Personnel: R. Archuleta*, F. Bonilla, A. Byers, J. Cadkin, M. Campbell,
P. Dioquino, G. Ely, G. Gurrola, D. Harris, J. Hartsfield, T. Hayes, A.
Hoffman, E. Keller, J. Lees, G. Lindley, B. Locke, R. Lucas, A. Martin, J.
McKenna, D. Oglesby, K. Olsen, R. Pizzi, M. Robertson, P. Rodgers, B. Seely, J.
Steidl, S. Swain, A. Tumarkin, A. Tumarkina, M. Watkins (*Agenda
coordinator)
The purpose of this project is to incorporate the CUBE (Caltech,
USGS Broadcast of Earthquakes) system into the curricula
of the Santa Barbara City high schools. The effort is to stimulate the
interest of students at each school in the study of earthquakes and the methods
of research. This project will develop a seismology curriculum for a two week
module to be integrated into the Physical Science, Earth Science, or Physics
course of study at the high school level. The project consists of: (1) the
installation of three CUBE systems at the participating high schools, (2) the
training of these participants in its operation, usage, and maintenance, and
(3) the incorporation of the CUBE data into the various curricula at each
school. The participating schools and personnel are as follows: Robert
Pizzi, Bishop Diego High School;Alan Hoffman, San Marcos High
School; Malcolm Campbell, Dos Pueblos High School. The CUBE project has
been established at each of the schools. The curriculum is in its final stages
of completion.
Peter Rodgers and Aaron Martin, along with engineers from other California
institutions, developed a procedure for calibrating electromagnetic sensors.
The method has been used to calibrate large numbers of sensors quickly for
experiments such as LARSE as well as network installations (LABNet) and
portable deployments (LA Microzonation) . Their paper describing the procedure
was recently published in the Bulletin of the Seismological Society of America,
Vol. 85, No. 3, pp 845-850, June 1995. Reprints are available on request from
the PBIC.
The PBIC WWW page underwent some development this past year. The updated page
includes more information about the PBIC as well as equipment status in a
timeline format. This allows users to see where the equipment is utilized and
by which projects.
Southern California undergraduates got some practical field experience during
the SCEC Los Angeles Regional Seismic Experiment (LARSE) 94 project. The PBIC
organized the UCSB contingent of student volunteers for this large cooperative
project aimed at imaging the LA Basin. Student volunteers Jennifer
Cadkin, Priscilla Dioquino, Geoff Ely, Todd Hayes,
Brian Locke, Braden Seely and Robert Lucas deployed and
retrieved seismic stations under the direction of ICS researchers.
The PBIC participated in several outreach presentations this past year.
Presentations were given at several Santa Barbara Elementary schools as well as
at UCSB as part of the Earthquake Fair. Several other presentations involved
use of PBIC equipment by other SCEC personnel.
The Garner Valley downhole seismographic array (GVDSA) project, installed under
the U.S. Nuclear Regulatory Commission contract NRC-04-87-108 in cooperation
with the French Commissariat à l'Energie Atomique (CEA), is designed to
improve our understanding of the effects of a shallow soil column on the
recorded ground motion at the surface of the column. The near-surface
geological site conditions have been shown to be the dominant factor in
controlling the amplitude and variation of strong ground motion, and the damage
patterns that result from large earthquakes. A unique set of data collection
makes it possible to advance two major areas of engineering seismology. First
problem is how weak motions scale to strong motions. The second one is how the
recordings at different soil types scale to each other, especially with respect
to a competent rock ("reference") site. The understanding of competing effects
of amplification and attenuation (including non-linearity) is of a vital
importance for seismic design studies.
We have found that the ground motion of small earthquakes can be scaled to
match that of large earthquakes. However, the necessary scaling implies that
the stress change from the small earthquakes must be multiplied by a factor of
10 to account for the differences in the level of ground motion. If this
difference in stress is borne out by future investigations it has critical
implications about the self-similarity of earthquakes.
