The vast and complicated landscape of western North America has been attributed to the long history of the ancient Farallon plate that began subducting underneath it about 70 Ma ago. Orogenies, volcano activities, and accompanying uplift and extension shaped the continent into different units of distinct tectonic features as seen today.
The Rio Grande Rift (RGR), one of the largest-scale topographic manifestations in the southwestern United States, extends approximately north-southwardly from the south tip of Rocky Mountains in Colorado, runs through New Mexico, gradually merges with Basin and Range Province in Chihuahua, Mexico and terminates in the westernmost part of Texas [Keller and Baldridge, 1999]. It is surrounded by the Rocky Mountains to the north, the Great Plains to the east, the Colorado Plateau to the west and the Basin and Range Province to the south and southeast (Fig. 1-1).
The opening of the RGR system started about 35 Ma ago when the crust and mantle lithosphere were pulled apart by the regional extension, inducing the upwelling of asthenospheric mantle to replace the thinning lithosphere and generate melts by pressure release melting. The upper crust have undergone brittle failure and formed a series of asymmetrical half grabens [Wilson et al., 2005]. Though being relatively dormant today, intensive igneous extrusion and volcanic activities have taken place during the rift-initiation period [Humphreys, 1995]. Heat flow measurements in the RGR show higher values than the surrounding Great Plains and Colorado Plateau [Slack et al., 1996].
In addition, the crustal thickness across the rift varies dramatically from on average 44.1 km under the Great Plain to about 35 km under the rift axis [Olsen et al., 1987; Wilson et al., 2005] (Fig. 1-2). The most recent volcanic activities occurred to the west of the main
(a)
(b)
Fig. 1-1. (a) is the topographic map of the southwestern United States. GP: Great Plains; RM: Rocky Mountains; CP: Colorado Plateau; BR: Basin and Range Province. Black lines outlined the Rio Grande Rift. Dashed line circles the Colorado Plateau. Red-shaded regions mark the late Cenozoic volcanic fields with the average ages less than 4.5 million years old. Stations used in this study are denoted by triangles with various colors that stand for different networks labeled on the bottom of (b). Geological features shown on the map are modified from [Spence and Gross, 1990].
fields, spanning about 800 km long and running from the Rayton Clayton volcanic field, northeastern New Mexico to the San Carlos volcanic field, Arizona, mark the transition zone between Colorado Plateau and Rio Grande Rift and obliquely intersect the rift axis [Spence and Gross, 1990] (Fig. 1-1).
Numerous geological and geophysical investigations have been conducted in this region, aiming at understanding the tectonic mechanism that influences the evolution and structure of continental rifts [Perry et al., 1988; Spence and Gross, 1990; Baldridge et al., 1991; Slack et al., 1996; Gao et al., 2004; Song and Helmberger, 2007]. Seismic tomographic imaging has been a useful tool to reveal underlying seismic velocity and attenuation structures, which provide important clues on the temperature, melt/fluid content and chemical composition in the lithosphere and upper mantle under the continental rift zones. Shear wave velocity models in North America obtained from large scale travel time and waveform tomography have imaged strong velocity gradients across
Fig. 1-2. Variations of the crustal thickness under the southwestern United States based on the
the margin of the stable cratonic keel and high wave speed anomalies in the upper mantle beneath the western United States associated with the remnants of the subducted Farallon plate [Grand, 1994; Lee and Nolet, 1997]. In the past two decades, the increasing temporary deployment of dense seismic arrays across continents has improved the tomographic images with unprecedented resolutions [Spence and Gross, 1990;
Humphreys and Dueker, 1994; Slack et al., 1996].
Different rifting mechanisms, either pure-shear or simple-shear deformation, result in distinguishable symmetric or asymmetric configuration of the crust and mantle lithospheric structure [Olsen et al., 1987; Wilson et al., 2005]. In order to construct the high-resolution seismic images of upper mantle velocity structure and topographic variation on the crust-mantle interface under the RGR, a NW-SE trending linear array of about 950 km long that consists of 54 broadband instruments spaced, on average, 18 km apart was deployed across the rift from July 1999 to May 2001. The Colorado PLAteau/Rio Grande Rift Seismic TRAnsect Experiment (hereafter referred as La Ristra and coded as XM in Fig. 1-1), with its two ends running through the Colorado Plateau and the Great Plains, was designed to record teleseismic wave arrivals from abundant earthquakes in the Aleutian and South American subduction zones which directly probe the crust and upper mantle beneath these three tectonic regions [Wilson et al., 2003].
