Seismic Warning Systems

Around 10am today, a M4.7 earthquake struck the Coachella Valley. The quake was felt throughout the valley, rattled nerves but caused no reported structural damage. The felt reports for this earthquake, which appear in the map to the left, cap out at about Intensity V, which corresponds to very light damage and moderate shaking felt. As of this writing, the earthquake has generated over 200 aftershocks, including one M3.0 and four more of M2.5 or greater.

The earthquake appears to have been associated with the Buck Ridge fault, a sub-fault of the San Jacinto fault zone. The San Jacinto is one of the major faults threatening the Coachella Valley (the other being the San Andreas fault), and this section of the San Jacinto has produced M5 earthquakes twice in recent memory: the 2001 and 2005 Anza earthquakes were both in this same area. There have also been a few earthquakes centered somewhat south of this location, closer to Borrego Springs.

This earthquake displays a pretty typical aftershock sequence, though there appears to have been a M2.7 foreshock, as well as several smaller quakes, about 20 minutes before the M4.7 mainshock. A good rule of thumb is that the largest aftershock should be about one magnitude unit less than the mainshock, so in this earthquake we would expect no aftershocks larger than about M3.7. For earthquakes like today’s, the aftershocks should subside to the point of not being noticeable after no more than a few days. For larger earthquakes, that subsiding takes a lot longer. Today is coincidentally the two-year anniversary of the M9.0 Tohoku earthquake in Japan, and they are still feeling aftershocks from that earthquake!

SWS has 16 stations installed in the Valley, and they all performed as expected for this earthquake. Because of local variations in geology and topography, an earthquake’s felt intensity can vary significantly in different parts of the Valley. Our sensors account for these variations. The QuakeGuard systems are tuned to trigger at Intensity V, and today’s event was borderline. That means most of the stations triggered and a few did not, but all stations took the correct action for their local site conditions.

To the right is a map showing the warning times that the Valley would have been afforded for this earthquake, had the CREWS system been in place today. Warning times range from just under 5 seconds, for parts of the Valley closest to the epicenter, to nearly 15 seconds at the North end of the Valley. The reason CREWS can provide so much warning is because we designed it to handle an earthquake in the Valley itself. That is a much more difficult earthquake, and we designed the system to provide warnings even in that worst-case scenario. Compared to that, today’s earthquake is so far from the Valley that it’s easy for CREWS to handle, and the warning times are as good as they are because of the effort that went into the design. In other words, planning for the worst-case scenario pays dividends even if the earthquake is not the worst case!

Gilead Wurman
Chief Seismologist

On Monday evening, Pacific Time, the Global Seismograph Network recorded a M 5.1 earthquake in North Korea. Turns out, this was not an earthquake at all, but a North Korean nuclear test. Although North Korea admitted to having conducted the test, we didn’t really need them to tell us, because we can use some basic seismological understanding to tell that it wasn’t an earthquake at all.

Since you’re reading this on an Earthquake Warning blog, you already know that earthquakes emit both P-waves and S-waves, the latter of which are both stronger and slower than the former. S-waves are slower than P-waves because of the fundamental physics of wave propagation in a solid body, so that doesn’t change regardless of whether we’re talking about seismic waves in the earth, or a hammer hitting an iron bar, or any other physical waves passing through a solid. But the fact that S-waves are stronger than P-waves is unique to earthquakes. Why is that?

Let’s for a moment imagine an earthquake as a result of two blocks sliding past one another. They stick for a moment because of friction, and then they jerk past one another in an uneven fashion. These “jerks” are earthquakes. The motion of the blocks when they slide past one another is a “shear” or sideways motion. Because of this, earthquakes generate S-waves, which are shear waves, much more efficiently than P-waves, which are pressure waves.

Contrast that for a moment with an explosion (doesn’t have to be nuclear, conventional explosions work the same way): the motion of the ground around the explosion is all outward, in a pressure wave. So explosions, nukes included, generate P-waves much more efficiently than S-waves. So the situation is the opposite of earthquakes: the P-wave is larger and the S-wave is smaller (well, as a practical matter they’re about the same size because of certain complications, but never mind). Here we have a simple, elegant means of discriminating earthquakes from explosions. What’s more, based on statistics gathered over many years of nuclear tests (at the Nevada Test Site as well as other locations around the world), we can make a rough estimate of the yield of the bomb from the apparent magnitude. This particular bomb yield was around 6 kilotons TNT equivalent. Not a large bomb by nuclear standards (Little Boy, the bomb dropped on Hiroshima, yielded around 16 kilotons), but larger than any of North Korea’s prior nuclear tests.

