The following material is excerpted from the USGS Open-File Report (96-69) titled: Global Volcanic Earthquake Swarm Database 1979-1989

The full Open-File Report (less the swarm data) may be downloaded here in PDF format.

Go directly to a list of volcanoes and links to swarm data in the GVESD


Earthquake swarms are pervasive at volcanoes, but have seldom been studied systematically. Most swarms that are described in the literature are those that occurred in association with eruptions; indeed, earthquake swarms are the most reliable method of forecasting eruptions. For the purpose of this report, a swarm is defined as many earthquakes of the same size occurring in a small volume. Swarms are different in these two ways from a mainshock-aftershock sequence or a foreshock-mainshock-aftershock sequence (fig.1). Swarms are especially common in volcanic areas.

Figure 1. The number of events (N) per unit time versus time is schematically shown for the three types of earthquake sequences. The mainshock (MS) indicates the sharp increase in rate for the upper two distributions.

Because swarms are such a common and important phenomenon, we undertook a systematic and comprehensive study of swarms at volcanoes using modern commercially available database software. We term the result the Global Volcanic Earthquake Swarm Database (GVESD). This report describes the database and how it is structured, and gives preliminary results of a study of swarm durations based on 11 years of data as reported in the Bulletin of Volcanic Eruptions (BVE) of the Volcanological Society of Japan.

We chose BVE as our primary data source for several reasons. First, it contains data on many swarms that were never reported in the open literature. This is because most papers report eruptions, and swarms are included only if they were associated with eruptions. In other words, the open literature is biased in favor of eruptions, whereas BVE more fully reports a variety of activity during times of no eruptions. Second, BVE is prepared once per year, which gives investigators the opportunity to summarize data on a broad time scale. Monthly reports, such as the Smithsonian Institution Global Volcanism Network Bulletin, often focus more narrowly on the necessarily short (one month) time scale. Third, BVE is organized by time, so the 1980 issue, for example, contains data on many volcanoes for 1980. This makes it easy to select a sample which includes known large (or small) eruptions, whereas the open literature often has a significant delay between an event and the report. Fourth, BVE data are organized systematically, which greatly aids the preparation of data for entry into the electronic database. Fifth, the BVE includes a section devoted to miscellaneous information. The miscellaneous information section includes many reports of seismic activity at volcanoes that were not in eruption that year. Finally, BVE includes a supplement, which is used to provide additional information on previous years’ activity. This systematic updating provides an additional element of quality control which is not found in most standard reports.

In spite of the generally high quality of BVE data, there are a number of limitations in those data as well as in the very nature of the problem we have chosen for study. In many respects this has been an exercise in the study of messy data. We have been faced with the difficult task of converting the judgments and measurements of many other scientists, which are often reported in words, into numerical data. We have thus been faced with a myriad of decisions and have struggled to maintain consistency and high quality control. A most basic decision is the start time of an earthquake swarm. In most cases this has been reported by a remote observer, but the units vary widely from “10:23 on July 24” to “late July.” In some cases we have had to read data from a graph or histogram, so we have had to decide whether a factor of 2 or 3 increase above background marks the onset of a swarm. The sections below on the structure of the database provide many examples of such decisions.

It was our initial intention to perform a full multi-parametric study of the database. However, it quickly became apparent that durations of earthquake swarms were the most widely reported parameter whereas many other parameters were poorly reported (e.g., energy, cumulative seismic energy release, detection threshold, b-values). Thus we have focused our initial efforts on understanding the distribution of the swarm duration with respect to eruptive activity.

Move on to the [NEXT] page.
Return to the GVESD or to the AVO home page.
This page is designed and maintained by John Benoit. Your comments and suggestion are welcome.