Imaging from Earthquake Sources, So. Calif.

Previous Page -- Outline -- Next Page

Crustal imaging, southern California, using earthquakes as sources

My supposition that I can image fault structure preferentially to non-tectonic structures becomes crucial in my examination of earthquake faults in southern California. A billion years of active tectonism has left a legacy there of many, many inherited structures. This is also the most controversial imaging method I will show you. Instead of vibrators or explosions I use aftershock sequences as source arrays, and instead of geophone cables I will use seismograms recorded on the widely distributed permanent earthquake detection networks. The stations are small squares here, and they are widely spaced.

Any method of locating unknown buried faults is crucial to earthquake hazard assessment in the Los Angeles area. Over the past 50 years almost all of the damaging earthquakes in southern California have occurred on previously unknown faults, including the 1987 Whittier Narrows and 1994 Northridge events on blind thrusts, that did not break the surface. The Northridge earthquake was the most economically costly natural disaster in U.S. history, with fifty thousand million dollars in property damage. Fortunately good building practice and some luck held the number of deaths under seventy.

The Northridge quake was not particularly large at 6.7 magnitude, but it hit directly below a very heavily populated area. Geologists in L.A. who examine the fault-bend-folding of sediments along the northern margin of the deep L.A. basin are concerned that a master detachment driving all the thrusts could rupture in a catastrophic magnitude seven-plus event directly below the urban center. The areas possibly exposed to severe ground shaking in this map are home to more than ten million people as well as nearly a trillion dollars of economic infrastructure.

This master detachment, known as the Elysian Park thrust, is controversial beacuse it is not exposed, and is inferred only from seismic stratigraphy and well data geometrically projected in balanced cross sections to ten times the data depth. The total width of the Elysian Park thrust in the upper 15 kilometers of seismogenic crust is not known and varies by a factor of ten among geologists, so we don't even know the maximum magnitude it is capable of to with 1.5 units. It is also unknown what the orientations and connections may be between the Elysian Park and its splays of blind thrusts that have produced so many events. We do not even have a count of the thrusts in the L.A. basin, hamstringing probabilistic seismic hazard assessments.

I will attempt to image crustal fault structure on two cross sections. My data source for each attempt will be small earthquakes recorded on the mostly high-gain, vertical-component, analog stations of the Southern California Seismic Network (SCSN), operated by Caltech and the U.S. Geological Survey. About half of the stations I use would appear on this map; each event I use is recorded on up to about 150 stations, at distances up to 150 km.

This map shows the 1991 magnitude 5 Sierra Madre earthquake and its aftershocks, along with other events in 1991. My image will be along the 50 km north-south line above, which you can see is close to and intersects the path of the 1996 Los Angeles Region Seismic Experiment (LARSE Line 1). LARSE Line 1 was a large reflection-refraction experiment sponsored by the Southern Califonria Earthquake Center (SCEC) to characterize the lithosphere. My Sierra Madre events section is the only one for which there is thus any corroborating seismic data, to test my results against.

Later I will show you another section aligned to the 1994 magnitude 6.7 Northridge earthquake sequence. It is to the west, near the 1994 mainshock, and will employ about 100 Northridge aftershocks, which are not shown above.

The sparse receiver geometry of the SCSN has many deleterious effects on the imaged section I will produce. Here I explore the effects of these limitations my migrating synthetic elastic finite-difference seismograms; using the real distribution of sources, receivers, and the image section that I will use for the Sierra Madre events. The map on top shows the locaitons of the stations and the image section, in a southwestward view.

In creating the synthetic seismograms I used a simple flat-layered model designed to show differences in imaging of the different component elastic properties of reflectors. The synthetic model has four interfaces; from the top: a delta-mu, S-velocity only interface; a delta-lambda, P-velocity only interface; a delta-rho, density only interface, and deepest the Moho as an interface having changes in all the properties. The dotted lines on the sections show the delths of these horizontal interfaces.

You can see that the delta-mu and delta-rho interfaces are less well imaged than the delta-lambda and Moho interfaces, despite differences in their crustal depth. Arcing migration artifacts cut through the whole section, forming the near-vertical features. Incomplete reconstruction as well as interrupting artifacts lead to discontinuous reconstructions even of the strong interfaces, especially in the area of the sources at the left.

The image is the result here of reconstructing both forward- and back-scattered reflections. Our result shows the very best reconstruction we can expect, given our reflection ray coverage.

Here is the migration of the real Sierra Madre events' data to the same cross section. I used 33 events, the M=5.4 Sierra Madre earthquake and 32 of its aftershocks, recorded on between a dozen and 110 stations each. When I talk about our more conventional imaging in Death Valley I will explain some of the tests used to separate real structure form artifacts we applied to this image.

I only want to suggest that the existence of the one strongest feature in this section is supportable - the reflector dipping from 15 to 19 km depth under the San Gabriel Mountains set off by the arrows above. The source region is largely by the leftmost arrow; and this stucture is imaged mostly by forward-scattered waves.

The Moho may or may not appear at about 35 km depth; work on LARSE-1 data by Clayton's students at Caltech suggests the Moho deepens by 5 km under the San Gabriel Mountains. Our statistical tests do not favor the true existence of the arc extending down from the surface trace of the San Andreas fault.

This slide overlays our Sierra Madre section across the San Gabriel Mountians (gray) with the result of the LARSE Line 1 explosion reflection experiment (colors). The 1996 stack published by Fuis and many others in EOS is plotted to scale and aligned with our section at Azusa, on the south end.

Clearly, the prominent mid-crustal reflection bright spot discovered by LARSE Line 1 we also found as the most prominent structure on our section. Our match with the location and depth found by LARSE is nearly perfect. The comparison in effect provides two completely independent lines of evidence for the bright spot. The LARSE result is from back-scattered reflections, and ours is from forward-scattered reflections.

The interpretation of the bright spot is problematic, since it does not intersect the surface. It may be the base of a regional Mesozoic subduction complex known in various southern California locations as the Pelona, Rand, or Orocopia Schist. The reflector may originate in the tectonic boundary with underlying oceanic crust, or it may have a metamorphic origin as the base of the seismogenic crust.

Now that we have seen a section with a definite but possibly non-tectonic structure, we turn our attention to the Northridge section. For this image we migrated traces from about 100 aftershocks, all within the mainshock source area. We used only shallow events, so this section represents mostly back-scattered reflections. Again, I will explain in another context how I identified the true structures, circled in yellow, against the remaining artifacts.

Here too I only want to support one main reflector, the 45-degree north dipping structure that cuts the entire crust from near the Northridge mainshock hypocenter to the Moho. This structure is entirely below the aftershock sequence, and is in the location of the Elysian Park master thrust proposed by L.A. Basin geologists such as Davis and Namson. Our image proposes, however, a much more thick-skinned model for the master thrust than Namson and Davis's mid-crustal, horizontal detachment. It appears to cut the Moho below the San Andreas fault, displacing other possibly relict structures.

The significantly non-horizontal dip on the Elysian Park thrust does limit its width through the seismogrenic crust to about that of the San Fernando and Northridge events. Thus a horrific event on a detachment more than 100 km wide, with a magnitude above 7, appears unlikely to originate from the Elysian Park system.

Previous Page -- Outline -- Next Page

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