"NEAR" REAL-TIME SEISMOLOGY
How often have we heard the term "real-time" applied to some sort of seismic analysis or process? "Real-time epicenter determination", "real-time moment tensor solution", and "real-time earthquake notification" are terms being used routinely to describe current seismic processing. Of course, this is usually a misapplication of the term "real time." According to Webster, real time means "the actual time during which something takes place", and according to the Oxford English Dictionary, it means "the actual time during which a process or event occurs, especially, one analyzed by a computer." In the strictest sense, for earthquake locations and mechanisms "real time" would mean during the actual faulting process. A more liberal interpretation might be analysis during the arrival of seismic waves. Currently seismologists seem to use the term "real time" to mean "automatic" or "pretty soon after." A 1991 report of the National Research Council entitled "Realtime Earthquake Monitoring" included discussions of early warning and rapid response. The former is truly real-time in that analysis and results take place during the arrival of seismic waves. The latter is more in the realm of "automatic" and might more accurately be called "near-real time." Many seismic recording systems now do "near-real time" processing; however, the Electronic Seismologist knows of only a few cases where final results of the processing were usefully available during the shaking.
Two examples of "early warning" systems are worth mentioning. The Urgent Earthquake Detection and Alarm System (UrEDAS) in Japan is currently designed to use the P-wave arrivals at a single station to predict the level of shaking to follow during the S- and surface-wave arrivals to warn critical facilities such as high-speed trains. While this system has been operational at some level for years in Japan, the Electronic Seismologist is not aware of its usefulness in actual large earthquakes (URL http://www.rtri.or.jp/rd/UrEDAS/UrEDAS_E.html [this URL is no longer valid]). During the aftershock sequence of the Loma Prieta earthquake, the USGS used a radio-telemetered seismic signal from the epicentral region to a site in Oakland, where an alarm was sounded to alert rescue workers on a damaged freeway structure several tens of seconds before the strong shaking arrived from an aftershock (URL http://quake.wr.usgs.gov/prepare/factsheets/Mitigation).
While such early warning projects are potential end goals of many seismic processing centers, few centers are seriously attempting such "real time" processing now. However, many do "near-real time" or automatic processing, providing useful information to others within minutes of an earthquake's occurrence. There is a large range of "nearness" to real time with which data and processed results are available. The previous two Electronic Seismologist columns reviewed the "far" end of "near-real time", covering seismic data acquisition over the Internet within many minutes to hours of an earthquake via the AutoDRM and IRIS SPYDERTM systems. This column reviews some of the seismic resources available on the Internet within seconds to minutes of "real time" and includes both raw data and the results of automatic processing.
Automatic, near-real-time processing of seismic data for basic earthquake hypocenter parameters is now being done at several global monitoring centers. The GSETT-3 (Group of Scientific Experts Technical Test 3) International Data Center continuously receives seismic data from approximately forty worldwide stations and automatically processes them in near-real time to detect and locate earthquakes and explosions based on a minimum number of observations. This system detects most events of magnitude 4 or greater over most of the globe. These parameters are passed on to various national data centers within an hour of real time, and an online catalog is periodically updated with these preliminary determinations (URL http://opt2.net/domres.shtml?cdidc.org). The U.S. National Seismograph Network also receives continuous data from U.S. seismograph stations (forty broadband and more than one hundred short-period) and over a dozen IRIS/IDA and Global Telemetered Seismic Network (GTSN) stations and automatically determines hypocenter parameters within tens of minutes of an event's origin time. These preliminary parameters are used to alert various agencies and cooperating groups via pager, e-mail, and fax. Manual review of triggered events is usually completed within one to two hours, and this catalog is made public via the "finger firstname.lastname@example.org" mechanism (URL http://neic.usgs.gov/neis/bulletin/). In both of these cases at least some of the delay between event origin time and hypocenter parameter availability is due to the seismic travel time inherent in global monitoring.