The common assumption that the nearby rock site represents the "reference"
motion to a soil site has to be questioned. This assumption does not seem to
hold in the case of Keenwild (KNW), Pinon Flat Observatory (PFO) and Borrego
Valley Downhole Array (BVDA). Spectral ratio estimates of amplification which
use these surface rock sites underestimate the amplification at frequencies
above 2 to 5 Hz. This is because the surface rock sites have a site response
at these frequencies, instead of the assumed flat response behavior associated
with a "reference" site. The rock site response is most likely due to the
weathered and fractured nature of the near surface which causes the velocity to
drop in the near surface. Even sites located on what appears to be competent
crystalline rock show this high frequency amplification when compared to
borehole data at the same site. These results suggest that one must be very
careful in the choice of a "reference" site for site-specific hazard analysis
and that the non-reference site techniques be examined in more detail.
Bedrock borehole recordings of ground motion can provide good reference motion
for soil sites even at distances greater than 20 km from the soil site. When
using the bedrock borehole as a reference the effect of the down-going
wavefield (surface reflection) and the resulting destructive interference must
be considered. This destructive interference may produce pseudo-resonances in
the spectral amplification estimates. If one is careful, the bedrock borehole
ground motion can be considered a good "reference" site for seismic hazard
analysis even at distances as large as 30 km from the soil site.
A two-year development effort culminated in the installation of several
instruments into a 520 meter deep borehole located in the Garner Valley,
California area. The string of instruments was lowered into the borehole and
set in place during the months of April and May, 1995. The new equipment began
operations on May 24, 1995.
The instrument string consists of two accelerometers, located at a depth of
500 meters; two dynamic pressure transducers, located at depths of 335 meters
and 419 meters; and six inflatable packers, which couple the accelerometers to
the borehole wall, and seal off zones in the borehole for pressure
measurements. Stainless steel tubing extends from the surface into the six
zones (defined by the top five of the six packers and the wellhead seal at the
top of the borehole) to monitor the quasi-static pressure and to provide
sampling lines for the water in the different zones.
This experiment is designed to look for correlations between fluid pressure
and seismic activity. The site is located near the Anza segment of the
seismically active San Jacinto fault which is expected to experience a large
earthquake of magnitude 6.5 or greater. This site is the location of the
separately-funded Garner Valley Downhole Seismic Array, which has been in
operation since late 1989.
During the months that followed the 17 January 1994 M6.7 Northridge, California
earthquake, portable digital seismic stations were deployed in the San
Fernando, Los Angeles metropolitan region to recover aftershock data. One of
the goals of this deployment was to examine the seismic hazard in this urban
environment by way of site-specific amplification factors. About 1365
waveforms from 39 aftershocks ranging from M3.0 to M5.1, and depths from 0.2 to
19 km were recorded and analyzed at 35 three-component stations. The
instrument response was removed from all the waveforms. We compared site
response estimates which use the two horizontal components of ground motion as
a complex signal with estimates which use only a single component or the
geometrical mean of the site response estimate from each component. We
compared whole record estimates of the site response with estimates which
window the s-wave and coda-wave portions of the data. We also compared
horizontal coda-wave with vertical coda-wave site response estimates. In
general, the coda method overestimates the site amplification factor with
respect to the factor obtained from the S-waves for all analyzed frequencies.
We found that the vertical coda-wave site response estimates tend to give
larger amplification factors than horizontal coda-wave estimates. We also
found that the vertical and horizontal coda-wave estimates produce larger
amplification factors by 2.0 and 1.6 respectively when compared to the
horizontal S-wave estimates.
In addition to the results presented in France, results were presented at the
1995 annual SCEC meeting, and the 1995 Fall AGU meeting. These results showing
the comparisons between different methods for estimating site-specific seismic
hazard have important implications for researchers attempting to mitigate the
hazard from future earthquakes. ICS researchers will also be providing
engineers with estimates of site response which use the response spectrum
instead of the Fourier spectrum, in order to better the communication between
seismologists and engineering communities. Another promising technique
developed here at ICS by Dr. Alexei Tumarkin estimates the site
amplification in the time domain. This new technique will also prove useful to
engineers who are not just interested in the frequency content of seismic
energy but also the duration of shaking. With a number of important
publications expected in the next few months based on the data analysis and
work done by the PI's and graduate student Fabian Bonilla, we hope to
appropriate additional funding to continue this important task of defining the
seismic hazard in the Los Angeles and surrounding Southern California region.
