Beginning in 2010, we have posted information here regarding errors in the Real-time Earthquake system that led to erroneous information getting posted on the website or distributed.
The USGS and networks contributing to the Advance National Seismic System (ANSS) take great effort to provide accurate and timely earthquake information. Occasionally our systems produce erroneous information that is released to the public via our web pages or Earthquake Notification System. These mistakes are generally promptly identified by seismologists, removed from our web pages, and “delete” e-mails are sent through ENS. In the interest of rapidly providing earthquake information to the public, most of the information about earthquakes that occur in the USA is automatically posted to the web and ENS if it meets quality standards. There is a trade-off between the speed of our earthquake notifications and number of false alarms in the same way that any kind of "breaking news" story may have substantial changes or corrections as more information is received. The faster we release earthquake locations and magnitudes, the more likely it is that the information may be erroneous. Experience demonstrates that imposing more restrictive quality standards prevents the release of legitimate earthquake information. Here we document serious errors that have resulted in the distribution of flawed information to response organizations and the public. Hopefully this discussion will provide our users with a better understanding of our system.
Commentary for Multiple "Phantom Events" in California - posted June 2, 2015
Automated notification systems are a convenient and often essential component of modern life. The USGS has invested heavily in developing automated systems that provide the public with timely and accurate earthquake information. On rare occasions the Earth throws a curveball and on May 29th and 30th, the USGS issued multiple alerts for false earthquakes in Northern California. The first, a M5.1 near Lewiston, CA, was distributed on Friday. More false alerts were distributed on Saturday, including a M5.5 near Ukiah and M4.7 near San Simeon.
These erroneous earthquake notifications were created by the seismic waves from large, distant earthquakes. On Friday, a M6.7 earthquake occurred at a depth of approximately 60 km, 111 km off of Chirikof Island, Alaska. It was this earthquake that fooled the automatic processing of the Northern California Seismic System to issue the first false alert. Just 28 hours later, a M7.8 earthquake off of Japan with a depth of more than 660 km – the deepest earthquake of its size to have occurred during our history of recording – spawned two more phantom events in Northern California.
Large earthquakes have created challenges for regional seismic monitoring in the past. This problem is particularly acute for deep earthquakes as they generate very impulsive seismic waves which may be misinterpreted as a local earthquake. The USGS and its partners have developed a number of methods to stop or screen these events from being distributed on the Web and through such mechanisms as the Earthquake Notification Service. The USGS will be implementing changes to improve the system and minimize the chances of this occurring in the future.
The erroneous events were deleted quickly by a duty seismologist. Unfortunately, a problem with the distribution software prevented the delete messages from being transmitted to recipients of the Earthquake Notification Service. The inability to transmit the information about the false events to the users of the Earthquake Notification Service caused significant confusion and the USGS regrets the problems caused by this failure. USGS staff have identified the problem in the distribution software and fixed it.
Errata for "Phantom Events" in Central and Northern California resulting from the M7.8 Japan Earthquake on 2015-05-30 11:23:02 UTC
Strong earthquakes generate seismic waves that spread across the entire globe. When the earthquakes are deep, the distant recordings are quite impulsive and are often mistakenly identified by automated systems as local earthquakes. On 2015/05/30 11:23:02 UTC, a 667 km deep M7.8 earthquake occurred offshore of Chichi-shima, Japam and swept across the seismic networks in northern California. The automatic earthquake detection systems recognized the arrival of seismic energy but misinterpreted it as several earthquakes, including a M5.6 event occurring near Brooktrails and M4.8 near San Simeon, rather than one large distant event. These "phantom events" were automatically released for public distribution on the Web and through the Earthquake Notification Service. These phantom events were deleted quickly by the duty seismologist.
For ths event and the events of 5/29, we have received reports that the delete messages were not distributed by ENS. We are investigating and will address the problem. We are also reviewing the tools used to screen large, distant earthquakes from creating phantom events.
Errata for "Phantom Events" in Central and Northern California resulting from the M6.7 Alaska Earthquake on 2015-05-29 07:00:29 UTC
Strong earthquakes generate seismic waves that spread across the entire globe. When the earthquakes are deep, the distant recordings are quite impulsive and are often mistakenly identified by automated systems as local earthquakes. On 2015/05/29 07:00 UTC, a 60 km-deep M6.7 earthquake occurred offshore of Chirikof Island, Alaska and swept across the seismic networks in northern California. The automatic earthquake detection systems recognized the arrival of seismic energy but misinterpreted it as several earthquakes, including an M 5.1 event occurring near Lewiston, rather than one large distant event. These "phantom events" were automatically released for public distribution on the Web and through the Earthquake Notification Service. All "phantom events" were cancelled by the duty seismologist within 15 minutes.