On a regional scale there are a number of seismic processing centers that provide automatic rapid notification and earthquake information within a few minutes of earthquakes in their monitoring region. One of the best known of these is the Southern California Seismic Network operated by Caltech and the USGS, whose "Information on Seismic Activity In A Hurry" (ISAIAH) seismic recording and processing system generates hypocenter parameter information within minutes of event origin time. This system produces notification messages for earthquakes determined to be large enough to be of interest and distributes this information via a variety of automatic means, including e-mail and a pager system called the Caltech-USGS Broadcast of Earthquakes (CUBE). Information about the latest earthquakes is also provided on a Web page (URL http://scec.gps.caltech.edu/cgi-bin/finger?quake). In northern California the seismic recording centers of the USGS in Menlo Park and the University of California at Berkeley combine efforts to produce notifications of significant events within about one minute, followed by updated parameters including moment-tensor solutions within several to a few tens of minutes (URL http://quake.wr.usgs.gov/recent/latest). This joint effort, called the Rapid Earthquake Data Integration (REDI) project, depends on Internet communications between the two institutions to exchange parameter data in near-real time. Other U.S. regional networks with automatic near-real-time notification and information include the University of Nevada at Reno (URL http://www.seismo.unr.edu) and the University of Washington (URL http://www.geophys.washington.edu/recenteqs/). Several additional U.S. regional networks are working on automated near-real-time systems which may be able to provide similar information soon. Also, there are regional networks in Europe and in Japan with varying abilities to produce automatic near-real-time hypocenter information.
While automatically processed seismic information is available within several minutes from many seismic recording centers, seismic waveform data are available, in a few cases, within even "nearer" real time. A nice view of a near-real-time seismogram from the IRIS station TUC is produced every five minutes by the Southern Arizona Seismic Observatory (URL http://saso.geo.arizona.edu/saso/seismograms/current.html). Using the same software the Seismograph Station at UC Berkeley can provide a custom seismogram from any of their fourteen digital stations which ends within only a few seconds of the time the request is placed (URL http://quake.geo.berkeley.edu/bdsn/make_seismogram.html). The Canadian National Seismograph Network provides similar, though much farther behind real time, seismograms in a compressed view showing the past hour for each of their stations on one line (URL http://www.seismo.emr.ca/cgi-bin/hplot_e.html). For those who miss the old helicorder view of the current seismic activity (with all its problems of broken pens and interesting signals on the back side of the drum) a clever presentation is available from the University of Nevada Seismo Lab. They take a digital picture of a real helicorder every few minutes and provide that "helicorder-cam" image on the Web (URL http://www.seismo.unr.edu/Webcam/webcam.html).
All of the above examples are static images of very recent seismic data. For a truly dynamic near-real-time display, the Southern California Seismograph Network provides a Java applet called "SeismoCam" which can be used to connect to a specific seismometer and produces a trace which is automatically updated once a second. Because of delays in the data collection and buffering system at the recording center the displayed data turns out to be from 30-60 seconds behind real time (URL http://www.scecdc.scec.org/waveforms.html). For a very near-real-time display the Pacific Northwest Seismograph Network provides an X-Windows client program which displays the trace of a selected station with no more than a two-second delay. This service requires that the end user be running an X-Windows server and so is not as generally available to users as is the Java applet from Southern California (URL http://www.geophys.washington.edu/SEIS/PNSN/CATALOG_SEARCH/tracedisplay.form.html).
The above seismogram viewing resources can be interesting and do provide ways for the public to view recent seismic data, but they are not particularly useful for research seismologists. Near-real-time data for seismologists consist of digital waveforms or phase data being available continuously for automatic processing. Of course, within any one laboratory or recording center it is assumed that data from one's own stations are immediately available. The interesting new developments are where data are automatically exchanged between recording centers in near enough to real time to be included in automatic processing. Such situations are usually arranged on a one-to-one basis with special procedures and limitations mutually agreed on and thus are not of casual interest to nonparticipants.