We updated the existing SCEC/ICS strong-motion database to include response
spectral ordinates and site-specific amplification factors for sites recording
events since the 1933 Long Beach earthquake up to and including the 1994
Northridge quake. Response spectral acceleration, velocity, and displacements
were first calculated for each event at each site in the database. The
attenuation relation of Sadigh et al (1994) was used as a reference model for a
site-specific comparison of the site's peak ground acceleration and specific
response spectral values. Strong-motion data is compared with this attenuation
relationship in order to find site amplification or de-amplification factors.
We also calculated amplification factors using the average rock response
spectral values as a reference. The results of these comparisons will be
available through the strong motion database. The records will consist of four
amplification factors at each site for all available events within the southern
California region. The four amplification factors are at periods of 0.1, 0.3,
1.0, and 3.0 seconds. The sites were chosen because of their structure type
and geographical position. Sites were limited to include only those located in
the southern California region (with a maximum latitude of 36.0 and a minimum
latitude of 32.0) and those that could be classified as free field. These
results will help to produce a more complete understanding of seismic hazards
for southern California and assist in further study of seismic events.
Research conducted by L.D. Gurrola and professor E.A. Keller has
yielded significant results providing a better understanding of the earthquake
hazards in the city of Santa Barbara and surrounding areas. The notable
results include:
The Ellwood marine abrasion platform and terrace surface are folded as a
result of deformation along the More Ranch fault. It is a gentle, asymmetric
fold and both flanks are observed in the Ellwood platform and terrace
surface.
The More Mesa marine abrasion platform and terrace are back-tilted and
probably represent the north flank of a fold in which the south limb and
associated axial surface are offshore.
A shallow, south-dipping reverse fault that juxtaposes Santa Barbara Formation
against Quaternary gravels has been identified, and probably is the south-east
extension of the San Pedro fault. This fault coincides with a linear ridge and
truncates an asymmetric, hanging wall fold. Paleomagnetic analyses of the
deformed sediments have yielded normal remanent magnetization and suggests
along with stratigraphic relationships, that they are mid-Pleistocene or
younger (<790 ka). Therefore, the fault is determined to be a
fault-propagation fold and is potentially active.
An east-west trending structure that exhibits reverse displacement of fluvial
sediments in Maria Ygnacio Creek has been identified.
A potentially active fault-propagation fold and associated anticline hill with
wind and water gaps has been identified in the hills west of downtown Santa
Barbara. The footwall anticline is truncated by the San Jose fault and
associated fault scarps several meters high have been identified in 1928 aerial
photos. Paleomagnetic results yield normal magnetization of the sediments that
are mid-Pleistocene in age.
Fossil deposits have been identified in the More Mesa and La Mesa marine
terrace deposits and will potentially yield solitary corals for dating of these
terraces.
The first year of a three-year project has just been completed to analyze data
from the U. S. National Seismograph Network (NSN), a network of
state-of-the-art seismic instruments that have been installed in the last
several years. The purpose of the project is to study seismic hazard in the
central and eastern United States. A data base of NSN earthquake recordings
has been collected and is stored on-line at the Institute for Crustal Studies.
The data set includes seismograms from 185 earthquakes recorded at 22 stations
in the central and eastern United States.
In order to study issues relevant to seismic hazard, Fourier spectra of the
data have been computed using a Fast Fourier Transform algorithm. The observed
spectra are modeled as a combination of site, path, and source effects, where
the source effect is the effect due to the earthquake source, the path effect
is the effect of the propagation of the seismic waves through the crust of the
earth, and the site effect is the effect on the seismic waves due to
propagation in the near-surface of the earth. The results of this modeling
will enhance our understanding of the ground shaking caused by earthquakes and
will allow us to better predict ground motion from future earthquakes in the
central and eastern United States.
The purpose of this project is to produce a realistic computer model for
earthquake ruptures. Many theoretical models for earthquakes have been
proposed and developed; however, most of these models are based on very
simplified assumptions of the earthquake source. For example, the earthquake
fault is generally assumed to be a flat, planar surface. In reality,
geological observations show fault zones to have geometrical complexities and
non-planar features at many length scales. One of the commonly observed
geometrical complexities is a fault jog or fault step, where two fault traces
at the surface of the earth come to within a few kilometers of each other, but
do not intersect.