The P-to-S ratio technique has been used for more than half a century to monitor nuclear tests. In fact, the Worldwide Standard Seismograph Network (WWSSN), which was the first global seismic network, was developed and deployed in the 1960′s, not for the sake of science, but for the express purpose of monitoring global nuclear testing! These days we have significantly more sophisticated means of discriminating nuclear bombs from earthquakes. These techniques involve gathering and integrating data from all over the world to arrive at a description of an event called a “moment tensor.” The most common way of looking at moment tensors is in the form of a “beach ball diagram”, three of which are shown in the figure to the left. The topmost beach ball looks just like a real beach ball, with two black panels and two white panels. This is the moment tensor for a typical earthquake, and the important thing is they are always half shaded and half white (sometimes other colors are used instead of black).

The beach ball in the middle is the moment tensor for an explosion: all black. The black color means that the initial motion of the earth was outward, away from the source. If the beach ball had been all white, that would indicate an implosion, with initial motion inward, toward the source. Seismologists can use these moment tensors to tell with a very high degree of accuracy whether a given signal is caused by an earthquake or an explosion. Or…

Something else! The bottommost beach ball looks pretty funny, kind of like a band of black goes around the middle. This is the beach ball from a very specific event: the collapse of the Crandall Canyon Mine in Utah, in 2007. This collapse led to the tragic deaths of six miners and three rescue workers in a subsequent cave-in, and led to a high-profile lawsuit against the owner of the mine. The owner claimed that a M 4 “earthquake” which coincided with the collapse caused the tragedy, and since it was an act of God he was not responsible. However, a team of seismologists, including Professor Doug Dreger of UC Berkeley, used the event’s moment tensor to prove conclusively that the M 4 event was the mine collapse itself and not an earthquake. The beach ball here is reproduced from a 2008 paper in Seismological Research Letters by Sean Ford (then Prof. Dreger’s grad student), Prof. Dreger, and Bill Walter at Lawrence Livermore National Lab.

EDIT: The beach ball at the bottom is not the entire moment tensor, but the part of the moment tensor that highlights the signal of the mine collapse. The complete moment tensor would be all white, like an implosion.

Gilead Wurman
Chief Seismologist

Last week saw a flurry of news articles covering a new paper in the journal Nature. The headline invariably went something like this:

“San Andreas Fault Capable of Producing a Megaquake!”

Well, they weren’t all that sensationalistic, but reading the articles one was left with this feeling of impending doom, as if Lex Luthor were about to knock California into the Pacific Ocean. The truth is, the Nature paper in question did not really address the San Andreas Fault at all, other than by way of saying “here’s a place where our work might be relevant” in the conclusion of the paper.

This paper was actually primarily about the 1999 Chi-Chi, Taiwan and 2011 Tohoku, Japan earthquakes. Both earthquakes were large, and both were notable for rupturing through regions of their respective faults which were not thought to support earthquakes. A fault that doesn’t cause earthquakes? How does that work?

It turns out that we see two types of fault behavior in nature. The most well-known type is called “locked” behavior, in which the fault does not slip at all, except during earthquakes. This is the behavior most people think of when they think of faults: long periods of nothing at all, interspersed with isolated large earthquakes that make the fault slip several feet at a time. The second type is called “creeping” behavior, and in this mode a fault does not cause earthquakes, but is constantly creeping along very slowly, about at the rate fingernails grow. The San Andreas Fault has a “creeping” section about at its midpoint, between San Juan Bautista and Parkfield, CA. It has been thought that this creeping section could not support earthquakes, so seismologists have historically considered only M7.8 “Northern San Andreas” and “Southern San Andreas” earthquakes likely, with very little probability of a M8.1 “wall-to-wall rupture” of the entire San Andreas Fault from Mendocino all the way to the Salton Sea.

