The deep creep plate interface extends from the down-dip edge of the seismogenic zone down to the base of the overlying lithosphere in subduction zones. Seismogenic/deep creep zone interaction during the earthquake cycle produces spatial and temporal variations in strains within the surrounding elastic material. Strain observations in the Nankai subduction zone show distinct deformation styles in the co-seismic, post-seismic, and inter-seismic phases associated with the 1946 great earthquake. The most widely used kinematic model to match geodetic observations has been a 2-D Savage-type model where a plate interface is placed in an elastic half-space and co-seismic slip occurs in the upper seismogenic portion of the interface, while inter-seismic deformation is modeled by a locked seismogenic zone and a constant slip velocity across the deep creep interface. Here, I use the simplest possible 2-D mechanical model with just two blocks to study the stress interaction between the seismogenic and deep creep zones. The seismogenic zone behaves as a stick-slip interface where co-seismic slip or stress drop constrain the model. A linear constitutive law for the deep creep zone connects the shear stress (σ) to the slip velocity across the plate interface (s′) with the material property of interface viscosity (ζ ) as: σ = ζ s′. The analytic solution for the steady-state two-block model produces simple formulas that connect some spatially-averaged geodetic observations to model quantities. Aside from the basic subduction zone geometry, the key observed parameter is τ, the characteristic time of the rapid post-seismic slip in the deep creep interface. Observations of τ range from about 5 years (Nankai and Alaska) to 15 years (Chile). The simple model uses these values for τ to produce estimates for ζ that range from 8.4 × 10 13 Pa/m/s (in Nankai) to 6.5 × 10 14 Pa/m/s (in Chile). Then, the model predicts that the shear stress acting on deep creep interface averaged over the earthquake cycle ranges from 0.1 MPa (Nankai) to 1.7 MPa (Chile). These absolute stress values for the deep creep zone are slightly smaller than the great earthquake stress drops. Since the great earthquake recurrence time ( T recur ) is much larger than τ for Nankai, Alaska, and Chile, the model predicts that rapid post-seismic creep should re-load the seismogenic zone to about (1/3) of the co-seismic change; geodetically observed values range from about (1/10) to more than (1/2). Also, for the case of (T recur /τ) ≫1, the model predicts that the slip velocity across the deep creep interface during the inter-seismic phase should be about (2/3) the plate tectonic velocity (R). Thus the deep creep velocity used in Savage-type models should be less than R. Even complex 3-D models with non-linear creep laws should make a similar prediction for inter-seismic deep creep rates. At present, it seems that geodetic observations at Nankai and other subduction zones are more consistent with a deep creep rate of R rather than (2/3) R. This discrepancy is quite puzzling and is difficult to explain in the context of a 2-D steady-state earthquake cycle model. Future observational and modeling studies should examine this apparent discrepancy to gain more understanding of the earthquake cycle in subduction zones.

Subduction zones generate most of the world's seismicity, and all of the largest earthquakes. This overview of large earthquakes in subduction zones consists of two parts: a review of the occurrence of large events in different tectonic regimes of subduction zones, and the timing of large interplate underthrust events. Our global review shows that large earthquakes have occurred in all intra-plate environments from the outer-rise down to 650 km depth, except for the fore-arc region of the upper plate in mature subduction zones. It seems that the seismogenic plate interface is an efficient concentrator of seismicity, though large earthquakes do occur just trenchward and also downdip of the interplate coupled zone. We focus on two aspects of the temporal occurrence of interplate events: a brief analysis of the composite global occurrence of great events, and then a brief review and analysis of the methodology of long-term earthquake forecasting, followed by a suggestion to improve the methodology. Occurrence times of the greatest interplate events in the 20th century are clustered more than expected from random occurrence. However, a "waiting time" analysis of the 40 great interplate events (M>8) in the 20th century shows that their origin times are consistent with a model of independent random occurrence. The key to earthquake forecasting methodology is the accurate determination of recurrence time for each plate boundary segment for the current earthquake cycle. Observations of large earthquake occurrence show great variability in rupture mode and recurrence times. Mechanical models that include interaction between adjacent plate boundary segments produce synthetic event catalogs with variable rupture modes and recurrence times, similar to observed earthquake sequences. One robust "rule" extracted from these simulations is that if the rupture mode changes from one great event to several smaller events, then the first smaller event will occur in the epicentral segment of the great event with a recurrence time that is shorter than the average time for that segment. This "rule" appears to explain four examples of sooner-than-expected large earthquakes: the 1942 event in Ecuador; the 1986 event in central Aleutians; the 1994 event in Sanriku, Japan; and the 1995 event in Kuriles Islands.

