by thermal variations on the length scales relevant for long-period body waves.

Long-period body wave traveltimes through the crust: implication for crustal corrections and seismic tomography

SUMMARY The traveltimes of P and SH waves through the crust depend strongly on wave period. At periods longer than 10–15 s, where traveltime measurements for indirect (e.g. surface reflected and core reflected) waves are typically made, the traveltimes are shorter than predicted by ray theory. Crustal corrections, often used in global mantle tomography to isolate the effects of the crust on teleseismic traveltimes, may have a complex frequency dependence and influence finite-frequency inversions. Crustal corrections for profiles of CRUST2.0 and the PREM reference model for the mantle may be several seconds larger or smaller than ray-theoretical values, depending on crustal thickness, crustal velocities and wave period. This variability is observed in the difference times between the seismic phases SS and S and between PP and P. It is therefore important to incorporate the effects of the crust on traveltimes in finite-frequency tomography, in order to take full advantage of the variable mantle sensitivity of body waves at different periods.

Comparison between different earthquake magnitudes determined by China Seismograph Network

By linear regression and orthogonal regression methods, comparisons are made between different magnitudes (local magnitude ML, surface wave magnitudes MS and MS7, long-period body wave magnitude mB and short-period body wave magnitude mb) determined by Institute of Geophysics, China Earthquake Administration, on the basis of observation data collected by China Seismograph Network between 1983 and 2004. Empirical relations between different magnitudes have been obtained. The result shows that: ① As different magnitude scales reflect radiated energy by seismic waves within different periods, earthquake magnitudes can be described more objectively by using different scales for earthquakes of different magnitudes. When the epicentral distance is less than 1 000 km, local magnitude ML can be a preferable scale; In case M MS, i.e., MS underestimates magnitudes of such events, therefore, mB can be a better choice; In case M>6.0, MS>mB>mb, both mB and mb underestimate the magnitudes, so MS is a preferable scale for determining magnitudes of such events (6.0 8.5, a saturation phenomenon appears in MS, which cannot give an accurate reflection of the magnitudes of such large events; ② In China, when the epicentral distance is less than 1 000 km, there is almost no difference between ML and MS, and thus there is no need to convert between the two magnitudes in practice; ③ Although MS and MS7 are both surface wave magnitudes, MS is in general greater than MS7 by 0.2∼0.3 magnitude, because different instruments and calculation formulae are used; ④ mB is almost equal to mb for earthquakes around mB4.0, but mB is larger than mb for those of mB≥4.5, because the periods of seismic waves used for measuring mB and mb are different though the calculation formulae are the same.

Seismological Aspects of the 2003 Bam, Iran, Earthquake and Its Aftershock Analysis

The source mechanism derived from the inversion of long-period body waves revealed that the earthquake occurred on a north-south trending strike-slip fault with a thrust component. According to the source model estimated in this study, the 2003 Bam, Iran, earthquake was a multiple event formed by two subevents. The rupture following subevent one started at a depth of about 8 km. However, the depth of subevent two is about 10 km. The total seismic moment estimated from inversion processes is 8.34×1018Nm. The pulse duration of subevent one and subevent two was determined from source time function as 1.7 s and 0.8 s, respectively. Corner frequency and source radius have been calculated by using major pulse duration. The corner frequency and source radius are 0.187 Hz and 5.47 km, respectively. The aftershock events distributed along a 30 km north-south striking fault. The focal depths of aftershocks distribution show a nearly vertical alignment of aftershocks located between 6 and 20 km depth. The focal mechanism solutions of aftershocks indicate right-lateral strike-slip faulting on a north-south trending fault, parallel to the previously known Bam fault trace in the east of Bam.

Active faulting and seismicity of the Dasht-e-Bayaz region, eastern Iran

The Dasht-e-Bayaz left-lateral strike-slip fault system in NE Iran ruptured over a length of ca 120 km during two earthquakes in 1968 and 1979. We constrain source parameters of the 1968 August 31 (Dasht-e-Bayaz) and 1979 November 27 (Khuli-Buniabad) earthquakes by analysing long-period body wave seismograms. Both earthquakes involved complex rupture processes, with at least two subevents in each case. Coseismic surface ruptures and cumulative scarps in alluvium indicate fault segmentation on numerous short (ca 20 km long) strands with small pull-aparts between them. The earthquake subevents seen in the seismograms probably relate to this segmentation. The total, cumulative, offset on the fault system is estimated at ca 4–5 km. This is small compared to the total amount of Late Tertiary deformation expected in this part of Iran, and indicates that the Dasht-e-Bayaz fault may be relatively young. Distributed strike-slip faulting is widespread in the region and there are indications that the Dasht-e-Bayaz fault is evolving from several short faults that are coalescing. These results are important not only for understanding of the regional tectonics but for the development and evolution of strike-slip faults in general.

