S21A-0251 0800h
Real-Time Prediction of Ground Motion from P-Wave Records
Real-time prediction of ground motion parameters, such as peak values, seismic intensity or spectral response, at a critical site or facility as well as on regional scale, can help to reduce losses by strong earthquakes by the automatic triggering of protective steps seconds prior to the arrival of high amplitude seismic waves. We present a concept for an earthquake early warning system designed for real-time processing of information obtained from the low amplitude P-wave that is recorded at a number of accelerographs installed in the epicentral area. To allow for the adaptation to rupture dynamics, the estimation of ground motion parameters starts as soon as the first station is triggered and is continuously up-dated with ongoing time. Our design is in contrast to other earthquake early warning systems that are based on the explicit knowledge of magnitude and hypocenter location for the application of empirical attenuation laws. The level of the cumulative absolute velocity (CAV) of the P-wave as well as relative values of CAV at different seismic stations give a good indicator for the severity of impending ground shaking. The site-specific prediction of ground motion parameters and interpolation for the generation of regional shake maps is achieved by the application of a three-layer feedforward neural network considering local site effects. The procedure is demonstrated for the Istanbul Earthquake Early Warning System in Turkey.
S21A-0252 0800h
Development of Displacement Monitoring System Using Force Balance Accelerometers
We selected the double integral method to measure the displacement on the deep sea floor, where it is impossible to have fixed referential points, even though we have had difficulties to obtain the amount of displacement from the double integral of an acceleration waveform. For the measurement, we proposed to fix force balance accelerometers on the sea floor and attempt to integrate acceleration waveform data from the sensors twice. To find a reason for difficulties of getting the amount of displacement from the calculation, we assembled a vibrator system using a pulse motor and a laser displacement meter. Using this system, we could compare the results of the double integral of the acceleration waveform data to the date of laser displacement meter. As a result, we successfully propose processes to receive the amount of displacement from the acceleration waveform data.
S21A-0253 0800h
The Virtual Seismologist (VS) method: a Bayesian approach to seismic early warning
The Virtual Seismologist (VS) method is a Bayesian approach to seismic early warning applicable to regions with distributed seismic hazard. It is modeled on ``back of the envelope'' methods of human seismologists for examining waveform data, in particular, in the use of the shapes of the ground motion envelopes and the relative frequency content of the observed ground motions to distinguish between small and large events. What differentiates the VS method from other proposed paradigms for seismic early warning is its capacity to assimilate different types of information that may be useful in arriving at quick and reliable estimates of magnitude and location. In addition to the observed ground motion amplitudes, the VS method uses prior information such as previously observed seismicity, the state of health of the seismic network, and station-specific amplification factors. These types of information are useful in resolving the trade-offs between magnitude and location when such trade-offs cannot be resolved by the limited available observations at the start of the earthquake rupture. We apply the VS method to various earthquake scenarios.
S21A-0254 0800h
Utilization of Station Clusters for Earthquake Early Warning
Earthquakes from three seismically active regions demonstrate a scaling relation between event magnitude and the predominant period of the P-wave, and confirm the potential of using P-wave arrivals to rapidly estimate event magnitude and issue a warning prior to the onset of damaging ground motion. Results show that using the scaling relations to calculate the magnitude produce errors that decrease as more stations record the earthquake, with an average error of 0.5 magnitude units once four stations register the predominant period. This suggests that the use of P-wave detection for earthquake early warning does not require a dense seismic network, and that a small cluster of broadband seismic instruments is sufficient to provide a useful warning of impending ground motion. Our study focuses on both rapid magnitude determination and earthquake location using P-wave arrivals at station clusters that may not encompass the epicenter. We begin by analyzing the capacity of the closest single station to determine the distance and depth of an earthquake, and investigate how the accuracy of these estimates change as more stations record the event. Not surprisingly, there is significant inaccuracy in earthquake location using only one station, but the ability to locate the epicenter, hypocenter, and origin time increases significantly as more stations record the event.