The data collected by the La Ristra array have been thoroughly explored in various seismological studies including body-wave and surface-wave travel time tomography [Gao et al., 2004; West et al., 2004; Wilson et al., 2005], shear wave splitting analysis [Gök et al., 2003], and receiver function imaging [Wilson et al., 2003; Wilson et al., 2005].
[Gao et al., 2004] present the 2-D, ray-based P- and S-wave models which yield the largest amplitude variations within the upper 200 km depth (Fig. 1-3). The slow
anomalies lie beneath the rift axis and become more deeply rooted under the closely-linked Jemez Lineament. Between the volcanic lineament and the Colorado Plateau exists a sharp velocity boundary. Under the eastern plateau shows a high velocity region, whereas beneath the central part resolves a significant low velocity zone extending from the surface to ~300 km depth which might contribute to the once active Navajo volcanic field at around 20-30 Ma ago. Further west to the verge, the fast anomaly,
Fig. 1-3. 2-D P- and S-wave velocity models of the upper mantle under the La Ristra array derived from body-wave travel time tomography [[Gao et al., 2004]
Fig. 1-4. 2-D shear velocity structurel of the crust (top panel) and upper mantle down to
~400 km depth (bottom panel) under the La Ristra array derived from surface wave inversion [West et al., 2004].
eastern end of the array the fast anomalies are imaged under the Great Plains at the depths above 200 km and between 300 km and 600 km in S-wave model. This sheet-like feature is interpreted as the downwelling limb associated with the small-scale convection. (Fig.
1-3)
Shear velocity structure imaged from surface wave dispersion reveals a broader low velocity zone in the uppermost mantle above 150 km depth beneath the rift axis [West et al., 2004]. It extends laterally about 150-200 km on both sides of the rift axis, ends abruptly under the Great Plains to the east, and gradually merges with deeper slow regions at depths greater than 150 km under the Colorado Plateau to the west. The rather
Fig. 1-5. Topographic variations of upper mantle seismic discontinuities from P-to-S conversions revealed from receiver function migration images [Wilson et al., 2003].
Fig. 1-6. Finite-difference synthetic waveform modeling using the S-wave model of Gao et al. [2004]. The model is amplified by one, two and three times to meet the actual observation of travel time anomalies and amplitude ratio. Model A is constructed by amplifying the fast anomaly by 2 and slow anomaly by 4 in order to get better fit. [Song and Helmberger, 2007]
warm mantle under the plateau from ~150 to 275 km depth has been interpreted as the asthenosphere which replaced the falling Farallon slab and was capable of providing the buoyancy needed to support the average 1.8 km elevation of the Colorado Plateau. The low velocity anomaly under the Rio Grande Rift does not appear to be fed by a deeper source, suggesting that the rift might be passively driven by the lithospheric extension [West et al., 2004]. The receiver function imaging from [Wilson et al., 2003] also shows negligible topographic variations on the 410 km and 670 km seismic discontinuity, indicating there is no localized thermal anomaly in the transition zone under the region. It thus concludes that the major thermal signature is predominantly confined to the uppermost mantle above 400 km depth (Fig. 1-5). Besides, the topography of the Moho discontinuity shows a 10 km rift-symmetric crustal thinning beneath the rift, suggesting a pure shear thinning mechanism which is also supported by surface wave results.
Because of the lack of low velocity anomaly extending down to the deeper transition zone and lower mantle [Gao et al., 2004; Wilson et al., 2005], the integrated seismic evidences support the scenario in which small-scale convection occurred in response to the lithospheric extension and thinning. Moreover, an elongated, relatively high wave speed feature was detected at the margin of well-resolved regions, which has been attributed to the fragment of the subducted Farallon slab [Gao et al., 2004].
