S33D-01
Near-field monitoring of seismic source behavior at South African deep gold mines
We introduce our SeeSA projects, as important as dense array monitoring According to a mining plan and a geological map detailing locations of faults or weakness, we can anticipate potential M > 2 seismic sources at depths of 2.0 - 3.6 km at South African gold mines. At such potential sources, we have installed instruments prior to an onset of irreversible process to monitor earthquake generation process. From the previous projects for periods of from a year to a few years, the possible widest dynamic range and resolution have revealed the finest detail of the process since 1995 in cooperation with ISS International Ltd and South African gold mines (Mponeng, Bambanani, Tau Tona, Buffelsfontein GM, and ERPM), Wits Univ., Geohydroseis CC., Seismogen CC., OHMS CC., GFZ, GMuG, CSIR. The talk summarizes examples of our successful monitoring and introduces some on-going projects. Highlighted are the following. Yamada et al. [05, 07] demonstrated that mine tremors have rupture process as complex as natural larger earthquakes and the scale dependency of rupture parameters is similar to that for natural larger earthquakes. We successfully recorded strain accumulations larger than 100 micro strain, followed by several hundreds of seismic events (-1 < M < 3; distance < ~ 250 m). The seismicity within about 100m from strainmeters caused frequent, seismic strain-steps; the largest recorded was greater than 100 micro strain by an M2.5 earthquake at a distance within ~100 m. One of the most important results were that no detectable accelerating precursors preceded strain-steps associated with several hundreds of the earthquakes (- 1 < M < 3) catalogued by minefs seismic networks (hereinafter Catalogued E/Q; Takeuchi 05), while significant post-seismic drifts followed some strain-steps by Catalogued E/Qs. Frequently observed were episodic strain changes with durations of much slower than strain-steps associated with the Catalogued E/Qs [Naoi et al. 06]. Striking were some examples of slowest strain-steps (with minutes durations), preceded by significant forerunners [Naoi et al. 06; Yasutake et al. 06]. With a clearest recording, in which the strain change was as large as 2 micro strains, a forerunner undergoes a linear change [Yasutake et al. 07AGU], not in a continuously accelerating manner that is predicted, e.g., by a typical rate-and-state-dependent friction-law. At a site in a ~ 50m-wide fault zone with multiple gouge zones ( < ~ 1 m thick) at 104 Level at Mponeng mine, Yasutake et al. [07AGU] found several examples of slow strain steps detected by a pair of strainmeters that was ideally spaced. With the recordings, we could constrain locations and magnitudes of the slower strain- steps, Mw being from -1.5 to -0.2. A seismic moment was proportional to a cube of a duration period, much as regular natural larger earthquake. The tendency was different from those proposed by Ide et al. [07] for slow events, much larger than those we found at Mponeng.
S33D-02 INVITED
Constraints on Seismogenesis of Small Earthquakes (-2.8 ≤ M\rm{W} ≤ 3) Defined by the Dense Instrument Network of the Natural Earthquake Laboratory in South African Mines (NELSAM)
The Natural Earthquake Laboratory in South African Mines (NELSAM) is designed to investigate the nucleation and rupture propagation of small (-2.8 ≤ M\rm{W} ≤ 3) earthquakes in quartz host rock at seismogenic depths. Acceleration, velocity, and strain meters are installed at a depth of 3.6 km within and surrounding the ancient Pretorius Fault zone in TauTona Mine, South Africa, which is a site of enhanced seismicity. Data from the 9 NELSAM instruments, recorded at 12 kHz, are supplemented by recordings from the mine-operated geophone arrays, which include ~25 stations in TauTona Mine and ~30 stations in nearby Mponeng Mine. This dense seismic network allows us to resolve detailed waveforms from earthquakes located meters to hundreds of meters from the seismometers and investigate their source properties. Additionally, we have obtained an extensive catalog from the mine-operated network that includes 733,000 earthquakes M\rm{W} ≥ -2 that occurred between 1995 and 2007. Using recent recordings from the NELSAM dataset in conjunction with catalog data from the mine-operated network, we address questions of how to distinguish between the multiple populations of mining-induced seismicity and whether a minimum magnitude, M\rm{min}, exists for the population thought to be most similar to tectonic earthquakes. During quiet hours in 2006-2007 with no ore production, we find tectonic-like earthquakes with magnitudes as low as -2.8 and a catalog completeness threshold of -1.2. Thus, we find that catalog detection limitations, not earthquake source physics, control the apparent M\rm{min} of "tectonic" earthquakes. Furthermore, assuming that laboratory results (Lockner and Okubo, 1983) from similar sized events in faults of similar composition can be directly applied to the mining environment, and that earthquake nucleation processes are dependent on the fault zone parameters, we estimate that earthquake nucleation in the mines entails a moment release equivalent to M\rm{W} ≤ -3. The nucleation patch radius and critical slip distance are estimated to be ~0.2 m and ~20 microns respectively, consistent with data from a magnitude -2.8 earthquake recently recorded in TauTona Mine. This study was supported by NSF Continental Dynamic grant 0409605.
