Early Sunday morning here (just after midnight on Monday morning in New Zealand), a very large Magnitude 7.8 earthquake struck South Island. The extent of the faulting is somewhat unclear because it was a complex event, but the active fault appears to extend from the epicenter, 100 km north of Christchurch, to the Cook Strait just southwest of Wellington. This earthquake appears to be part of the general subduction of the Pacific plate under the Australian plate. Although the earthquake did generate a small 3-6 foot (1-2 meter) tsunami, the tsunami effects were minimal. Remarkably, only two deaths have been reported in the town of Kaikoura as a result of this earthquake, though many more are displaced or injured. Keep in mind that M 7.8 is the estimated size of the 1906 San Francisco earthquake, which killed about 3000 people. The relatively minor impact in the face of such a large earthquake raises some interesting points about earthquake physics.

Another recent earthquake of comparable size is the Nepal earthquake of April 2015. This was also a M 7.8 thrust-type earthquake, and it killed nearly 9000 and displaced millions. In a long-past blog post I talked about the different effects of earthquakes in Chile vs Haiti, with a look at the pronounced effect that a country’s resilience has on the outcomes of earthquakes. We don’t need to belabor that point again. Suffice to say that yes, New Zealand’s building stock is generally much more seismically resilient than Nepal’s. But if we set aside the human effects of the two earthquakes and look just at the ground motions, we find that the Nepal earthquake generated much more intense shaking than the New Zealand event.

shakemap_comparisonLooking at the two shakemaps side-by-side, there’s a lot of dark orange-red in the Nepal shakemap on the right, representing MMI VIII-IX, while there’s no real red in the New Zealand shakemap on the left, and barely any deep orange areas. By the numbers, the maximum shaking level on the New Zealand map is about 60% g acceleration and 60 cm/s velocity, but there are several areas on the Nepal map where the observed shaking exceeded 100%, and even at times 120-130% g, and 100-110 cm/s! So yes, Nepal is much more vulnerable and the exposed population in Kathmandu alone (1.4 million) exceeds Wellington and Christchurch combined (about 800,000), but the earthquake itself was much more intense in Nepal than in New Zealand, despite the exact same moment magnitude.

slip_comparisonIn part, the difference in intensity is due to the depth of the earthquake: Nepal was 8 km deep while New Zealand was 23 km deep. Shallower faulting causes more intense shaking, all else being equal. However, there is another difference worth discussing: the earthquake rupture histories are very different. As mentioned in the discussion of directivity and its effect on ground motion, an earthquake occurs neither as a point in space nor in time. It is a slip event that takes place over a finite surface area, and it takes some time for the slip to occur as the fault “unzips” along the rupture path. That unzipping can occur smoothly or it can occur in fits and starts. In the Nepal earthquake, the slip model for the event (lower figure) shows relatively uniform contours of rupture time and a slip distribution that is smooth. The slip appears to have been mostly over by 50 seconds after the start of the earthquake. By comparison, the slip model for New Zealand (upper figure) shows a lot of unevenness both in the slip distribution and in the rupture speed. Note that the real slip doesn’t start until 50 seconds after the start of the earthquake, and nearly 100 km from the epicenter. So the rupture history for the New Zealand event is much more complicated. The moment release of this earthquake appears to have taken much longer than for the Nepal event. What’s more, the complex rupture might have absorbed more energy into starting and stopping the slip on different parts of the fault. These two effects combine to reduce the radiated seismic energy (which is felt as shaking) for this earthquake.

Note that the slip model for New Zealand is very preliminary and the story might change over the coming weeks as the data is analyzed more carefully. In the meantime though, we should be thankful that New Zealand appears to have largely dodged a bullet today.