The observation that led to the Nature paper, however, was that both the Tohoku and the Chi-Chi earthquake appeared to rupture through fully creeping fault segments, which had not been thought possible. To solve this mystery, the paper’s authors constructed a computer model of those faults and simulated earthquakes under a particular hypothesis involving water in the fault. Their computer models behaved in a way that matches the observations in Japan and Taiwan, so we have a plausible mechanism to explain how creeping segments rupture during an earthquake. Does this mean we need to re-evaluate the probability of a wall-to-wall earthquake on the San Andreas? I’d say, not quite yet.

There are a couple of lines of evidence that are still missing before I would feel comfortable hanging my hat on this scenario. First of all, the hypothesis put forward in the Nature paper requires rather specific geophysical properties on the fault, and we have not really sampled the creeping section of the San Andreas to look for those conditions. Second, we have a geologic record of earthquakes on the San Andreas Fault going back about 1000 years, and to my knowledge there is no evidence of a single earthquake rupturing both the Southern and Northern sections at the same time. Since this would be a very large event, it’s safe to assume that the evidence would be quite easy to spot if it were there, so I’m confident that such an event hasn’t happened in the past millennium.

Being a scientist means never having to say “never,” so I won’t say it’ll never happen. It is a remote possibility, and it doesn’t keep me up at night. Even if it did happen, the effect on San Francisco or LA would not be much different from a Northern or Southern scenario. The shaking would not be any worse in any given area, it would just be more widespread.

The difference is that both metro areas would be hit at once, so it does have some ramifications for our system design. For example, we are initially establishing two datacenters for the California-wide earthquake warning network: one in the San Francisco area, and one in the Los Angeles area. If one goes down the other can pick up the slack and the system will continue to function, but a wall-to-wall event could bring down both datacenters at once. So it’s prudent to place a third or even fourth datacenter away from the San Andreas Fault. That was our plan long before this paper came out, so in a sense nothing has changed. The system is designed to be robust, and it feels good to know that even if the doomsaying turns out to be true, we are prepared to handle it.

Gilead Wurman
Chief Seismologist

The original paper is (subscription required for the link):
Noda and Lapusta (2012), Stable creeping fault segments can become destructive as a result of dynamic weakening, Nature, doi:10.1038/nature11703

A couple of significant earthquakes happened in different parts of the world last week. Although they were both around M 7, and both caused a small number of deaths, their tectonics were very different.

First up, the M 7.4 off the Pacific coast of Guatemala, which sadly killed around 30 people. This earthquake is very simple: a vanilla subduction event on a pretty vanilla stretch of subduction zone known as the Middle America Trench, where the Cocos plate is diving beneath, here, the Caribbean plate. The only wrinkle in this area is the presence of the Motagua Fault, the strike-slip boundary between the North American and Caribbean plates, but that fault appears not to have been involved in this event. A few hundred kilometers north of this location, the Middle America Trench exhibits a seismic gap, known as the Guerrero Gap, where a major subduction earthquake (M 8+) is thought to be overdue, and threatens the Mexican coastline as well as the capital, Mexico City. But down here in Guatemala, there have been a number of M 7+ earthquakes and there is no particular sense of an overdue “big one.” In fact, the Motagua fault is much more hazardous to Guatemala, as it runs onshore unlike the subduction zone. If last week’s M 7.4 had occurred on the Motagua fault, the damage might have been much worse because the epicenter might have been much closer to Guatemala City. It is scenarios like that, as opposed to the earthquake that actually happened, that drive our system design for the Earthquake Warning System: large events, close to population centers, with limited communications and power infrastructure and sparse sensor networks. Our requirements for rapid processing, failsafe behavior, and single-station regional solutions mean that we can protect people even for a major event on the Motagua fault.