This chapter contains sections titled: Introduction Large Underthrusting Earthquakes Down-Dip Edge of Seismogenic Interface Does the Coastline Control the Downdip Edge of Subduction Earthquakes? Conclusions

Lithospheric density and thickness variations are important contributors to the state of stress of the plates. The relationship between the lithosphere's isostatic state, subcrustal structure and stress field, however, remains unresolved due to the uncertainties on its thickness, composition and rheology. To study the influence of lithospheric structure on intraplate stresses, we use a new model of global lithospheric structure (TDL) that accounts for the presence of depleted mantle to explore the effects of isostatic compensation, mantle density structure, lithospheric thickness (base depth) and mechanical coupling within the lithosphere on wavelengths >200 km. We compute the mean lithostatic stress (O) of 2 degrees x 2 degrees lithospheric columns and then solve for the resulting global tectonic stress field for a homogeneous elastic lithosphere with the finite element package ABAQUS. For a 100 km base depth, a historically common value for lithospheric thickness, tectonic stress patterns are largely insensitive to mantle density structure and match patterns in the world stress map, for both isostatically compensanted and non-compensated lithospheric structure. Increasing the base depth up to 250 km to account for thick continental roots, however, leads to sharp variations in the stress field between isostatic lithospheric structure models and TDL as the mantle portion of the lithosphere dominates O. Decreasing the model base depths up to 25 km as a proxy for vertical strength variations due to low viscosity channels within the crust or lithosphere as a whole, strongly alters stresses in magnitude, azimuth and regime, as the influence of topography and shallow crustal structure increases. We find that restricting spatial changes in O to a specified region to mimic lateral variations in strength also has a large effect on the resulting stresses, which leads us to conclude that regional models may not always be adequate for modelling the stress field. Strong deviations from long-wavelength patterns on the world stress map in models with a shallow (<<100 km) or deep (>>150 km) uncompensated model base depth likely reflect that the globally averaged lithospheric thickness is close to 100 km and large deviations from this depth generate unrealistic stress patterns related to uncompensated buoyancy forces. Because the stresses are so sensitive to base depth, we conclude that using O to represent spatial and vertical variations in lithospheric structure is not an adequate approximation. Our results suggest that future studies must incorporate the full 3-D variations in density and rheology of the lithosphere to elucidate the source and nature of the lithospheric stress field. These studies have become possible with the advent of modern computational tools and advances in our knowledge of lithospheric structure and rheology.

Elastic anisotropy of core samples

Underthrusting at subduction zones can cause large earthquakes at shallow depths, but is always accommodated by aseismic creep below a certain depth. This transitiofi depth is referred to as the depth of seismic coupling and can be directly observed in a subduction zone as the lower depth extent of significant moment release of the deepest large underthrusting earthquakes. In 1978, a large (Ms=7.5) earthquake occurred off the coast of Miyagi Prefecture in northern Honshu. Its focal mechanism represents underthrusting of the Pacific plate beneath Honshu. Since the hypocenter is located 150 km landward from the trench and there are no other large interplate earthquakes further landward from the trench axis, this event defines the maximtim depth of the coupled zone. The lower limit of significant moment release of the Miyagi-Oki earthquake is obtained by analysis of the longperiod P waves. The deconvolved source time function consists of a dominant single pulse with peak moment release at 12 s and a total duration of 18 s. The rupture extent of this dominant pulse does not extend deeper than 40 km, thus the transition from coupled to uncoupled in northern Honshu occurs at or above 40 km depth.

On May 23, 1989, a great (Mw=8.1) earthquake occurred in the Macquarie Ridge complex, south of New Zealand. The earthquake is one of the largest in the Macquarie Ridge complex this centmy, and represents the right-lateral strike-slip component of the motion between the Pacific plate and the Australian plate. Subduction ini•tion appears to be presently occurring in the Macquarie Ridge complex, and it has been suggested that plate boundary strike-slip earthquakes in a transitional tectonic env'n-onment have a high stress-drop. We have investigated tt,•e source rupture process of the 1989 earthquake, using teleseismic P and SH waves. The best point source depth is 12 kin below the ocean bottom. The deconvolved source ;urne function is dominated by a single pulse with a large •nt release (1.8x1021 Nm) and a short duration (20 s). The moment rate increases slowly in the first 10 s of the rupture process, and is suddenly truncated at 20 s. There is no resolvable direcfivity to this sharp truncation, which given the temporal resolution of the data set, means that the spatial location of the sharp nuncation is within approximately 50 km of the epicenter. The stress-drop and average displacement may be as high as 370 bars and 36 m. These values are unusually high, and strongly support he suggestion that high stress-drop earthquakes are characteristic for transitional tectonic environments.

