Alaska can be a shaky place. Earlier this week, at 10:15pm local time on Wednesday, there was an M8.2 earthquake in the region. It was the largest earthquake on Earth since the M8.3 in Chile in 2015. Let’s award the US the gold olympic medal for earthquaking (after all, the Chilean winner was from the previous olympic cycle). (There have been two other M8.2 shakes since 2015, one in 2017 in Mexico and one in 2018 in Fiji, so the gold should be shared between three nations.) The massive earthquake did little damage. But was this an exceptional event? Or a sign of things to come, so soon after the Anchorage earthquake?
Earthquakes this size occur somewhere on Earth on average once every one or two years. Since 2000, there have been 15 earthquakes of M8.2 or larger on Earth. They are plotted above. 10 of these were along the Pacific, and the remainder cluster along Indonesia in the Indian ocean. They include the Great Tohoku earthquake and the 2004 Sumatra earthquake. The Sumatra quake is particularly interesting since there are 5 M8+ events close together on the map: the M9.1, an M8.6 aftershock very close to it which happened a year later, a double M8.2/M8.6 8 years later, which were on the oceanic side of the subduction fault and presumably due to faults there which were stressed by the 2004 event, and finally an M8.7 much further along Sumatra in 2007 which may be due to stress transfer at the edge of the rupture zone of 2004. Seeing that kind of activity, people may almost be happy to live in Alaska instead. In contrast, the Great Tohoku earthquake had no aftershock above M7.
All but one of the 15 M8+ earthquakes since 2000 are on the edges of oceanic plates, and are located in subduction zones. Earthquakes of this size outside of subduction zone are very rare. The outlier was in Peru in 2019, an M8.0 120 km below the Andes. This is still in a subduction zone, but is much further downstream and is far underneath the continent.
Going back further in time, we do find some earthquakes this size which are not related to subduction. The largest ones are all caused by India’s collision with Asia, source of the Himalayas. Three earthquakes of M8.0 or larger occurred in the interior of China and Mongolia, in 1905, 1920 and 1957. Others were in Myanmar in 1946 and in Nepal in 1934. The worst by some margin was the Tibet earthquake of 1950. This event reached M8.6 and was so strong that it caused waves (‘seiches’) on lakes in Norway and England. That was 70 years ago. There were six such great earthquakes in the region between 1900 and 1957. But no comparable continental earthquake has happened here since 1957. It seems eerily quiet. If anything similar were to happen now in China or in the Himalayan nations, it would be an earth-shaking disaster. Subduction zone earthquakes tend to happen out at sea, or at great depth below the continent under which they subduct. This limits the damage they do (leaving tsunamis aside). Intracontinental earthquakes directly attack our cities.
Going back to 1900, it is possible to see which regions are most susceptible to earthquakes of this size. Almost all are along the Pacific ring. The coast from Canada to the Kamchatka has about one every 15 years. The Kuril islands just north of Japan have about the same frequency, but focussed in a much smaller area. North Japan, roughly from Sendai to Hokkaido, has one every 20 years, again in small area of no more than 500 km length. The region from southern Hokkaido to the Kurils has not had such an event since 1900. Tokyo had one in 1923, and the Pacific coast of Osaka had two in 1944 and 1946. Southern Japan has been quiet since. From here to the Philippines tends not have M8 earthquakes, until you get to the southern end of the Philippines where Davao has had three. The entire area of Indonesia is at risk, with about one M8 earthquake per decade. The same island of New Ireland has had three. The Tonga trench is a serial offender with seven. New Zealand is a notable gap: it did have an M8+ in 1855 but nothing since.
On the other side of the Pacific, Southern Mexico to Guatemala have had four, and there is a large concentration from Lima (Peru) to Chile (Santiago) which has one every 7 years. Other areas in the world have few, but not none. Pakistan had an M8.1 in 1945, and a few have occurred in the Atlantic ocean, one between the Azores and Portugal. But most non-Pacific-ring events are along the line from Myanmar to Mongolia.
Alaska in earthquakes
But let’s look at Alaska in more detail. After all, that is where this week’s winning event was. What can we say about the occurrence of large earthquakes in Alaska? Experience says that they are rare: this was the largest one in over 50 years. But it seems that they used to be less rare. Alaska has been rather calm since the major 1964 shock. Is it the calm before the storm?