Next up is a M 6.8 in Burma. This was a shallow strike-slip earthquake that unfortunately took about a dozen lives in the epicentral region. The tectonics of this part of Southeast Asia stand in stark contrast to the simplicity of Central America: it pretty much doesn’t get more complicated than this. The overarching story here is: blame it on India. See, about 25-50 million years ago, the Indian plate slammed into Asia at the blazing-fast speed of 20 centimeters per year. That may not sound very fast, but in plate tectonic terms that’s doing 100 mph in a 25 zone. The most obvious result of that collision (which is still ongoing, by the way!) is the Himalaya mountain range, the tallest on Earth.  But the collision was so vigorous, and the energy to be dissipated so great, that the impact has had an ongoing effect on all of Southeast Asia, which is being squeezed out of the way of the crash like so much plate tectonic gore. In effect, the entire Southeast Asian peninsula is a splatter of continental crust that is extending south and east away from the Himalayas. This can be seen quite easily in the figure to the left, which shows GPS-derived plate motions in this region. This phenomenon has been dubbed “escape tectonics,” and the net effect here is that the entire region is riddled with faults. In fact, it’s exceedingly hard to discern any real “plates” in this area, since the typical definition of a plate is a piece of crust that moves together as a unit. Since different parts of Southeast Asia are spreading in different directions, earth scientists have taken to referring to the tectonics in terms of “microplates” or even blocks, which are really small tectonic units.

Long story short, this area is a great big mess, tectonically speaking. There are a many known, large faults close together, and likely many more unrecognized faults waiting to rupture. The trouble with this is that you never know where the next big quake is going to come from. Well, you never know that anyway, but we know where the major threats are in places like California or Guatemala, and we can focus on protecting against those. This strategy won’t work in Southeast Asia, because there are just too many threats, known and unknown. Here again, SWS technology has an advantage over traditional networks. Although threat-based deployment strategies won’t work well here, SWS sensors can produce regional warnings from single stations. That means we can apply an asset-based deployment strategy without losing too much warning time, since we don’t have to wait for three or four stations to detect the shaking.

Although we are focusing on a warning system for California, the design decisions we’ve taken here make the system very versatile, and well-suited to deployments worldwide, even in relatively remote locations.

Gilead Wurman
Chief Seismologist

This evening, a Moment Magnitude 7.7 Earthquake struck the Queen Charlotte Islands off the west coast of Canada. Thankfully given the size, this earthquake struck a very sparsely populated region, and it appears that it did only minimal damage and there are no reported deaths as of this writing. The event was shallow (less than 20 km deep), and caused a small tsunami, so we will have to wait to see the effects of that tsunami in the coming hours.

The mechanism is consistent with a subduction zone earthquake, and it is near a major plate boundary. So… subduction earthquake, in the style of Sumatra and the Tohoku earthquake of last year, yes? Like the looming Cascadia earthquake south of this location, right? Stick a fork in, we’re done.

Er… not quite. Some things seem a bit off about this earthquake. The biggest one is that the Queen Charlotte Fault Zone (QCFZ), which extends from the Cascadia Subduction Zone in the south, to the Aleutian Trench in Alaska, is not a subduction plate boundary. It’s strike-slip, like the San Andreas Fault. The figure at the left (from Irving, et al., Tectonophysics, Nov. 2000) shows the plate motion of the Pacific Plate relative to North America (PAC and NAM in seismology lingo), and you can see it’s almost parallel to the QCFZ. Note: almost parallel. This will come into play a bit later.

The second issue is that, although this is a thrust earthquake, which is the type we expect to see in subduction zones, it is very steeply dipping (almost 45 degrees), whereas most subduction earthquakes have relatively shallow dips (more like 30 degrees or less). That is, most subduction zones have faults that are nearly horizontal, and this fault seems to be much more vertical.

Here’s what I think: this is a transform fault masquerading as a subduction zone! Perhaps it decided to dress up for Halloween. Possibly this is the West Coast’s answer to The Frankenstorm. Maybe nature just likes to keep us seismologists on our toes. The thing is, there is a small amount of convergent motion between PAC and NAM, as mentioned earlier. This convergent motion has to go somewhere. In some places, like Sumatra and Japan, the oblique motion partitions onto two large faults. In Sumatra the convergent motion is expressed on the Sunda Trench (where the 2004 earthquake struck) and the strike-slip motion is on the inland Sumatra Fault. Here, there is no one large subduction fault to take up the convergent motions, so they are accommodated on a series of thrust faults to the east of the main QCFZ.

I think it was one of these thrust faults that ruptured this evening, and that’s a good thing. Some people might wonder if this earthquake is a preview of an impending megathrust earthquake in the style of Japan or Sumatra. As we unfortunately learned thanks to an Italian court, a seismologist should never say never, but I think that the tectonic setting here is wrong for a megathrust. Surely a M 7.7 is nothing to sneeze at, but the likelihood of a larger, M 8.5+ earthquake in this region is pretty low.

