Tectonophysical Zoning of Seismogenic Faults in Eastern Anatolia and February 6, 2023 Kahramanmaraş Earthquakes

Abstract—The results of the tectonophysical reconstruction of stresses in the crust of Eastern Anatolia, obtained from the analysis of data on earthquake focal mechanisms, have shown that a significant restructuring of the stress state has occurred here in the last 20 years. It was largely confined to the southern and southwestern sectors of the region, covering hundreds of kilometers along the East Anatolian Fault. The data obtained from tectonophysical monitoring not only on the orientation of principal stresses, but also on their normalized values made it possible to calculate Coulomb stresses on faults. The results of fault zoning by intensity and sign of these stresses helped identify both hazardous sections close to the limit state and safe sections with negative Coulomb stress values. It has been established that in the region of the source of the first strong Pazarcık earthquake, which had a complex structure (three segments), there were extended sections with a critically high Coulomb stress level, separated by zones with low and even negative values of these stresses. The epicenter of this earthquake was located on the echelon fault within a section (first segment) with a high Coulomb stress level. The source of the second strong Elbistan earthquake was located on a fault with negative Coulomb stresses. The conducted analysis shows that this second Turkey earthquake may have been caused by stress changes that occurred in the crust of the region after the first strong earthquake. The research results show that Coulomb stresses in systems of closely located and differently oriented faults may be prone to sudden changes during the development of the earthquake on one of hazardous sections.

INTRODUCTION

The Kahramanmaraş earthquakes in Turkey occurred on February 6, 2023, on the East Anatolian Regional Fault System, which, with a strike from northeast to southwest, separates the Arabian and Anatolian microplates (Allen, 1969; Duman and Emre, 2013). In the middle, the East Anatolian Fault Zone (EAFZ) diverges into several main branches and extends westward and eastward in a “horse tail” pattern (Fig. 1). The first of the two strong Kahramanmaraş earthquakes, the Pazarcık earthquake with М_w= 7.8 (https://earthquake.usgs.gov), occurred at a depth of 10 km north of the city of Gaziantep and was about 350 km long. It started not on the main EAFZ branch, but to the south, on the echelon Narlı fault, then moved northeast along the main EAFZ branch (the Pazarcık fault) and southwest along one of the eastern EAFZ branches (the Amanos fault) starting near the town of Türkoğlu. Historical records show that this part of the EAFZ had a gap in strong earthquakes with magnitudes greater than 7.5 (Sunbul, 2019).

Fig. 1.

A map of faults in Eastern Anatolia that were active during the neotectonic period. The location of faults corresponds to the active faults of Eurasia database (Bachmanov et al., 2017). Faults: DTF, Deliler–Tecer; MF, Malatya; SüF, Sürgü; ÇaF, Çardak; SaF, Savrun; KF, Kozan; MiF, Misis; LF, Latakia; DSF, Dead Sea; IBF, Iskenderun Bay; KıF, Kırıkhan; AmF, Amanos; TüF, Türkoğlu; PF, Pazarcık; AhF, Ahirdag; AnF, Andırın; GöF, Göksun; KaF; Kayseri; OF, Ovacık; BF, Bingöl; BFF, Bozova Fay; EGF, Ergani Gungus; PüF, Pütürge; HLF, Hazar Lake; NF, Narlı; YeF, Yesemek. The gray filled bands correspond to the location of seismogenic fault sections for the two strongest Kahramanmaraş earthquakes.

The source of the second one, the Elbistan earthquake with М_w= 7.5 (https://earthquake.usgs.gov), which occurred 9 hours after the first one near the town of Elbistan, was located on the westernmost branch of the EAFZ, which runs along the Sürgü–Göksun–Misis fault system past the town of Çelikhan with a sublatitudinal strike from east to west.