The plot shows the local earthquakes of M8 or larger, since 1900. The 1964 Prince William Sound earthquake was by far the largest, at a staggering M9.2. There has been no earthquake of M8.0 or larger in Alaska since that time, and only one event (in 1986) along the Aleutian chain which reached M8. It shows how exceptional this week’s earthquake was.
Alaska is under attack by the Pacific ocean and its plate. This has always been true. In a way, Alaska’s entire existence is because of this battle. Te Pacific plate subducts along the Alaskan coast, forming a chain of islands and volcanoes in the process. As the plate sinks, it descends underneath Alaska. The red line shows the subduction fault.
The earthquakes are just on the other side of the trench, following the subducting plate. There is a tendency for the quakes here to be paired: three of the locations show two earthquakes in almost the same location, but far apart in time. Going through the events from left to right on the plot:
Adak. 1957: M8.6, 1986: M8.0. There was an M7.9 in between Rat Island and the Adak quakes, in 1996. In some papers the 1957 earthquake is assigned a size of M9.1.
Unimak. 1946: M8.6 There is a large gap between Adak and Unimak
Perryville. 1938: M8.2, 2021 M8.2. There were an M7.8 in 2020 in the same region, which in hindsight was a foreshock. There was also an M7.6 in the same year, so make that foreshock a double. To the right, there was an M7.9 in 1900 on Kodiak Island.
Prince William Sound. 1964: M9.2. This was a bit further from the subduction fault but clearly part of the same sequence.
These pairings show a pattern. The largest earthquakes (M8+) are repeat offenders, with recurrence times between 30 and 80+ years. In the area of the current quake (Perryville is the nearest settlement) an identical earthquake happened in 1938. In fact, looking at the plot, until this week it was the oldest of the unpaired M8+ earthquakes along the subduction zone. This had been noticed in the past. Already in 1981 there was a paper stating that the repeat period in the area of the current quake was 50 to 90 years. This meant that a large earthquake was expected to happen before 2030. Such a prediction should be used with caution, as determining recurrence times on just two events is very uncertain, but events decided to conform to the expectation, and deliver the earthquake within the range of recurrence time
Earthquakes of size 8 or larger are called ‘Great’ while an M7 is called ‘Major’. But what does ‘M8’ actually mean, and what is the difference between an ‘M9’ and an ‘M9’?
The magnitude scale of earthquakes can hide how quickly they increase with number. The magnitude is related to the amount of energy that is released in seismic waves. Each 0.1 step in magnitude corresponds to about a 40% increase in energy. A full magnitude increase (from 8.0 to 9.0) is a factor of 30 in energy. That is why an M7.9 can be a foreshock to a M8.2, as it was in this particular case: it releases only 25% of the total energy of both events. The first event left most of the stress intact.
For an M8.2, the total energy in seismic waves is about 1017 J. The 1964 earthquake released 4 x 1018 J. That was a lot. In an average year, al earthquakes together release around 4 x 1017 J. Obviously this varies a lot from year to year: the 1964 event released 10 times that much in just one event. But taking this number, we can estimate that this week’s Alaskan event accounts for a quarter of the world’s total seismic energy of 2021. Not bad! Alaska certainly knows how to produce energy, not only the human economy but the ground underneath as well.
To give some indication how much energy this is: the Alaskan quake released about the same as the 1980 Mount St Helens eruption. It was also almost the same energy as the largest nuclear bomb every detonated, but let’s not mention that. Comparing to human activity: Alaska produces around 500,000 barrels of oil per day (declining, because the oil fields are now ‘mature’). Each barrel contains around 6 x 109 J of energy, so oil in Alaska accounts for about 1018 J per year. This week’s great earthquake produced about as much energy in seismic waves as Alaska produces in oil each month. The Earth of course produces a lot more energy itself, mainly from internal radioactivity. Less than 0.1% of the Earth’s energy is used to produce earthquakes. The remaining 99.9% produces geothermal heat, amongst others driving volcanoes. Looking at it in this way, earthquakes are not so bad.