Gilead Wurman
Chief Seismologist

This morning Imperial County was shaken up by a series of earthquakes, including a M 5.4, another M 5.3, and, so far, six earthquakes between M 4 and M 5, as well as numerous earthquakes between M 3 and 4. This is a traditional swarm, with multiple large shocks with no obvious mainshock-aftershock relationship to the smaller events (as opposed to the double event on the Elsinore Fault earlier this month).

Such swarms are relatively common in the Brawley Seismic Zone, between the southern end of the Salton Sea and the Mexican border. In this region, the San Andreas and San Jacinto Faults disperse into a broad zone of deformation that can best be described as a series of blocks rotating clockwise (see the map on the left). Note that this is a very “cartoony” view of the situation, and in reality these blocks are not as clean and simple as that. Because they are rough and broken up, it is not thought that the Brawley Seismic Zone itself can support a major earthquake, and consequently the blocks generate swarms of small to moderate earthquakes where they rub past one another. A similar swarm to today’s occurred in 2005, with maximum magnitudes around M 4.5.

Although it is unusual to see M 5+ earthquakes in these swarms, it is not unheard-of. Today’s events caused some minor damage in the town of Brawley, but so far no reported injuries. Because these events occurred so far from the Coachella Valley, the warning times from CREWS for these events range from about 20 to 40 seconds in the Coachella Valley. However, an ICREWS system could have provided 0.8 seconds warning in Brawley, and up to 10 seconds to the rest of Imperial County.

Gilead Wurman
Chief Seismologist

Over the last 24 hours there have been over 30 minor and light earthquakes in Southern California, including two M4.5 events last night and this morning, as well as a M3.4 this morning. The two mainshocks of roughly equal size are suggestive of a swarm of activity, but to me it actually looks like two distinct (but related) events, each with its own aftershock sequence. It is possible that the first earthquake changed the stress on the fault in such a way as to trigger the second one.

This is not unusual behavior. Most often an earthquake will trigger other earthquakes which are smaller than the original one; we call these triggered earthquakes “aftershocks” since they are smaller than the “mainshock.” About 5% of the time, an earthquake will trigger an even larger event shortly thereafter on the same fault. We then call the larger earthquake the mainshock, and the earlier earthquake is labeled a “foreshock.” It so happens that in this case, the first earthquake triggered a second earthquake of almost the same size, so is it a foreshock-mainshock sequence, or a mainshock-aftershock sequence? The difference is perhaps semantic.

The earthquakes occurred on the Whittier Fault, which is the northernmost segment of the larger Elsinore Fault System. This fault is thought to be capable of a M 7 earthquake, and directly threatens the LA basin due to its proximity to heavily urbanized areas. Although there have been significant earthquakes in this area in the past (the 2008 M5.5 Chino Hills earthquake and the 1987 M5.9 Whittier Narrows earthquake), these occurred off the main trace of the fault. It appears based on location and source mechanism that today’s events occurred on the main trace, but one should not infer too much from this. The fault is considered relatively hazardous in the most recent USGS report, but the recent earthquakes neither increase nor decrease that hazard appreciably.

These earthquakes highlight an important aspect of Earthquake Warning System (EWS) design. When designing an EWS, it is tempting to consider the “most damaging” earthquake as the design case. In the Bay Area, this is a 1906 repeat M~8.0 event, propagating southward along the Northern San Andreas Fault from Mendocino into San Francisco and beyond. In Southern California, the most damaging earthquake (also M~8.0) would propagate northward from the Salton Sea to Palmdale, along the Southern San Andreas Fault. These earthquakes certainly are threatening, but there are two shortcomings to focusing on them:

1. They are rare. The EWS is much more likely to experience smaller (M6-7) earthquakes than a M~8 earthquake.
2. They are far away. Great earthquakes project their damage far from the fault and the epicenter, and in these scenarios the EWS has a lot of time to figure out what’s going on before the damaging waves arrive in LA or San Francisco.

