Author: Eric Bergman

The Salton Trough cluster now contains many more events and covers a wider area.

The Salton cluster is named for the Salton Trough in southeastern
California, a transitional region between sea-floor spreading in the
Gulf of California and the San Andreas transform fault system on land.
The largest event is 5.5 Mw for the July 7, 2010 earthquake. Only events
recorded to at least 10° epicentral distance were retained. Station
coverage for direct calibration is superb since about 2009 but events
prior to about 2007 could not be kept in the cluster because of poor
connectivity with the modern dense network in the area. Because of the
dense network coverage. Free-depth relocation was used to determine the
focal depth for most events (which were held fixed in the final runs),
but a few focal depths were adjusted manually to better fit the
near-source arrival time data. The cluster covers both sides of the
transform fault zone, as well as events in the fault zone itself, and
crustal heterogeneity is very evident in the arrival time dataset.
Because of the immense number of close-in data (~6000 readings for the
hypocentroid) the calibrated epicenters have quite small confidence
ellipses but it is likely that the true uncertainties are greater due to
the strong crustal heterogeneity. All events have depth control from
near-source and local-distance readings, and a few have depth estimates
from teleseismic depth phases which are in good agreement.

The El Mayor cluster now contains many more events and covers a wider area.

The elmayor cluster is named for a mountain named Sierra El Mayor, 60 km
south of Mexicali, Mexico on the Mexico-California border. The cluster
crosses the U.S.-Mexico border, about 30 km into California. The cluster
includes the 7.2 Mw El Mayor – Cucapah earthquake on April 4, 2010, but
no other events are larger than 5.9. All events are observed to at least
10° epicentral distance. The region is heavily instrumented and location
calibration is robust enough for depth control to be done with a free
depth relocation for most events. A few events were unstable in a free
depth relocation and are set manually for the final relocation.

On-line posting of clusters of seismic events with calibrated hypocenters (the so-called GCCEL dataset) began in early 2018. Despite our best attempts to keep things synchronized, some discrepancies have developed between the official GCCEL dataset hosted by the USGS/NEIC on the ScienceBase platform and the dataset posted at this website. Moreover, recent investigations of the details of the calibrated clusters of seismic events have revealed a few errors in the datasets at both sites. I am presently working with Harley Benz at the NEIC to eliminate these discrepancies and errors. If you intend to carry out any research on the GCCEL database it would be wise to get in touch with me to verify that you’re accessing the most reliable version of the dataset. When the process is completed I’ll post a note here.

The Bohol cluster is named for the island of Bohol in the southern
Philippines. It includes three larger earthquakes, 6.7 Mw and 6.6 Mw
(32 minutes apart) on February 8, 1990, and the 7.1 Mw earthquake on
October 15, 2013. Station coverage is quite good and the location
calibration is robust. All events have depth control, mainly from
near-source and local distance readings, but some are constrained by
teleseismic depth phases.

The Herat cluster is named for the ancient city of Herat in northwestern
Afghanistan. It is based on a remarkable sequence of strong earthquakes
in October 2023. Three earthquakes of magnitude 5.9-6.3 occurred within
about an hour on October 7, followed by a 6.3 Mw event on October 11 and
another 6.3 Mw event on October 15. An active aftershock sequence lasted
until the end of 2023. The source region has experienced very little
seismicity in recent decades, other than this sequence. The nearby
seismograph station HRA was, unfortunately, non-operational at the time
of these earthquakes so there is no depth control (or origin time
control) from near-source arrival times. The next nearest seismograph
stations are ~150 km to the west and 600-700 km to the east. Therefore
direct calibration, based on azimuthal coverage by stations at local
distance is not possible. The InSAR signals from this sequence have been
extensively studied by several groups and provide a basis for indirect
calibration. This analysis is based on the work presented by Zhao et al.
, who developed rupture
models for the five largest events. The rupture models derived in that
study help constrain the epicenters and focal depths. Waveform studies
and analyses of teleseismic depth phases also are used to constrain
depth. Origin times are calibrated by performing a preliminary direct
calibration analysis which includes the 6.0 Mw earthquake near Mashad,
Iran on April 5, 2017 and its aftershocks, which were very well recorded
at short epicentral distances. The Herat sequence is ~180 km away from
those stations, however, so unknown variations in the intervening
crustal structure likely bias the epicenters. This is also true for the
origin times, but there is no better alternative. By comparing the
results of the direct and indirect calibration the level of epicentral
bias in the direct calibration solution can be estimated as several
kilometers (to the south).