Over the weekend, a Magnitude 6.6 earthquake struck central Italy near the town of Norcia. This earthquake appears to be (at least we hope it is) the culmination of a sequence of moderate events that have hit this region, beginning with a M 6.2 event on August 24, about which I blogged already. Since that M 6.2, the region experienced a M 5.6 earthquake on October 24, and then M 5.5 and M 6.1 events within about two hours of one another on October 26, and finally yesterday’s M 6.6. In the mother of all silver linings, no-one appears to have been killed in this earthquake because the vast majority of the population had already evacuated to shelters following the August and October earthquakes. Nevertheless this event caused still more damage to many historic buildings that were already damaged from the preceding seismicity and were thus even more susceptible than normal. Why has this region been subjected to such an unusual rate of seismicity? Were the earthquakes last week and even the one in August merely foreshocks to yesterday’s event? And can we expect even more and larger earthquakes to come?


The clustering of earthquakes like this is not particularly unusual. Think of a fault not as a slice through the crust but as a string of firecrackers of varying sizes with different lengths of fuse between them. If you light one off, it’s going to cause the firecrackers around it to light off eventually (maybe days, maybe years in the future). Sometimes the details mean the whole string goes off at once (a big earthquake), sometimes only one firecracker goes off and that’s it (an isolated small earthquake), and sometimes they go off in bunches, at odd intervals, which is analogous to the series of unevenly-spaced moderate earthquakes we’ve seen since August.

As to whether the prior earthquakes were foreshocks of the current one, the answer seems to be a resounding “yes.” These events were all within about 20 miles of one another, aligned on the same fault in the Apennine thrust belt, and exhibiting essentially the same mechanism (pull-apart, or normal, faulting on a NW-SE trending fault). What is unusual about this sequence though, is that this most recent earthquake is NW of the August M 6.2 earthquake, and SE of the October M 6.1 earthquake. That is, the foreshocks are actually bracketing yesterday’s event. The M 6.1 on Oct. 26 was technically an aftershock of the August M 6.2 event (albeit a rather large aftershock), and we typically think of aftershocks as occurring on the edge of the ruptured patch of the fault from the earlier mainshock. In fact we often use aftershock locations to tell us how big the mainshock rupture was. In this case though, that doesn’t seem to be the case, and it appears that the Oct. 26 event “jumped” past an unruptured section of the fault—the section that ruptured yesterday. Going back to the firecracker analogy, the fire skipped one firecracker, popped off the one on the other side, and then backtracked to the firecracker it missed.

The implication of this is a bit disconcerting. To me, this sequence of events suggests that this particular fault is nearly critically stressed, and that the event sequence jumped around due to small-scale characteristics of the fault in different places: in one area the rock was weaker or slightly more stressed and that was induced to rupture first, while the area that ruptured yesterday had stronger rock or a less favorable geometry or stress level, and it took a second M 6+ earthquake to induce it to rupture. Normally the stress pattern from an earthquake is so concentrated around the edges of the rupture that these small-scale variations get overwhelmed, but if the fault is already close to failing everywhere, the small-scale variations may make the difference between failing or hanging on until the next earthquake. That means this fault may be ready for more seismicity, perhaps a series of M 6-class events or a single larger one.

Unsatisfying though it may be, we don’t have a good answer to whether there will a larger earthquake. We generally think of any given event as having a 5% chance of preceding an even larger event, though this is a blanket statistical assumption with no particular attention paid to the details of the tectonic regime. Several large events in recent history were preceded by foreshocks: Tohoku, Japan in 2011, Tarapacá, Chile in 2014, and the Kumamoto, Japan earthquake earlier this year come to mind. But a three-event sequence—a M 6.2 foreshock to a M 6.6 foreshock to some even larger event—would be an unusual and remarkable sequence. The experiences of several unfortunate Italian seismologists a few years ago, as well as common scientific sense, suggest one should “never say never.” It is certainly possible that central Italy is looking at an elevated seismic risk well beyond yesterday’s event.

brawleyseismiczoneBack in 2012, I wrote a blog post about a seismic swarm in Brawley, CA. In it, I had mentioned that a similar swarm had occurred in the same place in 2005. This week we are seeing yet another swarm in the same region, this time under the Salton Sea near Bombay Beach, northwest of the 2012 swarm. The activity began two days ago and to date the swarm has produced three earthquakes of Magnitude greater than 4, and 15 more in the Magnitude 3-4 range, with a total of some 250 earthquakes over Magnitude 1.

