Haijiang Zhang

Double-Difference Tomography: The Method and Its Application to the Hayward Fault, California

We have developed a double-difference (DD) seismic tomography method that makes use of both absolute and relative arrival times. By reducing systematic errors using the more accurate relative arrival times, the method produces an improved velocity model. Simultaneously, it yields event locations of a quality equivalent to those of the DD location method, while avoiding simplifying assumptions of that method. We test this method on a synthetic dataset and find that it produces a more accurate velocity model and event locations than standard tomography. We also test this method on a Hayward fault, California, earthquake dataset spanning 1984–1998. The earthquakes relocated by this method collapse to a thin line along the fault trace, consistent with previous results. The DD velocity model has sharper velocity contrasts near the source region than the standard tomography model.

Automatic P-Wave Arrival Detection and Picking with Multiscale Wavelet Analysis for Single-Component Recordings

We have developed an automatic P -wave arrival detection and picking algorithm based on the wavelet transform and Akaike information criteria (AIC) picker. Wavelet coefficients at high resolutions show the fine structure of the time series, and those at low resolutions characterize its coarse features. Primary features such as the P -wave arrival are retained over several resolution scales, whereas secondary features such as scattered arrivals decay quickly at lower resolutions. We apply the discrete wavelet transform to single-component seismograms through a series of sliding time windows. In each window the AIC autopicker is applied to the thresholded absolute wavelet coefficients at different scales, and we compare the consistency of those picks to determine whether a P -wave arrival has been detected in the given time window. The arrival time is then determined using the AIC picker on the time window chosen by the wavelet transform. We test our method on regional earthquake data from the Dead Sea Rift region and local earthquake data from the Parkfield, California region. We find that 81% of picks are within 0.2-sec of the corresponding analyst pick for the Dead Sea dataset, and 93% of picks are within 0.1 sec of the analyst pick for the Parkfield dataset. We attribute the lower percentage of agreement for the Dead Sea dataset to the substantially lower signal-to-noise ratio of those data, and the likelihood that some percentage of the analyst picks are in error.

Three-dimensional compressional wavespeed model, earthquake relocations, and focal mechanisms for the Parkfield, California, region

We present a new three-dimensional (3D) compressional wavespeed ( V p) model for the Parkfield region, taking advantage of the recent seismicity associated with the 2003 San Simeon and 2004 Parkfield earthquake sequences to provide increased model resolution compared to the work of Eberhart-Phillips and Michael (1993) (epm93). Taking the epm93 3D model as our starting model, we invert the arrival-time data from about 2100 earthquakes and 250 shots recorded on both permanent network and temporary stations in a region 130 km northeast–southwest by 120 km northwest–southeast. We include catalog picks and cross-correlation and catalog differential times in the inversion, using the double-difference tomography method of Zhang and Thurber (2003). The principal V p features reported by epm93 and Michelini and McEvilly (1991) are recovered, but with locally improved resolution along the San Andreas Fault (saf) and near the active-source profiles. We image the previously identified strong wavespeed contrast (faster on the southwest side) across most of the length of the saf, and we also improve the image of a high V p body on the northeast side of the fault reported by epm93. This narrow body is at about 5- to 12-km depth and extends approximately from the locked section of the saf to the town of Parkfield. The footwall of the thrust fault responsible for the 1983 Coalinga earthquake is imaged as a northeast-dipping high wavespeed body. In between, relatively low wavespeeds (<5 km/sec) extend to as much as 10-km depth. We use this model to derive absolute locations for about 16,000 earthquakes from 1966 to 2005 and high-precision double-difference locations for 9,000 earthquakes from 1984 to 2005, and also to determine focal mechanisms for 446 earthquakes. These earthquake locations and mechanisms show that the seismogenic fault is a simple planar structure. The aftershock sequence of the 2004 mainshock concentrates into the same structures defined by the pre-2004 seismicity, confirming earlier observations (Waldhauser et al. , 2004) that the seismicity pattern at Parkfield is long lived and persists through multiple cycles of mainshocks. Online material : 3D V p model and earthquake relocations.

High-resolution subducting-slab structure beneath northern Honshu, Japan, revealed by double-difference tomography

The high-resolution seismic velocity structure of the subducting slab beneath northern Honshu, Japan, has been obtained by double-difference tomography, capitalizing on the existence of two planes of seismicity. The upper plane lies in the region with average to high Vp/Vs ratios (1.72–1.85), which may correspond to the transformations of metabasalt and metagabbro to blueschist. The lower plane is associated with very low Vp/Vs ratios (1.6–1.7), in sharp contrast with high Vp/Vs ratios (∼1.8–1.85) in the region between the two planes. These features may be explained by forsterite-enstatite-H2O formation from serpentine dehydration in the lower plane and partial hydration of the region between the two planes. Our results provide direct constraints for petrologic modeling of the subducting slab.

