ALBERT

Notes: A significant correlation is found, in both space and time, between the intraplate (land) and interplate (sea, thrust zone only) earthquakes in Tohoku, NE Japan that has persisted since the times of reliably reported events in AD 1600. the correlation peaks at a land-lead of about 36 yr with an average correlation distance of 200km, with the implication of an average strain migration rate of 5.6 km yr-1. the correlation is highly significant (〉 99 per cent), both from formal statistics and from tests of random shuffles of the data. Additional analysis of the data, as a point process, confirms the results of the correlation analysis. the sharpness of the correlation peak, when compared to the individual times of occurrence of the land and sea events suggests a trigger mechanism.To explain the correlation, the general model ofsubduction-rupture-rebound is extended to include additional features; the buckling of the land plate from the force of the subducting slab, and the viscoelastic coupling of the plate to the underlying asthenosphere. A buckle produces a high-stress region in the continental plate where earthquakes are more prone to occur, thus producing the spatial correlation in the data. This may also explain the preferred location on land for the smaller modern-day seismic events in NE Japan. the viscoelastic coupling controls the interaction between the land and sea events, resulting in the temporal correlation in the data. Because of viscosity, the model equations are diffuse-like with strain pulses as solutions; thus from the inferred strain migration rate it is possible to estimate asthenospheric viscosity (η=7 × 1018Pa s) using this model. A large land shock generates a strain pulse that affects the locked fault at the thrust zone several decades later. As the continental plate tends to pull away from the subducting slab, the frictional force arising from the overburden pressure is reduced, thus unlocking the fault and triggering a sea earthquake.The viscoelastic model is also used to explain surface deformations measured by triangulation surveys in Japan in 1904 and 1964. Horizontal displacements, which we believe are surface manifestations of the strain pulse from the large 1896 Riku-U land shock (M= 7.5) in NE Japan, are fit well by the model and provide a viscosity estimate η= 13 × 1018 Pa s.

Notes: Observations of underside P-wave reflections from discontinuities deep within the upper mantle which precede the P'P’ phase have provided important constraints on the existence and physical properties of these dynamically critical features. In this study, we extend earlier work by examining for the first time, broad-band recordings of those earthquakes for which one would expect to observe P'P’ and its associated precursors P'dP’. An exhaustive search of 20 yr of Carnegie data and 11 yr of Gräfenberg data uncovered a number of fine P‘P’ observations, but in only one or perhaps two instances are there evidence of a precursor from near 670 km depth. These null results are consistent with what has been reported in earlier studies and provide us with an opportunity to estimate the variability of P velocity near 670 km or, alternatively, the scale length of discontinuity topography which acts to defocus the precursor. This has been done by using the reflectivity and Gaussian beam methods to compute synthetic seismograms for a suite of 1-D and 2-D earth models which vary slightly from PREM. We find that the null data may be satisfied either by (i) reducing the PREM impedance jump at 670 km from 8.5 to 4-5 per cent; (ii) smoothing this 8.5 per cent jump linearly over a zone 15-30 km thick or (iii) imposing deformations in the 670 km discontinuity as small as 10 km in amplitude and 300 km in wavelength. For the one case of an observed precursor, the relative size of the underside reflection and main P‘P’ phase suggests that, in this instance, the precursor's amplitude is correctly predicted by PREM.