When the last swarm occurred I showed a map (reproduced to the left) that cartoons the overall tectonic setting for these earthquakes. As a reminder, these are occurring at the extreme southern end of the San Andreas Fault, in a place where it is transitioning from the strike-slip fault mode we see in most of California, to the nascent mid-ocean ridge running down the Gulf of California. In this area the crust is thin, warm and pliable, and as a result the shear deformation across the San Andreas and San Jacinto Faults is expressed as a series of rotating tectonic blocks, as shown on the map. The Brawley swarm of 2012 occurred roughly on the boundary between the middle and upper-left blocks, and the 2005 swarm occurred on the northwest side of the upper-left block on this map. If I had done my homework, I would have put yet another block northwest of the upper-left block. This is where the swarm is happening this week, on the northwest side of that missing block. I might have guessed at its existence: there were small swarms in this same area in 2003 and 2009, but nothing as productive as we’re seeing now.

It’s been one week since the magnitude 6.2 earthquake that struck central Italy on 23 August. In that time the severity of this earthquake has become abundantly apparent: nearly 300 deaths and over 2500 people displaced in the epicentral region. The relatively small size of this quake relative to the damage wrought brings to mind the Christchurch, New Zealand earthquake of February 2011. That earthquake, a magnitude 6.3, killed 125 people. These quakes highlight a central philosophy we have long held: that moderate (M 5.5-6.5) earthquakes are far more common than the large earthquakes that preoccupy disaster planners, while wreaking as much havoc in the epicentral region as those larger earthquakes do. The fact that the damage is not so widespread as in large earthquakes makes these smaller quakes seem less dangerous, but that’s cold comfort to those caught near the epicenter.

The fact that the damage zones for these moderate earthquakes are relatively contained actually highlights the biggest problem with Blind Zones, which are the regions near the epicenter where the earthquake warning arrives after the shaking starts. That is to say, the Blind Zone is a region where an earthquake warning system has failed in any given earthquake, since no warning is provided in that region. While there is no public earthquake warning system in Italy today, we can estimate the performance of a hypothetical system based on current seismic networks and methodologies, with some educated guesses folded in.

Perugia_elarmsLet us assume the proposed Italian earthquake warning system uses the existing geophysical network operated by INGV (the Italian National Geophysical and Volcanological Institute). This is a common feature of all currently operating academic/government-run earthquake warning systems: they use existing seismometer networks which are optimized for seismic research rather than earthquake warning. Let us further assume the ElarmS algorithms out of UC Berkeley are implemented for this earthquake warning system. These algorithms require four seconds of P-wave data from the nearest four stations before declaring an event. We’ll also assume that as part of the implementation they upgrade the dataloggers at all the seismic stations to the latest generation, which would bring their telemetry times down to about one second, as they do in California. Finally, we’ll add three seconds of dissemination delay since the most common approach to earthquake warning in this community is dissemination via cell phones. A recent standards meeting was held in which the latency specification was set at 3 seconds (this is not possible today), so let’s give them that.

Perugia_damageThe outcome is the figure at top left. In this map, the colors represent Mercalli intensity, with orange corresponding to MMI VII, which indicates moderate damage to poorly built structures. Blue circles represent the three cities with the vast majority of fatalities in this earthquake. From north, they are Arquata del Tronto, Accumoli, and Amatrice. The red circle represents the Blind Zone for this hypothetical earthquake warning system. Anyone living within that red circle would have received no warning at all in this scenario. For reference, the bottom left figure shows approximately the same region, taken from a map produced by the European Civil Protection and Humanitarian Operations (ECHO) directorate. Here, yellow areas represent “slight damage” and orange represents “moderate damage”. Note that essentially all the areas that were affected by this earthquake are within the Blind Zone of the hypothetical earthquake warning system. In other words, such a system would fail entirely for a comparable earthquake, because the only people to receive a warning would be people who are unaffected by the earthquake!