Earthquake source time functions deconvolved from teleseismic broadband P wave recordings are used to examine rupture variations for 417 underthrusting earthquakes located on the interplate interface in circum‐Pacific subduction zones. Moment‐scaled duration of significant moment release varies with depth, with longer‐duration events occurring in the shallowest 20 km of the megathrusts. The source time functions are also used to estimate radiated seismic energy. Two estimates are obtained: a simple scaled triangle substitution for the moment release history provides a minimum estimate, while integration of the time function shape provides an estimate limited only by the bandwidth of the teleseismic deconvolutions. While these energy calculations underestimate total energy, they enable systematic comparisons of rupture process within and between subduction zones. We do not find significant depth dependence for radiated energy overall, but some regions do show mild trends of increasing energy/seismic moment ratios (E/Mo) with increasing source depth that correspond to rupture duration variations in those regions. The observations of longer rupture duration at shallow depth with moderate E/Mo may be due to heterogeneous friction and structural features on the shallow plate interface.

Ruff, L.J., 1992. Asperity distributions and large earthquake occurrence in subduction zones. In: T. Mikumo, K. Aki, M. Ohnaka, L.J. Ruff and P.K.P. Spudich (Editors), Earthquake Source Physics and Earthquake Precursors. Tectonophysics. 211: 61-83. Plate tectonics and the seismic gap hypothesis provide the framework for long-term earthquake forecasting of plate boundary earthquakes. Unfortunately, detailed examination reveals that earthquake recurrence times and rupture length vary between successive earthquake cycles in the same subduction zone. Furthermore, larger coseismic slip is commonly associated with larger rupture length. Hence, large earthquake occurrence in subduction zones is characterized by variability in: (1) recurrence times, (2) rupture length, and (3) coseismic slip. These facts, plus many other observations, indicate that there are significant spatial variations in the “strength” of the plate interface. One simple description of these variations and their role in the earthquake cycle is the asperity model, where the large strong regions of the plate interface are called asperities, and the large earthquakes occur when the large asperities break. The asperity model of earthquake occurrence is able to qualitatively explain several features of large plate boundary earthquakes. To go beyond general qualitative notions. I pose the following scientific test: are the observed asperity distributions and a simple model of their interaction self-consistent with the above three observed features of large earthquake occurrence? The distribution of the major asperities along plate boundary segments has now been determined for several subduction zones. Rupture process studies of adjacent large and great earthquakes have provided reliable estimates of the along-strike asperity lengths and separations for several adjacent asperities in the Kurile Islands, Colombia, and Peru subduction zones. The simplest mechanical model for asperity interaction is to idealize two adjacent asperities as frictional sliders that are connected by main springs to the upper plate, by a coupling spring to each other, and maintain frictional contact with a conveyer belt (the lower plate) that moves with a constant velocity. An “earthquake” occurs when the net force on the asperity frictional slider reaches some specified level. The failure force and spring constants are determined by the observed asperity distribution and simple models of elastic interaction. Two different macroscopic failure criteria are used. This simple mechanical model displays a remarkable range of behavior from simple to complex. When the two asperities are identical in all their properties, sequences of identical “earthquakes” are produced. For the more realistic case of non-identical asperities, “earthquake” sequences show great variety. Using system variables from the observed asperity distributions, the “earthquake” sequences typically display: (1) variable recurrence times, (2) variable rupture length, i.e. a combination of single-asperity and double-asperity failures, and for one of the failure criteria (3) larger coseismic slip for double-asperity failures. Statistical summaries of thousands of simulated “earthquake” sequences for asperity pairs in the Kuriles, Colombia, and Peru subduction zones are broadly consistent with the observed features of large earthquake occurrence in these subduction zones. The main conclusion is that the asperity model provides a self-consistent explanation for: fault zone heterogeneity, the rupture process, and recurrence times and rupture mode of large earthquake sequences cia a simple model for adjacent asperity interaction. In addition, a conclusion independent of any particular model for fault zone heterogeneity is that simple deterministic models of fault zone interaction can explain complex patterns of large earthquake occurrence in subduction zones.

