Previous Page -- Outline -- Details: Velocity Optimization -- Next Page


Imaging a listric normal fault, Death Valley

Calif. and Nevada location map
From a tectonic problem involving the imaging of active earthquake faults in the Los Angeles basin of Southern California for seismic hazard assessment, I would like to move north to the Basin and Range tectonic province. Geodetic work in the mojave Desert, and the 1992 magnitude 7.2 Landers earthquake, showed that about 12% of the Pacific-North America plate motion is diverted from the San Andreas fault eastward into the province in eastern California and Nevada. The seismic hazard there is thus about one tenth of the hazard in Los Angeles, and the population is very much lower still. I would like to explain how my imaging methods can reveal the very different tectonic character of this province, which has been a subject of much debate.

The Basin and Range is dominated by east-west extension, as you can see from the north-striking array of faults throughout Nevada. Against the Sierra Nevada mountains, on the west side of the province, the principal motion is not dip-slip but right-lateral along this fault system known as the Walker Lane.

Basin & Range extension models
At present there is a raging debate over the total Cenozoic extension experienced by the crust in the Basin and Range, especially in the Death Valley region (DV above). If the total extension is limited to 20-50%, then it could have occurred according to the ``Pure Shear'' model above. In a pure shear model, extension would be evenly distributed through all depths of the crust, demanding crustal thinning and Moho uplift. To match an apparently constant vertical travel time through the crust, the lower crust would have to have a significantly lower velocity in the extended region.

But pure shear is the only model that allows the pervasive normal faults of the Basin and Range to be straight and steeply dipping (60 degrees) in section. Such relatively thick-skinned extension, straight fault planes, and steeply-dipping normal mechanisms have been observed in Basin and Range earthquakes such as the 1954 Stillwater and Fairview Peak, and the 1983 Borah Peak events.

Geologic mapping has, on the other hand, led to the proposition by workers such as Brian Wernicke that the total Basin and Range extension is well beyond 100%. Such observations demand ``Simple Shear'' models, with the figure above giving two alternatives. Both concentrate extension to just a small depth range in the crust, by merging listric normal faults into regional detachment faults or decollements. If surface extension is truely large, say above 100%, then material has to be moved into the lower crust of the hyper-extended area from below the less-extended regions. Such mobile lower-crustal material must have comparatively high velocity, as in the lower section above.

I have been investigating many of the accessible manifestations of the possible extensional mechanisms, such as the history of basin development and the nature of mid-crustal velocities in this region (look here for papers on recent results). I will relate to you here just some analysis of reflection data for evidence of dip and listric geometry on one of the major normal faults, in Death Valley.

Death Valley location map
I look at reflection data collected by the COCORP consortium out of Cornell University. Their Line 9 crossed the southern Death Valley basin, and the basin-forming Black Mountains range-front normal fault. This fault has Holocene although not historic breaks, outlined in orange above. There is a significant right-lateral displacement component, dominating south of the 700,000-year-old cinder cone that is cut itself by 100 m dextral offset. I had to project my imaging analysis onto a 2-d section, from the tortuous path of the vibrator line.

COCORP proposed from Line 11 that a north-dipping fault connects a mid-crustal reflection bright spot to the cinder cone. Whether the north-dipping fault may be present or not, I will concern myself here just with the reflections from the west-dipping Black Mountains range-front fault, which has certainly generated at least magnitude 6 Holocene earthquakes, and is the principal source of down-dropping of the asymetric Death Valley basin.

Fault reflections and synthetics
The two shot records at left above provide the most direct evidence of Black Mountains range-front fault-surface reflections. The mapped surface location of the fault is at the apex of the diffraction atop the first arrival, direct evidence of a horizontally-propagating, negative-moveout reflection from a dipping fault surface. Note that one side of the diffraction is back-scattered (the right side, for the records above), and the other side forward-scattered (and usually larger in amplitude).