This project has modeled the affect of a fault step on the earthquake rupture.
The model includes two faults and includes the dynamic interaction between the
two faults during the earthquake rupture. Where two interacting faults are
present, the ground motion caused by the earthquake is observed to have an
additional pulse of energy corresponding to the earthquake rupture slowing down
when it encounters a fault step, then accelerating as it jumps from one fault
segment to the next. This research shows how geometrical complexities can
control earthquake ruptures and the ground motion that is caused.
ICS personnel assisted LLNL personnel in the formulation of a draft statement
of seismic hazard in the Santa Barbara Channel for presentation to the U.S.
Minerals Management Service. The role of ICS was to contribute towards a
report that was co-authored by LLNL personnel; "Technical Issues Relevant to
Seismic Hazard Analysis of the Eastern Santa Barbara Channel". The draft of a
co-authored report was filed April 27, 1995. Nine ICS scientists participated
in the writing of this document. We contributed sections on local geology,
structure, seismic activity, tsunami and slide hazard, and strong ground
motion. One feature we emphasized was the major unanswered questions regarding
seismic hazard in our region. These include lack of knowledge of the
controlling structures; whether thin or thick-skinned tectonics predominates in
the Channel. Blind faults have been proposed for the Channel but not generally
mapped. The maximum credible earthquake for our area is not constrained.
Another important issue is lack of offshore data for strong ground motion of
the Channel floor.
Aftershocks of the 17 January 1994 Northridge (M 6.7) mainshock recorded by
the Southern California Earthquake Center (SCEC) portable deployment,
TERRAscope and Southern California Seismic Network (SCSN) stations were
retrieved from the SCEC database and selected according to the following
criteria: (1)completeness (east-west, north-south, and vertical components) of
two stations: LA00 and SSAP of the Northridge portable deployment, (2)
proximity to the San Fernando Basin, and (3) quality of the waveforms. The
instrument response was then deconvolved from the data by dividing the Fourier
spectrum of the waveform by a transfer function constructed with pole, zero,
gain information specific to that instrument.
We developed a new program to interactively manipulate a broad range of
geophysical and geological data for exploratory analysis and presentation
[Nicholson and Lees, 1994]. The program is similar to GIS systems that allow
ASCII data-bases stored in memory to be accessed through a user-friendly
graphical interface, but differs in that it allows users to interact with data
in a third dimension. Data can be viewed in either map or cross section, and
at any strike or dip. Xmap8 was primarily designed to handle large sets of
earthquake related data. Color-coded geological and geophysical maps (or cross
sections) can be overlain with earthquake hypocenters, focal mechanisms and
station arrays. Special attention has been put into dynamic plotting of
earthquakes as time sequences, connecting related events derived from different
velocity models or phase data, and plotting earthquakes with a variety of
options related to hypocentral parameters. Several different views of
earthquake focal mechanisms are available including traditional beach-ball
plots, P- and T-axes, or single nodal planes with slip vectors [e.g., Seeber
and Armbruster, 1995], color-coded as a function of rake. Users are allowed to
select individual nodal planes from suites of focal mechanisms, that align with
seismicity trends in space and time, as a means of identifying structural
details of subsurface fault geometry. A contouring package is included for
plotting 2-D surface or subsurface field data in map view, or for projecting
contour slices in cross section. 3-D projection of deviated wells, dipmeter
logs, and well stratigraphy color-coded by lithology, in both map and cross
section, makes visual correlation of many diverse data sets intuitive.
Hard-copy output of graphic displays is all done in PostScript. Examples are
presented involving fluid injection and seismicity at the Coso Geothermal
Field, organizing focal mechanisms from the 1992 Joshua Tree sequence,
delineation of the magma chamber and seismicity at Mt. St. Helens, and relating
structural subsurface features in the Santa Barbara Channel.