According to the paper by (Reilinger et al., 2006), the lithospheric plate motion causes a sinistral strike slip along the EAFZ at a rate of 10 ± 1 mm/year. Studies cited in the paper by (Trifonov et al., 2018) show that the 12 km displacement of the Euphrates River since the mid–Quaternary is consistent with a geologic average slip rate of 12–15 mm/yr. These motions have caused the occurrence of many weak and frequent earthquakes in the upper crust down to depths of 20–25 km (Taymaz et al., 1991; 2021; Melgar et al., 2020), as well as a rather large number of destructive earthquakes in the last 2000 years. There are historical records of more than twenty M ≥ 6.1 earthquakes along the EAFZ from 1822 to 2023 alone (Sunbul, 2019; Chen et al., 2023).

The seismic hazard maps of Turkey (Fig. 2) (Giardini et al., 2018; Baltzopoulos et al., 2023) identify the EAFZ as the region of highest accelerations. At the same time, the sublatitudinal section of the EAFZ branch passing through the Sürgü fault corresponds to the predicted medium–level accelerations. This suggests that the Pazarcık earthquake is not just a trigger for the Elbistan earthquake, but the main factor that created the conditions for the second strong earthquake (Tikhotsky et al., 2023).

Fig. 2.

A seismic hazard map of Turkey in the form of expected maximum accelerations of the Earth’s surface in an earthquake. The stars are the epicenters of the two strongest earthquakes in Kahramanmaraş Province on February 6, 2023. The figure is taken from the paper by (Giardini et al., 2018) with additions.

The fact that there were no preconditions for such a strong earthquake on the section of the western EAFZ branch, which resulted from the analysis of previous earthquakes that occurred here, showed that the stress state of the Earth’s crust can change very quickly. Although less than half a year has passed since the Kahramanmaraş earthquakes, a large number of papers have been and are being published discussing the inhomogeneity of the crustal stress state in this region and the possible contribution of both gradual and rapid changes in the stress state to a greater seismic hazard in zones with complex fault geometry (Okuwaki et al., 2023; Kwiatek et al., 2023; Güvercin et al., 2023). The paper will further show that there have been significant changes in the stress state in Eastern Anatolia over the past 20–30 years. It can be assumed that it was these stress changes that prepared the ground for the strong Pazarcık earthquake, which was catastrophic in its consequences.

On the seismic hazard map of Turkey (Fig. 2), the zone of two strong earthquakes with a total length of more than 500 km occupies less than 5% of the length of all of the most seismically hazardous faults. Because the total length of hazardous faults in densely populated regions is large, it is impossible to effectively implement proactive preventive repair and construction measures to counteract the effects of earthquakes. It is necessary to develop methods to identify faults that will become more seismically hazardous within the next 3–5 years. Such zoning of faults by hazard level should be different from the methods used to predict strong earthquakes.

When predicting earthquakes, scientists study correlations between a set of earthquakes, not necessarily tied to a particular fault (Sobolev, 1993; Kossobokov, 2006; 2007; Zavyalov, 2006; Davis et al., 2012), that occur over fairly large areas (the extent of the predicted earthquake). This approach identifies all faults of the appropriate extent within the predicted earthquake zone as equally hazardous. In reality, different fault sections have different propensities for brittle fractures. It is quite obvious that the occurrence of an earthquake depends on the combination of a large number of factors leading to the appearance of a section on some fault where the level of shear stresses is close to the strength. This can happen due to either an increase in shear stresses (Reid, 1910) or a decrease in the fault strength (Richter, 1958), for example, due to an increase in fluid pressure in the fracture pore space of the fault zone or a decrease in compression normal to the fault.

Our studies (Rebetsky and Kuzikov, 2016; Rebetsky, 2018; Rebetsky, 2018; Rebetsky and Guo, 2021) show that such fault sections, whose state is close to the limit state, can be located discontinuously along the fault. They are sometimes separated by fault sections where shear stresses are substantially lower than the fault strength—a stable state, which can be considered as a barrier to the development of a brittle fracture—a barrier. Because of these barriers, any of the limit state sections can generate an earthquake with a magnitude that is not the largest for the region (e.g., M_w< 6.0). However, in some cases, the combination of parameters of such sections (their proximity to each other, their individual lengths, the level of relieved stresses) causes an earthquake occurring in any section with a length corresponding to magnitude 6.0 not to stop at such a barrier, but to overcome the fault section in a stable state. Thus, limit state sections can be combined to form a strong earthquake with a length and magnitude greater than those of the earthquakes that normally occur on the fault.