InSAR modeling suggest shallow depths (<8-10 km) for the ruptures in the Herat sequence mainshocks. In most cases we have found that focal depths estimated from close-in stations lie near the deeper edge of the rupture zones. Waveform depths from multiple studies range from 4-17 km for the 5 largest events, concentrating around 9-10 km. In the case of the Herat mainshocks and many aftershocks, teleseismic depth phases suggest similar depths. Events for which there is no depth constraint were fixed at 9 km. Indirect calibration is based on locating the hypocentroid of the cluster using teleseismic P arrivals between 30° and 90°and shifting the cluster in space and time so that the hypocenter of the 6.3 Mw event on October 15 (the one for which there is the clearest InSAR signal) matches the specified calibration hypocenter. The calibration hypocenter is chosen such that the cluster fits the pattern of rupture models in the manner thought to be most likely, i.e., concentrated in the deeper parts of the individual rupture models. The calibration shift needed to achieve this result is ~15 km at 340°, and -0.4 s in origin time.

The Acapulco cluster is named for the city of Acapulco on the coast of
Guerrero State, Mexico. This series replaces the acapulco5 cluster
previously published in GCCEL. The cluster includes a 7.0 Mw earthquake
on September 8, 2021 and four earthquakes with magniude 6.0 or greater
(6.8 Mw on 4/25/1989, 6.0 Mw on 4/13/2007, 6.2 Mw on 8/21/2013 and 6.5
Mw on 1/2/2026). Earthquakes located more than ~15 km off-shore or
deeper than ~60 km were not included, to avoid location bias. The
distribution of seismograph stations for direct calibration is quite
good. All events have depth control, mainly from near-source and
local-distance readings.

Further work on the Hubbard Glacier cluster involved adding 10 additional aftershocks and resolving an issue with many events wanting to go to zero depth or negative depth.

The Hubbard cluster is named after Hubbard Glacier, north of Yakutat,
Alaska. Nearly all events lie under an icesheet. The cluster is based on
the 7.0 Mw earthquake on December 6, 2025, and it includes 53
aftershocks through January 1, 2026. The cluster spans the U.S.-Canada
border. The local distance arrival time data are best fit with a
relatively high velocity at shallow depths (6.2 and 3.55 km/s) and a
single-layer crustal model with these velocities fits the data well.
Most aftershocks of the December 6 mainshock prefer a shallow depth,
with a peak of ~5 km. The location calibration is very robust, due to
good azimuthal coverage. The arrival time data for the December 6
mainshock are not consistent with a single, simple rupture process. The
initial rupture is reported consistently by most stations within about
600 km but at greater distances, especially teleseismic, many reported
first arrival times are delayed by 2-5 seconds from that initial source,
apparently representing one or more later pulses of energy release. To
avoid location bias, only data within 6.0° epicentral distance were used
to estimate the relative location for this event. At the northwestern
end of the aftershock sequence the epicenters spread orthogonal to the
main trend over ~15 km. All events have depth control from near-source
or local-distance readings and some events have teleseismic depth phase
data in general agreement with those depths.

The Sapporo cluster is named for the city of Sapporo on the island of
Hokkaido, Japan. The cluster covers the southeastern part of the island,
including events to a depth of ~80 km. The cluster lies in the “Hokkaido
Corner” at the intersection of the Japanese and Kurile trench systems.
It includes three major events, the 6.4 Mw Hikada earthquake on January
20, 1970, the 6.7 Mb Urakawa-oki earthquake on March 21, 1982 and the
6.6 Mw Eastern Iburi earthquake on September 5, 2018. The distribution
of seismograph stations is very good and the location calibration is
very robust. All events have depth control from near-source and
local-distance arrival time data and many events also have teleseismic
depth phases that are usually in close agreement. The distribution of
depths is bi-modal with a peaks at 32 and 60 km.