The 1992 M6.1 Joshua Tree earthquake occurred about two months prior to the
M7.4 Landers earthquake and was followed by nearly 6,000 M>1 aftershocks
recorded by the permanent regional network and an 11-element portable array
deployed by the Southern California Earthquake Center [Nicholson and Hauksson,
1992]. This sequence defined a complex set of subsurface faults that included
secondary structures that strike either sub-parallel to the Joshua Tree
mainshock rupture or on relatively short, left-lateral cross faults that strike
at high angles to the mainshock plane. Seismicity on this fracture network
ceased in the hours prior to the Landers event and did not resume. Instead, the
Landers mainshock appears to have caused the activation of a new fracture
network located farther west, that intersects the previous Joshua Tree activity
in the area of the Joshua Tree mainshock. Much of this later activity coincides
with a first-order discontinuity in 3-D velocity structure imaged by
tomographic inversion of P-wave arrival times [Lees and Nicholson, 1993]. The
Joshua Tree data thus provide important information on the pattern of
subsurface stress and strain, and how it changed with time, before and after
the Landers mainshock.
The large numbers of earthquakes, the wide variation in focal mechanisms
observed, and the complex pattern of subsurface faults involved, make the
Joshua Tree earthquakes an ideal data set to examine in more detail using the
SCEC portable digital data and a new enhanced, interactive 3-D color graphic
program - Xmap8 [Lees, 1994, 1995]. Comparison of arrival times hand-picked
from the portable data at Scripps (UCSD), ICS (UCSB) and Yale indicates that
~10% of the data are still susceptible to large timing errors. To remove large
systematic errors from the portable phase data, a cluster analysis was
performed. Any individual arrival times that moved an event epicenter beyond
the cluster radius of 0.7 km or increased the RMS error by more than 0.04 s
were removed from the relocation procedure. Relocated hypocenters are
typically deeper and located farther south and west, if both portable and
permanent network data are used. Polarities of the vertical components at the
portable stations were also found to be reversed. Revised focal mechanisms
with 15 or more first-motions were determined for 1,484 Joshua Tree events (23
April to 28 June). Relocated hypocenters and revised focal mechanisms were
then used to assess the pattern of subsurface active faults in the Joshua Tree
area. Faults were identified by alignment of nodal planes and hypocenters in
both space and time using Xmap8.
* Analysis of the SCEC portable digital data indicates that timing problems
can be largely overcome and that the data are extremely useful for increasing
model and structure resolution.
* Xmap8 proved to be an effective analytical tool in helping to organize large
numbers of hypocenters and focal mechanisms (that occurred in a relatively
small geographical area) in to recognizable patterns of subsurface faults.
* The Joshua Tree sequence is largely composed of predominantly strike-slip
events that typically strike either subparallel or at high-angles to the Joshua
Tree mainshock rupture plane.
* The main north-south structure responsible for the Joshua Tree mainshock is
composed of en echelon fault segments that typically strike slightly
west of north.
* Several of these strike-slip fault segments are actually curved when viewed
in cross section; this includes the Joshua Tree mainshock fault plane.
* In addition, second-order faults that dip at moderate-angles exist within
this fracture network; this includes structures with significant normal,
reverse, or oblique components of slip.
* Preliminary results indicate that rotations of local stress fields may have
occurred within the Joshua Tree area as a function of time prior to the M7.4
Landers mainshock.
An unusual effect of the 1992 Petrolia earthquake was that an accelerometer at
Cape Mendocino recorded a high frequency pulse of 1.47g, whereas the nearby
Petrolia station recorded a maximum of only 0.6g. Archuleta (1992) and Oglesby
and Archuleta (1993) have shown that extreme ground acceleration with high
spatial variability can be caused by source effects provided certain conditions
are met. If an accelerometer is located at a point of high symmetry with
respect to a rupture pattern on a dipping fault, seismic radiation can
interfere constructively at the station and greatly amplify ground
acceleration. In this model the geometry constrains the location of the
asperity on the fault plane, but the rupture of the asperity must also have a
degree of symmetry. In the present study, we derive a faulting model that is
consistent with both the high frequency pulse at Cape Mendocino and the overall
distribution of slip. First, we have inverted the strong motion data
(f<1.06 Hz) for the slip and rupture evolution of the earthquake
using the method of Cotton and Campillo (1995). The inversion shows an area of
high slip and high isochron acceleration on the fault where the asperity is
geometrically constrained to lie. We add to the slip model the rupture of a
circular asperity. By this procedure we arrive at a slip model that is
consistent with our low-frequency inversion results and correctly produces the
high-frequency pulse at Cape Mendocino. The results emphasize the hazard
associated with locations above the hanging walls of dipping faults, where the
geometry permits the production of extreme ground acceleration.