S21A-0255 0800h
Earthquake Early Warning Times Across Northern California
The Elarms methodology for earthquake early warning uses P-arrivals recorded by broadband velocity and strong motion sensors close to the epicenter to both locate the earthquake and estimate its magnitude. Given these parameters, attenuation relations can be used to map the predicted distribution of peak ground shaking across the affected region and traveltime curves provide the onset time. In this study probabilistic distributions of available warning times are calculated for northern California using Elarms. The set of sources used is that provided by the Working Group 2003 report on Earthquake probabilities in the San Francisco Bay Region: 2002-2031 and includes a total of 35 ruptures on the major identified faults each with an estimated probability of occurrence within the next 30 years. Warning times are estimated for each rupture based on a set of assumed operational parameters intended to represent a feasible implementation of Elarms in the region. Although only existing stations are used, it is assumed that telemetry upgrades provide data from the stations to a central processing site within 1 sec. An additional telemetry delay of 1 sec to transmit warning to users is also included. It is required that 4 sec of the P-arrival is available from two stations before a warning is issued; studies in southern California show that the average magnitude error when using two stations is 0.56 magnitude units. ShakeMaps of the 35 scenario events are also used to provide an estimate of the intensity (MMI) of ground shaking at any point. Thus warning time probability distribution functions can be estimated for each intensity of ground shaking. Results show that for any given location there is a much higher probability that a warning can be issued than not. While the intensity of ground shaking is dependent on the distance from the fault, the warning time is dependent on the distance from the epicenter. It is therefore the large lateral extent of strike-slip faults in the region that allow for useful warnings. For example consider the Marina District of San Francisco that suffered severe damage in the Loma Prieta earthquake with an epicenter 100 km southeast. For earthquakes that result in shaking at MMI $>$ IX it is more likely there will more than 20 sec warning than less. These observations for the Marina District are typical of any location across the Bay area.
http://www.geology.wisc.edu/~rallen/ELARMS
S21A-0256 0800h
The Northern California Earthquake Data Center: Seismic and Geophysical Data for Northern California and Beyond
The Northern California Earthquake Data Center (NCEDC) is an archive and distribution center for geophysical data for networks in northern and central California. The NCEDC provides timeseries data from seismic, strain, electro-magnetic, a variety of creep, tilt, and environmental sensors, and continuous and campaign GPS data in raw and RINEX formats. The NCEDC has a wide variety of interfaces for data retrieval. Timeseries data are available via a web interface and standard queued request methods such as NetDC (developed in collaboration with the IRIS DMC and other international data centers), BREQ\_FAST, and EVT\_FAST. Interactive data retrieval methods include STP, developed by the SCEDC, and FISSURES DHI (Data Handling Interface), an object-oriented interface developed by IRIS. The Sandia MATSEIS system is being adapted to use the FISSURES DHI interface to provide an enhanced GUI-based seismic analysis system for MATLAB. Northern California and prototype ANSS worldwide earthquake catalogs are searchable from web interfaces, and supporting phase and amplitude data can be retrieved when available. Future data sets planned for the NCEDC are seismic and strain data from the EarthScope Plate Boundary Observatory (PBO) and SAFOD. The NCEDC is a joint project of the UC Berkeley Seismological Laboratory and USGS Menlo Park.
http://quake.geo.berkeley.edu
S21A-0257 0800h
The Southern California Earthquake Data Center (SCEDC): Update for 2004
In 2004, the first deployment of the USArray transportable array (Big Foot) begins in Southern California. 40 stations of the Southern California Seismic Network (SCSN) will be used as the initial start. The 40 stations will be recorded by USArray as part of their archive, and they will also continue to contribute to the data archive at SCEDC to be part of an archive that extends back more two decades. Additionally there are 120 more broadband sites that are recorded. These data are available directly through the STP interface or through the NETDC interface. Recent SCEDC projects include: Implementation of IRIS products at the SCEDC: SeismiQuery, NetDC and DHI (Data Handling Interface). Improved Station information: The SCEDC and SCSN have cooperated to complete the production of station metadata in the form of dataless SEED volumes for the present configuration of all currently-active SCSN broadband stations. This effort is being expanded to provide complete station history in the SEED volumes distributed by the SCEDC.
