There is a bit of California here. The fault equals the length of the San Andreas, its shape, its movement, and its earthquakes. Even the ends are alike: where the San Andreas has formed the Salton Sea, its counterpart has embraced the Marmara Sea. Both are places where the crust is being pulled part, and the resulting extension basin has filled with water. The Salton Sea is an intermittent lake, with the latest fill-in being due to an engineering accident. The Marmara Sea is a much more permanent and much more impressive fixture. But whereas the San Andreas has had 2 major earthquakes with magnitude larger than M6.7 since 1900, the North Anatolian Fault has had 12. The North Anatolia fault is Asia’s San Andreas. But perhaps the San Andreas should really be known as America’s Anatolian.
Not many countries own their own plate. Turkey gets close. It is almost the sole occupant of the Anatolian plate, shared only with the northern edge of Cyprus. Owning your own plate should feel safe. It is a bit like owning a castle. But this is not quite true. Turkey also owns the edges of the plate, and part of the other sides. It has responsibility for the rifts that separate the plate from its neighbours. And the neighbours are a raucous lot. The microplate is sandwiched between the Eurasian plate, the African plate, the Arabian plate, and the Aegean Sea plate. All of these are converging in some ways, and Turkey is acting as the barrier keeping north and south (and east and west) apart. The Anatolian plate is being attacked from all sides. And it shows. Turkey is highly earthquake-prone. A map of earthquake risks shows the entire country as deep red (high risk) to orange (medium risk). Not a single square meter of the country is classed as low risk!
Most of the highest risk areas are found in a band across the northern part of Turkey, bending southeastwards towards Iran. There is another red area in the southwestern corner. Last week a major earthquake hit Izmir: this is located in the southwest high-risk zone. But our attention will be further north.
Plot the locations of the major earthquakes, and you will find a patchwork of faulty shakes across Turkey. One band of earthquakes extends across the north, one is in the east running towards the southwest, and an extended area of large quakes covers the southwest. The two main bands of earthquakes show the major faults on land. The southwest is a patchwork of small faults. It is an ancient area with famous cities, such as Smyrna (near Izmir), Troy and Ephesus. But few of those cities still exist. Between the 4th and 6th century, all major cities in this region were hit -and destroyed- by a series of large earthquakes. The geology shows that it is an area of extension, with many small grabens. It is very active, and very shaky.
There are also faults in the sea. Two separate subduction zones are active in the nearby Mediterranean, the first south of Cyprus and the second south of the Hellenic arc in the Aegean Sea. In both places the sea floor is subducting underneath the continental plate. Turkey is collateral damage.
The central area has fewer faults, and this is the ‘medium risk’ area. However, this does not mean that large earthquakes could not occur here. Major quakes may have repeat times that are much longer than a century, and absence of data may not mean absence of risk. No place in Turkey is fully safe.
There are four major faults in the region. The Bitlis-Zagros fault is a continental collision, extending into southeast Turkey. The Dead Sea Transform is where Arabia slides past Africa. This fault has given us such treasures as the Dead Sea. It is tectonically not very active, but should not be underestimated. In 750 BC, the fault ruptured with an estimated magnitude M7.8. Archeology shows widespread, catastrophic damage in Lebanon and Israel. We do not know the recurrence time of such massive earthquakes. Smaller, but still destructive, earthquakes happen about every 1000 years, the most recent one in 1202. This fault terminates inside Turkey, not far from Aleppo in Syria. Here it connects to the East Anatolian fault, which defines the south-eastern edge of the Anatolian plate. But the big one lies to the north, and is entirely within Turkey. This is the North Anatolian Fault. It begins at a triple junction, near Karliova, and runs close to the Black Sea. (The third arm of the triple junction is far from obvious. It is the wide crumple zone to east.) In the west it splits into two branches, called the southern strand and the northern strand, which run either side of the Marmara Sea. The strands end where they reach the Aegean Sea
The cause of Turkey’s tortured tectonics lies afar. Literally so. 30 million years ago a plume started rising underneath Afar. It caused a triple junction, and eventually led to the break-up of the African plate. The plume cannot have done this by itself: a plume can break through the crust, but to break a plate a pull is needed as well. This came from the subduction of the ancient Tethys plate underneath Asia. It pulled on Africa, and this allowed the plume to form a spreading rift which split Arabia from Africa. Arabia is still moving north, in the process rotating a little anti-clock wise.