Perugia_SWSBy comparison, let us take a hypothetical SWS earthquake warning deployment in this region. This is represented in the figure to the right. For fairness’ sake we’ll assume the same station distribution as the existing INGV network, just like the previous exercise. However, our system:

  1. takes only 1/2 second at one station to determine an earthquake has occurred, rather than four seconds at four stations,
  2. optimizes communications such that data latencies are 50 milliseconds or less, rather than 1 second, and
  3. uses optimized Internet protocols to ensure dissemination in less than 100 milliseconds to fixed devices, rather than 3 seconds to cell phones.

The consequence of this is the warning time map at right. As it happens, we too have a Blind Zone for this earthquake under the assumed station distribution because the nearest station is in Norcia, about 10 km away. In urban areas in California we target a maximum station spacing of 8 km, so no epicenter would be more than 4 km from the nearest station. Even so, the Blind Zone for this earthquake is very small, and does not contain any cities. The nearest town, Accumoli, would receive only about 1/2 second of warning for this event, but even that is better than nothing. Meanwhile, at the edge of what used to be the Blind Zone we would be providing nearly 10 seconds of warning!

Early Friday morning, a magnitude 5.2 earthquake struck just north of Borrego Springs in the Santa Rosa mountains. The quake was widely felt throughout the Coachella and Imperial Valleys and the Inland Empire. Because of the remoteness of the location and the relatively moderate size of the event, no major damage or injury has been reported from this quake. Nevertheless, the intensity was sufficient to activate most of the QuakeGuard installations in the Coachella Valley. In this event, Palm Springs received about 5 seconds advance notice of the shaking, and this was using the legacy on-site systems installed in their fire stations. With the CREWS network in place, this warning time would have been about double that.

CI_stations_Borrego2016As it happens, this earthquake was located in an area where the Southern California Seismic Network, operated by Caltech, is somewhat sparse (the cyan triangles in the map to the left). Dense station coverage is important for basic science, as good coverage in all directions from the epicenter is key to accurately measuring the earthquake. In addition, dense coverage means scientists are much more likely to record near-source or intermediate-distance data, which is rare and highly prized in seismology because it can reveal clues about the physics involved in the earthquake rupture itself. Since earthquake warning time depends strongly on the station density, this could also adversely affect any earthquake warning being issued off this network.

AZ_stations_Borrego2016 Luckily, there are two additional seismic networks currently operating in this area: the ANZA Network (shown in the map to the right) and the San Jacinto Fault Zone (SJFZ) Network. These networks are operated by UCSD/Scripps Institute of Oceanography. These networks cover one of the most significant gaps in the Caltech network’s coverage in the Santa Rosa and San Jacinto mountains, and instrument the San Jacinto Fault Zone, which was the source of Friday’s earthquake. Despite their critical location along one of the most active faults in Southern California, these networks’ funding was zeroed out a couple years ago. Without speculating as to the reasoning behind this, it highlights the fact that public money is not always a sure thing, especially on an ongoing basis. Political priorities change and long-term projects like seismic networks can be shelved in favor of more immediate or visible projects.

When this happens, there is an opportunity for the private sector to step in and fill in the void, which is what happened last year when Seismic Warning Systems agreed to fund the maintenance and operation of the UCSD seismic networks for five years. I’m very proud of our role in keeping this valuable geophysical network alive and providing data to the academic community for ongoing research. If we hadn’t stepped up to help, all the capital and labor that UCSD had already invested in the ANZA and SJFZ networks would have been lost, and the rich near-source data recorded by these networks on Friday would have been lost as well. This earthquake was thankfully small enough so as not to cause any harm, but significant enough to really highlight the importance of having seismic instrumentation along the SJFZ, which could easily generate a much larger earthquake. Friday’s outcome is a win for the research community, and a wake-up call to those of us in the private sector who depend on the data and research coming out of these networks, that we too bear the responsibility for ensuring that they remain open and available well into the future.

The 2014 M 6.0 South Napa earthquake caused significant damage in Napa, Vallejo and beyond, and killed one person almost two years ago. While that earthquake was obviously quite significant for those in the Napa area, the Bay Area came away relatively unscathed in the broader context. We seismologists do like playing what-if games, though, and occasionally nature provides us with a window into what our scenarios might look like in real life.

