Introduction

The most important long-term mission of the International Seismological Centre (ISC) has always been the production of the global definitive earthquake Bulletin (International Seismological Centre 2020a) which to this day remains in high demand for a wide range of geophysical studies. Selected references to hundreds of scientific articles that acknowledge use of ISC data can be found at the end of each annual ISC Director’s Report available at https://www.isc.ac.uk/docs/dirreport/.

Users of the ISC Bulletin benefit from its long-term coverage, the most comprehensive set of measurements from the majority of permanent seismic stations and networks, and homogeneity of included parameters of earthquakes and other seismic events worldwide.

As current data (hypocentre solutions, magnitudes, focal mechanisms, seismic phase arrival times, amplitudes and periods) from reporting agencies arrived (Willemann and Storchak 2001), they were integrated, re-assessed, reviewed and published to keep the Bulletin approximately 24 months behind real time. With the exception of a few cases, representation of earthquakes and other seismic events in the Bulletin did not change, once published.

To maintain homogeneity of data over the tens of years, the ISC followed a reasonably conservative approach (Adams et al. 1982). Nevertheless, the advances in global earthquake science have pushed their way into the ISC procedures that have changed most dramatically during processing of data years 2009–2011. With data year 2011, the current earthquake location and magnitude determination methodology (Bondár and Storchak 2011) finally settled, for however many data years to come. This methodology uses all reported seismic phases with a valid ak135 prediction and takes into account the correlated travel-time prediction error structure, which produces more reliable uncertainty estimates and reduces location bias for geometrically unbalanced networks. The need to re-compute the earlier part of the Bulletin (1964–2010) using the same current procedures was identified and has become a core task of the project of Rebuild of the ISC Bulletin.

Procedures

The first article on this project (Storchak et al. 2017) covered the results of the rebuild of the data period from 1964 to 1979, which was released to users in early 2018. That article described the history of operational changes in the production of the ISC Bulletin and details of procedures used during the Rebuild project, which included:

  • Removal of poorly constrained events;

  • Correcting seismic phase names in line with Storchak et al. (2003, 2011, 2013) and time stamp errors;

  • Adding previously unavailable bulletin data;

  • Computation of ISC hypocentres and magnitudes where appropriate, using ak135 velocity model and the location algorithm by Bondár and Storchak (2011);

  • Manual review of parameters of ISC solutions with either considerable departures in epicentre/depth from the old ISC solution or large travel-time residuals at individual reporting stations;

  • Final overall checks.

These procedures largely remained the same throughout the Rebuild project; therefore there is no need to cover them again in this article. Just one clarification we would make here is that the ISC hypocentre solutions were always based on station observations reported to the ISC by various agencies and not taken from waveforms at the ISC. The Rebuild project has not changed anything in this respect.

As to the general procedures that ISC uses, we would like to refer to the Appendix in any January–June issue of the Summary of the ISC Bulletin that contains a detailed explanation of the subject (International Seismological Centre 2020b). We shall now move to results for the data period from 1980 to 2010.

Summary of results for the 1980–2010 period

The ISC Bulletin contains information on seismic events, mostly earthquakes, reported by a large number of agencies around the world that follow their own policies and procedures. Smaller events, usually reported by single agencies, form part of the ISC Bulletin, but they are not reviewed by the ISC analysts and ISC hypocentre solutions are not attempted. For all other events, where solutions or station information is usually available from several agencies, we review the grouping of available information and, where certain criteria are met, we attempt to compute the ISC hypocentre and magnitudes.

In the course of the Rebuild project, we removed 76,938 events featured in the original ISC Bulletin (Fig. 1). These events generally fall into the following categories, listed in the order of frequency of occurrence:

  • Events supported only by data from hydro-acoustic networks (more detailed explanation later in the article): account for approximately 64% of the total number of deleted events; shown in Fig. 1 as a cloud of epicentres in the Pacific and Atlantic Ocean regions;

  • Poorly constrained events: usually reported by a single agency, where events are well outside the network with unacceptably small station azimuthal coverage and no stations at close distances;

  • Events removed by the original reporter: in some instances, following our request for additional data we received the entire network bulletins that contained much more detailed and reviewed information on events of the past, but, due to retrospective analysis, changes in procedures or the area of responsibility, were missing certain events reported in the past; an example is the network in New Zealand, where ~ 1200 events were either removed or relocated with a substantial departure in epicentre position or depth;