S33D-03 INVITED
Involvement of Overpressured Fluids in the Nucleation of High-Angle Reverse Ruptures: Evidence from Fault-Hosted Hydrothermal Vein Systems
Dips of near-pure reverse-slip M>5.5 ruptures are bimodally distributed with a dominant peak at δ = 30±5°, a subordinate peak at δ = 50±5°, and no ruptures with δ > 60°. Assuming horizontal trajectories for maximum compressive stress (σ1), the dominant peak corresponds to optimally oriented faults with Byerlee friction coefficients (μs = 0.6) for which frictional lock-up is expected at δ = 60°. In recent years, several compressional inversion earthquakes in the upper crust of Honshu, Japan (e.g. the 2003 Mw6.5 Northern Miyagi, the 2004 Mw6.6 Mid-Niigata Prefecture, and the 2007 Mw6.7 Noto-Hanto sequences) have involved high-angle reverse-slip with dips of 50-60° on inherited normal faults along the margins of Miocene extensional basins. Rupturing during these earthquakes thus took place on faults that were poorly oriented for frictional reactivation and close to lock-up. Frictional mechanics suggests that reshear of the steep reverse faults (in preference to the formation of new favorably oriented thrusts within intact crust) is allowable only under near- lithostatic fluid pressures with Pf approaching σ3, and that reshear of severely misoriented faults (δ > c.60°) requires Pf > σ3 (the hydrofracture condition). Notably, the 2004 Mid-Niigata sequence involved a criss-crossing network of high-angle and low-angle reverse ruptures, suggesting competition between reshear of steep inherited faults and the formation of more favorably oriented thrusts. A range of geophysical evidence, including local bright S-wave reflectors, indicates strong fluid overpressuring in the focal regions of these earthquakes. Mesozonal Au-quartz vein systems hosted in reverse faults exhumed from depths corresponding to the lower half of the seismogenic zone (P ~ 2-4 kbar; T ~ 250-400°C) occur throughout the geological record and provide additional evidence for the involvement of strongly overpressured fluids in reverse fault rupturing. Incrementally deposited fault-infill veins up to meters in thickness may extend for 1-2 km down-dip with comparable dimensions along-strike. For steeper faults, especially, these fault-veins are commonly in mutual cross-cutting relationships with arrays of flat-lying extension veins that are the product of hydraulic extension fracturing. Individual extension veins extend laterally for tens to hundreds of metres, tapering away from the reverse faults, but flat vein arays may extend over greater distances. The fault-related vein systems have been interpreted as the product of cyclical fault-valve action whereby failure on severely misoriented reverse faults (oriented at > 55-60° to σ1) is triggered by the accumulation of overpressure to near-lithostatic values, the ensuing fault rupture then allowing postfailure discharge upwards along the fault. Fluid inclusion studies support the cycling of fluid-pressure between ~lithostatic prefailure and sublithostatic postfailure values. The flat-lying hydrofracture arrays provide an explanation for the bright-spot reflectors observed around the base of the seismogenic zone, while the fault-vein complex may represent rupture nucleation sites on steep reverse faults where failure is predominantly fluid-driven. Under such circumstances, near-total shear stress relief may accompany rupture. Net fluid volumes involved in the formation of these vein systems may be of the order of 1 km3 per kilometer strike-length, but the fluid volume involved in each fault-valve cycle is likely to be 2-4 orders of magnitude lower.