If one really wants to design a warning system to save the most lives and the most money, the focus should be on earthquakes on the Whittier Fault as today’s events were, or on the Hayward Fault in the San Francisco Bay Area. These faults are right up against heavily populated areas, and are considered likely to generate moderate or large earthquakes that could cause much damage in the epicentral region. Because they are so close to people and property, earthquakes on these faults give the EWS very little time to react. So if you can design your EWS to deal with these earthquakes, the more distant great earthquakes become a more tractable problem just by virtue of having lots of extra time to figure them out.

SWS has been hard at work designing just such an EWS: one that reacts as quickly as technically and scientifically possible to any earthquake, so that warnings can be issued even for close-in epicenters like today’s event. The map to the left shows the LA basin and nearby faults, as well as the epicenter of the second 4.5 event as a red circle. The blue triangles are seismic stations in a “Los Angeles Regional Earthquake Warning System” or LAREWS, and the green square is the Los Angeles Emergency Operations Center. The black contours show how much warning time would be available at any point on this map, if LAREWS has been operational during these earthquakes. For example, the LA EOC would have received about 12 seconds of warning from LAREWS for this event. These warning times would be the same even for a larger earthquake with the same epicenter. As the map shows, there is some warning available everywhere on the map, even right at the epicenter!

This, then, is the benefit of designing an EWS from the ground up to cope with worst-case close-in earthquakes: we leave no-one behind. There is no “shadow zone” or “blind zone” in which clients would be left without a warning. Everyone is covered by this system.

Gilead Wurman
Chief Seismologist

A magnitude 7.4 earthquake occurred today on the border between Guerrero and Oaxaca states in Mexico. The earthquake hypocenter was 20 km below the surface of the Earth, on the Middle America Trench. This is a subduction fault, much like the one that caused the Tohoku earthquake in Japan a year ago. In this case, the fault is the boundary between the Cocos plate, underlying the Pacific Ocean, and the North American plate on which Mexico sits. The Cocos plate is subducting (diving under) beneath the North American plate, and this is exactly the kind of environment that can generate great earthquakes like the one seen in Tohoku last March.

Fortunately, today’s earthquake was only a large earthquake and not a great (M > 8.0) earthquake. In fact, today’s earthquake was less than 0.5% the size of the Japanese earthquake, and thankfully does not appear to have caused a significant tsunami or significant damage and loss of life in Mexico. There are currently no reported deaths from this earthquake, and the damage seems to be limited to on the order of 1000 collapsed houses near the epicenter. This is partly due to the fact that the earthquake was centered in a rather remote area, and partly due to the geometry of the faulting. In this region, the subduction fault is nearly horizontal, dipping very shallowly down to the northeast. Because the rupture propagated along this fault, sub-parallel to the surface of the earth, there was very little directivity effect on the surface from this earthquake. Instead, most of the shaking energy was absorbed in the crust, leading to relatively mild recorded ground motions even in the epicentral region (note, mild is a relative term: these regions still experience Mercalli Intensities up to VII, corresponding to very strong shaking).

The location of this earthquake is very interesting: it is just on the southern end of a portion of the Middle America Trench known as the Guerrero Gap. This region, extending from just south of Acapulco to around the city of Petatlán, has not experienced a large earthquake in over 150 years. It is considered by seismologists to be a region of high risk for a great earthquake. Because of this concern, the Mexican government has established an earthquake warning system for Mexico City, with seismometers arrayed along the Pacific coast and waiting to detect the Big One in the Guerrero Gap. The Sistema de Alerta Sísmica (SAS) detected this earthquake and issued a limited warning in Mexico City, giving schools, EMS agencies and transit agencies critical seconds to prepare for the imminent shaking. As it happened, the shaking in Mexico City was only around Mercalli Intensity VI, not enough to cause any significant damage, but if this had been a great earthquake in the Guerrero Gap the damage in Mexico City might have been greater.

SAS is a very good system and one of the first public warning systems ever deployed. However, EQW science has advanced significantly in the last few years and SAS has several limitations. First of all, it is based on S-wave measurements, so precious time is lost in waiting for the S-wave to arrive at the coastal stations before a warning can be issued. Second, due to the design of the system, it is really only capable of issuing meaningful warnings in Mexico City. The surrounding areas are left in the dark, so to speak. The SWS solution can perform better because it uses P-wave detection rather than S-waves, and because its distributed design and composite on-site and network-based capabilities are designed to maximize the warning times to all locations, including the epicentral region.