The movement is causing the collision that is traced by the Bitlis-Zagros fault. So far, so good. But it is also pushing the Anatolian plate, from the side. The triangle formed between Arabia and Eurasia acts like a squeeze on Anatolia. And Anatolia is responding to this pressure. To the west lies freedom: the subduction zone in the Aegean Sea is rolling back, and this creates space. Space for Anatolia to move into. The plate saw this, and it began to move west, out of the way of the collision between Arabia and Asia. The roll-back in the Aegean at the western end is pulling towards the south, and this adds to the movement: it helps Turkey to rotate its way into the Mediterranean. The rotation explains the bend in the North Anatolian fault, the bulge into the Black Sea, but also the gap between Turkey and Europe where the rotation causes the western side of Turkey to pull away a bit from Europe. This has caused extension, and formed a basin: the Marmara Sea.
The ultimate cause of Turkey’s westward escape route is the fact that Arabia is moving north faster than Africa is. The Earth really is a battle ground between many warring parties. The word ‘balkanization’ comes to mind.
The North Anatolian fault is a right lateral strike-slip fault. It separates the Anatolian block (Turkey) from the Eurasian plate, and runs from the Karliova Triple Junction to the Marmara Sea at Izmit. There it dives underneath the sea and re-emerges on the Gelibolu peninsula, along the Dardanelles Straits. The total length exceeds 1200 km. This is the northern strand. The southern strand splits off near Düzce and runs a little further south, via Bursa south of the Marmara Sea. Here it again splits in two, with one side running along the Marmara Sea before petering out, and the other side running overland to Edremit, near the coast. The fault accommodates about 24 mm/yr of transform movement. The numbers are almost identical to the San Andreas.
The North Anatolian Fault is quite young. This will not be a surprise, seeing that Afar didn’t start its action until 30 million years ago. The Earth needs its time. Turkey’s journey of escape started about 11 million years ago when the southern side of the fault began to move west.
The effect is seen in the rivers in this region. An example is the Elmalı, a tributary to the famous Euphrates. It runs from the mountains to the south, but on the way crosses the fault. As the transform motion took hold, its downstream bed was moving west, and the river had to follow. Now it runs for some 70 kilometers along the fault before escaping through its old bed. Further west, other rivers do the same but here they flow north rather than south.
The detached river beds show the amount of movement that there has been on the fault. The shift varies between rivers, perhaps because some rivers are younger than the fault. The largest shifts are around 85 km. To the west, the shifts become smaller, in places as little as 30 km. The fault appears to be younger in the west. How young is disputed. Along the Marmara Sea, the offset has been measured as only 4 km.The fault through the Marmara Sea may only be a million years old – one paper even claims 250 thousand years. But not everyone agrees: others find offsets as much as 30 km, and ages of a few million years. Along the Dardanelles fold, at the westernmost end, the offset again is 70 km. It is a confused region. It can be difficult to align the two sides of a transform fault. A river crossing is very helpful but is not always available. Other land forms can sometimes be mismatched: all hills and troughs look alike.
All evidence does point at the fault becoming younger towards the west. Especially the northern strand seems young. One to a few million years is probably a safe guess. The southern strand is older, up to 5 million years. The northern strand seems a more recent addition, and it may have formed quite quickly. The age discrepancies can perhaps be resolved if it started out as a north-south extension zone, and only recently changed to become a transform fault. Once the transform motion was initiated, the new fault grew fast. The current growth rate is estimated at 11 cm per year.
The slip rate also seems to have increased over time. The slip rate along the fault ranges from 25 mm per year in the east to 20 mm per year in the west. In the west, 70% of the slip is taken up by the northern strand. The southern strand account for only 30%. At the western end it may be moving towards the slip rate of the Aegean sea which is 30 mm per year and which pulls in the Turkey block.