The above summary of tectonophysical studies of hazardous faults agrees well with how the sources of the two strong Turkey earthquakes evolved. According to data available on the U.S. Geological Survey website (https://earthquake.usgs.gov), the first of them—Pazarcık earthquake—did not start on the main EAFZ branch, but on the adjacent echelon fault about 50 km long (segment 1 in Fig. 3a). Probably, the stress changes it caused on the main EAFZ branch were the trigger, and instead of an earthquake with a magnitude of about 7, there was an earthquake with a strength of almost an order of magnitude greater in terms of released energy.

Fig. 3.

The seismogenerating segments of the sources of the strongest Kahramanmaraş Province earthquakes which occurred on February 6, 2023 and the coseismic displacements at the points of permanent GPS observations from data of the U.S. Geological Survey (https://earthquake.usgs.gov). The images are taken from the GS USA website (June 1, 2023).

The strikes of the seismogenerating faults of the first and third segments were close to each other, but differed from the strike of the second segment by an angle of about 45° (Okuwaki et al., 2023; Kwiatek et al., 2023), which may contribute to instability on previously stable faults. There are horizontal sinistral strike–slip components along all the three segments of the source. The largest surface shocks observed during this earthquake were located along the third seismogenerating segment of its source (https://earthquake.usgs.gov). In the Hatay Province of Turkey, where the southern, end part of this source segment was located, in its eastern wing, near the village of Tepehan (36.158299 N, 36.230053 E), in the region of an olive grove, a huge 30–meter–deep downwarp appeared, stretching for several kilometers. Based on data of remote sensing of the Earth’s surface, the paper by (Taftsoglou et al., 2023) identified sections with indications of ground liquefaction, confined to the end zones of the seismogenic fault and sections with an abrupt change in its strike. All of this indicates large–scale changes in the stress state.

Three seismogenerating segments were also found in the source of the second, Elbistan, earthquake (https://earthquake.usgs.gov), located on the Sürgü, Çardak, and Savrun faults of the westernmost EAFZ branch (Barbot et al., 2023). The epicenter of this earthquake was located in the middle of the source—segment 1 (see Fig. 3b). However, the first two segments of the source were located on a fault section with a strike that is significantly different from the strike of the eastern EAFZ branch (about 60°). The third seismogenerating segment of the source of this earthquake was located on a fault that branches off the sublatitudinal Çardak fault to the northeast, i.e., with a strike nearly parallel to the main EAFZ branch. Thus, the source of the Elbistan earthquake was formed in the western wing of the northern—end section of the second seismogenerating segment of the Pazarcık earthquake source, which could have contributed to the change in the stress state on the closely located Sürgü and Çardak faults.

The strongest earthquakes that have occurred in Turkey over the last century have raised a number of questions: 1). The first segment of the Pazarcık earthquake source was not a direct continuation of the other two segments. How did the first segment relate to the other two? Could an earthquake of the same magnitude have occurred if it had not started from the first segment? 2). Would the Elbistan earthquake have occurred if the Pazarcık earthquake had not occurred? 3). Could it have been predicted that an earthquake of such an energetic magnitude was being prepared here? 4). Were there any data that might have indicated the hazard of specific fault zones in the region of the future sources of the strongest earthquakes?

SEISMIC HAZARD PREVENTION ISSUES

From a geomechanical point of view, a fault that regularly generates seismic events is near the stress limit state. There are different approaches to explaining how a strong earthquake can occur on such a fault, with a level of released energy an order of magnitude higher than in previous earthquakes that have occurred here over a long period of time (Rice, 1982; Kocharyan, 2017). They attribute such a transition from moderately strong and weak earthquakes to very strong earthquakes to the specific characteristics of a slow slip on faults resulting from an elastic–brittle–plastic deformation of the geomedium. So far, these geomechanical theories that are based on the study of the limit stress state in the crust do not provide us with guidelines that would help formulate the concept of deterministic prediction of when a strong earthquake will occur.