  • Events with 1 min (hour, day, month or even a year) errors in origin time; suspicious almost identical pairs of events were identified in the original Bulletin; most of these events occurred during the period of analog instrument recordings where readings were manually taken from seismograms, at times erroneously; for example, there were two suspiciously similar events in the original ISC Bulletin: a magnitude ML2.6 event was reported by KOERI observatory in Turkey (ISK) directly to the ISC, approximately 2 years after event occurrence, with the origin time 11h 10min 55s on February 20, 1992; at the same time, a very similar event with no magnitude estimate was reported to ISC by NEIC/USGS with the origin time 11h 10min 55s on March 20, 1992 and referenced to KOERI; the P&S arrival times at all four reporting stations (IZI, DST, YLV and KCT) in the two events were identical to one-tenth of a second; this discrepancy was caused by an unfortunate mistake in either the preliminary fast report by KOERI to NEIC or by the interpretation of this report at NEIC; this mistake, somewhat typical for the time of analogue instruments, has now been corrected and the event in March has been deleted.

  • So-called “split” events, usually caused by one of the agencies reporting a severely mis-located hypocentre solution that, in turn led to the formation of two separate events in the original ISC Bulletin, where in fact, there was just one physical event; such instances have been identified and rectified.

Fig. 1
figure 1

A map of 76,938 poorly constrained or phantom seismic events (1980–2010) removed from the ISC Bulletin during the Rebuild project

Full size image

At the same time, we added a large number of additional reports based on both permanent and temporary seismic deployments. Figure 2 demonstrates a list of these contributions by 90 agencies with the timeline showing the period that each individual contribution covered. The Appendix below lists these agencies with explanations on the nature of each additional contribution and references to relevant scientific publications. As a result of the integration of these reports into the ISC Bulletin, ~ 489,000 new events have been added (Fig. 3). The majority of them are small events (such as those in Eastern Russia reported by MSUGS, based on the data of the local agencies), although these additional reports also contributed useful additional station measurements of already known moderate to large events.

Fig. 2
figure 2

Timeline of additional bulletin datasets introduced into the ISC Bulletin during the Rebuild project; 90 agency codes are listed in the Appendix; on each line, the first agency code belongs to the reporter of the data and, if different, the second (in brackets)—to its original author

Full size image
Fig. 3
figure 3

A map of 488,570 earthquakes and other seismic events (1980–2010) added to the ISC Bulletin during the Rebuild project

Full size image

With these additional reports, the Bulletin was enriched with ~ 15 million arrival times and amplitude/period readings from ~ 2600 new stations, not present in the original Bulletin. These stations belonged to both permanent and temporary deployments. They helped to improve overall station coverage and thus source location accuracy in the Rebuilt bulletin. They also created alternative source-to-station paths useful in tomographic studies. Figure 4 shows the map of these stations. Some groups of stations reveal temporary station deployments such as those in Europe, Central and Eastern Africa, Tibet and the South Pacific. The map also shows additional stations of regional networks in Europe, Russia and the USA that did not report to the ISC directly.

Fig. 4
figure 4

A map of all seismic stations that contributed arrival time data (1980–2010) to the original ISC Bulletin (in grey) and ~ 2600 stations that were added to the rebuilt ISC Bulletin (in red)

Full size image

All ISC hypocentre solutions have been recomputed using the current location procedures and ak135 velocity model. In some instances, where the data did not warrant an ISC solution, the old ISC solutions were dropped and a local network’s hypocentre was re-assigned as primary hypocentre. Figure 5 shows analysis of departures in ISC epicentre and depth between corresponding matching events in the original and rebuilt ISC Bulletins. It appears that in approximately 80% of cases, both epicentre and depth departures are within 20 km and in approximately 95% of cases the departures are within 40 km.

Fig. 5
figure 5

Cumulative statistics of departures in location and depth of origin between matching events in the original and rebuilt ISC Bulletins (1980–2010)

Full size image

Only 2698 events (~ 0.6%) show an epicentre difference in excess of 120 km. Out of those, just 10 events had an epicentre difference in excess of 1000 km.