The fault is surrounded by a shear zone, crossed by Riedel faults. This shear zone becomes wider towards the west. This may also be caused by the younger age. When the transform movement starts, the movement can be taken up in its entirety by Riedel faults, crossing the shear zone at an angle. As the lateral motion takes hold, the Riedel faults rotate with it and the shear zone narrows. Eventually the Riedel faults lock, and a through-fault now forms. This process is complete in the east, but is still in an earlier phase in the west.
The transform fault split the Anatolian plate from Eurasia. But why did the fault form here, in this particular location? It turns out that the fault follows the line where the ancient Tethys ocean disappeared. The rocks on either side of the fault are different, both of Jurassic age but of different composition and origin. This left a weakness which allowed the fault an easier passage. Faults always take the easy way. The Earth can be very human.
There are remarkably few images of the fault published. It is a far less touristic destination than the San Andreas is. And of course, it has to compete with amazing attractions elsewhere in Turkey. This country has history – and has the monuments to show it.
This is a photo from near Gerede, published on https://blogs.agu.org/mountainbeltway/2010/12/06/gerede/ The fault runs from the cement factory towards the lower right of the photo. It seems a fairly typical view. The fault sometimes forms a narrow valley, as is seen here. In other places it is not so obvious from the landscape. The valleys are interspersed with basins, evidence of some pull-apart action, perhaps related to the 35 degree bend. One of the largest is the Erzincan Basin in the far east of the fault. Uniquely, it shows evidence of volcanic activity around 250 thousand years ago. The pull-apart motion was apparently strong enough to allow some melt to occur.
In many places there are sag pools and pressure ridges. These form around a step-over where one segment ends and the next starts some distance to the side. Depending on the direction of the sideways shift, the region in between is either pulled apart or pushed together. The first case gives a depression, the second gives a pressure ridge.
This image is of Gerede, a town destroyed in the 1944 earthquake but enthusiastically rebuild. The ridge in the middle is a pressure ridge. The valley in front – that is the fault. The red roofs are the new town. You do wonder whether this was the wisest place to rebuild. But then, I have also seen beautiful houses build right along the San Andreas. Humanity has a peculiar attitude to danger.
Locking and shaking
The fault is locked for much of its length. This makes it prone to large earthquakes. Only in the central region is there evidence for some aseismic slip, where the fault moves by intermittent creep.
The area of the 1944 earthquake has shown extended creep since the event. The picture shows a wall in the area which was built in the 1950’s, after the earthquake. On the left is a photo taken in 1969, and on the right in 2014. On the left a bit of movement is already visible. The difference becomes much more notable by 2014. The creep happens over a distance of several meters. It has been measured at 1 cm/yr, and it happens episodically. It may still be related to 1944 earthquake in this region: post-quake creep is also known from the San Andreas.
But generally, this fault is solidly locked, and the slip is accommodated by violent events whenever the tension becomes too much.
The North Anatolian Fault was discovered in the 1940’s, after some major earthquakes. The lack of clear evidence in the landscape may have made it more difficult to discover. But in hindsight, there was a record of the fault in the list of historic earthquakes. This fault has plagued Turkey, and especially Istanbul (Constantinople) for a long time. The records are most complete in the western part of Turkey, and more fragmentary further east. The historical record in the west is unique, though, being complete for large events for the past 2300 years. It shows almost 70 earthquakes estimated at M7+ over this period, or one every 30 years. This record is certainly incomplete for the central and eastern parts! (And it only applies to the North Anatolian Fault. The southwestern zone of Turkey is not included!)
The records show that on the western side, M7+ earthquakes are frequent. In any particular location the recurrence times of M7.4 earthquakes is around 200 to 400 years. The real number of such earthquakes is much larger, of course, as location more than 100 km apart tend to have their own earthquakes. The central region has fewer but larger events: recurrence times for M7.5+ quakes per location is suggested to be around 350-400 years. It was hit in 1944, 1794, and 1668. The eastern region has similar recurrence times, and also shows very large earthquakes, up to M8. However, this region is not as densely populated and the earthquake history is not well known. A series of large quakes appear to have have taken place in the late 1500’s, culminating in the complete destruction of Erzincan in 1583.