The mechanism of great Banda Sea earthquake of 1 February 1938: applying the method of preliminary determination of focal mechanism to a historical event

A moment tensor solution is derived for the great 1938 Banda Sea earthquake, one of the largest events for which no definitive focal geometry has been published. We apply the method of preliminary determination of focal mechanism, which consists of inverting for a moment tensor using exclusively the spectral amplitudes of mantle Rayleigh and Love waves, while discarding the phase information which can be affected by uncertainties in relative timing between historical records. From a dataset of 17 records at seven stations, we obtain a robust solution, featuring a very large moment of 8.37×10 28 dyn-cm, the centroid depth being constrained around 60 km by minimizing the rms residual of the inversion. The method involves an inherent two-fold 180° indeterminacy on the orientations of strike and slip, which is lifted by the examination of the polarization of long-period body waves read individually on original records, resulting in a mostly thrust-faulting mechanism (strike: 276°; dip: 63°; slip: 70°). This approach, coupled with more traditional relocation efforts, allows a rare glimpse of this exceptional event, confirmed here to rank among the 10 largest moments ever published. Within the extremely complex regional plate tectonics pattern, the 1938 event took place in a region of sparse seismicity, and away from the presumed locations of block boundaries. The deeper than normal focus of the event agrees well with the generally minor level of reported damage, and with the generation of a relatively benign tsunami. Available modern focal mechanisms in its vicinity are too few and too scattered to allow any meaningful comparison, but the 1938 event shares its compressional axis with that of the smaller and deeper 1963 shock to the southwest, which probably expresses coherence in the regional contortion of the subducting Australian plate lithosphere.

Born scattering of long-period body waves

The Born approximation is applied to the modelling of the propagation of deeply turning long-period body waves through heterogeneities in the lowermost mantle. We use an exact Green's function for a spherically symmetric earth model that also satisfies the appropriate boundary conditions at internal boundaries and the surface of the earth. The scattered displacement field is obtained by a numerical quadrature of the product of the Green's function, the exciting wavefield and structural perturbations. We study three examples: scattering of long-period P waves from a plume rising from the core-mantle boundary (CMB), generation of long-period precursors to PKIKP by strong, localized scatterers at the CMB, and propagation of core-diffracted P waves through large-scale heterogeneities in D?. The main results are as follows: (1) the signals scattered from a realistic plume are small with relative amplitudes of less than 2 per cent at a period of 20 s, rendering plume detection a fairly difficult task; (2) strong heterogeneities at the CMB of appropriate size may produce observable long-period precursors to PKIKP in spite of the presence of a diffraction from the PKP-B caustic; (3) core-diffracted P waves (Pdiff) are sensitive to structure in D? far off the geometrical ray path and also far beyond the entry and exit points of the ray into and out of D?; sensitivity kernels exhibit ring-shaped patterns of alternating sign reminiscent of Fresnel zones; (4) Pdiff also shows a non-negligible sensitivity to shear wave velocity in D?; (5) down to periods of 40 s, the Born approximation is sufficiently accurate to allow waveform modelling of Pdiff through large-scale heterogeneities in D? of up to 5 per cent.

Moment tensor inversion of the November 6, 1988 M S=7.6, Lancang-Gengma, China, earthquake using long-period body-waves data

Moment tensor inversion was carried out to invert the source mechanism and source time function of the M S=7.6 November 6, 1988, Lancang-Gengma, Yunnan Province, China, earthquake. Waveforms of long-period bodywaves recorded by China Digital Seismograph Network (CDSN) were used in the inversion. The inverted result shows one nodal plane of right-lateral strike-slip faulting and another of left-lateral strike-slip faulting and a simple source time function of a duration of about 15 s and scalar seismic moment of 6.4×1020 N·m. From the geological data and tectonic settings and also from field observations and epicentral distribution of aftershocks, the nodal plane striking in the azimuth of 313° is preferred as the fault plane. The pressure axis lies almost horizontally in north-south direction.