On 23 June 2001, a Mw = 8.4 underthrusting earthquake occurred in the southern Peru subduction zone, followed by several large aftershocks, including 26 June (Mw = 6.7) and 7 July (Mw = 7.5). Broadband analyses of seismic data for the largest of these earthquakes show southeastward rupture of 180 km along the portion of the subduction zone previously ruptured in 1868 (Mw 8.8–9). Moment release distributions determined are consistent with aftershock location patterns. Earthquake rupture mode varies along southern Peru, from larger multi‐segment rupture in 1868 to several smaller segment ruptures in 2001, similar to rupture variation observed in northern Peru and other subduction zones. Based on models of subduction zone segment interaction, the 2001 earthquake sequence may suggest a shorter recurrence time for future earthquakes along this portion of the Peru‐Chile subduction zone.

Many people are familiar with the shaking experience of earthquakes, but very few people have experienced a tsunami\---|even a small one. Can we safely watch miniature tsunami waves and learn something about wave propagation? Breaking waves at the beach are a familiar sight and the basis for wave knowledge for most people. Tsunami waves are not a gigantic vertical wall of water that crashes directly onto land, but that persistent notion captures our fascination and awe. While few of us have seen tsunami waves, the pictures of the destruction left behind when the water recedes amply portray the hazards. Public education can mitigate the tsunami hazard from a nearby earthquake: If you are at the beach and you feel earthquake shaking, move away from the water! Great subduction earthquakes can generate tsunami waves that propagate across the ocean and then grow to dangerous heights on distant shores—this hazard underscores the importance of tsunami early warning systems (see http://www.prh.noaa.gov/pr/itc for links to international sites for tsunami warning and information). Of course, one must be at a Pacific Ocean beach to worry about tsunamis, right?While tsunami waves represent a terrible hazard, they also provide an excellent educational vehicle to understand waves. It is a constant challenge to explain seismic wave generation and propagation to students. To help explain basic concepts, it is useful to employ more familiar wave systems such as light waves and sound waves. But our familiarity with seeing and hearing does not translate into an easy grasp of light and acoustic wave propagation. Surface waves on a pond …

We apply a moment-tensor inversion algorithm to microseismic events generated by hydrofracturing. Our dataset consists of inicroseismic events associated with a hydraulic fracturing experiment conducted at the GRYDOE M-site near Rifle, Colorado. The event recordings were from a vertical string of 3-component accelerometer pods that are cemented in a borehole. The accelerograms clearly show P and S waves, which can be rotated into the SH and SV components.To invert the P and S amplitudes for moment tensors, we use just the first two peaks after the first arrival. Using this amplitude information and the average pulse width, we formed a symmetric, stylized pulse for integration and we inverted for the moment tensor.Resolution tests revealed that with the source (microseism) to receiver geometry we have at the M-site, where all receivers are in one borehole and at the same azimuth from the source, we can extract the five deviatoric components of the moment tensor using combinations of P, SH and SV amplitudes (for example, a minimum data set of 3 P-wave, 2 SV-wave, and I SH-wave amplitudes). To obtain the sixth component, which is the isotropic component, requires amplitude information from receivers at another azimuth.We studied seven high-quality events in detail and tested their results against two alternative source geometries. These microseismic events tend to have a major-double-couple nodal plane aligned parallel to the main hydrofracture plane, rather than have their tension (T) axes aligned parallel to the regional minimum principal-compressive-stress direction. The events have seismic moments (M-0) on the order of 10(-5) Nm and moment magnitudes (M-w) of about -2.5. (C) 2001 Elsevier Science B.V. All rights reserved.

Good problem sets for seismology students—now that is something we can all appreciate and use! As I browse the Web looking for educational materials, I am impressed with the number of sites that offer earthquake information, basic descriptions of earthquakes and plate tectonics, and even complete lectures. Most of this material is suitable for the K-12 level or college survey courses. At the other extreme, many researchers now post Web pages that contain research programs, projects, results, resources, and published papers—there is no shortage of information for the professional seismologist. But at the level between introductory and advanced, there is a shortage of seismological educational resources. When we teach college-level seismology to undergraduates and first-year graduate students, we need problem sets that let students work with seismograms to construct and use travel-time curves, group velocity curves, instrument responses, wave attenuation, hypocentral locations, …