J. R. Cleary

Evidence for scattering of seismic PKP waves near the mantle-core boundary

Abstract The hypothesis of scattering of PKP waves near the mantle-core boundary provides a comprehensive interpretation of the observed precursors to PKIKP, certain features of which have not been adequately accounted for by any alternative interpretation. These features include the variation with distance of the times and slownesses of precursor onsets, and the variations in amplitude, azimuth and slowness observed in precursor wavetrains. The observed times and slownesses of the earliest precursor onsets are in close agreement with the theoretical least time curve for singly scattered waves. Amplitudes and slownesses of scattered waves have been calculated for earth models which are spherically symmetrical except for random variations in density and elastic parameters in a layer 200 km thick at the base of the mantle. The calculations show that observed precursor amplitudes and slownesses can be accounted for by random variations of about one percent having a correlation distance of about 30 km in this layer. In particular, it is shown that scattering structures up to 900 km above the mantle-core boundary inferred by Doornbos and Vlaar (1973) are not required by their data. There is a suggestion that the main scattering may actually occur inside a layer much less than 200 km thick at the base of the mantle.

Azimuthal variation of the longshot source term

The P times from Longshot are strongly biased by the presence of azimuthal variation in the source term. This bias has produced systematic error in the travel-time curve derived by Chinnery and Toksoz. When the azimuthally dependent source effects are removed, the times are in good agreement with those predicted by Cleary and Hales. The azimuthal variation is apparently caused by a siab of high-velocity material dipping steeply beneath the Aleutian arc. It appears likely that the material in the slab consists of olivine exhibiting strong velocity anisotropy. If this kind of structure is common to arc regions, it represents a potent source of error in the determination of earthquake locations.

The D″ region

Abstract Two very different types of models are currently being proposed for D″, the lowest region of the earths mantle: (a) those in which the P and S velocities vary smoothly down to the core-mantle boundary, without any extreme change in gradient; (b) those in which the velocity gradients decrease fairly abruptly at a height of 100 km or so above the core-mantle boundary, and maintain a value close to the critical gradient down to the boundary. Type (a) is represented by model UTD124A′ of Dziewonski and Gilbert (1972) and model B1 of Jordan and Anderson (1974). Both models are in good agreement with most travel time and free oscillation data. Their validity rests on the supposition, supported in part by theoretical studies, that data which suggest the presence of a low velocity zone in D″ result from distortion of seismic waves by the core-mantle boundary. On the other hand, slowness and amplitude data from short period P waves indicate a fairly rapid decrease in velocity gradient at a depth corresponding to an epicentral distance of about 92°, and it is very unlikely that these data can be interpreted as interface phenomena. The measured P and S times at distances beyond about 96° also indicate reduced velocities in D″. The suggestion that the measured velocities are in error as a result of interface effects is weakened by the fact that the results are apparently not wavelength-dependent. Type (b) is represented by model B2 of Jordan (1972), Bolts (1972) model, and a new model designated as ANU2. All models have high density gradients indicative of inhomogeneity in the region. Model B2 fits the oscillation data reasonably well, but has an unjustifiably low S velocity at the core-mantle boundary. In Bolts model the P and S velocities at the top of D″ are based on the models of Herrin et al. (1968) and Jeffreys (1939), whereas in ANU2 the values are taken from Hales and Herrin (1972) and Hales and Roberts (1970b). The velocities at the core-mantle boundary in Bolts model and ANU2 are based on observations of “diffracted” P and S. Both of these models were designed to produce flattening of the P curve at about 92°. Both may require some modification in order to be compatible with free oscillation data.

P wave travel-time gradient measurements for the warramunga seismic array and lower mantle structure

Abstract Over six hundred values of the P wave travel-time gradient (dT/dΔ) have been derived from 224 events recorded at the Warramunga Seismic Array over the distance range 28° to 99°. These dT/dΔ measurements have been corrected for local structure beneath the array, and analysed on a regional basis to try and establish the presence of velocity anomalies in the lower mantle. The smoothed dT/dΔ data have been inverted by the Herglotz-Wiechert method to obtain a velocity model over the depth range 700 to 2700 km. The anomalies in the dT/dΔ curve correspond first to fairly low velocity gradients between depths of 800 and 850 km, 1070 and 1110 km, 1260 and 1330 km, 1750 and 1850 km, and 2460 and 2600 km; and secondly to high velocity gradients between 1160 and 1220 km, 2180 and 2370 km, and possibly between 2700 and 2750 km. These results have been shown to agree satisfactorily with recent travel-time and array dT/dΔ investigations, once the sources of systematic error in such studies have been isolated and compared.