Large earthquakes rupture a large length of fault. This length will now need time to rebuild the stress, for the next earthquake olympics. During the rebuilding, another segment along the fault may rupture. Turkey’s Anatolia fault is a case where the rupture sequence can be followed nicely, with adjacent segments failing one but one, in a sequence heading towards Istanbul. That is a transform fault where a rupture causes movement along the fault and adds significant stress to the next segment. In a subduction fault, the movement is perpendicular to the fault and therefore there is much less (or no) stress transfer to the next segment. Each segment ruptures in its own time, independent of what the neighbour does.
The rupture length depends on the size of the earthquakes. This is why repeating earthquakes may have similar size: they rupture the same length segments. On the other hand, a monster quake such as that of 1964 may rupture a series of segments in one go, thereby inhibiting the usual recurrence of earthquakes there. This regularly happens on the San AndreasVolcano at fault: Neenach and the art of moving mountains where the great earthquakes rupture great lengths of the fault, and do not confine themselves to their own segment. That makes it next to impossible to predict the size or location of the next one.
The plot above shows that the 1964 Prince William Sound event has silenced a large region of he subduction fault, all the way to the current event. Perhaps there was also a long silent period before 1964, which allowed the stress to build up to the level where this earthquake became possible.
How far did the 1964 M9.2 rupture the fault? It is thought to have extended to the western end of Kodiak Island. The length of the ruptured segment can normally be measured from the locations of aftershocks. A plot of the various segments is shown below, taken from Davis et al 1981, Journal of Geophysics Research, Vol 86. It shows the great length of the 1964 rupture, ending only at the segment that ruptured in 1938 and which had not yet build up enough stress to take part in the great quake. The 1938 segment acted as a roadblock and the 1964 rupturing ended there.
1964 is a long time ago. The stress along this section is slowly building up again. It won’t always be this quiet. But nothing like it is on the cards for now, and next time it may rupture in smaller sections, a a series of smaller events. Not small, but smaller.
The plot of the segments contains a few threateningly looking black regions called ‘GAP’. This is not the UK clothing store which did not survive the pandemic, but it indicates areas which have not ruptured for a worryingly time. The one with the big black arrow pointing at it (actually this arrow the plate movement direction) is called the Shumagin gap. To the left of it is a tentative gap, the Unalaska gap. Nowadays, the 1946 eruption is assigned M8.6: it may have ruptured a much larger area than shown in this plot. In fact, the two gaps seem to fit this earthquake rather nicely. The two gaps therefore present a region which last ruptured 75 years ago.
As an aside, the plot from the Davids et al paper assigns a much smaller magnitude of M7.4 to the 1946 earthquake. Their number represents the surface magnitude, and it was estimated from the short duration (20 seconds) and the small aftershock area. The value of M8.6 is the estimated moment magnitude of the earthquake. The large difference is a bit of a mystery. The event generated a large tsunami, which suggest the larger magnitude may be right. A more recent study (Lopez & Local 2006, https://academic.oup.com/gji/article/165/3/835/555752) suggests that the aftershocks had been misplaced in the original measurements, and that a much larger segment ruptured than was thought at the time.
There is also a dark gap on the far left which has not yet been filled in. A fourth gap on the far right along the Canadian coast has been filled since, with a M7.8 in 2012 and a M7.5 in 2013.
Which area ruptured this week? It is clearly seen in the USGS earthquake plot, shown below, which contains the earthquakes following the main shock. The main one is indicated by the large blue star. The aftershocks follow a 150 km section, mainly east of the main event. The area closely agrees with the rupture of the 1938 event. Not only were they the same magnitude, they were also the same size and covered the area. This was a copycat event. I confidently predict that a very similar earthquake will happen in 2104. You read it here first. (It is pretty safe to predict this. No one will remember unless it happens to be correct.)
What about the foreshocks? It turns out, they were just at the western end of the current series of aftershocks. They are the two smaller dark blue stars, with on the right the M7.8 of July 22, 2020, and on the left the M7.6 of Oct 19, 2020. Being smaller, they will have ruptured a smaller section. They went off around the eastern border of the Shumagin gap, or in other words at the edge between the 1938 and presumed 1946 ruptures.
After the July 2020 earthquake, an update to this plot was published (Liu et al. 2020, Geophysical Research Letters, sorry but access is blocked to the public) (I am not sure why scientists publish work where it can’t be read). It is in much prettier colours but with the same dark message that the area of the 1946 earthquake (for which they assume the low magnitude) may be next in line.
