The southern shores of Europe have seen their share of geological conflict. Africa is moving north, and Europe is in the way. There have been many skirmishes, which have left scars in southern Europe. Africa is an irresistible force and Europe an immovable object. The next phase of this battle will be at Cyprus where the African continental plate is approaching.
About 10 million years ago, Turkey began to be affected by the battle of the plates, albeit indirectly. It had started 30 million years ago when Africa lost Arabia. The Afar hot spot split the continent, Arabia separated, formed its own plate and began to move north while the Red Sea formed behind it. Arabia also rotated a bit anti-clockwise. (Younger readers may need to look up how a clock can rotate.) But it had too little room for these manoeuvres and Arabia soon collided with Asia. A transform fault formed along the Jordan Valley, known as the Dead Sea Transform, where the Arabian plate slid past the African plate. Because of the slight rotation, in the south the fault was widening and the valley dropped well below sea level.
The head-on collision of the Arabian plate with Eurasia was focussed on Iran, itself a fractious nation, and the eastern region of Turkey. This thrust threw up the Zagros mountains. In Turkey, this range is known as the Southeast Taurus mountains, but geologically it is an extension of the Zagros suture zone. Westward, the mountain range bends from northwest to southwest, following the outline of the Arabian plate. The southwest-oriented section joins up with the north-south Dead Sea Transform: this completes the border of the Arabian plate.
The southwest-oriented section is the East Anatolian Fault. This 500-kilometer long fault was only discovered in the 1970’s. The movement along the fault is complicated with both compression and transform motion. Turkey is being pushed north, but the Eurasian plate provides strong resistance and Turkey can’t go that way.
There is a way out. To the west is the fractured southern shore of Europe. And because Arabia is moving north faster than Africa is, the pressure in the east is greater than the resistance in the west. 11 million years ago, Turkey began to move west. A fault formed which split the region from the Eurasian plate, and the Anatolian (mini)plate had formed. This fault is the infamous North Anatolian Fault. It is still forming but is already responsible for many of the strongest earthquakes in Turkey. Turkey is also rotating a little, and this motion has created the Marmara sea, often seen as the dividing line between east and west.
The next large earthquake in Turkey had been expected to occur on the North Anatolian plate, possible near the Marmara sea. But instead the other major fault failed: the East Anatolian Fault. And its failure was quickly followed by another one, on a nearby fault that had not been considered as particularly dangerous. The first earthquake had been predicted, as a seismic gap between active segments. The second one came as a surprise.
The East Anatolian Fault consists of some 10 separate segments, separated by step-overs. There are many other faults in the region as well, some running parallel to the master fault, some going off on angles. There are fault fragments in many places and it can be hard to know which one is the real thing! Precise locations are also not always easy to determine. This region hasn’t yet organised itself into a well-designed fault system. The connection to the Dead Sea Transform Fault is particularly complex.
The East Anatolian fault is the main source of the devastation of last week which will haunt the region for years to come. In the M7.8 shock it failed over a length of 300 km, more than half its total length. The second quake ten hours later hit a westward fault (Sürgü-Cardak Fault – the names refer to two segments of this fault) which failed over its full length of 170 km. This fault was known but not well studied.
The map below (adapted from USGS) shows the known faults (red). It also shows the two sets of recent earthquakes (purple and pink), and historical earthquakes (red). I have indicated in black the faults that failed, and in dashed lines a very approximate indication how they might extend towards and into the Mediterranean Sea – assuming the sea is better organised than the land! The dashed lines should be taken with significant caution. The connection to the Dead Sea Transform Fault is also quite unclear – in fact there isn’t an obvious connection in this map.
Why were these faults not better known? The East Anatolian Fault was recognized but it had not moved in this region for 200 years. People had forgotten it was there. The other fault is just one of the jumble of separate faults and fault-lets in a mountainous region. The land here is deeply fractured.
A more detailed map was published 10 years ago. It shows a (geologically) broken land riddled with faults. In this map, the East Anatolian Fault runs on the western edge of the Karasu Trough, and the Yesemek Fault runs on its eastern edge. The latter is seen as an extension of the Dead Sea Transform fault. Instead of joining forces, the two faults are competing, with both potentially active. In the future a triple point may yet form in the Amik Basin (AM) but at the present time, the fault system near Gaziantep makes a disorganised impression. It is a mess.
Faults of failure
The Sentinel spacecraft from the European Space Agency provide monitoring of the world in a variety of ways. One of these involves radar to measure distance between ground and satellite. Each area of the world is mapped perhaps once a week. The strip of land covering the quake zone was mapped on the Frday following the quake. There were clear and present changes.