In this event, SWS could have provided about 3-4 seconds of warning time to Ometepec and Pinotepa Nacionál, two cities in the epicentral region totalling over 40,000 citizens. The city of Oaxaca, with over 260,000 people, and Acapulco with over 650,000, could have received 15-16 seconds of warning time from an on-site system. With a network-based warning, these warning times would have increased to 35-40 seconds. This is more than enough to get under cover and protect yourself in the event that your building starts to come down. Had this been a great earthquake of magnitude greater than 8, Mexico City would have been provided with more than one minute of warning.

Gilead Wurman
Chief Seismologist

On Sunday a magnitude 7.2 earthquake struck near Van, Turkey. This earthquake was a reverse-mechanism or “thrust” earthquake, meaning it resulted from two blocks of the crust being compressed together. This is consistent with the overall tectonic setting of this earthquake, in the heart of the Taurus and Zagros Mountains. This broad mountain belt, extending from the Mediterranean to the Persian Gulf, is the result of the slow, counter-clockwise rotation of the Arabian Peninsula into the Eurasian continent (see figure). The thrust fault where the Arabian and Eurasian Plates collide here is called the Bitlis Suture, but this particular earthquake occurred on one of the many smaller thrust faults in the broad mountain belt to the north of the suture itself. The flip-side of events like this, which build mountains in Turkey and Iran, are so-called “normal” earthquakes, which are the result of two blocks or plates being pulled apart. In the case of the Arabian Plate, this is expressed as the Red Sea, which is opening up along the southwest edge of the plate.

This event has, to date, claimed almost 300 lives. The death toll will likely rise somewhat over the next few weeks, but this is a comparatively low death toll. While Turkey has had stringent building codes for some time, there is a significant compliance problem on the part of contractors. Consequently many buildings are not very earthquake-resistant, and collapse is common. This earthquake was only 10 miles from the city of Van, with a population of some 400,000. Under the circumstances one might expect death tolls in the thousands. For example, the M 7.6 Izmit earthquake in 1999 killed 17,000 in a city about three-quarters the size of Van.

One possible explanation is that the fault may have ruptured from south to north and from the bottom of the fault to the top. This would have directed the bulk of the seismic energy away from Van and toward the smaller town of Erciş, and this is consistent with the extent of damage in Erciş, which was some 30 miles from the epicenter.

This event highlights some of the major challenges involved in implementing earthquake warning. The event was in a relatively remote part of the country, and very close to the largest city in the region. In these circumstances there is very little time to issue a warning. Because of the remote location, the distance between stations is likely to be very large, meaning that one cannot rely on multiple stations right near the epicenter for all events. Consequently, the important thing is to implement a warning system like QuakeGuard which has robust single-station behavior, and to minimize the processing time to get to the warning. This is difficult to do if one tries to estimate the magnitude of the earthquake, but if the system estimates shaking intensity directly from the P-wave, the warning can be issued much faster.

Although this earthquake was indeed tragic, Turkey has to contend with the knowledge that the next major earthquake is likely to occur right underneath Istanbul, the capital and largest city. The Turkish government has been working hard to implement an earthquake warning system for Istanbul, and we can only hope that this system will be in place before that disastrous event.

Gilead Wurman
Chief Seismologist

6.4 Vancouver Island Earthquake – September 9^th

Sept. 9 Vancouver BC, EQ ShakeMap (USGS)

Scotts Valley, California (September 9th, 2011) – A strong earthquake initially reported at a (magnitude 6.7) occurred today at 19:41:34 UTC off the western coast of Vancouver Island. Thankfully, the island is sparsely populated on its western shore. Reports of shaking have been received from as far away as Seattle, WA.

“Had a similar size event struck an urban area, significant shaking and potential damage may have occurred.

Regional Earthquake Warning networks, like the one we’re deploying in California, will provide users a level of protection and heads-up … automating system responses at businesses, fire stations and schools ,” states Scott Nebenzahl, Director of Government Affairs, Seismic Warning Systems, Inc.

A regional earthquake warning system would have provided approximately 19 seconds warning in Ucluelet, BC and over 60 seconds warning to central Vancouver, BC.