Each earthquake ruptures a different part of the fault. The rupture length varies between earthquakes. The fault has a 10-km-wide step-over at Niksar, and this tends to stop ruptures and prevents the entire fault from failing on one event. Ruptures are normally unable to cross step-overs of more than 5 km.
Very few records exists for the central and eastern sections for much of the Middle Ages. It has been argued that the North Anatolian Fault had few earthquakes here in this period but this is unproven. It is most likely due to a lack of written records.
Some of the past earthquakes of the North Anatolian Fault have become famous enough to have a wikipedia page, albeit with limited information.
The earthquake happened at midnight of Dec 14, 577. Reports say that the city was ‘almost completely razed’. This was probably overstating things, but there was major damage. The walls of the city were badly damaged, and were unable to hold back the Huns the following year. (This was shortly after the apocalypse of 540, and the known world was badly weakened and in turmoil.) The dome of the Hagia Sophia was so badly damaged that eventually it collapsed. There had been two notable foreshocks over the preceding 10 months.
This was among the most destructive earthquakes to hit the city. The name ‘Little Apocalypse’ has become attached to the event. But the accounts are sometimes confused, and probably overstated.The earthquake happened on 10 Sept 1509, during the ‘4th hour of the night’ (1-2 hours before midnight). The size has been estimated at M7.2, and the location was in the Marmara sea close to Istanbul. There was major damage to the city, with several newly build mosques badly damaged. The Sophia lost its newly acquired minaret. Many other minarets of the city succumbed. The walls of the city were badly damaged. A tsunami flooded the coast over a long length. The number of fatalities is not known. Some reports claim 12,000, but others state a few hundreds. Perhaps 1% of the houses in the city were destroyed.
On 17 Aug 1668, the central-east part of the North Anatolia fault ruptured. This is probably the largest earthquake known from the North Anatolian Fault with a magnitude estimated at M8, similar to the largest event the San Andreas can do. The 1668 rupture may have been after a long time of quiescence but this is under discussion. The 1668 event ruptured a third of the entire fault, along a length of 500 or 600 km. It may even have moved across the Niksar step-over. The earthquake caused a slip estimated as 7 meters.
1766 Marmara earthquakes
This was the last time an earthquake caused major damage in Istanbul. (Less major damage has happened, for instance in 1894 and 1999).There were two events. The first earthquake happened on 22 May 1776, shortly after sunrise. It was followed by a large earthquake in the Dardanelles in August. Both earthquakes measured M7.2 or higher.The first earthquake caused major damage in Istanbul and nearby regions. The palace and many mosques were damaged. Some 4000 people are estimated to have died. The second was further away and has less impact on Istanbul.
This earthquake is sometimes overlooked. It happened on 10 July, and badly affected the region between Izmit and Istanbul with significant damage in Istanbul itself. The magnitude is estimated as M7.0, and the location was along the gulf of Izmit.
The North Anatolian Fault is broken up into segments, separated by step-overs. In the east and central regions the segments are around 90 km long, but in the west the typical length is 30 km. And as mentioned, there is a notable difference in earthquake size. While in the east earthquakes are recorded as large as M8, in the western part the 2300 year record shows no event larger than M7.5.
This difference is related to the age. The fault first forms as individual segments, leading to many step-overs. Over time, the segments connect and the step-overs are smoothed over. The segment length now becomes longer. This process is compete in the east but still at an early stage in the west. Longer segments allow for larger earthquakes.
The marching quakes
The North Anatolian Fault came to life again in the 1930’s. The unlocking began in 1939 with an M7.8 event on the eastern side, where a large length of the fault failed. It was followed in quick succession by five further events, ranging from M6.9 to M7.3, which broke the adjacent regions. By 1951, almost the entire eastern half of the fault had failed, in a series of quakes which seemed to move westward along the North Anatolian fault. And it didn’t stop. In 1957 the next segment failed, and the following segment went in 1967. In 1992 the Erzincan earthquake filled a gap. Now things finally calmed down.