Scattering of spherical elastic waves from a small-volume spherical inclusion

SUMMARY A theory is presented for the scattering of spherical elastic waves by a small, spherical inclusion. The domain of validity of the treatment is ka′« 1, where k is the wavenumber and a′ is the radius of the spherical inclusion. The treatment is an extension of that of Gritto, Korneev & Johnson (1995), who derived exact expressions for the scattered wavefield from a plane P wave incident on a spherical inclusion in the low-frequency limit. Since neither treatment assumes small perturbations, they are also valid for an arbitrary material property contrast between the spherical inclusion and the background medium. For spherical elastic wavefields propagating through an infinite, homogeneous medium and interacting with a single small, spherical inclusion, the scattered wavefield is obtained directly as a sum of spheroidal motions of degrees 0, 1 and 2 in scatterer-centred coordinates, with nine independent amplitude coefficients which depend on the contrast in material properties and are proportional to the volume of the inclusion. If the inclusion is embedded in a stratified and bounded medium, then these scattering coefficients provide a valid local description of the scattered wavefield in the vicinity of the inclusion. Exploiting this fact, source equivalents in the frequency domain (six moment tensor elements and three single forces) are derived from the scattering coefficients. To first order in material property contrasts, equivalence is established between these source equivalents and the coupled mode theory developed for surface waves and long-period body waves using perturbation theory.

Inversion of long-period body-wave data for the source process of the Gonghe, Qinghai, China earthquake

Long period body waves data recorded by the China Digital Seismograph Network (CDSN) are inverted for the seismic moment tensors of the April 26, 1990, Gonghe, QinghaiMS=6.9 earthquake and itsMS=5.0 after-shock occurred on May 7, 1990. In the inversion, the generalized reflection-transmission coefficient matrix method is used to generate Green’s function. From the inversion it is obtained that the rupture process of theMS=5.0 aftershock is relatively simple, and that of the main shock is rather complex. There are at least two events during main shock rupture process with an interval about 35 seconds. The focal mechanisms of two events are roughly the same as that of the aftershock, all of them were mainly reverse dip-slipping faulting with minor left-lateral strike-slip motion. These results indicate that the Gonghe earthquake was the result of the farther extension of one NWW-SEE striking buried fault on the southern margin of Gonghe basin from shallower depth to deeper depth and from NW to SE under the action of a nearly horizontal NE direction compressive stress.

Automated near‐realtime CMT inversion

The Earthquake Research Institute (ERI) of the University of Tokyo has completely automated the process of determining source mechanisms of earthquakes using waveform data. The solutions are automatically distributed using the Internet without being checked. The process is initiated by receipt of e‐mail giving the origin time and location of an earthquake. To determine the source mechanism, we apply the Harvard centroid moment tensor (CMT) inversion method to long‐period body wave data (the portion of the seismogram from the first arrival to the initial surface wave). The method has been successfully applied since October 1993 to all of the major earthquakes in the world for which we received event e‐mail. 108 of the 242 CMT solutions in 1994 were determined within 6 hours of the event, with an average delay of 3.8 hours. For local events (e.g., events around Japan), it is also possible to determine CMT solutions within 30 minutes from the occurrence of the earthquakes, using an efficient dialup data retrieval system operated by ERI.

Formalism for traveltime inversion with finite frequency effects

SUMMARY A simple modification of the waveform inversion formula, based on the normal mode perturbation theory, is shown to lead to a formula for traveltime anomalies. The kernel which is derived can be used for traveltime inversion with automatic inclusion of finite frequency effects. Inversion for Earth structure with such kernels will lead to better resolution estimates than ray-theoretical traveltime inversion. Examples of kernels for transverse component seismograms are shown for direct S waves, ScS, Love waves and diffracted S waves. A measure of finite frequency effects is also proposed by comparing our formula with the one from ray theory. A quantity which should be 1 in the case of ray theory is computed for the finite frequency kernels and is shown to have deviations up to about 30 per cent from 1. Therefore, the use of ray theory for long-period body waves applies incorrect weight along a ray path and may introduce a small bias to an earth model.

ML:M0 as a regional seismic discriminant

AbstractThe mb:MS ratio determined by teleseismic observations has proven to be an effective discriminant, for explosive sources tend to be significantly richer in short-period energy than are earthquakes. Unfortunately, this method is limited by the detection threshold of teleseismic surface waves. However, recent advances in instrumentation allowing low amplitude surface wave measurements coupled with new analytical techniques make it feasible to use regional waveform data to determine the long-period source excitation level of low magnitude events. We propose using the ratio of ML (local magnitude) to M0 (scalar seismic moment) as an analogous regional discriminant. We applied this criterion to a data set of 299 earthquakes and 178 explosions and found that this ratio seems to be diagnostic of source type. For a given M0, the ML of an explosion is more than 0.5 magnitude units larger than that of an earthquake. This separation of populations with respect to source type can be attributed to the fact that ML is a short-period (1 Hz) energy measurement, whereas seismic moment is determined from long-period body wave phases (period > 4 s) and surface waves (10 to 40 sec). Using regional stations with sources 200 to 600 km away, the effective threshold for magnitude measurements for this discriminant is found to be ML = 3.1 for earthquakes and ML = 3.6 for explosions. This method does require the determination of regional crustal models and path calibrations from master events or by other means.