At the right above are 2-d synthetics from a finite-difference solution of the acoustic wave equation. These show the diffractions in the data do arise at the Black Mountains fault surface. Also, the match in slopes and locations of the diffraction hyperbolas' asymptotic tails and apexes proves that the fault I am imaging is west-dipping and not a projection of a north-dipping fault from a sideswipe reflection.

Since the fault-surface reflection has negative moveout in these gathers, any conventional seismic processing method based on the normal-moveout (NMO) correction and stacking will scramble the data from the dipping fault.

Serpa's COCORP results
The COCORP results from Serpa, presented by her as a post-stack migrated section, illustrate the limits of standard analysis. Near-horizontal stratigraphy of the asymetric basin stacks in well, but any direct image of the fault surface is lost. Serpa can interpret fault location and dip from stratigraphic truncations.

Optimized velocity models
To acurately interpret fault dip and geometry by imaging the fault-surface reflections in a prestack migration, I must have an accurate model of the lateral velocity variations along Line 9. This made done by picking the first arrivals from the reflection data, and then using Sathish Pullammanappallil's simulated-annealing optimization methods. The refined velocity model for the Death Valley basin part of Line 9 is above, along with the coarser model for all of Line 9. The accurate velocities allow the migration to fully account for reflection ray bending through the strong lateral velocity variations at the basin edges.

Details on velocity optimization

Migrated COCORP and synthetic records
Migrating only one shot gather at a time through the optimized velocity model shows how the prestack Kirchhoff-sum migration method works. The migrated data are on the left above, and the migrated synthetics are on the right. Among the many artifacts of migrated source-generated noise and multiple reflections, we can see the fault-surface image forming inside the red circles.

The image includes both the back-scattered and forward-scattered reflection components. For the data it is building up an image of the delta-lambda, Vp reflectivity. For the acoustic synthetics it is building an image of the reflectivity resulting from the change in incompressibility, which has isotropic scattering for the acoustic case.

Migration focusing criteria
The many artifacts present in the migration leads us to a statistical test of the final image section, above. On top is the migrated reflectivity section, derived directly from the data by Kirchhoff summation through the laterally variant optimized velocity model. It is cut by many near-vertical migration artifacts of poor reflection ray coverage, and spatial truncations of reflections due to near-surface variations or surface-consistent amplitude and phase statics.

Following the work of Harlan and Claerbout, we perform an identical migration to yield the next image down, but after scrambling the shot records by sign-reversing about every other trace. This gives an image of artifacts that result from the migration of noise that is not coherent from trace-to-trace in the prestack shot gathers. For the focusing image, the next one down, we statistically compare the amplitude histograms of the two images above using Harlan's signal expectation measure. Real reflections and diffractions, which should have more trace-to-trace coherency in the prestack data, give high signal expectation and identify the real reflective structures in the coherency image.

Using the coherency image to mask the original migration (from the top), we get a coherency-enhanced focused migration as the bottom image. The near-vertical artifacts are less prominent, and the west-dipping Black Mountains range-front fault is the most salient feature. Some of the basin stratigraphy, basin bottom, and an unknown structure within the hanging-wall bedrock remain.

Listric fault interpretation
The original migration, without statistical enhancement, we interpret to show likely features and artifacts at two times vertical exaggeration. The migrated image is also cut away to show the elevation of the ground surface (and shot and receiver points) within the image. Lowest elevation is about sea level. The Black Mountains range-front fault is clearly dipping at 40-60 degrees within 1 km of the surface, and takes a shallowing bend of at least 20 degrees below that depth. Dip of the deeper section of the fault may well be less than 20 degrees.

This image provides the first-ever accurate depth profile of a basement-involved listric normal fault. Listric normal faults entirely within clastic delta sediments have long been imaged by the petroleum industry. Here the combination of prestack Kirchhoff migration, to allow the use of forward- as well as back-scattered reflections, with accurate velocity modeling, was able to correctly place basement reflectors within a depth section. Having the correct fault geometry was crucial to proving the listric nature of this fault.


Previous Page -- Outline -- Details: Velocity Optimization -- Next Page


Presented by invitation to the Geophysics Section of Science Wellington, New Zealand, on September 17 1998.