An improved automatic time-of-flight picker for medical ultrasound tomography

OBJECTIVE AND MOTIVATION Time-of-flight (TOF) tomography used by a clinical ultrasound tomography device can efficiently and reliably produce sound-speed images of the breast for cancer diagnosis. Accurate picking of TOFs of transmitted ultrasound signals is extremely important to ensure high-resolution and high-quality ultrasound sound-speed tomograms. Since manually picking is time-consuming for large datasets, we developed an improved automatic TOF picker based on the Akaike information criterion (AIC), as described in this paper. METHODS We make use of an approach termed multi-model inference (model averaging), based on the calculated AIC values, to improve the accuracy of TOF picks. By using multi-model inference, our picking method incorporates all the information near the TOF of ultrasound signals. Median filtering and reciprocal pair comparison are also incorporated in our AIC picker to effectively remove outliers. RESULTS We validate our AIC picker using synthetic ultrasound waveforms, and demonstrate that our automatic TOF picker can accurately pick TOFs in the presence of random noise with absolute amplitudes up to 80% of the maximum absolute signal amplitude. We apply the new method to 1160 in vivo breast ultrasound waveforms, and compare the picked TOFs with manual picks and amplitude threshold picks. The mean value and standard deviation between our TOF picker and manual picking are 0.4 micros and 0.29 micros, while for amplitude threshold picker the values are 1.02 micros and 0.9 micros, respectively. Tomograms for in vivo breast data with high signal-to-noise ratio (SNR) ( approximately 25 dB) and low SNR ( approximately 18 dB) clearly demonstrate that our AIC picker is much less sensitive to the SNRs of the data, compared to the amplitude threshold picker. DISCUSSION AND CONCLUSIONS The picking routine developed here is aimed at determining reliable quantitative values, necessary for adding diagnostic information to our clinical ultrasound tomography device--CURE. It has been successfully adopted into CURE, and allows us to generate such values reliably. We demonstrate that in vivo sound-speed tomograms with our TOF picks significantly improve the reconstruction accuracy and reduce image artifacts.

Joint inversion for Vp, Vs, and Vp/Vs at SAFOD, Parkfield, California

We refined the three-dimensional (3-D) Vp, Vs and Vp/Vs models around the San Andreas Fault Observatory at Depth (SAFOD) site using a new double-difference (DD) seismic tomography code (tomoDDPS) that simultaneously solves for earthquake locations and all three velocity models using both absolute and differential P, S, and S-P times. This new method is able to provide a more robust Vp/Vs model than that from the original DD tomography code (tomoDD), obtained simply by dividing Vp by Vs. For the new inversion, waveform cross-correlation times for earthquakes from 2001 to 2002 were also used, in addition to arrival times from earthquakes and explosions in the region. The Vp values extracted from the model along the SAFOD trajectory match well with the borehole log data, providing in situ confirmation of our results. Similar to previous tomographic studies, the 3-D structure around Parkfield is dominated by the velocity contrast across the San Andreas Fault (SAF). In both the Vp and Vs models, there is a clear low-velocity zone as deep as 7 km along the SAF trace, compatible with the findings from fault zone guided waves. There is a high Vp/Vs anomaly zone on the southwest side of the SAF trace that is about 1–2 km wide and extends as deep as 4 km, which is interpreted to be due to fluids and fractures in the package of sedimentary rocks abutting the Salinian basement rock to the southwest. The relocated earthquakes align beneath the northeast edge of this high Vp/Vs zone. We carried out a 2-D correlation analysis for an existing resistivity model and the corresponding profiles through our model, yielding a classification that distinguishes several major lithologies.