The project is in progress and preliminary results include simulations of 3-D
ground motion in the LA Basin from hypothetical M 6.75 constant-slip
ruptures on three faults on the Los Angeles fault system. These simulations
are a part of the SCEC ground motion scenario predictions. The three scenario
simulations are for the Palos Verdes and Elysian Park faults, and the January
17 Northridge event. The ground velocity is computed on a grid of points with
400 m spacing throughout a volume that is 155 km x 134 km x 34 km (11.1 million
grid points) covering the entire greater Los Angeles area. The maximum
frequency is 0.4 Hz. The simulations show significant 3-D basin effects,
including edge-generated waves and prolonged durations above the basin.
Compared to more typical earth models where the velocity varies only in 1-D
(depth), the ground motion is amplified by factors of 2 to 6 throughout the Los
Angeles basin.
Together with Aaron Martin, Michelle Robertson and Mary Hsu of
USC, and Dave Harris of LLNL, we developed a new method by which
electro-magnetic seismometers may be easily calibrated, and with unprecedented
accuracy. This is now the standard method by which all SCEC electromagnetic
seismometers are calibrated. Unlike previous methods, the only information
required from the manufacturer is the mass of the inertial element. The method
involves removing a step of current from the signal coil of the seismometer,
and simultaneously switching the signal coil to a recorder to capture the
response. No calibration coil is required. A theory was developed which
obtains the damped generator constant, resonant frequency, and damping ratio of
the seismometer from the output of a system identifier used to analyze the
response. Only the seismometer mass (from the manufacturer) and the applied
current (measured) need be known for a complete calibration. The coil and
damping resistances are not required. The method was confirmed by comparing
this signal coil method with weight lift and calibration coil calibrations.
For a GS-13 V seismometer, these results were within 1.3% of each other. The
undamped generator constant obtained by the signal coil method matched the
generator constant given by the manufacturer to better than 1%. Calibration of
nine new L-4C components resulted in undamped generator constants all within 3%
of the values given by the manufacturer.
The project objective is to collect seismic data from sites throughout the Los
Angeles metropolitan region for seismic hazard analysis. It has long been
known that each soil type responds differently when subjected to ground motion
from earthquakes. Usually the younger softer soils amplify ground motion
relative to older more competent soils or bedrock. Our goal has been to
instrument different sites in the Los Angeles area to quantitatively measure
this amplification of ground motion and produce a data base of amplification
factors. These amplification factors are then used to help distinguish regions
where the seismic hazard is greatest due to amplification from the surface
geology and sub-surface structure.
In the 94/95 year we received 10 instruments from the PASSCAL instrument center
and deployed these in addition to the 3 SCEC and 2 Caltrans instruments that we
have been operating since March 1993, for a total of 15 portable stations in
the Los Angeles metropolitan region. Undergraduates Mike Watkins,
Robert Lucas, Priscilla Dioquino, and Geoff Ely were
involved in the instrument calibration, the site deployment and maintenance.
The 15 sites were operating from December of 1994 to July of 1995, when the
PASSCAL instruments were returned. Currently we are still maintaining the 3
SCEC and 2 Caltrans stations.
Many of the sites chosen for deployment in this project are co-located with
permanent strong motion stations. In addition to collecting weak motion data
at these sites, we will also be able to compare amplification factors derived
from weak motion data to those derived from strong motion records. An
important question to engineering seismologists is how significant is the
non-linear effect on strong ground motion, and at what amplitude level can we
expect to see it? The Northridge mainshock and aftershock data recorded by
this study, along with the permanent strong motion stations will provide ground
motion data with the amplitude range to address these questions. In addition
to the busy year of field work collecting data in the Los Angeles basin, we
also have studied the data previously collected from the January 1994 M6.7
Northridge aftershock sequence.
New methods of site-specific ground motion prediction in the time and
frequency domains were developed. A large earthquake is simulated as a
composite (linear combination) of observed small earthquakes (subevents)
assuming various functional models of the source time functions (spectra).