The most extreme case was an earthquake that was originally believed to have occurred on July 9, 1983 in North Pacific Ocean with ISC mb 5.5, based on the amplitude/period pairs from 15 stations. This event was not reported by an agency but built at the ISC, based on the un-associated station arrival time reports sent separately by several agencies—a so-called “search” event. Its ISC hypocentre solution in the original Bulletin was based on the readings of 47 stations at recorded distances in the range of 27–101° which had an azimuthal coverage of 143°. The stations included MTN (Australia), TRT (Indonesia), BJI (China) and 44 European stations that were interpreted as direct P-arrivals. In fact, the S–P time difference reported at MTN was inconsistent with the station’s epicentral distance and thus indicated a possible mis-association of this reading to this event, but it was sadly overlooked at the time.

As a result of the current analyst’s review during the Rebuild project, we found in the ISC database a set of other reported readings that fitted along with all original station readings in Europe (if interpreted as PKP arrivals) and BJI. These readings included P and S pairs from NOU and KOU (New Caledonia), and PVC (Vanuatu); direct P-arrivals at BRS, CTA and CTAO (Australia), KRP, MNG, MSZ (New Zealand) and SPA (South Pole, Antarctica). The originally associated readings of MTN and TRT have been disassociated. The whole new event was found to occur Southeast of the Loyalty Islands (a staggering move of 6197 km) with 56 associated stations in the distance range of 3.9–156.2° and the station azimuthal coverage of 192° (Fig. 6). Sadly, there was only one amplitude/period pair at SPA that fell into a legitimate distance range for mb. All amplitude/period pairs from Europe, now treated as PKP amplitudes, were out of the range. Thus, no new ISC mb was included in the Rebuild Bulletin for this event and the original mb 5.5 has been removed. All station codes above are part of the International Seismograph Station Registry (International Seismological Centre 2020c).

Fig. 6
figure 6

The epicentre (denoted with a star) and stations that supported an alleged mbISC 5.5 event in North Pacific Ocean on July 9, 1983 in the original ISC Bulletin (left); following a re-association of all available station data during the Rebuild project, the epicentre was moved by 6197 km to Southeast of Loyalty Islands (right). Pink and orange arcs indicate the primary and secondary station azimuthal gaps

Full size image

Only 3260 events (~ 0.7%) show a depth difference in excess of 120 km. The extreme cases of depth difference were usually attributed to subduction zones, such as Kermadec, Fiji-Tonga, Solomon and Vanuatu where (due to lack of depth constraint) different default depths have been selected in the original and the rebuilt ISC Bulletins. The current system of default depths, used in the rebuilt Bulletin is more consistent because it is based on general statistics of well-located events in each geographical area/cell. If such statistics are insufficient in some regions of low seismicity, then general defaults of 10 or 35 km are used. For details, please see the Appendix in each January–June issue of the Summary of the Bulletin (International Seismological Centre 2020b).

All ISC magnitudes were re-computed in line with new more robust procedures that require at least three station amplitude/period measurements and use alpha-trimmed median to compute a network magnitude and uncertainty. This means that the network magnitude is determined as the median of station magnitudes after discarding the lower and upper 20 percentiles, and its uncertainty is given by the SMAD around the median (Bondár and Storchak 2011). Alpha-trimming does not change the median, but it does affect the magnitude uncertainty by avoiding the outliers.

Notably, many events lost their original mb or MS estimates where amplitude/period pairs were available at less than three stations. Figure 7 shows the timeline of this reduction. With a much larger overall number of stations closer to the end of the period, it was inevitable that a lot smaller proportion of events per year lost their mb magnitudes in recent years as compared to the 1980s and mid-1990s. With MS magnitudes this effect is not obvious, partly because of the large original number of events in recent years, where just one or two automatic surface wave readings were provided by PIDC (during the period 1995–2000) and then by IDC/CTBTO (from 2000–02-26 to present) driven by their important requirement to provide data for mb/MS event type discrimination. Original versus rebuilt numbers of both mb and MS are very similar during 2009 and 2010. This is because current magnitude computation procedures have already been used in the original ISC Bulletin for these 2 years. We included the 2009–2010 period in Rebuild as a precaution, because during this data period certain teething problems have been identified and fixed, and we wanted to re-run the updated procedures systematically over these 2 years.