We know rather little of earthquakes before 1900. However in the region of this week’s event we know of two major earlier events, accompanied by tsunamis: one in 1847 or 1848, and a double event on July 22, 1788 – the same day as the 2020 event – and Aug 7, 1788. If those are the previous ruptures, it indicates a recurrence time of around 80 years. This would predict that the next event will be around 2018. 2021 is indeed not far off – I wish I had predicted this last week. The 1788 and 1847 event may have ruptured the Shumagin gap as well, in which case it will have a similar recurrence time. The 1946 event would have been 20 years ‘late’. It seems plausible the Shumagin gap will rupture within the next 25 years. The main uncertainty in this is that we do not understand the 1946 earthquake well enough.
What causes these subduction earthquakes? The cause is the Pacific Plate, but the earthquake itself is in the continental plate. The plot below shows the Pacific plate moving from right to left while descending into the deep. The continental plate gets stuck to it by friction, and it is forced to also move to the left. This causes it to buckle, and to be pushed up. At some point the stress gets too much, the friction fails and the lock gives way: the continental plate whips back in place. The movement can be 5-10 meters for large earthquakes.
The Perryville quakes all have a depth around 30 km and occur some 100 km from the trench. As the oceanic place goes deeper, the continental plate gets hotter and more ductile, and is finally replaced by the mantle. Here the friction is much less and so the two don’t get stuck as badly. The earthquakes here are frequent but are very much smaller. The green dots show low frequency earthquakes which are slow-slip events: they occur in this region
A bit deeper the descending rock begins to melt, and the melt starts to percolate up through the continental crust. On the surface a volcano forms. In the case of Perryville, that volcano is Mount Veniaminof, site of a VEI 6 eruption around 1750 BC. Alaska does geology rather well.
This week’s event generated a tsunami warning. Tsunamis can be caused by large subduction earthquakes and this event was large enough to raise the alarm. However, nothing appeared: the tsunami did not come. Twenty years ago we had become quite complacent about tsunamis. Now, after two disasters with 300,000 dead, we know better and we take warnings seriously. Even science deniers do not deny tsunamis. Large tsunami are often associated with the largest earthquakes (M9) but smaller ones can also cause them. The 1946 tsunami is an example, as the earthquake itself appeared to be relatively small – but the wave wasn’t.
A tsunami is a tidal water wave. The origin is simple: a large displacement of rock or sediment under water displaces the same amount of water, and sets the wave going. The displacement can be caused directly by the earthquake. As mentioned above, the buckling, locked plate gives way and unbuckles itself. If the buckle extends to the surface, then the surface will suddenly drop. Water rushes in to fill the void. Further upstream the plate moves forward, and this pushes the water out of the way. And this happens over the entire length of the rupture.
The wave of water ripples out, at speeds approaching that of an airplane. In the open ocean the wave may only be centimeters high, but 100 kilometers wide, with a length initially equal to that of the ruptured segment, slowly expanding as it moves out into the ocean. A wave that size is invisible, but contains a lot of water. When approaching the coast, the wave slows down. This makes the wave less wide but it still contains the same amount of water. The wave rises up to compensate. In extreme cases it can reach 10 meters or more. At that point, Lurking’s maxim ‘don’t be there’ applies. And this is what the tsunami warnings attempt to achieve. It gives people a chance.
This particular earthquake apparently did not have much effect: there was no significant tsunami. But this is not known until the wave fails to arrive. This can be annoying, but far worse would be if no warning was given and the tsunami did come.
Many of the great Alaskan earthquakes have caused tsunamis. The 1946, 1957 and 1964 all did, impacting not only Alaska but locations around the Pacific. It seems common. In 1788, a 30 meter high wave was reported (this should be taken with some caution). There has been no significant tsunami in Alaska since 1964. But there had also been no great earthquake since that time. Nothing happened this time – but the next great earthquake remains a tsunami risk.
There are no final words on earthquakes. They have happened before and will happen again. Some areas of the world have been hit hard: Indonesia and Japan come to mind. Others have had time off. But in a way, that can be more dangerous. They will come again, and people may no longer be sufficiently prepared. This event, large but without damage, was a useful warning sign. The years of quiet may be coming to an end. Other places are more at risk, especially places such as China and Pakistan which similarly have not seen anything on this scale for 50 years or more, but have had them before. After the tsunami warning has passed, the earthquake warning should remain.
Albert, July 2021