http://www.data.scec.org/
S21A-0258 0800h
Renewal of K-NET (National Strong-motion Observation Network of Japan)
The National Research Institute for Earth Science and Disaster Prevention (NIED) operates K-NET (Kyoshin Network), the national strong-motion observation network, which evenly covers the whole of Japan at intervals of 25 km on average. K-NET was constructed after the Hyogoken-Nambu (Kobe) earthquake in January 1995, and began operation in June 1996. Thus, eight years have passed since K-NET started, and large amounts of strong-motion records have been obtained. As technology has progressed and new technologies have become available, NIED has developed a new K-NET with improved functionality. New seismographs have been installed at 443 observatories mainly in southwestern Japan where there is a risk of strong-motion due to the Nankai and Tonankai earthquakes. The new system went into operation in June 2004, although seismographs have still to be replaced in other areas. The new seismograph (K-NET02) consists of a sensor module, a measurement module and a communication module. A UPS, a GPS antenna and a dial-up router are also installed together with a K-NET02. A triaxial accelerometer, FBA-ES-DECK (Kinemetrics Inc.) is built into the sensor module. The measurement module functions as a conventional strong-motion seismograph for high-precision observation. The communication module can perform sophisticated processes, such as calculation of the Japan Meteorological Agency (JMA) seismic intensity, continuous recording of data and near real-time data transmission. It connects to the Data Management Center (DMC) using an ISDN line. In case of a power failure, the measurement module can control the power supply to the router and the communication module to conserve battery power. One of the main features of K-NET02 is a function for processing JMA seismic intensity. K-NET02 functions as a proper seismic intensity meter that complies with the official requirements of JMA, although the old strong-motion seismograph (K-NET95) does not calculate seismic intensity. Another feature is near real-time data transmission. When a K-NET02 detects a strong-motion, it can automatically connect to the DMC in 2 to 5 seconds and then transmits seismic data. Furthermore, the full-scale is improved from 2000 gals to 4000 gals and the dynamic range of AD conversion is more than 132 dB. Strong-motion records of the new K-NET are available at: http://www.kyoshin.bosai.go.jp/
http://www.kyoshin.bosai.go.jp/
S21A-0259 0800h
Prompt Assessment of Global Earthquakes for Response (PAGER): An Automated System to Estimate Impact Following Significant Earthquakes Worldwide
The US Geological Survey's National Earthquake Information Center (USGS/NEIC) is developing a system to rapidly assess societal impact immediately following significant global earthquakes. NEIC's near realtime earthquake solutions are being monitored to automatically identify quakes that likely caused human suffering or damage to infrastructure, or that will attract significant media attention. Our goal is to help the USGS fulfill its mission to provide critical earthquake-related information to emergency response agencies, government agencies, the scientific community, the media, and the general public. Currently, it takes several hours to days for the media and other organizations to provide an assessment of a damaging earthquake. Our system, known as Prompt Assessment of Global Earthquakes for Response (PAGER), will estimate the severity of damage caused by an earthquake immediately following its location and magnitude estimation (minutes to an hour). PAGER will assess the situation based on estimated and any observed ground motions, total population exposed to varying degrees of shaking, and vulnerability of the affected region. We expect that an automated summary impact statement and associated alarms can be deployed within seconds of computing the ground-motion estimates, well before ground truth damage estimates arrive. The USGS is collaborating with the US Agency for International Development (USAID) to develop a prototype system. The prototype will estimate ground motions using modifications to the methodology developed for ShakeMap, extended to the entire globe. Since strong-motion recordings will rarely be available for global earthquakes in realtime, we will rely on predicted rather than observed ground motions. Site corrections will be approximated using a combination of elevation and topographic slope (see Wald et al. this meeting) and the exposed population will be determined using Oak Ridge National Lab's Landscan2002 global population database. PAGER will be an iterative system with new alarms issued as better estimates of magnitude, location, fault orientation, finite fault effects, and felt reports become available. We will present details of the assessment algorithm and examples from the prototype system.