Fig. 7
figure 7

Annual distribution of the original and rebuilt ISC mb (top) and MS (bottom) magnitudes

Full size image

Figure 8 shows the change in the global magnitude frequency distributions for mb and MS. These graphs also show a considerable drop in the number of mb and MS estimates for moderate-to-small events where less amplitude data are available. By design, only poorly constrained estimates were lost in order to make the ISC magnitudes generally more reliable. We avoid making any conclusions on the apparent drop in completeness of the overall catalogue. Completeness of the ISC Bulletin is highly variable throughout the world and global graphs can often be misleading if used for this purpose. Regional study of completeness is an important topic in its own right and is therefore outside the scope of this article.

Fig. 8
figure 8

ISC mb (top) and MS (bottom) magnitude frequency distributions of the original and rebuilt ISC Bulletins (1980–2010)

Full size image

Having finished with the analysis of the rebuilt ISC Bulletin in two major time periods, released to users separately (1964–1979 and 1980–2010), we now move to an overall analysis of changes in the Rebuilt ISC Bulletin for the entire rebuilt period (1964–2010).

Analysis of the entire rebuilt period (1964–2010)

The ISC Bulletin is a collection of information about unique seismic events, mostly earthquakes. Events come as reviewed (generally larger) or not-reviewed (generally smaller), depending on whether the event parameters and corresponding station recording information have been manually assessed by ISC analysts. Each event is characterised by one or a few hypocentre solutions, one of which is set as the primary (prime). If an ISC hypocentre is available, then it is most likely to be the prime solution and the event is always considered as Reviewed. Each event is accompanied by a number of magnitudes, fault plane solutions, felt reports, comments and seismic phase arrival time information at a variety of seismic stations, supplied by different reporters, including the ISC itself.

As a result of the Rebuild project, the ISC Bulletin within the 1964–2010 period has experienced a large transformation. In particular, a large number of events and hypocentre solutions have been deleted or added. Figure 9 shows this transformation in before/after style, demonstrating:

  • Removal of poorly constrained and, at times, phantom hypocentres (panel row 1) or entire events (panel rows 2 and 3) reported to the ISC (for example, the former artificial arc crossing Africa that consisted of single low aperture array solutions (LAO) or events in the Pacific and Atlantic constrained solely by hydro-acoustic stations);

  • Removal of many ISC hypocentre solutions (but not events) where hypocentre solutions of local agencies were considered to be sufficient to serve as primary and ISC had no further station reports in addition to the station reports of that local agency (for example, for small events in Japan and South Africa, panel row 4);

  • Addition of many previously unreported events, mostly of small-to-moderate magnitudes (for example, in Russian Far East, panel row 2) based on additional bulletin contributions collected during the project, such as those shown in Fig. 2.

Fig. 9
figure 9

Global distribution of earthquakes and other seismic events in the ISC Bulletin during 1964–2010 period before (left) and after the Rebuild project (right); individual rows show all and primary hypocentres in the entire Bulletin as well as in its reviewed part

Full size image

Figure 10 shows regional details of seismic events (1964–2010) deleted from and added to the rebuilt ISC Bulletin per Flinn–Engdahl (FE) regions (Flinn et al. 1974). Only ~ 2.5% of events have been deleted, usually due to being poorly constrained. The large fraction of them were those reported solely by hydro-acoustic networks in the Pacific and Atlantic oceans (FE-regions: Galapagos Area, SE and Antarctic Pacific Ocean, Eastern South America, Atlantic Ocean). Although many of these events were real, their location was poorly constrained. The main mission of the ISC, to present the most accurate account of world seismicity, dictated these events be removed. The number of events has increased by ~ 14.2% with the addition of many previously described reports from temporary and permanent networks. The most prominent was the addition of a large number of small events in Siberia and Far East (Mackey et al. 2010).

