S23E-01
Earthquake early warning across California: Performance of ElarmS on the existing seismic networks
The ElarmS methodology is designed to provide seconds of warning prior to damaging ground shaking in earthquakes. The algorithms are currently being tested as part of a state-wide effort to assess the likely accuracy and timeliness of warnings should the system be fully implemented. This is part of a collaborative study by the California Integrated Seismic Network (CISN) and funded by the USGS. The algorithms have been running in an automated but delayed fashion since February 2006 on all earthquakes with magnitude greater than 3.0 in northern California. The processing of this data shows the sensitivity of system performance to the density of stations in the region. In the greater Bay Area where station density is greatest the system performs very well, providing accurate locations, magnitude estimates and ground shaking predictions within 2 or 3 seconds of event detection. To the north of Santa Rosa and south of Hollister the station density is lower but still sufficient for ElarmS. The lower station density means that a few additional seconds are required to obtain good predictions. The station density in the Mendocino Triple Junction region and the Sierras is not sufficient to provide useful information about earthquakes in these regions. The ElarmS development team has been porting the algorithms to the realtime system in northern California. The system is now running in a test mode 15 seconds behind realtime. We will present the performance of the realtime system and comparisons with the automated system at the meeting. http://www.ElarmS.org
S23E-02 INVITED
Seismic Activity: Public Alert and Warning: Legal Implications
As science and technology evolve in ways that increase our ability to inform the public of potentially destructive seismic activity, there are significant legal issues for consideration. Even though countries and even states within the United States have differing legal tenets that could either change or at least re-shape the outcome of specific legal questions that this session will be pondering, there are fundamental legal principals that will permeate. It is often said that the law lags behind society and in particular its technological developments. No doubt in the area of warning the public of impending destructive forces of nature or society, the law will need to do some catching up. The law is probably adequately developed for at least some preliminary discussion of the key issues. No matter the legal scheme, if there is a failure or perceived failure in the system to warn people of a pending emergencies, albeit an earthquake, tsunami, or other predictable event, those who are harmed or believe they are harmed will seek relief under the law. Every day there are situations wherein the failure to warn or to adequately warn is key, such as with faulty or defective consumer products, escaped prisoners, and police high-speed vehicle chases. With alert and warning systems for disaster, however, we have a unique set of facts. Generally, the systems and their failures occur during emergencies or at least during situations under apparently exigent circumstances when the disaster's predictability is widely recognized as less than 100 percent. The law, in particular United States tort law, has been particularly lenient when people and organizations are operating during compressed timeframes and their actions are generally considered necessary to address circumstances relative to public safety. The legal system has been forgiving when the actor that failed or appeared to fail was government. The courts have liberally applied the principal of sovereign immunity to governmental actions during emergency situations. At a minimum, the courts have shown a high degree of deference and provided immunity protection for discretionary governmental actions. For example, government organizations are often protected from legal redress for making basic policy decisions such as whether or not to implement an early warning system for emergency actions. Some national and state governments, however, have gone further to provide a legal shield of immunity through specific statutory enactments. Statutory protections generally extend to both the governmental organizations and the decision makers therein. In contrast, these protections are not always extended to third parties such as private businesses, which are often part of the chain of people and organizations that are critical for providing emergency notifications to the public. These businesses include the warning devices manufacturers, the communications systems installers, the software developers, and many other non-governmental parties essential to notifying the public. It can be argued that the legal risk in providing these private sector products or services serves to ensure their quality. But these businesses' real or perceived risk of liability could dissuade their participation in the notification system, or at least chill their innovation. Those involved in designing, developing, implementing, and operating emergency notification systems must consider how their unique situation will be impacted and potentially altered by the legal environment, or in some cases how they should affect change to that legal environment in order to have successful warning systems.
S23E-03
Calibrating and Implementing the Virtual Seismologist Approach for Earthquake Early Warning in Switzerland
The Virtual Seismologist (VS) method for earthquake early warning is a Bayesian approach to estimating earthquake source parameters (magnitude, location) and predicting the spatial distribution of peak ground motion in real-time. Among the critical components of the VS method are a suite of relationships describing the magnitude, distance, and site-dependence of P- and S-wave amplitudes in various frequency bands and sensor orientations, as well as relationships between the ground motion ratios (between acceleration and displacement) and magnitude. The VS method was originally developed using a southern California database, and hence, its ground motion models are primarily valid for southern California. However, recent attenuation studies by Stafford et al (2007) have suggested that the differences between ground motion attenuation characteristics in areas characterized by shallow, crustal seismicity are small, and that the ground motion models Next Generation Attenuation (NGA) project are valid in Europe. This would suggest that the southern California ground motion models for the VS method (which are essentially attenuation relationships) would be valid in Switzerland, a region of dispersed seismicity and moderate seismic activity. To test this hypothesis, we derive a new set of VS ground motion models, using records from M&>&2.5 events recorded on the Swiss Digital Seismic Network (SDSN) and the Swiss Strong Motion Network (SSMN), and compare these with the southern California relationships. Implementation efforts of the Virtual Seismologist method in real-time currently under way in California, as part of the CISN Early Warning implementation project, and in Europe, as part of the EU-funded SAFER (Seismic eArly warning For Europe) project. We describe the current status of these implementations, and illustrate the method's performance and limitations on Swiss and California waveforms.