Today we got a glimpse of what the Napa earthquake would have looked like if it had been centered near San Francisco (population over 800,000) instead of Napa (population about 80,000). That’s because, this morning, a M 6.2 earthquake struck the city of Kumamoto, population about 800,000, the capital of Kumamoto Prefecture in Japan. At the time of this writing there are nine reported fatalities and nearly 800 injuries, as well as some 45,000 people displaced by this earthquake. The USGS PAGER assessment estimates the economic damage from this earthquake to be on the order of $1-10 billion compared to perhaps $300 million in the Napa earthquake.

The scale of the destruction from this earthquake is remarkable, because Japan is one of the most well-prepared nations on Earth when it comes to seismic hazard. Looking at the reasons this earthquake was so damaging, one sees some striking parallels to California (whether it’s LA or San Francisco, which I will focus on here):

1. The earthquake was very shallow, only 10 km deep, and pretty much right beneath the city. Seismicity in California has a median depth of only 8 km (the Napa quake was 11.3 km down), and the San Andreas Fault runs right through Daly City and the southwestern corner of San Francisco, where it goes offshore.

2. Kumamoto City is situated on a broad plain on Shimbara Bay. The soft alluvial sediments the city is founded upon may have amplified the shaking and the consequent damage, and may have suffered liquefaction because of the likely high water table. Some parts of San Francisco (the Marina and Embarcadero) are built on similar soft sediments right on the San Francisco Bay, and may also liquefy and amplify shaking in a comparable event.

3. Kumamoto as a city was founded in the 19th century (it has centuries of history before that, including a 16th-century castle), and although building codes are very stringent in Japan, any city of 800,000 is likely to have some older building stock that is vulnerable to earthquakes. The same can be said of San Francisco.

This earthquake highlights the importance of our approach to earthquake warning, which focuses on the hazards posed by moderate (magnitude-6-class) earthquakes near urban centers, which are much more likely than large (magnitude-8-class) earthquakes. The latter will spread their damage much wider, across half the state, but to people right near the epicenter the damage ends up being nearly the same as in the moderate events. Although, even under ideal conditions, we couldn’t have provided more than a second or so of warning to much of the city in this earthquake, even that second or fraction thereof could have been useful. In this video from a newsroom in Kumamoto, we can see that the earthquake managed to cause most of its damage before anyone in the room could react, and even afterward they did not respond by drop-cover-and-holding in anticipation of more shaking. Had they heard a warning tone a second before, they would have already been on their way down under their desks when the shaking began.

Yesterday a moderate M 6.4 earthquake struck southern Taiwan. As of this writing, the death toll from this quake stands at 15, all in the nearby city of Tainan, 50 km from the epicenter. Most of the deaths occurred in the collapse of the Weiguan Jinlong apartment complex. This is unusual inasmuch as the earthquake occurred in a relatively remote part of a country with very stringent building codes. Why did this happen?

Weiguan JinlongThe city of Christchurch in New Zealand was struck by an earthquake of similar magnitude in 2011 and suffered significant damage. Despite the stringent building codes in New Zealand, Christchurch had a large stock of historic buildings that were not up to modern standards. This is certainly a possible factor here, as Tainan is an old city and probably has some historic buildings, but the Weiguan Jinlong complex was built relatively recently. In some countries (Turkey comes to mind) stringent building codes are undermined by lax oversight and contractors that cut corners to lower their costs. That could be the case here as well: news footage from the wreckage shows reinforced concrete with embedded rebar which has been torn to shreds. A dozen or so buildings collapsed around the city, whereas with endemically lax construction we would expect more widespread damage.

I think one possible clue comes from the way the buildings collapsed. They did not fall in place, as is common in unreinforced construction where the floors pancake on top of one another. Rather, they appear to have toppled to one side. That is, the upper floors are mostly intact apart from lying on their side. This suggests a low-frequency excitation of the building’s fundamental period, such that the building swayed back and forth and ultimately failed at the ground level. The thing is, such a (relatively) small earthquake shouldn’t excite such low-frequency energy very efficiently.