Errata for false posting of M6.0 earthquake in Puerto Rico on February 27, 2015
The M6.0 earthquake that was posted on the USGS website for about 10 minutes at 2015-02-27 14:21 UTC resulted from an error during a test at the NOAA Pacific Tsunami Warning Center. The problem has been fixed.
Errata for "Phantom Events" in Central and Northern California resulting from the M7.7 Sea of Okhotsk Earthquake on 2012-08-14 02:59:42 UTC.
Strong earthquakes generate seismic waves that spread across the entire globe. When the earthquakes are very deep, the distant recordings are quite impulsive and are often mistakenly identified by automated systems as local earthquakes. On 2012/03/20 18:02 UTC a 626km-deep M7.7 Earthquake occurred offshore of Poronaysk, Russia and swept across the seismic network in northern California (NCSN). The automatic earthquake detection systems recognized the arrival of seismic energy but misinterpreted it as 3 M 5.1 events occurring near King City, San Martin, and Chico, rather than one large distant event. These "phantom events" were automatically released for public distribution on the web and through the Earthquake Notification Service. All 3 events were cancelled by the duty seismologist within 12 minutes.
Errata for "Phantom Event" offshore of Northern California resulting from the M5.9 Blanco Fracture Zone Earthquake on 2012/04/11 22:41 UTC.
An earthquake occurred offshore of Oregon at 43.59 -127.56 with a magnitude of 5.9. The earthquake was reported to the USGS recenteqs website by the Alaska Tsunami Warning Center at 22:46:57. The Northern CA Seismic Network (NCSN) also detected the earthquake with its automated software, but mislocated the earthquake because it occurred outside the network of seismic stations monitored by the NCSN. The erroneous NCSN location was posted to the recenteqs website at 22:46:46 UTC with Ml 3.9 and coordinates 39.36 N, 124.05 W. Because the mislocation was more than 100km from the ATWC location, both solutions were displayed on the web.
The NCSN automated system then updated the Ml magnitude with an Mw of 5.34 at 22:50:17 UTC, which underestimated the event magnitude because of the incorrect location. The NCSN duty seismologist responded at 22:53:34 UTC, but didn�t recognize that the NCSN information was a mislocation of the Blanco Fracture Zone event. Consequently, instead of deleting the incorrect information, the duty seismologist issued a reviewed version of information, but with the previous ML value. The mistake was quickly recognized, and the event was deleted at 23:04:20 UTC.
Errata for "Phantom Events" in Southern California resulting from the M7.4 Mexico Earthquake on 2012/03/20 18:02 UTC.
Strong earthquakes generate seismic waves that spread across the entire globe. As the seismic waves generated by the M7.4 Earthquake near Ometepec, Mexico on 2012/03/20 18:02 UTC swept across the seismic network in southern California (SCSN) the automatic earthquake detection systems recognized the arrival of seismic energy but misinterpreted it as six local events rather than one large distant event. Three of these "phantom events" were of sufficient quality to be automatically released for public distribution. All were cancelled by analysts within 5 minutes. In addition an explosion occurred as part of quarrying operations in eastern San Diego County as the distant earthquake waves passed. Such quarry explosions are common in California. This event was correctly detected and located by the SCSN but an incorrect magnitude (M4.2) was calculated because of the coincidence of high amplitude waves from the Mexican quake.
Errata for Temporary Deletion of Earthquake Information for Magnitude 3.5 San Leandro, California on 2011 August 24 16:57:44 UTC
The Northern California Seismic System located a magnitude (M)3.5 earthquake 5 km from San Leandro on 2011 August 24 16:57:44 UTC. Automated event processing was normal. A preliminary automatic solution with a ML3.6 was available 3 minutes, 40 seconds after the origin and was distributed via the USGS Earthquake Notification System (ENS). Initial seismologist review and confirmation occurred at 17:19 UTC. At 17:44 an update was released with an Mw3.5. Other products including ShakeMap and Did-You-Feel-It, and mechanisms were also properly generated. At 20:19 an analyst performing routine data review accidentally deleted the event. This error caused the issuance of a "delete" message from the ENS to all subscribers who received the original notification of the event. Restoration of web information occurred at ~20:58 UTC.
ENS logic prevents issuing of updates to subscribers unless there is a significant change in the event magnitude. In this instance, identical information was re-posted to the web, so no subsequent ENS message was issued to ENS subscribers. We regret that our systems were unable to issue a correction through ENS and apologize for any confusion that resulted.