The picture here shows how the stress was resolved along the fault. The blue areas show where the stress went down, following an earthquake. The red areas show where the stress increased, in part because of the neighbouring event. This model from 1997 (Stein et al. Geophys. J. Int. 128, 594-604) shows how neighbouring regions had a significantly increased chance of rupture. The model ended with high stress along the eastern edge, in the Izmit region.
Two years later, in 1999, this eastern segment failed. There were two M7.2 and M7.4 quakes, exactly where the model showed enhanced stress. It devastated the city of Izmit. Since that time, the fault has been quiet. Some would say, eerily quiet.
The most destructive of the earthquake sequence were the 1939 event in Erzincan, and the 1999 quakes in Izmit. Together some 50,000 people died.
Izmit is not the end of the fault. It continues through the Marmara Sea and beyond. The region on the far side of the Marmara sea, along the Dardanelles Strait, had failed in 1912. That only leaves one stretch of the fault unbroken: the segment that runs underneath the Marmara Sea. There is one major city near this segment. The metropole of Istanbul lies just 20 km from the fault.
The chance of failure of this segment in the next 20 years has been put as high as 70%. (The number seems on the high side given the difficulty of predicting earthquakes.) Earthquakes on the western end of the faults are not as large as on the eastern side, as the rupture lengths are smaller. Whilst the eastern end can do events up to M8, the Marmara Sea is probably limited to M7.5. But this could still be devastating to Istanbul.
The Marmara Sea itself last ruptured in 1766, while the Izmit region failed in 1509 and 1719, before the 1999 event.
This plot shows the timing of major earthquakes in the Istanbul area, and it gives a stern warning. After an Izmit quake (black), the Istanbul area tends to follow within a few decades. We are now 2 decades after the 1999 Izmit earthquake. However, the 1894 quake is missing from this plot, and that was not followed by an earthquake in the Marmara Sea. There are exceptions, and this plot shows a risk, not a certainty. But it is a risk that needs to be taken seriously.
What does this mean for Istanbul? There is clearly an expectation of a major earthquake near it, but no certainty on either the location or the time. 20 years have passed since the Izmit quakes. Those will have added stress to the Marmara fault, but apparently 20 years ago it wasn’t yet ready to fail. Is it now?
The Marmara Sea remains as a gap in the earthquake sequence, and gaps have a tendency to become filled. How long is the unruptured segment? That depends on the 1912 earthquake. This M7.3 event happened on the western side of the Marmara Sea and ruptured 100 km from the Gulf of Saros to the Marmara Sea. But did the rupture extend into the Marmara Sea? Recent studies suggest it may have extended as far as the central basin of the sea. This would leave a 100 km gap between the 1912 and 1999 ruptures, sufficient for an M7+ earthquake but perhaps not M7.5. Istanbul is located midway along this gap.
The last major earthquake in the Marmara Sea was in 1766. This length of time is enough for 5 meters of slip to have accumulated. It has been suggested that the western part of the gap is not fully locked, and much of the slip is accommodated by small events, but the eastern part is locked – and primed.
How would such an earthquake affect Istanbul? The destruction in Izmit in 1999 does not bode well. Models suggests that an M7.2 earthquake on the main fault would severely damage 3% of the buildings in Istanbul.
In addition to the shaking, tsunamis are possible. Over the past 1900 years, 40 tsunamis have been recorded in the Marmara Sea. The earthquakes of 1509 and May 1766 caused tsunamis in Istanbul. But the August 1766 earthquake did not. This was located in the western basin, much further from the city. The 1509 event was reported to have caused the waves to top the city walls, 6 meters high. (But see below.)
