April - June 2001


ISC Seismologists Eliminate Publication Delay
Data Collection Described
Missing Amplitudes
ISC-Related Activities at IASPEI 2001
Progressively Improving Data Management and Processing

ISC Seismologists Eliminate Publication Delay

Working "faster than real time" for more than two years, Dmitry Storchak has guided ISC seismologists in catching up to the target publication schedule while maintaining high standards and accommodating changes to the data management and processing. "I remember former Senior Seismologist Wayne Richardson being rightly concerned even having a week of a delay. We had to eliminate 7 months. I did not realise back in 1998 how long would it take," says Dmitry. "It certainly is a landmark at least during my work in the ISC. It became possible due to the hard work of all ISC team and new developments of the ISC software."

Mamy Andrianirina from Madagascar has been with the ISC since late 1998 and is the longest-serving member of Dmitry's team. "It was a team effort and I acknowledge the hard work and dedication of my colleagues," says Mamy.

Alison Bird, now with the Geological Survey of Canada, and Esmeralda Banganan, on leave from the Philippine Institute of Volcanology and Seismology, also made substantial contributions to the effort. As usual, Melda thinks that there should be a party to acknowledge the success, and perhaps more holidays.

Data Collection Described

"Data Collection at the ISC" by Ray Willemann and Dmitry Storchak has been published in the July/August issue of Seismological Research Letters. It includes information about stations and agencies contributing data for the Bulletin and how ISC manages data, which may interest users who compare data from different years or regions.

Missing Amplitudes

In January-April 1999 printed Bulletins, surface wave amplitude and period are missing for many phase readings. The amplitudes are all available in the on-line Bulletin (from the ISC web site or AutoDRM server) and will be in files on the 1999 Bulletin CD. All amplitudes, including those not printed, were used to compute surface wave magnitudes.

ISC-Related Activities at IASPEI 2001

The Governing Council will meet on Monday 20 August, the first day of the IASPEI Assembly in Hanoi. The provisional agenda has been mailed to Governing Council members and is posted on the ISC web site at http://www.isc.ac.uk/doc/misc/GC_agenda_2001.html. Topics to be discussed this year include analysis thresholds, phases and travel times used to compute hypocentres, and the possibilities for computing ISC magnitudes for more events.

Dmitry Storchak will present a poster titled "The 1998 annual summary of the ISC Bulletin". In his poster, Dmitry plans to describe regional completeness, statistics of data reports, comparisons with other global seismic bulletins, and new events found by the ISC.

Symposium S8.3, "Practices and Standards at Seismic Stations and Data Centres" includes posters by Executive Committee member Akio Yoshida and Governing Council member John Adams. The oral session includes talks on location accuracy by Governing Council member Andrea Morelli, recent visiting fellow at the ISC Chen Qi-fu, and Ray Willemann. There are also talks on how feasible and useful it would be for seismic bulletins to include new types of measurements. A discussion period will feature a panel that includes Governing Council chairman Adam Dziewonski and Executive Committee chairman Jens Havskov, as well as Paul Richards and Bob Engdahl.

Peter Bormann's symposium on the new Manual of Seismic Observatory Practice has papers on standard practices, many of which are appropriate for preparing bulletins contributed to the ISC. The symposium includes a poster by Willemann, Luckett, Presgrave and Havskov on IASPEI Seismic Format, which is the preferred format for contributions to the ISC.

Progressively Improving Data Management and Processing

Since joining the ISC in early 2000, Richard Luckett has written new parsers for dozens of different formats in which data are contributed to the ISC. The new parsers insert data directly into the ISC relational database, without first storing them in intermediate formats. This helps in saving the complete content of the contributions and attributing each datum to a particular report. Working in parallel, Peter Dawson has developed a program to copy data from the database to formats used by existing ISC programs, while James Harris has developed a program to use the output files from ISC processing and analysis to update the ISC database.

Peter Dawson, James Harris and Richard Luckett are working together to create a system for the ISC to manage the ever-growing volume of data flexibly.

With these jobs substantially completed, the ISC development team is focusing on programs to automatically process data in the relational database. The new programs will be able to take advantage of data that cannot be accommodated in the files used by the old programs. Some processing will be done soon after insertion in the database so that the results can be made available as a preliminary bulletin on the Internet.

Standardising Comment Formats

Some types of data previously saved as character strings are stored in the database numerically. These can be formatted for easier use in the CD data files and the on-line Bulletin. Intensities are now formatted consistently regardless of the source, for example, making it easier to compare damage reports from different agencies.

Grouping Hypocentres

Each time a hypocentre is received and inserted in the database a triggered procedure searches for hypocentres already in the database that are close in time and space to the new one. If such hypocentres are found then the new one is grouped with them in the same ISC event. Previously, hypocentre grouping was carried out just before ISC analysis, at least 22 months after the earthquakes occurred.

The ease of repeatedly selecting hypocentres from the database with different time and location criteria has made it feasible to implement a more sophisticated grouping algorithm. The improvement on the once-through algorithm used previously is especially important for closely spaced earthquakes, such as aftershocks, where an initial grouping hypothesis is most likely to be in error. In the first month analysed after hypocentres were grouped in the database, ISC editors manually moved only half as many hypocentres between events as in previous months.

Identifying Duplicate Phases

Duplicate readings recognised in the database are deprecated currently only if they are identical, i.e. the same readings made by the same agency. Such duplicates may be sent to the ISC either by the recording agency or via a third party. Promptly identified duplicates will be hidden from ISC web site users, simplifying the results of their queries.

There are also near-duplicates, readings that differ only slightly, which may represent two agencies' picks from the same time series or one agency's repeated measurements for the same event. Approximate duplicates are recognised and recorded in the database to guarantee that both are associated with the same event, which ensures that ISC editors will spot them.

Currently, the database algorithm does not deprecate near-duplicates, but leaves the editors to decide which to use in the Bulletin. In the future, however, we plan further development that will let the editors to accept or override preferences between near-duplicates.

Automatic Association Based on Reports

Associating phase arrivals with hypocentres is one of the ISC's most important jobs, partly because even users who re-compute hypocentres rarely re-associate arrivals with different events. Association has become more challenging as more agencies report regional and local arrivals along with teleseismic phases.

But comprehensive reporting also helps to address this challenge, since more agencies now report locally-determined associations. For earthquakes from 1999 July, the ISC's processing has been modified to use reported associations, and to seek the best-fitting event only for readings with no reported associations.

As a test, the modified system was used to re-determine associations for readings from a month that had already been analysed at the ISC. Associations were reported for 77% of the 81,000 readings in the 1999 January Bulletin. These reported associations differed from those in the Bulletin only 0.6% of the time, while ISC seismologists have typically changed 3% to 4% of the associations that fit "best" according to the algorithm used at the ISC since last year.

The test also revealed that associations with events in the Bulletin were reported for more than 2000 further readings. It is likely that ISC editors would have confirmed many of these associations if they had been included in initial processing. Thus, by making full use of reported associations, the new software will make the Bulletin more complete as well as quicker to edit.