Waveforms of Long-Period Body Waves In A Slightly Aspherical Earth Model

Summary Considering the coupling of normal modes of different dispersion branches, we present a formulation for synthesizing waveforms of long-period body waves in a slightly aspherical Earth model. A general asymptotic result, which is valid when angular degrees of modes are much higher than the maximum harmonic degree of lateral heterogeneity, is derived for arbitrary source mechanisms and receiver components. If we restrict the coupling to the modes along a single dispersion branch, this result reduces to the pre-existing theory for surface waves. For body waves, we demonstrate numerically that cross-branch coupling is important to set up the correspondence between normal mode and ray theories. Actually, the coupling of the modes which interfere constructively in the spherically symmetric reference model represents the largest effects of heterogeneity on the waveform of a body wave packet. This would lead to a significant reduction in computation time. Numerical tests show that the formulation is able to bring out the ray character of body waves: only the structural perturbation in the neighbourhood along the geometrical ray affects the seismogram.

Waveform modelling of the May 20, 1990 Sudan earthquake

Very-broad-band seismic waveforms were modelled to derive the moment tensor, the source time function and the depth of the May 20, 1990 Sudan earthquake. Long-period body waves, surface waves and broad-band displacement, recorded by the M edN et regional network, were merged with digital data from global seismic networks, to compare fault geometries obtained by different methods and assess associated uncertainties. Our analysis indicates that the May 20 earthquake had a seismic moment of 5.3 × 10 26 dyne · cm and a depth of about 10 km; the focal mechanism shows a mixture of transcurrent geometry, with one plane striking NW-SE, and a smaller dip-slip component with the T axis oriented in the N–S direction. The fault geometry agrees with models calling for the recent re-activation of the NW-SE trending Aswa transform zone, marking the northern termination of the western branch of the rift. In contrast, the N–S tensional component, which also dominates the focal mechanisms of the aftershocks (Dziewonski et al., 1991), supports the evidence for the rotation of the regional Quaternary stress field to the north, in contrast with simple two-plate models of rifting in East Africa. The Juba event allowed us to test the capability of the M edN et regional seismic network to provide an accurate real-time description of the seismic source.

The Ms = 8 tensional earthquake of 9 December 1950 of northern Chile and its relation to the seismic potential of the region

The only known great ( M s = 8) intermediate depth earthquake localized downdip of the main thrust zone of the Chilean subduction zone occurred landward of Antofagasta on 9 December 1950. In this paper we determine the source parameters and rupture process of this shock by modeling long-period body waves. The source mechanism corresponds to a downdip tensional intraplate event rupturing along a nearly vertical plane with a seismic moment of M 0 = 1 × 10 28 dyn cm, of strike 350°, dip 88°, slip 270°, M w = 7.9 and a stress drop of about 100 bar. The source time function consists of two subevents, the second being responsible for 70% of the total moment release. The unusually large magnitude ( M s = 8) of this intermediate depth event suggests a rupture through the entire lithosphere. The spatial and temporal stress regime in this region is discussed. The simplest interpretation suggests that a large thrust earthquake should follow the 1950 tensional shock. Considering that the historical record of the region does not show large earthquakes, a ‘slow’ earthquake can be postulated as an alternative mechanism to unload the thrust zone. A weakly coupled subduction zone—within an otherwise strongly coupled region as evidenced by great earthquakes to the north and south—or the existence of creep are not consistent with the occurrence of a large tensional earthquake in the subducting lithosphere downdip of the thrust zone. The study of focal mechanisms of the outer rise earthquakes would add more information which would help us to infer the present state of stress in the thrust region.