http://gldresponse.cr.usgs.gov/pager/
S21A-0260 0800h
A Method for Estimation of Death Tolls in Disastrous Earthquake
Fatality tolls caused by the disastrous earthquake are the one of the most important items among the earthquake damage and losses. If we can precisely estimate the potential tolls and distribution of fatality in individual districts as soon as the earthquake occurrences, it not only make emergency programs and disaster management more effective but also supply critical information to plan and manage the disaster and the allotments of disaster rescue manpower and medicine resources in a timely manner. In this study, we intend to reach the estimation of death tolls caused by the Chi-Chi earthquake in individual districts based on the Attributive Database of Victims, population data, digital maps and Geographic Information Systems. In general, there were involved many factors including the characteristics of ground motions, geological conditions, types and usage habits of buildings, distribution of population and social-economic situations etc., all are related to the damage and losses induced by the disastrous earthquake. The density of seismic stations in Taiwan is the greatest in the world at present. In the meantime, it is easy to get complete seismic data by earthquake rapid-reporting systems from the Central Weather Bureau: mostly within about a minute or less after the earthquake happened. Therefore, it becomes possible to estimate death tolls caused by the earthquake in Taiwan based on the preliminary information. Firstly, we form the arithmetic mean of the three components of the Peak Ground Acceleration (PGA) to give the PGA Index for each individual seismic station, according to the mainshock data of the Chi-Chi earthquake. To supply the distribution of Iso-seismic Intensity Contours in any districts and resolve the problems for which there are no seismic station within partial districts through the PGA Index and geographical coordinates in individual seismic station, the Kriging Interpolation Method and the GIS software, The population density depends on whether the districts are more urbanized or not. As the present researches are concerned, there were not a good and reliable relationship between the mortality and the characteristics of ground motions. We propose the concept of Equal Population Gaps to resolve the influence of mortality in a rural or urban district and decision of the weighting function to each district. The relationship between PGA Index and the mortality determined in this study can be expressed as:\[{\it M}=28.9/[1+exp{(1.67-0.0029 \times {\it PGA})}] \] Here {\it M} is mortality in %, and {\it PGA} is PGA Index in gals. The corresponding curve matches the data reasonably well, with R$^{2}$=0.91. We process the estimation for districts in different scales to verify the feasibility of the method. The mortality-based on PGA Index is particularly useful in real-time application for death tolls prediction and assessment--a piece of information most critical for post earthquake emergency response operation.
S21A-0261 0800h
A Study on Near-Fault Mortality from the 1999 Chi-Chi, Taiwan Earthquake
A new approach for estimating the relations between mortality and strong shaking from the 1999 Chi-Chi, Taiwan earthquake is introduced. We have finished the database giving the attributes of victims through a compilation of various after-earthquake survey documents. This survey was a comprehensive filed visit confirming exact locations of victims and their possible cause of death. Among the total 2492 victims of the Chi-Chi earthquake, 2039 victims (81.8% of the total) were located by using GPS. Through the attributive database of victims, digital maps and Geographic Information Systems, we can easily map the spatial distribution and the cause of death data of victims with accuracy of the smallest administrative districts in Taiwan. Moreover, a regression analysis gives correlated equations for the mortality as functions of the distance to the Chelungpu fault. We find that the percentage of the mortality {\it M} can be expressed as:\[{\it M}=\frac{(1.626+0.0007\times{\it d})} {(1+0.0125\times{\it d})}\] Here {\it d} is the closest distance to the fault in meter. As expected, the shorter distance to the Chelungpu fault causes the higher mortality. We device three disastrous levels and then suggest the orders and scopes of the earthquake disaster rescue according to the regression curve of the mortality and the closest distance to the fault. The difference in mortality between hanging-wall and footwall areas is also discussed and different regression curves of hanging-wall and footwall areas are suggested. In near-fault regions, the mortality for the residents lived in hanging-wall block (1348) is significantly higher than that in the footwall block (557). The death ratio of hanging-wall vs. footwall block is approximately 2.4:1. Finally, we gathered the data of the peak ground acceleration (PGA) data for the mainshock of Chi-Chi earthquake at the 63 seismic stations on both sides of the causal fault. The regression curve of the PGA and mortality as functions of the closest distance to the fault can be determined. We clearly find that the mortality is nearly zero in areas experiencing a PGA below 220 gals. On the other hand, the mortality increases dramatically from 0.2% up to 2% of the total population when the PGA exceeds 400 gals. This rapid increase at about 400 gal also shows up in the building damage. This is not surprising as earthquake death victims by-and-large are caused by building collapse.