Australian crustal structure

There have been eight large-scale refraction experiments in Australia during the last fifteen years. P1 velocities derived from these experiments are significantly higher in the Precambrian shield region than in eastern Australia. Pn-velocities are also higher beneath the shield, and appear to increase systematically from east to west across the continent. There is good evidence for an intermediate layer in all parts of Australia, with an average depth of about 20 km to the Conrad discontinuity. The crustal thickness has an average value of about 40 km, and the observed variations in thickness are apparently unrelated to topography in most cases.

Scattered PKKP: Further evidence for scattering at a rough core-mantle boundary

Abstract PKKP signals from Novaya Zemlya recorded at LASA at distances around 60° show consistent anomalies in both slowness and azimuth. The observed anomaly suggests that the signal is a BC branch arrival, although the arrival time corresponds to the DF branch. The BC branch, however, does not extend back to this distance. The azimuth of approach is in the range 229–245°, instead of the expected 186°. These anomalies are associated only with PKKP; analysis of the core phases PKiKP and P′P′ (BC) from the same events show that they arrived at LASA with the appropriate slownesses and azimuths. The PKKP signals can be interpreted as “scattered” PKKP; the scattering occurs on underside reflection at the core-mantle boundary and is probably caused by topographic irregularities on the boundary itself. The calculated scattering region has a surface projection at about 60°S, 134°E, which is outside the diametral plane through source and receiver, and about 21° from the expected PKKP reflection point at 76°S, 95°E. Both the “direct” and “scattered” arms of the PKKP signal have a PK path close to that of the “C” end of the BC branch. The unexpectedly large amplitude of the arrival suggests that there may be a focusing of energy at C, which would indicate a change in velocity gradient just above the inner core boundary. The observations nevertheless require, on the scattering interpretation, lateral variations in the topography of the core-mantle boundary and a region of relatively large topography responsible for the anomalous PKKP observations.

Evidence for seismic wave scattering in the D″ layer

Measurements of slowness variations within PKIKP precursor wave trains provide crucial evidence for scattering near the core-mantle boundary.

Comparison of the 1968 P tables with times from nuclear explosions (I) Longshot and Greeley

Abstract The chief differences between recent determinations of the P travel time curve beyond about 24° may be expressed, in each case, as a deviation in the slope of the curve which is approximately constant throughout the range. If one of these curves is used to calculate the residuals of P times from a seismic event, with station corrections applied, then it is possible to solve simultaneously for the value of this deviation and for the source correction. This has been done for the nuclear explosions Longshot and Greeley , using the 1968 P tables and the Herrin-Taggart station corrections. In both cases the calculated deviation was close to that of the Cleary-Hales curve. The analysis of the Longshot data was repeated using the times and station corrections derived by Lilwall and Douglas, and essentially the same result was obtained.

COMPARISON OF THE 1968 P TABLES WITH TIMES FROM NUCLEAR EXPLOSIONS. II. THE MARSHALL ISLANDS AND SAHARA SERIES.

Abstract P times from Longshot and Greeley have recently been analysed by Cleary and Muirhead, using 1968 P times and Herrin-Taggart station corrections to derive source corrections and travel time corrections (the latter being assumed to vary linearly with distance). The analysis is here extended to times from the series of nuclear explosions in the Marshall Islands and in the Sahara. Combination of the results for the four explosion sites results in a value of −0.023 ± 0.004 sec/deg for the slope of the line representing the error in the 1968 P curve. By comparison, the mean slope of the Cleary-Hales deviation curve is −0.022 sec/deg. The effect of the error in the slope of the 1968 P curve would be to overestimate focal depths of seismic events by about 70 km.

Estimation of PKP times from ISC data

Abstract The International Seismological Centre (ISC) treats reports of the earliest observed core phase arrivals from earthquakes as PKIKP observations and provides residuals based on Jeffreys-Bullen PKIKP times. An analysis of some 30 000 such data from ISC bulletins for the year 1967 has shown that the data include, in addition to PKIKP, many observations of PKP1 and precursors to PKP. By the use of simple truncation and trimming techniques, travel-time curves (expressed as differences from JB), with standard errors, have been obtained for PKIKP in the ranges 110 to 139° and 152 to 173°, and for PKP1 in the range 144 to 154°. The PKIKP curve in these ranges agrees quite closely with the curve obtained by Cleary and Hales, and with one derived from model 1066B of Gilbert and Dziewonski, except that the Cleary-Hales curve is up to 0.7 s earlier at distances below 120°, and the 1066B curve is up to 1 s later beyond 152°. The PKP1 curve is in good agreement with that derived from 1066B. A plot of the number of observations in each 1° distance interval has been used to estimate the positions of the cusps B, C and D on the PKP curve as 144, 152.5 and 115°, again in reasonable agreement with 1066B.