S33D-04
Deep structure of the strain concentration zone in northern Miyagi Prefecture, NE Japan
A moderate-size (M6.4) intraplate earthquake (the 2003 northern Miyagi earthquake) occurred on 26 July 2003 in the northern part of Miyagi prefecture, NE Japan. In this area, high seismic activity forms a belt in which another moderate-size earthquake (M6.2) occurred in 1962. From GPS analysis, large strain concentration was observed along this seismic belt (Sato et al., 2004; Miura et al., 2005; Hasegawa et al., 2005). We call this zone of large strain concentration and high seismic activity the northern Miyagi strain concentration zone. We applied the double-difference tomography method (Zhang and Thurber, 2003) to determine the three- dimensional velocity structure and relocate the hypocenters simultaneously. Travel time data is from the Japanese universities joint seismic observation in the Tohoku Backbone range (1997-1998) and the aftershock observation for the 2003 Northern Miyagi earthquake (Umino et al., 2004; Okada et al., 2004). We also used data from stations of Tohoku University, JMA, and Hi-net during the period from 1997 to 2003. Note that we have included events within the subducting Pacific slab in the tomography inversion to increase the resolution of the lower crust structure. In the focal area of the 2003 northern Miyagi earthquake, we obtain a steeply westward-dipping aftershock alignment, which shows the high dip angle of the reverse faulting (dip ~ 50 degree). We also obtain the low-velocity hanging-wall and the high-velocity footwall, consistent with the previous results of Okada et al. (Tectonophysics, 2007, doi:10.1016/j.tecto.2006.11.001). We infer that the 2003 northern Miyagi earthquake occurred along a fault that acted as a normal fault in the Miocene and has been reactivated as a reverse (unfavorably oriented) fault under the present compressional stress regime. The large slip area by the main shock rupture (asperity) corresponds to an area with relatively high velocity. We also obtain a distinct low-velocity area in the lower crust just beneath the focal area of the 2003 northern Miyagi earthquake. This low-velocity zone appears to continue downward to the upper mantle. The lateral extent of this low-velocity zone in the lower crust roughly corresponds to the spatial distribution of the northern Miyagi strain concentration zone. This low-velocity zone might correspond to the upwelling flow of fluid originating from the mantle wedge. This observation supports the hypothesis by Hasegawa et al. (2005) that anelastic deformation due to fluids forms the strain concentration zone. This observation also supports the hypothesis by Okada et al. (2007) and Sibson (1990) that high fluid pressure reactivates the unfavorably oriented fault of the 2003 northern Miyagi earthquake. http://www.aob.geophys.tohoku.ac.jp/~okada
S33D-05
Structural Heterogeneity Along the San Andreas Fault at Parkfield, California, from Seismic Imaging: Velocity, Anisotropy, and Attenuation
Seismic imaging is an essential tool for extending surface and borehole observations of crustal heterogeneity into three dimensions. Borehole and surface site investigations carried out for the SAFOD project provide an unparalleled opportunity to examine the detailed structure of a fault zone, at multiple scales. Our long-term deployments of PASSCAL and USArray seismic instruments combined with data from the USGS Northern California and UC Berkeley HRSN networks, SAFOD borehole logs, borehole seismometers, and several active- source projects have yielded an extensive dataset for seismic imaging analysis, including seismic velocity, anisotropy, and attenuation. Not surprisingly, the three-dimensional heterogeneity of all three of these seismic parameters in the upper crustal volume surrounding SAFOD is extraordinarily complex. The widely variable seismic wave sampling of this volume motivated the development and application of an adaptive-mesh double- difference tomography technique to obtain higher-resolution structural information in the earthquake source regions. Key features are the high-velocity granitic rocks on the southwest side of the fault, a complex steeply dipping low-velocity zone beneath and southwest of the surface fault trace, and an upper-crustal region of low velocity (below 4.5 km/s) extending to at least 5 km depth on the northeast side. Properly accounting for the extreme seismic velocity heterogeneity at km and larger scales is crucial for unraveling the highly variable shear- wave splitting observations. When projected to depth along the ray paths, average fast-polarization directions fall mainly into either fault-parallel or fault-normal clusters, with localized transitions of oblique orientation. We also use the velocity model to back-project shear-wave splitting delay times to crudely image zones responsible for the observed delays, which concentrate in the upper 3 km of the fault zone and extend somewhat deeper in portions of the granitic basement. Our attenuation (Q) tomography results provide further evidence for the strong heterogeneities surrounding SAFOD. The Qp and Qs models correlate reasonably well with the seismic velocity models, but they show no clear relationship to the strength of the inferred anisotropy.