A California Statewide Three-Dimensional Seismic Velocity Model from Both Absolute and Differential Times

We obtain a seismic velocity model of the California crust and uppermost mantle using a regional-scale double-difference tomography algorithm. We begin by using absolute arrival-time picks to solve for a coarse three-dimensional (3D) P ve- locity (VP) model with a uniform 30 km horizontal node spacing, which we then use as the starting model for a finer-scale inversion using double-difference tomography applied to absolute and differential pick times. For computational reasons, we split the state into 5 subregions with a grid spacing of 10 to 20 km and assemble our final statewide VP model by stitching together these local models. We also solve for a state- wide S-wave model using S picks from both the Southern California Seismic Network and USArray, assuming a starting model based on the VP results and a VP=VS ratio of 1.732. Our new model has improved areal coverage compared with previous models, extending570kmintheSW-NEdirectionand1320kmintheNW-SEdirection.Italso extends to greater depth due to the inclusion of substantial data at large epicentral distances. Our VP model generally agrees with previous separate regional models for northern and southern California, but we also observe some new features, such as high-velocity anomalies at shallow depths in the Klamath Mountains and Mount Shasta area, somewhat slow velocities in the northern Coast Ranges, and slow anoma- lies beneath the Sierra Nevada at midcrustal and greater depths. This model can be applied to a variety of regional-scale studies in California, such as developing a unified statewide earthquake location catalog and performing regional waveform modeling. Online Material: Smoothing and damping trade-off analysis, a priori Moho depth, resolution tests, and map-view slices and cross sections through the 3D VP and VS models.

Passive seismic tomography using induced seismicity at a petroleum field in Oman

A borehole network consisting of five monitoring wells monitored the induced seismicity at a producing petroleum field for about 11 months . Nearly 5400 microseismic events were analyzed and used to image the reservoir based on a new double-difference (DD) seismic tomography. The DD tomography method simultaneously solved for event locations and VP , VS , and VP ∕ VS models using absolute and differential P, S, and S-P arrival times. Microseismicity in the field was caused primarily by compaction of the reservoir in and above the gas-bearing formation and was distributed along the two major northeast-southwest faults in the field. The model resolution analysis based on the checkerboard test and the resolution matrix showed that the central part of the model was resolved relatively well for the depth range of 0.7 to 1.1 km . Clear velocity contrasts were imaged across most parts of the two northeast-southwest faults. The VP ∕ VS ratio estimates from the tomographic inversion were low (<1.75) in the shallow...

Aftershock distribution and 3D seismic velocity structure in and around the focal area of the 2004 mid Niigata prefecture earthquake obtained by applying double-difference tomography to dense temporary seismic network data

A destructive large earthquake (the 2004 mid Niigata prefecture earthquake) sequence occurred in the central part (Chuetsu district) of Niigata prefecture, central Japan on October 23, 2004. We have deployed a temporary seismic network composed of 54 stations for aftershock observation just above and around the focal area of the earthquake for about a month. Using travel time data from the temporary seismic network and surrounding routine stations, we obtained precise aftershock distribution and 3D seismic velocity structure in and around the fault planes of the earthquake and four major (M ≥ 6) aftershocks by double-difference tomography. The results clearly show three major aftershock alignments. Two of them are almost parallel and dipping toward the WNW. The shallow and deep aftershock alignments correspond to the fault plane of the mainshock and that of the largest aftershock (M6.4), respectively. The third alignment is almost perpendicular to the WNW-ward dipping planes and perhaps corresponds to the fault plane of the M6 aftershock on October 27. General feature of the obtained velocity structure is that the hanging wall (western part of the focal area) has lower velocity and the footwall (eastern part of the focal area) has higher velocity. Major velocity boundary seems to shift westward in comparison to in northern and southern parts at a location near the central part of the focal area, where the main shock rupture started. Some parts of the fault planes were imaged as low velocity zones. This complex crustal structure would be one of possible causes of the multi-fault rupture of the 2004 mid Niigata prefecture earthquake sequence.

Ductile Gap between the Wenchuan and Lushan Earthquakes Revealed from the Two-dimensional Pg Seismic Tomography

A high-resolution two-dimensional Pg-wave velocity model is obtained for the upper crust around the epicenters of the April 20, 2013 Ms7.0 Lushan earthquake and the May 12, 2008 Ms8.0 Wenchuan earthquake, China. The tomographic inversion uses 47235 Pg arrival times from 6812 aftershocks recorded by 61 stations around the Lushan and Wenchuan earthquakes. Across the front Longmenshan fault near the Lushan earthquake, there exists a strong velocity contrast with higher velocities to the west and lower velocities to the east. Along the Longmenshan fault system, there exist two high velocity patches showing an “X” shape with an obtuse angle along the near northwest-southeast (NW-SE) direction. They correspond to the Precambrian Pengguan and Baoxing complexes on the surface but with a ~20 km shift, respectively. The aftershock gap of the 2008 Wenchuan and the 2013 Lushan earthquakes is associated with lower velocities. Based on the theory of maximum effective moment criterion, this suggests that the aftershock gap is weak and the ductile deformation is more likely to occur in the upper crust within the gap under the near NW-SE compression. Therefore our results suggest that the large earthquake may be hard to happen within the gap.