On the other hand, the geomechanical criteria for a brittle fracture themselves tell us the difference in the stress state between a weak earthquake and a strong earthquake—it is the length of the limit stress state section. Of course, everything is not as simple as it was thought 50–70 years ago, when a region at risk for a strong earthquake was understood to be a fault zone with a high level of stress (Gzovsky and Belousov, 1954; Gzovsky, 1954; 1956; 1957; Dobrovolsky, 1991). It turned out that not only the level of shear stresses shifting the fault sides is important, but also the level of frictional force on it, which increases at a high level of stresses normal to the fault (Rebetsky, 2003). Therefore, it is more correct to speak not of a high stress level, but of a stress limit state with a high level of shear stresses relative to the normal stresses that determine the level of friction on the fault. Their relationship is characterized by the value of Coulomb stresses.

As a result of all this, general considerations suggest that for a deterministic prediction, it is necessary to zone faults according to how close some of their sections are to the stress limit state in terms of the Coulomb stress level and then infer the magnitudes of the earthquakes being prepared according to the length of such sections.

Obviously, this problem is quite different from the one considered in the theory of strong earthquake prediction. This problem cannot be solved by purely statistical and mathematical methods; it requires the use of either geomechanics—mathematical modeling of stresses in the crust and on faults (Garagash, 2006; Makarov et al., 2007; Stefanov and Bakeev, 2015), or tectonophysics—study of stresses directly in a natural object (Rebetsky, 2003). The paper will further discuss the issues of zoning hazardous faults under tectonophysical monitoring of the stress state of the Earth’s crust.

At present, the primary data for seismic risk assessment of areas are data on the strength (energy level) and the location of epicenters of past earthquakes, which can be obtained both by instrumental seismological methods and by seismotectonic methods based on geological, engineering and geological, historical and archaeological, dendrochronological and radiocarbon dating principles of paleoseismogenic deformations. These prediction methods use a principle similar to that of actualism, only with a change in the time sequence of the actual data and the predicted event. The possible magnitude of a future strong earthquake of a crustal section is determined by the magnitude of the strongest earthquake that has previously occurred there.

In this case, there are two main approaches to identifying the hazard of strong earthquakes:

1. Develo** seismic zoning maps based on the analysis of data on surface seismotectonic manifestations in areas of the past strongest earthquakes in the region (paleo, historic and modern).

Under this approach, maps of the Earth’s surface shaking intensity in points (expected accelerations) are created, which serve as a basis for determining special design requirements and load increase factors used in the design of industrial and civil facilities in accordance with the Construction Norms and Regulations (SNiP). Periodic revisions of such maps (once every 10–30 years) lead to a gradual increase in the red area on them, characterizing the most hazardous sections of regions and seismogenic faults. This is due to an increased modern instrumental observation period and also to new information on strong paleoearthquakes of the distant past obtained by seismotectonics from field observations. These new data allow us not only to detect the strongest earthquakes that have occurred in the study region, but also to establish/clarify the recurrence period of strong earthquakes. However, seismic zoning maps lack information that could be used to determine which epoch, of a greater or lesser seismic hazard, the current period belongs to. At the same time, there is evidence showing a significant fluctuation of the seismic process in time (Gusev and Shumilina, 2004).