Source models incorporate basic scaling relations between source and spectral
parameters. Ground motion predictions are consistent with the entire observed
seismic spectrum from the lowest to the highest frequencies avoiding deficiency
in the vicinity of the target corner frequency. These methods are designed to
use all the available empirical Green's functions (or any subset of
observations) at a site. Thus a prediction is not biased by a single record,
and different seismic wave propagation paths are taken into account. Any
procedure of adding subevents in the time domain requires knowledge (or
determination) of rupture times of subevents. Joyner and Boore [1988]
recognized a major problem with using a uniform distribution of rupture times:
the natural assumption of a constant rupture velocity leads to a significant
underestimation of the main event's spectrum in the vicinity of the target
corner frequency (by producing a local minimum of energy instead of a global
maximum). As the spectral corner frequency acts as a source resonant frequency,
any misfit to the spectral amplitudes near the corner frequency significantly
affects the total energy in the computed time- series. This problem can not be
overcome by allowing for different subevent sizes, but only by imposing a
specific variation of the rupture velocity or of the stress drop. Our
time-series prediction algorithm is based on determination of a specific
distribution of rupture times of subevents. This approach is an extension of
the method proposed by Wennerberg [1990]. The method is completely empirical.
It requires only four input parameters for the simulated large event: i)
seismic moment; ii) size of the rupture area; iii) location of the hypocenter;
and iv) direction of rupture propagation. There are no other free parameters.
We applied this method to predict ground motions in the Los Angeles Basin from
scenario earthquakes on San Andreas and Elysian Park faults.
We study the source complexity of the Northridge mainshock by a dual approach
to inversion for the slip and rupture histories using both theoretical and
empirical Green's functions. One of the most promising ways to incorporate both
the high-frequencies and the site response into the inversion scheme is to
utilize empirical Green's functions (EGF) - recordings of aftershocks. The
problem here is how to correct these recordings to obtain true Green's
functions, i.e. what source time-functions should be deconvolved from the data.
Without doing this deconvolution we should use EGFs only below corner
frequencies of the aftershocks. Another well-known feature of aftershock
distribution is that aftershocks tend to avoid the area of the major slip on
the fault. We propose a way how to use a single aftershock record to improve
quality of inversion with synthetic Green's functions. In this hybrid approach
we first simulate an observed aftershock using a plane layered structure and
then deconvolve synthetic ground motions from the recording at a site. This
transfer function accounts for a possible mismatch between the model and real
propagation and site effects. In performing the inversion we apply this
empirical correction for each site. Having the digital data from the
instruments co-located with permanent CDMG and USGS sites which recorded the
mainshock we are able to perform inversions in different frequency ranges and
compare results from different methods.
SMDB is used by scientists and engineers from 33 institutions in the US,
Canada, France, Germany, Italy, Japan and UK. During the last year we added
Northridge data from USGS, CDMG and USC as well as response spectral ordinates
at 0.3, 1 and 3s. The Web page for SMDB is at:
http://quake.crustal.ucsb.edu/scec
/smdb/smdb.html.
We are now putting data into EGFL. EGFL allows to search through parameters of
seismic records, access unclipped and low-noise records and process them with
SAC, plot selected earthquakes and stations and interactively obtain additional
information from the maps. We have already processed all TERRAscope data and
started working on SCSN data.
Performance of EGFL and SMDB was demonstrated on a stand-alone SUN workstation
during the Annual SCEC Meeting (September 1995, Ojai).
We have used 2-D elastic finite-difference methods to simulate
0-3 Hz P, SV and SH waves for three Northridge aftershocks with epicentral
locations near the northern edge of the San Fernando Valley. We used a
vertical, approximately N-S striking, cross section (Vs > 0.5 km/s) taken
from a geological model assembled by Magistrale and others which includes the
structure of the Los Angeles and San Fernando basins. The simulations show
multiply reflected phases and dispersive surface waves propagating southward,
generated at the northern edges of the San Fernando and Los Angeles basins.
The aftershock simulations mostly underpredict the durations observed, and the
fit between observed and synthetic peak velocities varies significantly.
However, these comparisons may be useful to improve the models of the 2-D basin
structure.