Developing an On-line Editing System

Tom Bayliss joined the ISC in December to start work on an on-line system for editing the Bulletin. Tom recently earned an M.Sc. from Keele University in computational geophysics, a programme that prepares geophysicists to develop professional-quality software. Tom concentrated on seismology in his B.Sc. and M.Sc. programmes and has experience developing graphical user interfaces (GUI). One of his GUIs is used in an EU-funded project to map seismic hazard in Greece.

Development of an on-line editing system is partially funded by the UK Department of Trade and Industry and is expected to take about take two years. The project will be carried out jointly with University of Reading computer scientists who are expert in human-computer interaction (HCI). HCI studies have shown that a computer screen is not the best tool for presenting very large quantities of parametric data; people are able to review printouts of such data far more quickly. In his first year, therefore, Tom will develop a system for ISC seismologists to use printouts and light pens to automatically generate edit files as they work.

Later passes involve review of fewer data, but more trial and error. This type of work can be aided by immediately computing new results, such as arrival time residuals if a hypocentre is fixed at a different depth. To help this part of editing, in his second year Tom plans to develop a program for interactively testing hypotheses about associating readings, identifying phases, and constraining hypocentral parameters.

Fellowship Completed

In December Chen Qi-fu completed his Royal Society / BP-Amoco Fellowship. During his year at ISC, Qi-fu modified the ISC's location program to use alternative travel-time models. He linked the program with functions to compute travel times of the one-dimensional models PREM and ak135. He also linked in travel time functions contributed by Adam Dziewonski and Rob van der Hilst to compute travel times from their tomographic models.

Qi-fu's work should help the ISC to compute more accurate locations for the Bulletin, which is planned by the end of 2001. He found that hypocentres computed with either of the tomographic models were better than those from any 1-D model, but that ISC location accuracy would be degraded if regional and teleseismic data were neglected.

Qi-fu's results are summarised in a paper submitted to BSSA; an abstract is posted at http://www.gxjydl.com/doc/analysis/2001p03/

Historical Seismicity

Imperial College Press has published a new book by Nick Ambraseys and Robin Adams, ISC's senior seismologist from 1978 to 1995. Following a summary of methods of consistently estimating hypocentres and magnitudes from historical and seismic records, there is a comprehensive catalogue and detailed descriptions of larger events. Unfortunately, magnitudes determined by other agencies were omitted from the main catalogue; the complete table is available by anonymous ftp from milne.isc.ac.uk in directory pub/adams.

Changing Seismicity

Ray Willemann's AGU poster, "What's Happening to mb?" suggests that some apparent changes in the rate of mb 4 to 6 worldwide seismicity are real. The Harvard CMT catalogue contains few large earthquakes for 1997-1998, so it is likely that aftershocks are few. Based on analysis of inter-event times, Ray concluded that there were sufficiently fewer aftershocks in those years to perceptibly reduce the global seismicity rate. Figures and text from his poster are posted at http://www.gxjydl.com//doc/analysis/2000p07/poster.html.

Analysis of Seismic Events from 1998 Completed

The ISC has finished analysing events from 1998. Bulletin data were posted to the web as each month was completed. The printed Bulletins and Catalogues were available within several weeks of analysing the last data for each issue, while CDs were produced within a month of completing the year.

Big Catalogue for a quiescent year

Even in comparison with its relatively quiet predecessor, 1998 lacked great earthquakes and prolific aftershock sequences that can be especially challenging to sort out. Nevertheless, continuing improvement in collecting data resulted in more than 98% as many events as 1995, which is the ISC's busiest year to date.

With few great earthquakes of late and more small events, the average number of readings associated with each event is falling, from 23 per event in 1995 and 1996, to 21 in 1997, and 18 in 1998. This still exceeds the 16 readings per event of 1993, when the Bulletin included many small Japanese quakes.

Wider use of thresholds

The continued growth in the number of events in the Bulletin results from desirable improvements in monitoring and data exchange. But it also threatens the accuracy of analysis by ISC seismologists, who must review an inexorably larger set of data each year.

Implicit thresholding has taken place for many years, as national agencies exclude small events from contributions to the ISC. An explicit threshold was first used at the ISC when events of 1994 were analysed. From then on, shallow m<3.5 earthquakes reported by the JMA are excluded from the Bulletin.

For earthquakes of 1998, the ISC excluded small events from more contributors, including Bergen, Helsinki, and Jakarta. The full set of criteria is posted at http://www.gxjydl.com/doc/misc/1998thresholds.html Use of thresholds frees the ISC to encourage contribution of complete Bulletins, allowing the Centre to compile a more comprehensive database of global seismicity. At the same time, it helps ISC seismologists to focus on the events where re-analysis has the most benefit.

New data contributions

In recent years ISC has received data of the Philippine Institute of Volcanology and Seismology only through NEIC. PHIVOLCS resumed reporting both station readings and hypocentre solutions directly to the ISC from 1998. The hypocentres were introduced to the Bulletin from September 1998, which immediately improved completeness of the ISC Bulletin in the area and cut the number of new events "found" by the ISC from unassociated readings.

Also in the year, New Delhi and Pretoria replaced their printed contributions with electronic versions, while JMA added fault plane solutions to its electronic contributions. In each case, the ISC had previously keyed only selected data, and is now able to re-publish more.

Catching up continues

Progress continued in correcting the publication delay that accumulated during 1997-1998. Thanks to a fully staffed team of three experienced editors, as well as several changes in processing, it took only 9 months to edit this year's Bulletin. Dmitry Storchak, ISC's senior seismologist, is confident of catching up analysis to the schedule during 2001.

New association algorithm pays off

A new association algorithm was implemented from April that takes more account of the likelihood that a station will report readings for small, distant events. The new algorithm also adds consideration of later phase times and inconsistency between event magnitudes and phase amplitudes.

In the past editors usually needed to manually re-associate readings for 30% or more of the events each month. Since implementing the new algorithm, automatic association has erred on only about two-thirds as many events.