S33D-06
Crustal Imaging Using Earthquake Waveforms From Precision-located Earthquake Sequences.
Our ability to accurately locate crustal earthquakes has markedly improved with recent improvements in the density of temporary and permanent arrays, and with new approaches for relative earthquake location, using cross-correlation analysis and double-difference location techniques. If these approaches are applied to earthquake swarm sequences, such as commonly observed in the active Taupo Volcanic Zone of New Zealand, waveforms from closely-related earthquakes in a swarm sequence can be subsequently used for crustal imaging, once errors in relative locations are much less than the shortest wavelength. Here we use waveforms from more than 700 earthquakes (2 < ML < 4) in the 2005-2007 Matata earthquake swarm to image crustal structure in the Taupo Volcanic Zone. We first relocate the sequence using both absolute and waveform-based differential time measurements, simultaneously inverting for the local velocity structure using double-difference tomography. The relocated events are clearly aligned c.NE-SW (strike 35 degrees) on several sub-parallel planes, with events clustered at 2-6 km depth, just northwest of the rupture area of the 1987 M6.2 Edgecumbe earthquake and southwest of deep bright spot' seismic reflections observed at 10-11 km depth. Super-receiver gathers of the earthquake waveforms, processed using coherency and frequency filtering, show excellent consistency out to 110 km+ from some of the event clusters. Coherent phases with strong amplitude are observed in the S-wave coda, 2-4 seconds after direct S, especially at offsets between 23 and 52 km. These may relate to lower crustal changes in S-wave reflectivity associated with intrusions. This interpretation is being examined using elastic wave finite difference synthetics and the observed 3-component data.
S33D-07
Shear-wave splitting analysis at the Niigata-Kobe Tectonic Zone using the seismic network of the Joint Seismic Observations at NKTZ
In central Japan, high strain rate zone which was called Niigata-Kobe Tectonic zone (NKTZ) was detected by GPS study. Many historical large earthquakes have occurred inside the NKTZ. A large right lateral Atotsugawa fault is located in NKTZ. The understanding of the cause of NKTZ will be important to know the accumulation mechanism of the stress and strain in Japan. It is well known that elastic anisotropy of the earth is closely related to mantle dynamics. The studies of shear-wave splitting will be important to know the mantle dynamics beneath NKTZ. We conducted seismic observation at the area with a spatially high dense seismic array. The seismic network was used for the analysis of shear-wave splitting to detect the heterogeneous structure at the crust and upper mantle at NKTZ. The deep earthquakes with depth of deeper than 280 km are used. The deep earthquakes are located beneath the seismic network. The large lateral variation was found on the polarization direction in the results of our study. The research area was divided into 4 regions based on the observed polarization directions. One is the northern part of the research area. In this area, the polarization direction was NW-SE. The eastern part of this research area indicated NE-SW directions. In the central part of the research area, the polarization direction was NNE-SSW. The E-W polarization direction was obtained at the seismic stations located in the southern part of the research area. The large shear-wave splitting, which is larger than 0.6 sec on time lag data, could not be explained by the crustal anisotropy. The cause of the shear-wave splitting should be located in the mantle wedge. Lack of the spatially distribution of the seismic stations prevents high resolution study in this area. Recently, seismic tomography studies using spatially high dense seismic network indicated high resolution velocity structure in this area. The configuration of the subducting Philippine Sea slab was clearly delineated. The observed polarization directions are consistent with the subduction direction of Philippine Sea slab. The observed shear-wave splitting can be explained by the preferred orientation of the olivine crystal which was caused by the mantle flow.