Fig. 10
figure 10

Regional distribution (FE—Flinn and Engdahl, 1974; Young et al., 1996) of seismic events deleted and added to the ISC Bulletin (left) and the FE-distribution of the ISC hypocentres in the original and rebuilt ISC Bulletins (right)

Full size image

Figure 10 also shows the change in the overall number of ISC hypocentres in each FE-region. Before data year 1999, the ISC aimed to compute its hypocentre for each event in the Bulletin. This practice has now changed to give prominence to local network hypocentres, where additional station data (if any) available to ISC do not present a reasonable chance of location improvement. This is why we deleted many ISC hypocentres for small events in Europe, Mediterranean, Japan and Chile. We obviously kept these events in the Bulletin and assigned the status of prime solution to hypocentres reported by established local agencies.

Figure 11 shows the magnitude frequency distribution plots for ISC mb and MS during the 1964–2010 period. As explained before, many unreliable ISC mb estimates have been removed. The MS curve experienced a greater transformation for both large and small magnitudes because, originally, no ISC MS magnitudes were available before 1978 and, on the other hand, there were many MS estimates in the past, based on just one or two station measurements.

Fig. 11
figure 11

ISC mb (top) and MS (bottom) magnitude frequency distribution in the original ISC and rebuilt ISC Bulletins (1964–2010)

Full size image

Because the ISC magnitudes have essentially been recomputed, it is important to state the overall difference between the original (old) and rebuilt (new) magnitudes. Figure 12 shows the density-distribution plot for ISC mb for those events that had a magnitude in both the original and rebuilt Bulletins. During the period 1964–2010, the mean difference between the old and new mb is −0.05 ± 0.11, i.e. on average, the rebuilt mb is 0.05 units of magnitude larger than the original mb. There can be many contributing factors to this small change. One of them is the change in the influence of large outliers (sometimes in excess of  ± 1.0 magnitude unit with respect to the mean): in the past, the ISC mb was calculated as a mean value of all station magnitudes, whilst now it is based on the median value. Figure 13 shows similar information for ISC MS for the shorter period from 1978 to 2010. This is because the IASPEI practice of MS computation, so-called “Prague formula” (Vanĕk et al. 1962) has been introduced at the ISC only from data year 1978. In fact, the rebuilt Bulletin now contains many newly added MS estimates for the period from 1964 to 1977 and beyond (Storchak et al. 2017). During the period 1978–2010, the mean difference between the original and rebuilt MS is −0.01 ± 0.11, i.e. on average, the rebuilt MS is 0.01 units of magnitude larger than the original MS.

Fig. 12
figure 12

Comparative analysis of ISC mb (1964–2010) for matching 391,570 events in the original and rebuilt ISC Bulletins

Full size image
Fig. 13
figure 13

Comparative analysis of ISC MS (1978–2010) for matching 69,231 events in the original and rebuilt ISC Bulletins

Full size image

Figure 14 shows the timeline of improvement in the number of seismic arrivals associated to events in the ISC Bulletin. The year 1999 appears most “fruitful”, where the number of arrivals has increased by ~ 100%. The latter is a result of several additional reports received at the ISC, with the largest single new contributor in that year being the meteorological bureau in Taipei (CWB) that reported to the ISC the entire catalogue of the 1999 MW7.6 Chi-Chi earthquake sequence. In the past, CWB reported seismic arrivals just for those earthquakes that were felt. We have now included into the rebuilt ISC Bulletin the entire CWB catalogue complete with station arrivals from data year 1991 to present. Two other large additions for data year 1999 were some station arrival time data from Japan’s JMA and Russia’s GS RAS that appear to have been accidentally left out of the original Bulletin due to teething problems whilst introducing the relational database into ISC operations. Data year 1999 was in fact the first data year produced using a database. Binary electronic files (so-called working tape files) were used in the past to process and store the ISC Bulletin information.

Fig. 14
figure 14

The annual numbers of seismic station arrival times: original (light grey bars) and those added to the ISC Bulletin (dark grey bars) during the period 1964–2010; the total number of stations in the original (light grey triangles) and rebuilt (dark grey triangles) ISC Bulletins

Full size image

The annual station numbers (also in Fig. 14) show an improvement in the rebuilt ISC Bulletin as compared to the original Bulletin for every single data year. An apparent drop in the number of arrivals in the original Bulletin during the period 1997–1998 appears to be “compensated” by the additional arrivals added to the rebuilt bulletin. This compensation though is not quite equivalent, as a general loss of reports from ~ 500 stations worldwide in the original Bulletin merely coincided with the addition of station reports to the rebuilt Bulletin in entirely different areas, often as part of temporary deployments (Fig. 15).