The answer may lie in the location of the city itself. I started this blog by discussing the Three D’s that can impact the intensity of shaking: Distance, Direction and Dirt. It may be time to add a different “3-D”, which in this case refers to the three-dimensional structure of the earth beneath the city. Tainan is built on the edge of a large sedimentary basin, the Tainan Basin. Apart from basins being full of soft sediments which by themselves amplify the shaking, basins are bounded by hard rock. The sharp contrast between soft sediments and hard bedrock creates a reflecting boundary that can contribute to bad outcomes in a couple of ways. First, the low-frequency energy of an earthquake can get trapped in a basin and slosh back and forth for a lot longer than it otherwise would. Think of a bowl full of jello that, when you whack the bowl, jiggles back and forth long after the bowl itself has stopped vibrating. The second effect is that, near the edge of the basin where the sediments get really thin, a critical amplification can take place that generates much larger motions than you see anywhere else. These basin effects have been shown to be a factor in a number of earthquakes. They were responTainan_intensitysible for the significant damage to Santa Rosa during the 1906 earthquake, and for many of the building collapses in the 1995 Kobe earthquake. Indeed, looking at the ShakeMap for this event, it’s pretty clear that the intensity increases in the flat topography of the Tainan Basin, especially right at the edge of the basin near the mountains.

This earthquake is something of a wake-up call for people who think that the only earthquakes worth worrying about are the “big ones” of M 7 or larger. Even the USGS earthquake forecasts typically look at M 6.7 and up, not considering the moderate events in the M 5.5 to 6.5 range that can easily cause damage and loss of life as well. And the thing is, these earthquakes are much more common than the larger ones.

A couple of days ago a story went up on a Portland, OR news station’s website trumpeting the imminent arrival of an earthquake warning system for the Pacific Northwest. Among other things, the story quoted John Vidale of the University of Washington saying that the proposed ShakeAlert system could provide up to several minutes of warning before you feel the shaking. This isn’t the first time he has made that claim, and he’s not the only one to make it. The same claim has been made for the proposed ShakeAlert system in California, where the typical number is “over a minute of warning” for San Francisco or Los Angeles.

So, first of all, I want to be clear: yes, such generous warning times are theoretically possible. Typical quoted numbers are a little over a minute in San Francisco, a minute and a half in LA, three minutes in Portland, five minutes in Seattle. They are possible if:

  1. The city in question is close to a very large fault, and
  2. The epicenter of the earthquake happens to be very far from the city, but
  3. The earthquake is large enough (think magnitude 8+) that the city experiences significant shaking, despite the distance to the epicenter (remember way back to my very first post on this blog: The First D), and
  4. The recipients of the warning don’t mind getting low-ball or speculative ground motion estimates, since the true size of such a massive event won’t be known for at least a minute or two.

Notice that the list above doesn’t include any question of the technical feasibility of ShakeAlert itself. I’ll concede that the system can be built to the specification that is being publicized. Even though there is no false advertising here, it is irresponsible to promote any earthquake warning system as providing “minutes of warning” for several reasons. First of all, “minutes” sounds like a long time, and that time has a tendency to grow in the retelling, especially if a specific number is not given. I’ve personally heard public officials claim that the warning system being developed will provide 15 minutes of warning! The scientists who keep making the “minutes” claim may be disciplined in sticking strictly to what is technically possible, but members of the public do not have that discipline. What’s more, when you talk about “seconds” of warning that makes it easier to connect in people’s minds that this is not an earthquake prediction system. That is, the warning does not come before the earthquake. That fact is much harder to convey when the timescale is “minutes” because most people think of earthquakes as more or less instantaneous.