The satellite was moving roughly south to north, on a track to the east of the quake zone (and several hundred kilometers higher, obviously). While it is moving, it measures distance along the track and perpendicular to the track. Note that this is the distance to the satellite. When the satellite is straight overhead, changes would be up or down – that is what we want to know when a volcano is involved. When it is far away, changes are mainly horizontal. But in practice, we are seeing a combination of the two and it is difficult to know what movement is up-down and what is sideways. Do keep this in mind when seeing such radar (or INSAR) maps.
The INSAR data above shows the changes from the earthquakes. These are the changes along the direction of the orbit of the satellite, which is almost north to south. Red is towards the satellite, i.e. northward. Blue is the opposite. The Arabian plate shifted north by around 2 meters, as shown by the orange colours. This motion occurred between the East Anatolian fault and the Yesemek fault, roughly 30 km to the east, and a fault zone that continues northeast from here. The indicated epicentre of the M7.8 earthquake was near the Yesemek fault. The main event was however on the East Anatolian Fault, as shown by the Sentinel images. (There is another fault zone closer to Gantiazep which could even have been involved.) The movement was mainly within the region between the two main faults. Interestingly, the Anatolian plate on the other side of the East Anatolian fault seems to have moved a bit in the opposite direction (south or down), as shown by the light blue colour.
It is a bit of guess work on precisely what happened. The East Anatolian Fault was locked. The bulk of the Arabian plate is still moving, in a continuous creep northward, driven from below. This puts an enormous stress on the area close to the locked fault. Several meters of movement had build up. A bit of the stress was transferred to the Anatolian side of the fault which had been dragged a bit north. When the fault failed, the stressed zone suddenly moved north along the East Anatolian Fault. This also released the stress on the Anatolian side, which moved back south a little bit. Do note that this is guess work. It also ignores the role of the Yesemek fault, any up/down movement, and the depth profile of the fault.
Note that there is no north-south motion visible on the Sürgü-Cardak Fault.
The second plot shows the movement perpendicular to the orbit of the satellite, i.e. roughly west-east. Now the picture is very different. The orange colours show movement to the east. Eastward movement is seen in the Arabian plate but only north of the bend to the northeast. The plate shifted as much as 2 meters in this direction. The combined motion is about 3 meters.
Now a strong effect is seen on the Sürgü-Cardak Fault. North of the fault the colours show a large westward movement, and south of this fault there is a smaller eastward shift. The westward motion is that of the Anatolian plate which is accommodated by the North Anatolian Fault. In this region, it seems to be distributed over other faults, one of which failed. The reflex motion shows that this was a locked fault where the north side was pulling the reluctant southern side along. The cause lies in the bend in the East Anatolian Fault. South of here, the push is to the north. North of here, the bending also causes a push to the west. This gradient in the force direction has led to this fault developing.
Photos near the East Anatolian Fault line after the quake show offsets of around 2 meters, consistent with the numbers above. The image below is from Hassa, around 30 km northeast of Iskenderun. The second image, from the same general area, shows vertical offset as well.
Further south, several faults run either side of the Gulf of Iskenderun (also known as the Gulf of Iskanderoon, Gulf of Alexandretta, Gulf of Issus and in the past as Armenian Gulf – it is a bay of many names). This is part of the Levantine Sea. The gulf-of-many-names seems to be a small pull-apart basin, caused by the enforced westward movement of Turkey. The East Anatolian fault runs though the Karasu Valley, and has pushed up a small mountain range between it and the bay, called the Amanos mountains. They are also known by other names.) Iskenderun is on the east side of this basin: it sunk by a meter or so during the earthquake, perhaps due to the pull-apart effect. This explains the photos of sea water flooding the edge of the city. Iskenderun is known for industrial air and water pollution: it could do without this additional problem.
The older earthquake history of the region is not as widely known as it should be. There are comprehensive catalogues but these are not publicly available. A few of the largest events are well known even to wikipedia, but just glancing at the wikipedia page shows how incomplete this is. Otherwise the main information is in research papers that use part of the main catalogue. The most comprehensive list is from Sbeitani et al, 2005, in the Annals of Geophysics, in a paper on historical earthquakes in Syria.
The information originally comes from historical documents. They may have been written long after, can exaggerate things, and be vague on details such as precise locations, dates or even the year. Nowadays, news travels fast. In those days, it could take weeks, months or longer for the knowledge to spread. Of course, there was no talk of epicentres in those days. There are descriptions of damage in various cities, but we need to guess which fault was at fault from which cities are mentioned and how severe the damage was. (Name confusion is also common. And Tripoli in Lebanon was commonly confused with Tripoli in Libya, giving rise to some very strange earthquake stories.) The damage may be overstated, especially if the documents were written long after the events, and damage reports may have a local bias: because a city was not mentioned does not mean that there was no damage there. At the extreme end, it could just mean there were no survivors! Neither do the documents give the magnitude of the earthquake. Where a paper claims that an earthquake say in 1300 had a magnitude of 7.2, this number is approximated from the area over which damage was mentioned and should be taken with care.