A quake in the eastern Marmara Sea could cause a tsunami reaching the Istanbul shore within 5 to 10 minutes. Modeling suggest a height up to 2 meters, with little impact on the historic centre. An earthquake in the western part has little effect on Istanbul: the waves mainly affect the opposite, southern shore of the sea. The models suggests that there is a tsunami risk, but an extreme tsunami (more than 6 meter run-up height) is unlikely. But in that case, why the 1509 tsunami? This could have been the result of an underwater landslide in the Cinarcik basin, the deep basin alongside the North Anatolian Fault that lies opposite Istanbul. The modeling only took the movement of the fault itself into account. However, much more likely is that the event never happened. Although there was major flooding, no contemporary document states that the walls were overtopped. It is only found in later documents, and may have been an addition to the story, a bit of fake news. Not all scare stories are true.
Volcanoes on the Anatolian fault
This is a volcano blog, and I can’t leave Turkey without a comment on its volcanoes. Indeed, Turkey has some impressive volcanoes on its eastern border. Mount Ararat is particularly well known. But is any volcanic activity associated with the North Anatolian Fault?
There is, but we need to look a bit beyond the fault.
At the southeastern end, the North Anatolian Fault terminates at the Karliova triple point where Arabia, Eurasia and Anatolia meet. But earthquake activity shows that the fault continues a bit further southeast, beyond the triple point and on the Arabian plate. (At this point, it runs east of the East Anatolian fault which may be confusing). The activity peters out after about 50 km, towards Lake Van. There is not much surface expression here. The fault is known, and it called the Varto Fault Zone. In 1966, there was a M6.8 earthquake here. And here is where volcanoes begin.
Lake Van is the largest lake in the Armenian highlands (this is called the Eastern Turkey High Plateau in Turkey, but it extends across geographical borders) that make up the border region with Iraq and Armenia. It is also the largest lake in Turkey. Lake Van is both tectonic and volcanic. It is entirely tectonic: the lake originated as a subsidence basin. There is no outlet, so that water collects from streams but can’t escape. This causes an extreme environment: the water is alkaline with a pH approaching 10, with strongly fluctuating lake levels. But in the past there was an outlet. It was blocked by an eruption from Mount Nemrut, west of the lake. Thus, this tectonic lake also has a volcanic origin.
Look for Mount Nemrut, and it does not look impressive. Another volcano north of the lake, Mount Süphan, looks far more impressive, and is in fact the second tallest volcano in Turkey, after Ararat. But it plays no role here. Perhaps Süphan deserves its own post. Mount Nemrut is our villain. It rises as a flat-topped block about 1 km above the surrounding fields. (This is a highland, and the fields themselves are around a mile above sea level. Expect cold and snowy winters.) From below, it does not look particularly volcanic. To see the real thing, you need to look from above.
The flat top is in reality a large caldera which shows beautifully in satellite images. It is near-circular, and 7 km across. A large VEI-7 eruption from here (58 km3 of ignimbrites) formed a 60-km long lava flow, which dammed Lake Van. This plinian eruption collapsed the mountain and formed its 50 km2 circular caldera, 270,000 years ago. Nemrut remains active, with three reported eruptions (all small) over the past millennium, the most recent one in April 1692.
Mount Nemrut lies beyond the end of the Varto Fault Zone. there is no direct connection, even though the line of the fault points at it. But there is a second, extinct caldera along the fault, called Bingöl (or Varto) caldera, some 20 kilometers southeast of the triple point. This 10-km wide caldera is named after the town of that name. This town in a previous era was known under third name, as Karliova. And so we come back to the North Anatolian Fault, which begins at the Karliova triple junction. The fault not only shares the name of the start of its journey with a town, but also with its caldera. And it has taken more than just the name.
The 3 million year old Bingöl caldera is dissected by the Varto fault. The caldera sits north of the fault. But is only half a caldera: the Bingöl caldera is also known as a half-caldera. (It is more like 2/3rd, though.) Where is the other half (or third)? A match was found with the Turna mountains, 50 km away on the other side of the fault. This is another case of a volcano on the move. (It is another similarity with the San Andreas.) (I should say that this is not universally accepted. It has also been suggested that they are two different volcanoes which were active at the same time in each other’s vicinity.)
The Turna mountains, though, are not on the Varto fault. They are on the other side of the triple junction, on the North Anatolian Fault. And so not only did the North Anatolian Fault take the name of the town for its triple junction, it also took half its volcano. What a fault.
Albert, November 2020