The arrival time dataset of the 2018 Eastern Iburi earthquake is quite
complex and the normal procedure of using data at all distances to
estimate the relative location resulted in significant location bias;
the mainshock epicenter was shifted several tens of km away from its
aftershock cloud. From a detailed analysis of the aftershock sequence,
Katsumata et al. concluded
that the main rupture consisted of reverse faulting on two fault segments
separated by a short step-over on which the rupture initiated with
strike-slip motion and magnitude ~5. There is an unspecified time lag
(likely several seconds) between the initial rupture and the two
subsequent larger subevents. The arrival time data at short epicentral
distances consistently appears to be from this triggering subevent, but
at greater distances the arrival times display significant scatter,
apparently as the arrivals from one or the other of the two larger
reverse faulting events dominate. In the calibrated relocation all
arrival times beyond 8.0° for this event were skipped, so the calibrated
epicenter represents the triggering subevent near the middle of the
overall aftershock pattern.

The version of the Hubbard Glacier cluster uploaded on December 14 (hubbard4.12) contained a biased location the M7.0 mainshock on December 6, 2025. That has been corrected in hubbard5.7.

The Hubbard cluster is named after Hubbard Glacier, north of Yakutat,
Alaska. Nearly all events lie under an icesheet. The cluster is based on
the 7.0 Mw earthquake on December 6, 2025, and it includes 42
aftershocks over the next 4 days. The cluster spans the U.S.-Canada
border. The region contains extreme variations in elevation, from near
sealevel to 3 km above sealevel, and this seems to make difficult the
specification of a crustal velocity model that is broadly satisfactory.
More than the usual number of events prefer extremely shallow focal
depths and such events are especially common among the aftershocks of
the December 6 earthquake. The arrival time dataset has few observations
from very close range so the shallow depths are driven by negative
residuals at distances greater than 50 km, likely at least partially due
to an inadequate crustal model. Most aftershocks of the December 6
mainshock prefer a shallow depth, less than ~5 km. Despite some
uncertainty in focal depths, the location calibration is very robust,
due to good azimuthal coverage by seismic stations. The arrival time
data for the December 6 mainshock are not consistent with a single,
simple rupture process. The initial rupture is reported by most stations
within about 600 km but at greater distances, especially teleseismic,
many reported first arrival times are delayed by 2-5 seconds from that
initial source, apparently representing one or more later pulses of
energy release. To avoid location bias, only data within 6.0° epicentral
distance were used for this event. At the northwestern end of the
aftershock sequence the epicenters spread orthogonal to the main trend
over ~15 km. All events have depth control from near-source or
local-distance readings and some events have teleseismic depth phase
data in general agreement with those depths.

The Hubbard cluster is named after Hubbard Glacier, north of Yakutat,
Alaska. Nearly all events lie under an icesheet. The cluster is based on
the 7.0 Mw earthquake on December 6, 2025, and it includes 42
aftershocks over the next 4 days. The cluster spans the U.S.-Canada
border. The region contains extreme variations in elevation, from near
sealevel to 3 km above sealevel, and this seems to make difficult the
specification of a crustal velocity model that is broadly satisfactory.
More than the usual number of events prefer extremely shallow focal
depths and such events are especially common among the aftershocks of
the December 6 earthquake. The arrival time dataset has few observations
from very close range so the shallow depths are driven by negative
residuals at distances greater than 50 km, likely at least partially due
to an inadequate crustal model. Most aftershocks of the December 6
mainshock prefer a shallow depth, less than ~5 km. Despite some
uncertainty in focal depths, the location calibration is very robust,
due to good azimuthal coverage by seismic stations. The mainshock
epicenter, with a preferred depth of 8 km, consistently lies near the
southeastern end of the SE-NW trending aftershock pattern, but always
5-10 km south of that trend, suggesting some geometrical complexity in
the rupture. At the northwestern end of the aftershock sequence the
epicenters spread orthogonal to the main trend over ~15 km. All events
have depth control from near-source or local-distance readings and some
events have teleseismic depth phase data in general agreement with those
depths.