S21A-0262 0800h
ShakeMap Implementation for the Upper Mississippi Embayment
ShakeMap is a tool using basic seismological concepts for the rapid generation of maps of various types of ground motion and shaking intensity following significant earthquakes and is based on both observed and modeled data. ShakeMap assists with emergency operation plans and informs emergency responders, emergency management personnel and engineers of potential damage due to strong shaking. The media and public are informed of strong motion by simple map formats, with downloadable images and files also available. To date, ShakeMap has been used in several regions of the western U.S. We adapted the ShakeMap operational requirements to eastern North America (specifically, the Upper Mississippi Embayment) by customizing the aspects of amplification, attenuation, and instrumental-intensity correlation. Grids with site-specific soil velocity based amplification factors were constructed from data compiled by Bauer (2001), resulting in a four-degree grid centered on the New Madrid Seismic Zone. Soil types were delineated to give velocities (Vs30) for each point on the grid. Equations from Borcherdt's (1994) four-step process to determine amplification factors were used to determine the scaling of peak ground motions at short and mid periods. For attenuation, the multiple weighted relations used in the CEUS portion of the 2002 National Seismic Hazards Maps (Frankel, 2002) were chosen, and we plan to upgrade ShakeMap to allow multiple regressions as soon as possible. In the meantime, Atkinson and Boore's (1995) relation was implemented and scenarios were run to determine its appropriateness. Next, the western U.S. regression relationship between Modified Mercalli Intensity and horizontal peak ground motions was compared to observed intensities for recent central United States. Finally, with this regionalized implementation, scenarios were conducted for the 1811-1812 New Madrid earthquake sequence and the results were compared to the observed isoseismals.
S21A-0263 0800h
Progress Toward Quantifying CISN ShakeMap Uncertainty
We are developing and testing algorithms to quantify uncertainties associated with ShakeMap ground motions through efforts by the California Integrated Seismic Network (CISN) ShakeMap Working Group. There are multiple sources of uncertainty in producing a ShakeMap, including sparse ground motion measurements, approximate representation of fault finiteness and directivity, empirical ground motion predictions, numerical interpolation, and site corrections. Although significant contributors to the uncertainties can be reduced through the release of successively more accurate versions of ShakeMap that contain improved fault representation and newly recovered data, it is imperative to provide a clear mechanism to convey the estimates of uncertainty for each successive map. To this end, we quantify the uncertainties of the maps on a point-by-point basis, by combining the separate, but related, contributions of uncertainty using the following sequence. For a small (point source) event, uncertainty at any point on the map away from a station is controlled by the (inter- and intra-event) variability associated with the ground motion attenuation estimates. If there are stations close by, we further reduce the total sigma using a spatial variance equation as a function of inter-station spacing to account for spatial correlations. For extended sources, and for sites in the near-fault region, additional complexity is required. The total sigma is adjusted upward for an event without defined fault dimensions due to the range of ground motions possible given the range of potential distances a given site can be from the source (ranging from epicentral to the closest station-to-fault measure). At several source dimensions, these distance measures converge and this additional uncertainty is small; likewise when the source dimension is specified (source modeling, aftershocks, etc.), this contributor to the total sigma is removed. Again, any nearby stations significantly reduce the total sigma locally, and as expected, the potential uncertainty associated with sites in the near-fault region can be huge when stations are sparse. From this exercise we are converting ShakeMap ground motion predictive estimates to a median distance measure when the source dimensions are unconstrained; the associated median motions then more accurately reflect the range of potential peak motion values there. Finally, we show results of estimates of uncertainty for ShakeMap for both real and scenario earthquakes (weak and strong ground motions) in California and with/without defined fault traces. We discuss future developments and plans for integration of these uncertainty measures, both quantitative and qualitative, into the online system and user interfaces.
S21A-0264 0800h
Earthquake Detection and Location Capabilities of the Advanced National Seismic Network
We have computed minimum earthquake moment magnitude, Mw, detection thresholds for a 1x1 degree grid across the US using the existing backbone stations of the Advanced National Seismic System (ANSS). For every grid point we compute the minimum Mw for which the P phase should be detectable by at least five ANSS stations. Detection is declared at a station when body wave power levels produced for a given Mw are above the frequency dependent 80th percentile noise level for the station. Noise levels were determined in a previous study from probability density functions of noise spectra computed for each ANSS backbone station (McNamara and Buland, 2004). To model event power levels, earthquake moment, Mo, is computed as a function of apparent corner frequency using the source scaling formulas of Brune (1970, 1971). The apparent corner frequency is the frequency at which body wave spectral amplitudes are maximum as a result of attenuation and short period filters applied during NEIC phase picking. The corresponding moment magnitude, Mw, is computed after Kanamori (1977). Body wave amplitudes are then computed for each station depending on the distance and attenuation along each raypath. Amplitude is then converted to power (dB) and compared to station noise levels. The fifth lowest power, above station noise levels then corresponds to the minimum earthquake magnitude for that particular grid point. Our theoretical minimum Mw threshold compares favorably to magnitude thresholds determined from USGS PDE catalogs. We also model the regional variation in event location improvement with the installation of planned ANSS backbone stations. Results from this study are useful for characterizing the performance of existing ANSS broadband stations, for detecting operational problems, and should be relevant to the future siting of ANSS backbone stations. Results from this analysis are also used to optimize the distribution of ANSS regional network stations.