Errata for ANSS/NEIC double posting of Arkansas Earthquake Feb. 17, 2011
A minor earthquake occurred on February 17th at 10:49 UTC (4:49 AM at the epicenter) 4 miles from Guy, Arkansas. Eleven minutes after the earthquake, the Advanced National Seismic System (ANSS) released an initial location and preliminary magnitude of 4.0 for the earthquake. Forty-two minutes after the earthquake, a software problem resulted in a double posting of the same earthquake with an incorrect magnitude of 4.7. These errant postings were removed within 16 minutes. The earthquake was updated to a refined magnitude of 3.8 an hour after the event.
The release and double posting of the magnitude 4.7 event was an error. However, magnitude updates (in this case from a preliminary magnitude of 4.0 to its current magnitude of 3.8) are standard procedure. Magnitudes are often updated as more data become available and time-intensive human analysis is conducted on the seismograms.
Errata for UUSS False Alarms of Feb. 11 & 12, 2011
Over the course of approximately two days, February 11 and 12, 2011, the automatic seismic processing software of the University of Utah Seismograph Stations (UUSS) reported a large number of earthquakes occurring across the Yellowstone region. The overwhelming majority of these reports were false alarms, caused by malfunctioning telemetry equipment. This includes the report of an M 4.6 earthquake on Friday, February 11 with an origin time of 7:36 pm (MST). The automated report of this event was broadcast at approximately 7:40 pm (MST), and within four minutes, at 7:43 pm (MST), a seismic analyst had reviewed the report, determined it was a false alarm, and sent delete requests to the UUSS and USGS/NEIC web pages.
Although the automatic seismic processing system used by UUSS is enormously useful, it is not sophisticated enough to produce robust, definitive results without human oversight. Earthquake reports posted to UUSS and USGS web pages should be viewed cautiously if they are labeled as automatic reports, i.e. not labeled as having been reviewed by a seismologist. UUSS technical personnel are currently working to address the Yellowstone telemetry problem, but more false alarms are possible until the problem is fixed.
See additional information on the Volcano Hazards Program-Yellowstone Observatory website.
Errata for UUSS False Alarm of August 4, 2010
On Wednesday, August 4 shortly after 6:00 pm (MST) the automatic seismic processing software of the University of Utah Seismograph Stations (UUSS) reported the occurrence of a magnitude 4.1 earthquake in northern Utah. The nominal origin time was 6:04 pm with an epicenter of 41.8 N 112.1 W. This information was automatically posted to the UUSS and USGS web sites. Within approximately 10 minutes of this automatic alert, UUSS staff seismologists had reviewed the relevant data and determined that it was in fact a false alarm, and that no earthquake had occurred in northern Utah. Subsequently, the event was “cancelled” and removed from the web pages.
The false alarm was created by a magnitude 3 foreshock that occurred 14 seconds prior to a magnitude 4.8 earthquake that occurred near Jackson, Wyoming at 6:04 pm (MST). The UUSS automatic processing software was confused by the combination of seismic waves from these two earthquakes and inadvertently created the fictitious Utah earthquake. Although the automatic seismic processing system is enormously useful, it is not currently sophisticated enough to always produce robust, definitive results without human oversight.
Errata for UUSS False Alarm of August 18, 2010
On Wednesday, August 18 shortly after 6:50 am (MDT) the automatic seismic processing software of the University of Utah Seismograph Stations (UUSS) reported the occurrence of two nearly simultaneous events: one event near Cedar City, UT (MC 4.1, ML 3.3), with approximate coordinates of 37.63 N 113.24 W, and a second event in the mining region of Utah (MC 3.5, ML 3.0) with approximate coordinates of 38.97 N 111.59 W. Within approximately 15 minutes, a UUSS staff member reviewed the second of these events and confirmed it as a true seismic event. A second staff member reviewing the other automatic detection (near Cedar City) determined that it was in fact two events: a magnitude 3.0 foreshock followed by a magnitude 3.8 main-shock, both in the Cedar City area. The two staff members were in communication, and soon determined that the initially confirmed event in the mining region of Utah was in fact a false alarm created by the automatic seismic processing software. It was officially cancelled.
In summary, UUSS detected and located two seismic events in the Cedar City region (Event IDs uu08181251 and uu08181252). The main event (ML 3.8) was located within approximately 45 minutes of origin time and a press release was issued. At that same time, UUSS mistakenly confirmed a false alarm event (Event ID uu00007301), before canceling it approximately 35 minutes later. Although the automatic seismic processing system used by UUSS is enormously useful, it is not currently sophisticated enough to always produce robust, definitive results and can occasionally generate misleading information.