This area of Turkey has an amazing history, as so vividly described in the previous post. The cities are ancient. The two main affected cities in the region, Gaziantep and Aleppo, are among the oldest continuously inhabited cities on Earth. We should be able to find out when they were badly damaged in the past.
Gaziantep is a large city with 2 million inhabitants. It also has a long history with a large variety of rulers. There was a large Armenian community here but they left (or were exterminated) in 1915. Old documents refer to the city as Hantab or later as Aintab. It is at risk most from the East Anatolian Fault, but also from the Dead Sea Transform and even from the distant Bitlis-Zagros Suture. But there is little direct information available on past destruction. Reading papers on the history, it almost seems like the city was mysteriously spared. We know that cannot be true, but complete destruction seems to have been avoided. There were many research papers on the expected impact that a large earthquake would have on the city (one as recent as 2022), exploring earthquakes with assumed magnitudes typically from 6.5 to 7. The relaxed attitude in building standards that was evident last week was done in both full knowledge and denial of the danger.
Aleppo also is a city with much history. It may be 8000 years old. The city was called Halab; the current name is of Italian origin. For Aleppo we do have an extensive record of earthquake damage including a hugely destructive event in 1138. It is closer to the Dead Sea Transform, so is at danger from two sides: in many cases the source of its shaking is not in Turkey. Documents about damage in Aleppo sometimes mention damage in Gaziantep as well: these are the events we are looking for.
Antakya (Antioch) is yet another old city. It is further south and is affected mainly by the Dead Sea Transform Fault, with shaking amplified by the old lake bed.
Looking for earthquakes which badly affected both cities, the two top candidate is the event of 13 August 1822. It devastated the region between Gaziantep and Aleppo, including both cities themselves. It is well documented. It started as a series of weak foreshocks in the week leading up to August 13. On the 13th there was a larger shock, followed 30 minutes later by the main event. Aleppo was in ruins: 2/3rd of the houses were destroyed. It lost the wall of the citadel. Lattakia and Iskenderun were badly damaged. Fissure were reported in the area. The death toll is estimated as between 20,000 and 60,000, including 7000 death in Aleppo. The earthquake was felt across the Middle East. This earthquake was either on the East Anatolian Fault or on the Yesemek Fault.
There was another large earthquake in the region in 1872. This seems to have been centred further south, in the Amik Basin.
Going further back, less is known. There was a damaging earthquake in 1513 or 1514 but it seems to have been focussed further west, perhaps on the southern extension of the Sürgü-Cardak Fault. This however is disputed.
An earthquake in November 1114 is often mentioned. There were two strong foreshocks, on 10 August (perhaps off-shore in the Gulf-of-many-names) and on 13 November closer to Antioch. The main shock followed on 29 November 1114. It may have been similar to last week’s events. This earthquake completely destroyed the ancient city of Marash (modern Kahramanramas) with damage as far away as Aleppo. Contemporary reports claim 40,000 dead in Marash. That number seems a bit excessive as the population is estimated at only 5,000! The ruler, Richard of Salerno (a crusader), evidently also died in the earthquake. The city was an important fortress, but disappeared from records for at least 10 years. The earthquake caused extensive damage over a wide east-west region but cities like Gaziantep and Aleppo were apparently less affected.
Evidence for ground slip from the 1114 event has been found along the same Sürgü-Cardak Fault that failed last week: this fault was therefore the likely culprit in 1114. This may have been the last time this particular fault failed! The last evidence for ground slip here which preceded 1114 was from the 3rd century.
(There were other similarities with 2023: ‘When the shocks ceased, snow began to fall and the country was buried under thick blanket.’)
There is more to be found in historical records, but the problem is always to assign magnitudes and to know exactly which fault was involved. The segment of the East Anatolian Fault north of Kahramanramas, where it bends more to the east, had been quiet for a long time – we don’t know how long. The section south of this city may have failed in 1822, however it is possible that this was the parallel fault. The quietness of this section had been noted, and after the sequence of 20th century earthquakes further up the East Anatolian Fault, marching along the fault, the warning flags were up. An earthquake northeast of Kahramanramas had been expected. That section indeed failed last week, but the earthquake also broke the southern segment, and it triggered the Sürgü-Cardak Fault. That was unexpected.
Based on the limited evidence, the section of the East Anatolian Fault closest to Gazantiap may fail once every 200 years in large events, the Sürgü Fault once per 900 years, and the northern end of the Dead Sea Transform Fault perhaps once every 500 years.