S21A-0265 0800h
ANSS Backbone Station Installation and Site Characterization
During 2004 several new broadband seismic stations have been deployed as a part of the USGS's Advanced National Seismic System (ANSS) backbone and regional networks. New stations include: ERPA, MNTX, OGLA, AMTX, NATX, KCCO, BMO, MARC, TZTN, LAO, DGMT, REDW, KSU1, MOOW, TPAW, LOHW, RAMW. Permanent station locations were chosen to minimize the local noise conditions by recording continuous data and using a quantitative analysis of the statistical distribution of noise power estimates. For each one-hour segment of continuous data, a power spectral density (PSD) is estimated and smoothed in full octave averages at 1/8 octave intervals. Powers for each 1/8 period interval were then accumulated in one dB power bins. A statistical analysis of power bins yields probability density functions (PDFs) as a function of noise power for each of the octave bands at each station and component. Examination of earthquake signal, artifacts related to station operation and episodic cultural noise in the PDFs allow us to estimate both the overall station quality and the level of earth noise at each potential backbone site. The main function of a seismic network, such as the ANSS, is to provide high quality data for earthquake monitoring, source studies, and Earth structure research. The utility of seismic data is greatly increased when noise levels are reduced. A good quantification and understanding of seismic noise is a first step at reducing noise levels in seismic data and improving overall data quality from the ANSS backbone network.
S21A-0266 0800h
Automatic Phase Picker for Local and Teleseismic Events Using Wavelet Transform and Simulated Annealing
Since recent years, various automatic phase pickers based on the wavelet transform have been developed. The main motivation for using wavelet transform is that they are excellent at finding the characteristics of transient signals, they have good time resolution at all periods, and they are easy to program for fast execution. Thus, the time-scale properties and flexibility of the wavelets allow detection of P and S phases in a broad frequency range making their utilization possible in various context. However, the direct application of an automatic picking program in a different context/network than the one for which it has been initially developed is quickly tedious. In fact, independently of the strategy involved in automatic picking algorithms (window average, autoregressive, beamforming, optimization filtering, neuronal network), all developed algorithms use different parameters that depend on the objective of the seismological study, the region and the seismological network. Classically, these parameters are manually defined by trial-error or calibrated learning stage. In order to facilitate this laborious process, we have developed an automated method that provide optimal parameters for the picking programs. The set of parameters can be explored using simulated annealing which is a generic name for a family of optimization algorithms based on the principle of stochastic relaxation. The optimization process amounts to systematically modifying an initial realization so as to decrease the value of the objective function, getting the realization acceptably close to the target statistics. Different formulations of the optimization problem (objective function) are discussed using (1) world seismicity data recorded by the French national seismic monitoring network (ReNass), (2) regional seismicity data recorded in the framework of the Corinth Rift Laboratory (CRL) experiment, (3) induced seismicity data from the gas field of Lacq (Western Pyrenees), and (4) micro-seismicity data from glacier monitoring. The developed method is discussed and tested using our wavelet version of the standard STA-LTA algorithm.
S21A-0267 0800h
Improvement of 3-D Automatic Surface-Wave Tomography Picks Using Windows Defined by the 1-D Refraction-Microtremor Method
We present a method to improve the automatic arrival-time picking of small source-receiver-offset surface waves. We use very narrow bandpass filters and the Hilbert transform to find the instantaneous amplitude at discrete frequencies. We then make the assumption that the maximum instantaneous amplitude of the waveform represents the fundamental mode surface-wave arrival at that frequency. This assumption can be violated in a number of ways, chiefly when a mode other than the fundamental is dominant over a given frequency range. To better recognize subsurface geometries that give rise to these situations, we first gather the data from all source-receiver pairs, sort by offset, and process them using the Refraction-Microtremor (ReMi) method. In this way, we make a 1-D approximation over the entire region sampled by the tomography data. The resulting ReMi slowness-frequency plot allows us to simultaneously identify not only fundamental-mode, but also higher-mode, surface waves. We then define windows representing an acceptable range of slownesses for each mode for each frequency. If the automatic tomography picks fall outside these windows, the value is not used and is flagged for further analysis. This process has been tested on a 330-element seismic array near Albuquerque, New Mexico, and resulted in better agreement when compared to body-wave tomography of the same site.