It should be noted that the same principle of actualism also serves as the basis for the system of predicting the locations of catastrophic strong earthquakes for the northern flank of the Pacific subduction region developed by Academician S.A. Fedotov (Fedotov, 1968; 2005; Fedotov and Solomatin, 2017). This approach essentially assumes that the accumulated elastic energy is approximately uniformly distributed along the plate boundaries and can be most effectively released only by strong earthquakes. In accordance with this concept, regions of seismic gaps of the first kind are distinguished, where earthquakes with a magnitude greater than 7.7 have not been observed for a long time. A strong earthquake has been predicted near Avacha Bay since the late 1970s. In the last 40+ years there have been many earthquakes of similar and even greater magnitude in other regions of the Kuril Islands and in Japan (Kunashiri in 1994, Tokachi–Oki in 2003, Middle Kurils in 2007 and 2008, Tohoku in 2011), but the prediction of the Avacha earthquake has not come true.

2. Predicting the location, time, and magnitude of a future strong earthquake from instrumental earthquake data and/or from various precursor data.

Three stages of prediction are considered: the long–term stage, which determines the location and strength of the earthquake; the medium–term stage, which specifies the location and strength and determines the time of the earthquake—weeks, months; and the short–term stage, which determines the time of the event—hours, days. Under this approach, there are statistically significant omissions of strong earthquakes, false—unrealized predictions, and possibilities to remove a prediction if an earthquake does not occur for a long time. This approach has not been widely adopted in practice by the government to ensure the safety of people’s lives and the preservation of industrial and civil facilities in seismically active zones. At the same time, in some regions, there is already an established system of interaction between seismological groups and local authorities in the form of regular reports on observed seismic activity and risks of its increase (Kamchatka, Baikal). The Institute of Earthquake Prediction Theory and Mathematical Geophysics RAS (IEPT RAS) has developed and long been operating an algorithm for the global prediction of strong earthquakes with magnitudes greater than 7.5 (Ismail-Zadeh and Kossobokov, 2020).

It is important to note that seismic zoning maps (Ulomov, 1999; Ulomov and Shumilina, 1999) are based on seismogenic faults where the strongest earthquakes were observed. At the same time, in earthquake predictions, as implemented in almost all approaches, faults are not of fundamental importance, since the main source of data is low and moderate intensity earthquakes, which can occur over a rather large area and are not tied to a single system of faults. This system of analysis essentially determines that it is not the stress limit state of the fault section that is observed, but rather large volumes of the Earth’s crust are identified as having reached the energy saturation limit. As a result of these predictions, their authors are sometimes not even able to say for sure on which fault the earthquake will occur. Therefore, even with a long–term prediction of a strong earthquake, it is impossible to identify in advance the zones that will be subjected to high intensity shaking—high accelerations.

ON POSSIBILITIES OF ESTIMATING COULOMB STRESSES ON FAULTS IN THE EARTH’S CRUST

In geomechanics (Nikolaevsky, 2010a; 2010b; 2012), Coulomb stresses (${{{{\tau }}}_{C}}$) are defined as the difference between shear stresses (${{{{\tau }}}_{n}}$) acting on the fault and frictional stresses caused by stresses normal to the fault (${{\sigma }_{{nn}}}$), given the softening effect of fluid pressure in fractures and pores (${{p}_{{fl}}}$). The increase in Coulomb stresses to positive values indicates that the stress state is approaching the critical state, which can lead to a brittle fracture of rocks:

$$\begin{gathered} {{\tau }_{C}} = {{\tau }_{n}} + {{k}_{f}}\sigma _{{nn}}^{*} \leqslant {{\tau }_{f}}, \\ \sigma _{{nn}}^{*} = {{\sigma }_{{nn}}} + {{p}_{{fl}}} < 0, \\ \end{gathered} $$

(1)

where ${{k}_{f}} = {\text{tan}}\varphi$ is the coefficient of internal friction (φ is the angle of internal friction), ${{\tau }_{f}}$ is the strength of internal bonding, and $\sigma _{{nn}}^{*}$ is the effective normal compressive stress reduced by the fluid pressure. The equality in formula (1) is satisfied for shear fractures with the maximum strength (${{\tau }_{f}} - {{k}_{f}}\sigma _{{nn}}^{*}$). Hereinafter, the sign convention adopted in continuum mechanics is used—extension stress is positive.