The real reason this is dangerous though, is that it gives the public a completely unreasonable expectation of the performance of any earthquake warning system. It’s one thing to say “give me $120 million and I’ll give you five minutes of warning,” but quite another to say “give me $120 million and I’ll give you 15 seconds.” It’s much much harder to get public buy-in with the latter claim, but the fact is that yes, a moderate event right beneath our feet (say a M 6.5 to 7 on the Hayward or Newport-Inglewood Faults) is far more likely to occur than a great M 8+ event. And even if we do see a M 8+ on the San Andreas or Cascadia Faults, the likelihood of the epicenter by chance being as far from the major cities as possible is extremely small.

Allen_2006_Fig2Just how much more probable is a warning time of a few seconds than one of a few minutes? I’d have to do a probabilistic study of all the available scenario earthquakes for California and calculate warning times to tell you. Fortunately, I don’t have to do that because it happens that Richard Allen, a big advocate of the ShakeAlert system (and full disclosure, also my former Ph.D. adviser) published just such a paper in 2006. The main figure from that paper is reproduced here. The data is nearly 10 years old now, and both the networks and the earthquake science have advanced since then, but the conclusion is largely unchanged: for earthquakes that cause moderate or greater shaking in San Francisco (all the colors in the figure except for gray), the probability of warning times greater than 15 seconds is small, and the probability of getting more than 20 seconds is tiny. A careful reading of graph B in the figure shows about a 7% chance in 30 years of experiencing MMI 5 (V) or stronger shaking in San Francisco with more than 20 seconds of warning, but about a 72% chance of experiencing that shaking with less than 20 seconds of warning. So it’s more than ten times as likely that the average San Francisco resident, who willingly paid into the $120 million bill for ShakeAlert, will experience a warning time that is significantly less than they were expecting.

I have asked the academic community in the past to tone down the wild claims and do a better job of managing expectations. This is in part self-interest, because we have a hard time making more realistic claims of several seconds’ warning, when the people we talk to are used to hearing about several minutes’ warning. But beyond that it’s really important to be honest with the public, who are after all being asked to foot the bill for this system.

This evening just before 8pm Chile Time, a great earthquake, Magnitude 8.3, occurred off the coast of the Coquimbo Region of Chile. This earthquake, like many other great earthquakes before it, occurred on the enormous Peru-Chile Trench, the same fault that was responsible for the M 9.5 Valdivia Earthquake of 1960, the largest ever recorded. Today’s earthquake spawned a significant tsunami that inundated large coastal areas of Chile and will, in the coming hours, be observable all around the Pacific Rim. No significant flooding is expected beyond Chile itself.

While the damage in Chile is likely to be significant, the death toll at this time stands at five people, a remarkable number. As I have mentioned before, Chile is a shining example of the dividends paid by a concerted, prolonged campaign of hazard mitigation. First-rate building codes (that are scrupulously adhered to by contractors) and a system of tsunami sirens spared the people of Chile a much higher death toll tonight, aided in no small part by a thorough campaign of public education about the proper response to earthquakes and tsunami.

When this earthquake first occurred, global seismometer networks detected it and were temporarily confused about the location. Automated location programs called out two simultaneous M 7.9 earthquakes: one at the actual location, and one farther north, very near to the location of the 2014 M 8.2 event off the coast of Iquique, Chile. This sort of thing happens on occasion, and the automated system was never designed to replace human oversight (that is a big part of what makes earthquake warning so much harder: we have to get it right, right away). Within half an hour or so a seismologist had reviewed the data and concluded that the northern event was spurious and had deleted it. But for a few minutes there I was pretty excited (I know, ghoulish). See, back in 2014 I blogged about the Iquique Earthquake, and in that post I mentioned that there was room for another M 8.2 in the Iquique gap. For a few golden minutes I thought I could build a career as an earthquake prognosticator. Ah well…

Chile_2015-09-16_intensityThis afternoon a M 8.3 earthquake occurred on the subduction zone off the coast of Chile. The earthquake caused significant shaking in Coquimbo and Valparaíso Regions, and was moderately felt in the capital, Santiago. PAGER forecasts suggest 10 to 100 fatalities and up to $1 billion in damages. This is from the earthquake itself, and not from any ensuing tsunami.

At this time there has been confirmation of a significant (~4 meter) tsunami from at least one tide gauge. I will update this post later tonight as more information becomes available.