Depth extent of rupture of the 1981 Chilean outer-rise earthquake as inferred from long-period body waves

We have investigated how well the depth extent of rupture can be determined for a shallow dip-slip earthquake from long-period WWSSN body waves. The 16 October 1981 Chilean outer-rise earthquake ( MW = 7.2) is analyzed as an example. Calculation of the resolution matrix for the vertical distribution of displacement rate on the fault plane shows that the depth extent of rupture cannot be resolved better than 20 km, even when the data set consists of both P and S waves. Although the depth extent of rupture of the 1981 earthquake is too small to be resolved, it is possible to infer some constraints from the point-source depth in combination with the source duration. While inversion for the depth extent of rupture can be formulated as a linear problem, determination of the point-source depth is nonlinear, and a formal resolution study cannot be conducted. The point-source depth estimate for the 1981 earthquake is 28 to 32 km (80% confidence interval); the source duration is 8 sec. Assuming a rupture velocity of 3 km/sec, we have explored several rupture models in a synthetic experiment to shed light on the relation between point-source depth and centroid depth. The experiments show that point-source depths agree with the centroid depth. Given that there is poor resolution of the depth extent of rupture of the 1981 earthquake, we conclude from the synthetic experiments that the earthquake may have ruptured updip to a depth as shallow as 17 km or downdip to a depth as deep as 43 km.

Seismic body waves in a 3-D, slightly aspherical Earth - I: testing the Born approximation

SUMMARY The first stage of a method for calculating long-period body wave seismograms in a laterally heterogeneous Earth is presented. First, the elastodynamic equation of motion in spherical coordinates, which is valid for weak lateral heterogeneity, is established using a generalized spherical harmonic expansion of the displacement and traction fields and the elastic parameters. It is shown that the equations for the displacement-traction components can be simplified by using high-wavenumber asymptotic approximations to the first-order coefficients, which are appropriate for 5s - 20s body waves in the mid- to upper mantle. The Born approximation is then used to develop a computationally efficient method for calculating me first-order scattered field in terms of the high-frequency asymptotic solution for a laterally homogeneous Earth. This efficiency is increased further by decomposing the wavefield into its P, SV, and SH components. Finally, the numerical integration of the displacement-traction components is performed for a few selected frequencies and a restricted band of angular degrees, in order to test the validity of the Born approximation. Comparison of these results with those calculated using the Born approximation indicates that the latter is well suited to models of weak lateral heterogeneity in the Earth, for which the elastic parameters can be expressed as a sum of low-order spherical harmonics. It is demonstrated that, for this particular class of models, the Born approximation yields a good representation of the scattered field for frequencies within the long-period body wave band of 0.05–0.2 Hz.

Are long-period body wave coda caused by lateral heterogeneity?

SUMMARY Data from two broad-band arrays in western Europe (NARS and GRF) are used to study the character of long-period body wave coda. The events studied are at epicentral distances of 40 to 60, in the Hindu Kush region and on the Mid-Atlantic Ridge, sampling the upper mantle to a depth of about lo00 km. The periods studied are 5-50 s. The long-period coda of P at GRF (interstation distance about 10 km) are strongly coherent, whereas the long-period coda of PP-, PPP- and S-waves are incoherent. This indicates that the latter coda consists of scattered waves. To investigate the nature of the scattering process, the data of GRF are analysed for slowness and azimuth variations in the coda intervals. A new beamforming algorithm is presented to deal with the low frequences and relatively short time intervals. The method is based on Backus-Gilbert inverse theory. The results show that the incoherent long-period coda intervals almost entirely consist of surface waves; these waves are scattered from the preceding body waves. Some calculations with linearized theory for body wave to Rayleigh wave conversion at topography at the surface or at the Moho show that realistic scatterers can account for the observed (constant) coda level. The beamforming results show that the phases in the P coda all arrive along the great circle. As scattering calculations point out that body wave to body wave scattering is inefficient, it is concluded that the long-period P coda does not contain a significant amount of scattered energy. Synthetic seismograms obtained with the reflectivity method show that spherically symmetric upper mantle models can explain these coda waves. For events in the Hindu Kush region, an upper mantle with a thick lid overlying a pronounced low-velocity zone (LVZ) is necessary to explain the character of the P coda at GRF. Such an upper mantle agrees with previous studies of similar great circle paths. The strong coherence of the P coda at GRF is lost on the scale of NARS (station separation about 200 km). This suggests lateral variations in the upper mantle at a scale of about 200 km. It appears from previous studies of the upper mantle under Europe that these variations must be sought in the LVZ. It is shown that the long-period P coda is sensitive to variations in the P velocity structure of the LVZ. This suggests the P coda (i.e. PdP phases) as a tool for monitoring lateral variations in the LVZ and possibly to prove the existence or absence of a LVZ in the P velocity.