The first number is most uncertain because of the limited record and the two overlapping faults. It has also been argued that either of the southern East Anatolian Fault and the Yesemek Fault fail only once per thousand years, as the slip rate is divided between them. That seems optimistic, though. The recurrence time in more northern parts of the East Anatolian Fault is around 190 years, based on a long record of lake deposits deriving from large events with liquefaction. This central region last failed in 1893, in an event estimated at M7.3.
The 2023 earthquake was exceptional because a much longer section of the East Anatolian Fault failed. The fact that two separate faults went in close succession is common in the region. The succession often involves faults further south, in the complex region where the Dead Sea Transform diverges into fault segments going off in various directions. That didn’t happen this time. Looking at the historical record, after last week there is possibly an enhanced risk of a follow-up event along the northern Dead Sea Transform, over the next several years. But that could happen in offshoots of the East Anatolian fault as well.
The Karasu Trough
The main earthquake followed the Karasu trough, a deep, wide valley filled with sediment. The cause of this trough is still disputed, but the fact that it is bounded on both sides by major faults must have something to do with it! One explanation is that both faults are slightly deviating from each other, causing a basin to form in between. (This is called ‘transtensional’.) The Yesemek Fault on the east side of the trough is seen as an extension of the Dead Sea Transform, a fault that has formed several deep basins further south, including the one that gave it its name.
The Karasu trough is about 20 km wide and 150 km long. The western side is steeper, and therefore the East Anatolian Fault on the west appears more active than the Yesemek Fault on the east. The Amanos Mountains are on this western margin and rise to 2 km above the trough. The east margin has a height of up to 800 meters.
Having two such major but independent faults run side by side seems unusual. It is possible that the East Anatolian Fault is still developing and feeling its way south, and may eventually take over the show.
Both margins of the trough show evidence for volcanic activity, with most eruption centres on the western margin. They are mainly cinder cones and basalt flows. Most of the volcanic activity is dated to around 300,000 to 400,000 years ago, but two of the cones near the central region are only around 50,000 years old. The lava composition is quite uniform, basalt with little crustal contamination.
The volcanic activity has been put to good use in the region. The Yesemek fault is named after the town of that name, in the Hatay region. Yesemek town is known for an ancient quarry where large stone sculptures were made. This quarry dates to the Hittites, and is 8000 years old! It used the basalt stone on the side of the valley for the sculpture. The quarry is now a monument. More than 300 of these sculptures have been found here.
Predictions are best made in hindsight. Still, this section of the East Anatolian Fault was a known risk: it was a notable gap in the series of past earthquakes. A 2021 paper optimistically predicted that the strongest earthquake in the period 2020-2030 would likely not exceed M5.3. That was based on a time series analysis going back to 1900. But as James Jackson had noted for this region, already back in 2010: ‘The last 100 years are simply not representative for the longer period and give a completely misleading and unrepresentative view of the true earthquake hazard’. Other papers had predicted a significant earthquake in this region, and even mentioned the chance that the largest earthquake in Turkey for the 21st century would be on this section of the East Anatolian Fault.
I will quote Duman et al., Mediterranean Geoscience Reviews volume 2, pages 411–437 (2020):
Although, the individual fault recurrence intervals might be relatively long (e.g. 1000–3000 years), there are multiple–fault segments that can produce earthquakes at this rate (0.001–0.00033 per year). Therefore, the combination of activity rates from all of the low activity rate structures in the SMF system results in a significant rate of large magnitude earthquakes (0.007–0.0023 per year) in the Gulf of İskenderun region, which poses a significant hazard to infrastructure in the area.
The faults in this area of Turkey are strongly segmented and each segment can cause a strong M6 to low M7 quake. But several segments can fail together, causing a much larger event. This happened last week.
Both faults that failed will take a long time to rebuild the stress. Duman may be optimistic about the recurrence times, but for the East Anatolian Fault, it may be 200 to 300 years, or even much longer. For the Sürgü-Cardak Fault, it may indeed be a millennium. The same event will therefore not repeat any times soon. But the chance of earthquakes further west and south continues as before. There are many faults in this region. Even when each individual fault only has a small chance of failure, there are many other faults in this fractured country – some well known, others seemingly dormant. A fault that has been locked for centuries may be innocuous, or it may be ready to fail. The risk analysis should take the combination of all these faults and their individual segments into consideration.
Turkey has now had two major quakes with tens of thousands of fatalities in the last 25 years. Both of these earthquakes had been predicted. For Turkey’s faults, failure is not an option – it is a certainty. It will happen again. And again.
It is not only Turkey. There are other places in the world at risk of slow recurrent failure. Turkey has an ancient history, which helps in recognizing the dangers, although apparently knowing the danger was not enough. Other places may lack the long memory. One of my most sobering drives was along the San Andreas fault in southern California. Primed and ready to go – and still people build houses only meters from it. Humanity is like that.
Albert, February 2023