Fig. 15
figure 15

Background information for Fig. 14: the upper map shows (in red) the stations that reported to the original Bulletin in 1995 but not during 1997–1998; the bottom map shows (in green) the station that were added to the rebuilt ISC Bulletin during 1997–1998. Regions of lost and gained stations do not coincide

Full size image

Overall, data from 2742 new stations (an increase of ~ 21%) have been added into the rebuilt ISC Bulletin during the period 1964–2010.

Finally, Table 1 summarises the changes in overall numbers in the rebuilt ISC Bulletin for the period 1964–2010. The overall numbers of events, seismic arrivals and stations have increased. The numbers of ISC hypocentres and magnitudes have gone down, whilst many local agency hypocentre solutions have taken their place as primary.

Table 1 Comparative analysis of rebuild versus original ISC Bulletin (1964–2010)
Full size table

Conclusions

This has been a unique project. Never in its entire ~ 55-year history has the ISC been able to critically review, update and extend its already published Bulletin. The work to obtain additional funding, design criteria and procedures, plan the details, run automatic procedures and manually review the results has taken more than 11 years and a large fraction of the ISC staff time.

The ISC Bulletin available from the ISC website today has been completely rebuilt for the 1964–2010 period (Fig. 16). As a result:

  • A considerable number of poorly constrained or phantom events has been removed from the ISC Bulletin;

  • All ISC hypocentre solutions are now based on the ak135 velocity model (Kennet et al. 1995) where Jeffreys-Bullen travel-time tables (Jeffreys and Bullen 1935, 1940) were previously used for decades;

  • All ISC hypocentres and magnitudes currently available to users are now mutually consistent (1964-present) because they have been computed using the same technique currently used in ISC operations; the ISC hypocentres are less biased by the uneven distribution of stations in source-to-station azimuth;

  • The variety of general default depths, used in the original ISC Bulletin have been modified to those region-specific default depths described by Bondár and Storchak (2011);

  • The rebuilt ISC mb and MS magnitudes and their uncertainties are more robust;

  • A considerable number of ISC magnitudes based on only one or two stations have been removed from the Bulletin;

  • MS magnitude estimates are now available for the period 1964–1977;

  • We noted small changes in average mb and MS that should be taken into account in studies of seismic hazard;

  • A considerable amount of previously unavailable data from 90 permanent and temporary seismic deployments have been added to the Bulletin;

  • We critically reviewed relocated hypocentres and recomputed magnitudes of seismic events where they deviated to a considerable degree from the original (historical) reviewed ISC solutions or where no good fit to reported station data was detected; all other ISC solutions remained as determined by automatic procedures.

Fig. 16
figure 16

An update on the overall status of the ISC Bulletin Rebuild project (Storchak et al. 2017) which shows the data available to ISC users in the past, today and in the future. Jeffreys-Bullen—seismic travel-time tables by Jeffreys and Bullen (1935, 1940); ak135—1D velocity model by Kennett et al. (1995); Reviser and ISCloc-I are the operational earthquake location programs used at the ISC before the most recent program ISCloc-II (Bondár and Storchak, 2011) was introduced into operations

Full size image

Future work

This was the first, but not the last, rebuild of the ISC Bulletin. Similar projects will no doubt follow in the future as better location techniques, more adequate global velocity models and better data become available.

The improvement in the ISC Bulletin will soon be translated into other ISC products, such as the IASPEI Reference Event List (GT), ISC-EHB dataset, ISC-GEM catalogue and ISC Event Bibliography (International Seismological Centre 2020d, e, f, g).

As the next step (Fig. 16), we now aim to extend the ISC Bulletin into the early instrumental period (1904–1963), based on the existing data in the ISC database and data from BAAS, ISS, BCIS and other sources, collected and digitised during the work on the ISC-GEM catalogue (Di Giacomo et al. 2015, 2018).

Remarks

We admire the work accomplished by generations of ISC seismologists and data-reporting agencies since 1964, when patterns of seismicity were not known as accurately as they are today. Most of the work, accomplished recently, was in fact almost impossible to accomplish even 20 years ago, before powerful computers, database software and modern ways of exhibiting the catalogue data graphically and interactively became available.