[Song and Helmberger, 2007] employed 2-D finite-difference waveform modeling to calculate synthetic P and S waveforms in 2-D heterogeneous velocity structures beneath the La Ristra array resolved from previous tomographic studies. The predicted travel time and amplitude variations of P and S waves across the array were too weak to match the observed results, suggesting that either the regularization invoked to suppress the noisy short-wavelength fluctuations has sacrificed the power spectra of the tomographic models or finite-frequency effects on observed travel time delays was not properly taken
into account in classical ray-based tomography. Stronger velocity heterogeneity by increasing a factor of 2 for fast anomaly and 4 for slow anomaly yields significantly improved fits of observed travel time and amplitude anomalies (Fig. 1-6).
To account for intrinsic wavefront healing effects on finite-frequency seismic waves, the tenet of seismic tomography has moved on from ray theory in the infinite frequency limit to physically-realistic finite-frequency theory [Dahlen et al., 2000]. Considering the destructive interference of scattered waves from velocity heterogeneities at different frequencies, the travel time shift of a finite frequency phase arrival is no longer sensitive to seismic wave speed perturbations along its infinitesimally narrow ray path; rather, it is mostly influenced by 3-D volumetric structure surrounding the path. Whenever the scale length of heterogeneity is on the order of the cross path width of the sensitivity zone that depends on the dominant wavelength and propagation distance of the wave, finite-frequency wavefront healing becomes significant and ray theory will overestimate observed travel time anomalies and thus underestimate the strength of velocity heterogeneity. Benefiting from the great leap in computing power and the availability of high-quality data from dense broadband seismic networks, finite-frequency theory has been extensively employed to better resolve the images of upwelling mantle plumes [Hung et al., 2004; Montelli et al., 2004] and fine scale structures in the lowermost mantle [Hung et al., 2005].
Alternatively, due to uneven data coverage, data uncertainty and approximation of forward theory, different model parameterization and regularization schemes imposed on the tomographic inversion can lead to the non-uniqueness of the resolved velocity model.
Multiscale model construction through the wavelet decomposition and synthesis was then introduced to better achieve the spatial resolution in densely-sampled regions while securing the long-wavelength spectral resolution [Chiao and Kuo, 2001; Chiao and Liang,
2003]. The merit of the multiscale tomography lies on the automatic determination of the tomographic images with spatially varying, data-adaptive resolvable scales.
In this study, all the important elements in seismic travel time tomography, including frequency-dependent travel-time observations, finite-frequency theory, and multi-resolution parameterization are essentially combined to reconstruct 3-D high-resolution images of P and S velocity perturbations beneath the southwestern United States. The region of interest spans from 28° to 42° in latitude and -114° to -100° in longitude. All the accessible teleseismic data recorded from the permanent and temporary broadband stations distributed in the area are collected through Data Management Center (DMC) of Incorporated Research Institutions for Seismology (IRIS). In addition to data from the linear La Ristra array and a few stations nearby to the east deployed in early 90s, 170 available USArray stations covering the west of the study region are also incorporated in this study. As a part of the ambitious NSF-funded EarthScope Project, the USArray has been underway since the fiscal year of 2005 and planned to sweep the entire United States starting from the west coast with 400 neatly pinned transportable seismometers of 50-70 km spacing over a 10-year period. Using seismic waves as probes, the experiment is aimed at investigating the structure and dynamics of the continental lithosphere under North America and the earth’s deep interior. The IRIS DMC provides the timely and open data access, which substantially augments the dataset used in this study.
I reexamine finite-frequency travel time shifts measured by inter-station cross correlation of waveforms band-pass filtered at high- (0.3-2 Hz for P and 0.1-0.5 Hz for S waves) and low-frequency bands (0.03-0.125 Hz for P and 0.03-0.1 Hz for S waves).
Employing more accurate finite-frequency sensitivity kernels for relative travel-time measurements and the wavelet-based multi-scale parameterization in the inversion, I
anticipate to yield more robust 3-D lithospheric and upper mantle structures under the RRGR and adjacent regions and gain more insights into the regional tectonics and mantle dynamics that characterize the surface geology and underlying velocity structure.