Experiments on samples show the dependence of ${{k}_{f}}$ and ${{{{\tau }}}_{f}}$ on stresses. In tectonophysics (Rebetsky, 2003a), due to the large scale of stress averaging and the impossibility of obtaining real data on the strength of large massifs of fractured rocks, the dependence of the strength parameters on stress is ignored, ${{k}_{f}}$ and ${{{{\tau }}}_{f}}$ are assumed to be constant, and the coefficient of internal friction is considered to be equal to the coefficient of friction at faults (Fig. 4).

Fig. 4.

A graphical scheme for determining Coulomb stresses on Mohr’s diagram with Mohr’s large circle tangent to the outer envelope as a straight line of the strength limit—limit state. The vertical lines are shear stresses and the horizontal lines are effective normal stresses (negative values of normal stresses are plotted to the right). The dotted line is the minimum frictional strength for the faults, $\sigma _{i}^{*}$ (i = 1, 2, 3) are the effective principal stresses, the gray filled areas show the normal and shear stresses on an arbitrarily oriented plane for a given stress state. The curly bracket shows the value of Coulomb stresses for stress state A on the fault.

Following the work by (Stein et al., 1992; King et al., 1994; Harris et al., 1995), the zoning of fault sections according to the Coulomb stress increments $\Delta {{\tau }_{C}}$ produced by changes in the stress level as a result of a strong earthquake that occurred in the region under study began to be used to identify zones of higher seismic activity. Such calculations were typically based on an elastic shear failure model corresponding to the source of the earthquake. Initially, analytical or numerical calculations were performed for areas with the geodynamic regime of the horizontal shear stress state—the San Andreas Fault region (USA) (King et al., 1994; Mallman and Zoback, 2007), later this approach was applied to crustal sections with other regimes (horizontal compression or extension) (Ganas et al., 2006).

$$\begin{gathered} \Delta {{\tau }_{C}} = \Delta {{\tau }_{n}} + {{k}_{f}}\Delta \sigma _{{nn}}^{*}, \\ \Delta \sigma _{{nn}}^{*} = \Delta {{\sigma }_{{nn}}} + \Delta {{p}_{{fl}}} < 0. \\ \end{gathered} $$

(2)

Where $\Delta {{\tau }_{n}}$ and $\Delta {{\sigma }_{{nn}}}$are the increments of shear and normal stresses on the fault, and $\Delta {{p}_{{fl}}}$ is the increment of fluid pressure. Positive values of $\Delta {{\tau }_{n}}$ and $\Delta {{p}_{{fl}}}$correspond to an increase in the level of shear stresses on the fault plane and fluid pressure in the fracture pore space of rocks, while negative values of $\Delta {{\sigma }_{{nn}}}$ correspond to an increase in the level of compressive stresses normal to the fault.

To calculate the changes in Coulomb stresses, it is necessary to formulate the boundary conditions for the loading of the crustal section under study and to specify the structure (faults) and properties of the medium. For shift zones (geol.), it is common to specify the direction of action of the maximum horizontal compressive stress, its ratio to the minimum horizontal compressive stress, and the coefficient of friction on the seismogenic fault.

It should be noted that in the shear fracture problem (mech.), the medium is an elastic single–phase solid; therefore, it is impossible to calculate fluid pressure changes. For this reason, in calculations of Coulomb stress increments, it is essentially assumed that the fluid pressure in rocks does not change before and after a strong earthquake,$\Delta {{p}_{{fl}}} = 0$,which is an oversimplification of the real natural process.

The approach proposed in the paper by (Stein et al., 1992) was initially used to explain why aftershocks had a higher density in crustal sections for which positive values of Coulomb stress increments were obtained. Recently, the same approach has also been used to explain the migration of strong earthquakes along fault zones (Pang, 2022). It should be noted that for the EAFZ, Coulomb stress changes were investigated in the papers by (Sunbul, 2019; Chen et al., 2023) before and after the Kahramanmaraş earthquakes, respectively, using data from a series of about 20 historical earthquakes.