S23E-04
Neural Network Methodology for Earthquake Early Warning - first applications
PreSEIS is a method for earthquake early warning for finite faults (Böse, 2006) that is based on Artificial Neural Networks (ANN's), which are used for the mapping of seismic observations onto likely source parameters, including the moment magnitude and the location of an earthquake. PreSEIS integrates all available information on ground shaking at different sensors in a seismic network and up-dates the estimates of seismic source parameters regularly with proceeding time. PreSEIS has been developed and tested with synthetic waveform data using the example of Istanbul, Turkey (Böse, 2006). We will present first results of the application of PreSEIS to real data from Southern California, recorded at stations from the Southern California Seismic Network. The dataset consists of 69 shallow local earthquakes with moment magnitudes ranging between 1.96 and 7.1. The data come from broadband (20 or 40 Hz) or high broadband (80 or 100 Hz), high gain channels (3-component). The Southern California dataset will allow a comparison of our results to those of the Virtual Seismologist (Cua, 2004). We used the envelopes of the waveforms defined by Cua (2004) as input for the ANN's. The envelopes were obtained by taking the maximum absolute amplitude value of the recorded ground motion time history over a 1-second time window. Due to the fact that not all of the considered stations have recorded each earthquake, the missing records were replaced by synthetic envelopes, calculated by envelope attenuation relationships developed by Cua (2004).
S23E-05
An Algorithm for the Real-Time Recognition of Foreshocks: Application in Greece
Foreshock activity is one of the most important precursory phenomena occurring before the mainshock. The number of foreshocks increases with the inverse of time while the b-value usually drops with respect to the b- value of the background seismicity. Earthquake satatistics in Greece indicates that at least 50% of the mainshocks of M ¡Ý 5.5 are preceded by foreshocks and that foreshock sequences occur as a rule within a time interval no longer than four months before the mainshock. The recognition of the onset of a foreshock activity in near real-time is of importance to issue warning for a forthcoming stronger mainshock. To this aim we developed a computerized procedure for the automatic recognition of foreshock activity. The system consists of an earthquake data basis, an algorithm which performs the statistical tests, and a decision matrix which indicates the level of alert. The data basis is automatically updated from the results of the daily analysis of the national monitoring seismograph system. The algorithm updates calculations regarding seismicity rate and b-value changes and performs statistical tests of change significance. The decision matrix calculates the level of alert by taking into account the level of significance for both the seismicity rate change and the b-value change. Actual application of the system on real-time basis is scheduled to start after the phase of testing which is expected to terminate by the end of 2007. This is a contribution to the EU research project SAFER, contract n. 036935, FP6- 2005-Global-4, Reduction of seismic risks.
S23E-06
Seismic Intensity Magnitude for Earthquake Early Warning
A reliable prediction of the shaking intensity from s-waves is the primary purpose of any earthquake early warning (EEW) system. To date, empirical regression formulae have been used to estimate this intensity from conventional seismic magnitude (e.g.,Mjma), which is the only parameter regarding earthquake size currently broadcast by an EEW. Others have claimed that the magnitude of large earthquakes can be estimated from just a couple of seconds of p-wave data, thus signifying a deterministic earthquake process; results from analysis of data from the Japanese Hi-net, Kik-net and K-NET arrays, however, do not support these claims. Here, we introduce a new parameter, seismic intensity magnitude MI, which is defined directly from observed seismic intensity Ip: MI = Ip/2 + log(r) + a*t + b - c, where r is the hypocentral distance, a = pi*f/(ln(10)*Q), and t, f, and Q are the travel time, predominant frequency and Q for the p-waves, b is a constant and c is a site correction term. The intensity Ip is determined from three- component seismograms of the p-wave in which the filtered signal exceeds a given level for a certain amount of time after the p-arrival. We are able to determine MI in real-time and can therefore estimate the seismic intensity for s-waves (Is) at any site by using a pre-established relation between Is and Ip, and allowing for distance, s- wave travel times and Qs structure. To investigate effectiveness, we determined MI for 127 earthquakes recorded by Hi-net, and estimated the s-wave intensity at Hi-net sites for 11242 cases. A comparison of the observed versus predicted seismic intensity at each site reveals that the estimation error, on average, when using MI is 22% smaller than the error from an empirical method using Mjma. Most important for an EEW system, we find that MI can estimate shaking intensity more rapidly than conventional methods for larger earthquakes, which makes it a very valuable and reliable parameter in an EEW system. http://www.bosai.go.jp/e/index.html