Chunquan Yu
SeismoLab, California Institute of Technology


1) Virtual Deep Seismic Sounding (VDSS), Moho depths and crustal buoyancy

a. VDSS method

VDSS utilizes the prominent SsPmp phase, which undergoes S-to-P conversion at the free surface and P-to-P total reflection off the Moho. The differential timing between the SsPmp phase and the direct S phase, Ss, is closely related to the crustal thickness H and average crustal P-wave velocity VP.


where p is the incident S-wave ray parameter (horizontal slowness).

b. Application to the North China craton

We applied VDSS to study the crustal structures of the North China craton along a 1000 km-long, east-west trending seismic profile. Results of VDSS, combined with traditional receiver function (RF) analysis, suggest that the crust under the eastern Ordos is thicker (at least 60 km) than expected from previous studies and from its modest elevation (about 1500 m above sea-level). The thick crust under the Ordos plateau is characterized by a distinct layer of lower crust, likely of mafic composition and reaching a maximum thickness of about 20 km.

Insofar the current configuration of the lithosphere under the Ordos plateau might serve as a proxy for the initial condition prior to reactivation of the eastern part of NCC - where a cratonic keel no longer seems to exist - our results support the hypothesis that lower crust foundering was due to transformation of a thick mafic lower-crust to a garnet-rich assemblage (possibly caused by hydration associated with subduction during and/or before mid-Mesozoic times). For more details, please refer to Yu et al. (2012).

c. Earthquake source deconvolution

Original applications of VDSS rely on deep earthquakes as sources of illumination to circumvent strong, near-source scattering (e.g. depth phases) and are, therefore, limited by the uneven distribution of deep seismicity. To extend both the applicability and the quality of VDSS, we developed a method to effectively remove earthquake source signatures. It involves two steps. First, based on analyses of particle motion, we separate 'pseudo-P' and 'pseudo-S' wave trains from the vertical and the radial component of ground motion. The latter is then used as the appropriate reference time-series for the deconvolution of the vertical and the radial component of ground motion.

d. Applications to the Hi-CLIMB seismic array

The method is verified from a series of synthetic tests, and is further validated using data recorded by the Hi-CLIMB array from both deep and shallow earthquakes. Since shallow earthquakes are much more abundant (and geographically distributed more widely) than deep seismicity, the approach presented here greatly extends the applicability of VDSS, including many geologically important regions where crustal isostasy and dynamic topography are yet to be constrained. For more details, please refer to Yu et al. (2013).

2) Imaging upper mantle structures with SS precusors

a. Principles

The teleseismic SS phase bounces off the free surface roughly at the middle point between the earthquake epicenter and the recording station. Its precursors, for example S410S and S660S, reflect at deeper interfaces, and therefore arrive before the main SS phase. The differential travel time between the SS phase and its precursors has been widely used for mapping upper mantle discontinuities, especially the so called "410" and "660".

b. Simple stacks

If we stack all seismic traces (aligned by the SS phase) with respect to epicentral distances, the S410S and S660S phases, which usually can not be seen on a single trace, show up clearlly.

c. SS precusor enhancement, coherent/incoherent noise suppression with curvelet transform

Traditionally, researchers limit the event-station epicenter distance to > 110o to avoid contamination with multiples of S and Sdiff at receiver side, e.g. Ss410s and Ss660s. However, this data selection criteria significantly reduces the number of dataset for many regions, e.g. Hawaii. On the other hand, 110o-170o is not really a perfect range. In fact, in this range SS precursors are also contaminated by ScSScS precursors and others.

We applied curvelet transform to remove coherent/incoherent noise from SS and its precursors. Curvelets are localized so that features with different size can be automatically captured with different scales. The directional information of curvelets are particularly useful to isolate coherent noise, e.g. Ss410s and Ss660s. Tests on synthetic and field seismic data show excellent recovery of SS precursors and removal of contamination phases on stacked images. Currently, I am applying this technique to image the mantle transition zone in the Pacific region.

3) Crustal deformation using GPS measurements

a. Block models vs. Continuum models

In the late 1960s, plate tectonics was introduced by geophysicists as an internally consistent theory of relative motions of vast rigid plates. While oceanic plates with narrow plate boundaries fit this model remarkably well, numerous examples reveal dispersive continental deformations over thousands of kilometers.

Various models have been proposed to describe intracontinental deformations. These models can generally be divided into two categories: (1) continuum models, and (2) microplate or block models. Continuum models assume that the continental lithosphere deforms as continuous ductile material. They aim to address the problems of both kinematic description of lithospheric deformation and its underlining geodynamic driving forces. Microplate or block models, however, only focus on the kinematic aspects of crustal deformation. They divide the brittle/elastic part of the crust into many blocks and study their relative movements. There is still no consensus on which approach is better.

b. Block models of crustal deformation in the Tien Shan area

The Tien shan is the most prominent orogenic belt in Central Asia. Widely spread thrust faulting and significant N-S shortening suggest that active intraplate deformation is still going on. In the last few decades, geodetic, geological, and geophysical surveys have accumulated abundant field dataset. Thus it is an ideal place to study the kinematic and dynamic processes of intracontinental deformation.

GPS Measurements, Data fitting, and Shortening rates

4) Earthquake locations and focal mechanisms.

a. Method

We relocate earthquakes using double-difference method. Focal mechanisms are determined using P-wave polarities.

b. Applications to the Wenchuan earthquake sequence

We relocated 1376 earthquakes with Ms>3.5 and obtained 83 well-determined earthquake focal mechanisms with Ms>4.0 in the Wenchuan earthquake sequence. The results show some distinct features.

i) The compressional axis (P-axis) is subhorizontal for almost all large-magnitude earthquakes.

ii) The major fault is dominated by thrust events but there are two exceptions. Around Xiaoyudong in the southern portion and Qingchuan in the northern portion, focal mechanisms are dominated by strike-slip type. Combined with their spatial distribution, we suggest that the Xiaoyudong cluster is located on a left-lateral NW-SE trending fault, and the Qingchuan cluster is located on a right-lateral NW-SW trending fault.

iii) There are two groups of thrust-type focal mechanisms. The first group, including to the main shock, has a P-axis perpendicular to the strike of the main fault. The second group, on the other hand, has a P-axis parallel to the strike of the main fault