There are a number of prototype or experimental rapid data exchanges already in development. An early one was the sharing of automatic phase picks between the USNSN, the Northern California Seismic Network (NCSN), and the Pacific Northwest Seismic Network (PNSN) using Internet e-mail in the ISC telegraphic format (Malone et al., 1993). Real-time pickers in each of the network recording centers send picks to the other centers based on detected event size and location. These picks can then be combined in each center for improved coverage.
A similar near-real-time exchange of picks has been part of the Earthworm project being jointly developed by several groups and coordinated by the USGS in Menlo Park (Johnson et al., 1995). In this case picks generated by the Earthworm system in Menlo Park are exchanged over the Internet with those generated by the ISAIAH system in southern California. In fact, the Earthworm system itself is ideally suited to the near-real-time exchange of seismic data at several different levels, from raw waveforms to phase picks to hypocenter solutions. Its modular design is conducive to an Internet-protocols-based system. The parts of this data collection and processing system could be quite remote from one another.
The simple off-the-shelf components and Internet communications of the Earthworm design are in contrast to the U.S. National Seismograph Network (USNSN), which depends on its own custom satellite telemetry system to retrieve data in near-real time from its own as well as cooperating seismograph stations. The delay of the data from real time in this system is from several tens of seconds to minutes because of buffering the data into compressed packets for efficient transmission. To reduce the need for expensive leased telephone lines from cooperating regional network centers and to obtain data from non-USNSN broadband stations, the USNSN has developed two alternatives. A hardware front-end processor is used to digitize up to sixteen analog data channels into USNSN format and to send these data via the satellite back to the national center in Golden. Alternatively, a software package called a Virtual Data Logger (VDL) has been written to emulate a USNSN field data logger on a SUN SPARCStation. Local time-tagged waveform data arriving at a SPARCStation located at a regional network center are reformatted into USNSN format and sent back to Golden either via a serial connection to the satellite telemetry system or through a virtual circuit over the Internet.
Besides receiving and processing waveform data at the national center, the USNSN satellite system also can provide data back to regional networks, either from their own stations or from others. In this way the USNSN is starting to be used as an important part of the communications infrastructure of a national seismic system. Further integration of the National Seismograph Network with regional networks for both waveform data and processed picked phases is planned using the Earthworm model.
While there is not enough space in this column to include all of the other near-real-time data exchanges the Electronic Seismologist is aware of, there is one final project worth including here because of its home-grown uniqueness. The People's Seismograph Network in northern California not only runs a few of its own seismograph stations, but, by tuning into the radio frequencies of USGS-operated analog stations, they capture these stations as well. All stations are digitized and their data provided in a binary format on an Internet Web server. With a Web browser one can recover near-real-time seismic data as well as programs to display the waveforms on one's own PC (URL http://psn.quake.net/request.html).
Near-real-time data exchange between national, regional, and even local hobbyist seismic networks is bound to increase and improve over time. The Electronic Seismologist envisions a time in the future where data from almost any seismograph in the U.S. could be available in a continuous stream to almost anyone. A back-of-the-envelope calculation with some liberal assumptions regarding numbers of stations and variety of data quality indicates that we could generate about 2 MB/sec of data for the whole national seismic system. This type of load is almost doable on the Internet now for short periods of time. In the future this will be a small fraction of its capability. Now, we had better think about what we might do with all of these data.
Johnson, C. E., A. Bittenbinder, B. Bogaert, L. Dietz, and W. Kohler (1995). EARTHWORM: A flexible approach to seismic network processing, IRIS Newletter XIV(2), 1-4.
Malone, S. D., R. Buland, B. Presgrave, W. Ellsworth, A. Michael, and T. Ahern (1993). Rapid exchange of seismic data between international, national and regional networks using the Internet, Eos 74(16), 216.
SRL encourages guest columnists to contribute to the "Electronic Seismologist." Please contact Steve Malone with your ideas. His e-mail address is email@example.com.
Posted: 5 February 1999