S21A-0268 0800h
Finite Fault Modeling in Near-Real Time for Tsunami Warning Applications
The Pacific Tsunami Warning Center (PTWC) provides notification to its clients in the Pacific basin of large earthquakes with an assessment of their potential for tsunamigenesis. It is our goal to issue our message products as soon as possible after the detection of a Pacific basin earthquake. Typically, our first official messages are issued within 15 minutes of an earthquake. The Achilles heel of the warning system has been its inability to predict the size of the waves when they arrive on shore. Due to taking a conservative posture, this has resulted in false evacuations. The tsunami modeling community has developed tools that give the tsunami warning system a nascent ability to forecast wave heights. This study is a step towards using a more realistic source, rather than rule of thumb source prescriptions based on Mw, which may be inadequate for complex events resulting in modeling errors in the near and regional fields. We are experimenting with computing slip distributions in near real time in an effort to provide a more realistic source for the purpose of tsunami modeling. The PTWC currently receives broadband data from as many as 100 stations at any given time. Hence we are well situated to do the fault modeling in near real time. The procedure requires a fault plane solution. We use a variant of the method used by Mendoza (SRL, Vol. 67, 19-26, 1996) to perform the inversion. Tests on synthetic earthquakes illustrate the limitations and parameter trade-offs inherent to bodywave inverse solutions. We have so far applied this technique to the Mw 8.4 2001 Peru earthquake and have obtained results that are broadly consistent with those found by previous studies Giovanni et al., ( GRL, Vol. 29, 14-1 - 14-4, 2002), Bilek and Ruff (GRL Vol. 29, 21-1 - 21-4, 2002) and Kikuchi ( wwweic.eri.u-tokyo.ac.jp/EIC/EIC\_News/105E.html ). The resulting source time function as found by other studies is primarily bimodal, with one peak occurring at 15 sec, and a much larger peak at 75 sec. The rupture propagates almost exclusively to the Southeast. The first peak in the source time function corresponds to a section of fault in the immediate vicinity of the epicenter, and the main moment release over a region 130-200km Southeast of the hypocenter. The slip in this region tends to be distributed over more of the vertical extent of the fault. The solutions display some sensitivity to the sub-event source time function, however the moment is usually ~2 \ast 10^{28} ergs.
S21A-0269 0800h
Progress Towards Near-Realtime Seismic Moment Tensors at the Alaska Earthquake Information Center
A near-realtime seismic moment tensor inversion routine has been operational at the Alaska Earthquake Information Center (AEIC) in a test mode for over a year. The AEIC real-time earthquake detection system, based on the Antelope software package, triggers the automatic moment-tensor inversion routine. It is based on a software package developed at the Berkeley Seismological Laboratory and performs a time domain inversion of three-component seismic data for the seismic moment tensor. We use a library of precalculated Green's functions for a suite of regional velocity models and a range of source depths (from 5 to 200 km with 5 km interval) to compute synthetic seismograms. The resulting moment tensor inversion information is distributed via the web. The Alaska seismic network in its current configuration includes 45 broad-band sites. Stable inversion results can be obtained for events with magnitude 4.0 and greater in the network core area (southern and central Alaska) and 4.5 and greater in the rest of the state including the Aleutian Islands. We will present a catalog of nearly 200 regional moment tensor solutions for Alaska and Aleutian Islands starting from October, 2002 through the present including, the 2002 Denali Fault earthquake sequence.
S21A-0270 0800h
Automated Moment Tensor Solution for the Southern California Seismic Network
Automatically generated moment tensor solutions have recently been added to the suite of real-time products produced by the Southern California Seismic Network (SCSN/CISN). The moment magnitude, Mw, and the moment tensor are available within minutes for all regional earthquakes that trigger the network with Ml>4.0, and in special cases for events between Ml 3.5-4.0. The method uses the 1-D Time-Domain INVerse Code (TDMT\_INVC) software package developed by Doug Dreger, which is routinely used in real-time by the UC Berkeley Seismological Laboratory. Green's Functions are determined for various velocity profiles in Southern California, which are used in the inversion of observed three component broadband waveforms (10s-100s) for a number of stations. The duty seismologists will review the automatically generated solution before distribution. A web-interface has been developed to evaluate the quality of the automatic solution, and determine whether it meets the minimum requirements for an immediate distribution. Simple modifications to the stations selected for the inversion are possible, and the inversion can be re-run to optimise the solution. The Mw determined with this method will be the official SCSN/CISN Mw solution for the event. Comparisons of the moment tensors determined using this 1-D model are made with 3-D models generated for larger earthquakes in the Southern California to facilitate calibration of the automated algorithm.
S21A-0271 0800h
Fast Teleseismic Bodywave Inversion Of Intermediate And Deep Earthquakes Source Parameters: Combining Simulated Annealing With Linear Inversion
We develop an inversion algorithm to extract source parameters: Moment Tensor and Source Time Function, from a set of teleseismic bodywave records for intermediate and deep earthquakes. The method proceeds in two stages. First, a non linear inversion is performed (a) to align in time the set of waveforms, (b) to determine a common source time function and (c) to recover a collection of observed station amplitudes. The nonlinear inversion is made with the simulated annealing technique. Those results are then used, in the second stage as \textit{secondary observables} and a standard linear inversion is made to estimate the complete moment tensor as well as the corresponding errors. All of the calculations are extremely simple and, in particular, it is not necessary to compute synthetic seismograms. The present state of the method requires well isolated phases at different stations, which restricts its application to intermediate and deep events. The algorithm is applied to the FDSN broadband records corresponding to the period 1993-2003 of worldwide intermediate and deep seismicity. The moment tensors thus obtained are then systematically compared to the Harvard-CMT solutions.
S21A-0272 0800h
QuakeML - An XML Schema for Seismology
We propose an extensible format-definition for seismic data (QuakeML). Sharing data and seismic information efficiently is one of the most important issues for research and observational seismology in the future. The eXtensible Markup Language (XML) is playing an increasingly important role in the exchange of a variety of data. Due to its extensible definition capabilities, its wide acceptance and the existing large number of utilities and libraries for XML, a structured representation of various types of seismological data should in our opinion be developed by defining a 'QuakeML' standard. Here we present the QuakeML definitions for parameter databases and further efforts, e.g. a central QuakeML catalog database and a web portal for exchanging codes and stylesheets.
http://quakeml.ethz.ch
S21A-0273 0800h
GSAC - Generic Seismic Application Computing
With the success of the IRIS data management center, the use of large data sets in seismological research has become common. Such data sets, and especially the significantly larger data sets expected from EarthScope, present challenges for analysis with existing tools developed over the last 30 years. For much of the community, the primary format for data analysis is the Seismic Analysis Code (SAC) format developed by Lawrence Livermore National Laboratory. Although somewhat restrictive in meta-data storage, the simplicity and stability of the format has established it as an important component of seismological research. Tools for working with SAC files fall into two categories - custom research quality processing codes and shared display - processing tools such as SAC2000, MatSeis,etc., which were developed primarily for the needs of individual seismic research groups. While the current graphics display and platform dependence of SAC2000 may be resolved if the source code is released, the code complexity and the lack of large-data set analysis or even introductory tutorials could preclude code improvements and development of expertise in its use. We believe that there is a place for new, especially open source, tools. The GSAC effort is an approach that focuses on ease of use, computational speed, transportability, rapid addition of new features and openness so that new and advanced students, researchers and instructors can quickly browse and process large data sets. We highlight several approaches toward data processing under this model. gsac - part of the Computer Programs in Seismology 3.30 distribution has much of the functionality of SAC2000 and works on UNIX/LINUX/MacOS-X/Windows (CYGWIN). This is completely programmed in C from scratch, is small, fast, and easy to maintain and extend. It is command line based and is easily included within shell processing scripts. PySAC is a set of Python functions that allow easy access to SAC files and enable efficient manipulation of SAC files under a variety of operating systems. PySAC has proven to be valuable in organizing large data sets. An array processing package includes standard beamforming algorithms and a search based method for inference of slowness vectors. The search results can be visualized using GMT scripts output by the C programs, and the resulting snapshots can be combined into an animation of the time evolution of the 2D slowness field.