Concurrent with this approach developed in seismology, the ratio of shear stresses on faults to the modulus of effective normal stresses began to be used in petroleum geophysics (Morris et al., 1996):

$${{T}_{d}} = \frac{{{{\tau }_{n}}}}{{\left| {\Delta \sigma _{{nn}}^{*}} \right|}},$$

(3)

which was defined as the slip tendency of faults, since the growth of the values ${{T}_{d}}$ really corresponds to the increase in the level of Coulomb stresses (1). Later, the parameter ${{T}_{d}}~$ began to be associated with the ability of rocks to exhibit dilatancy (Ferrill and Morris, 2003).

The calculation of the parameter ${{T}_{{\text{d}}}}~$is now part of the IAEA standard for assessing the hazard of faults found at nuclear power plant construction sites. An example of this is the Hanhikivi-1 nuclear power plant in Finland, where such work was carried out in 2018–2019. In this case, in–situ stress measurements obtained in mining (Kivinen and Varis, 2009; Pohjatekniikka, 2018a; 2018b; 2018c) were used as input data.

The papers by (Rebetsky and Kuzikov, 2016; Rebetsky et al., 2021) show the possibility of zoning hazardous fault sections based on the results of tectonophysical reconstruction of stresses in a natural object. The data source for stress reconstruction is seismological or geological indicators of fault deformations. These data should be supplemented with information on the location and geometry of faults that have been found by geological methods to be active in the neotectonic stage. This set of initial data allows us to calculate Coulomb stresses on the fault surface, the absolute or relative values of which can be used to zone fault sections with different hazard levels. This approach to seismic risk assessment should be considered deterministic, which allows monitoring the state of active faults and assigning different magnitude levels of expected strong earthquakes to their different sections. In this paper, the tectonophysical approach was used to zone faults according to their hazard level.

TECTONOPHYSICAL APPROACH TO ESTIMATING COULOMB STRESSES

As can be seen in the analysis above, the calculation of Coulomb stresses requires data on the shear (${{\tau }_{n}}$) and normal (${{\sigma }_{{nn}}}$) stresses on fault sections. To obtain these stresses, it is not enough to have data on the orientation of the principal stress axes (${{{{\sigma }}}_{i}}$, i = 1, 2, 3) and the ratio of their values characterized by the Lode–Nadai coefficient (${{\mu }_{\sigma }}$). In addition, it is necessary to know the maximum shear stresses $\tau$ and the confining tectonic pressure $p$ in order to determine stresses on fault sections:

$$\begin{gathered} {{\tau }_{n}} = \tau ~{{{\tilde {\tau }}}_{n}}\,\,\,\,{\text{at}}\,\,\,\,{{{\tilde {\tau }}}_{n}} = ~\,\,\left( {1 - \frac{{{{\mu }_{\sigma }}}}{3}} \right){{l}_{{n1}}}{{l}_{{s1}}} \\ + \,\,\left( {\frac{{2{{\mu }_{\sigma }}}}{3}} \right){{l}_{{n2}}}{{l}_{{s2}}} - \left( {1 + \frac{{{{\mu }_{\sigma }}}}{3}} \right){{l}_{{n3}}}{{l}_{{n3}}}, \\ {{\sigma }_{{nn}}} = - p + ~\tau ~{{{\tilde {\sigma }}}_{{nn}}}\,\,\,\,{\text{at}}\,\,\,{{{\tilde {\sigma }}}_{{nn}}} = \left( {1 - \frac{{{{\mu }_{\sigma }}}}{3}} \right){{l}_{{n1}}}{{l}_{{n1}}} \\ + \,\,\left( {\frac{{2{{\mu }_{\sigma }}}}{3}} \right){{l}_{{n2}}}{{l}_{{n2}}} - \left( {1 + \frac{{{{\mu }_{\sigma }}}}{3}} \right){{l}_{{n3}}}{{l}_{{n3}}}, \\ \end{gathered} $$

(4)

where ${{l}_{{ji}}}$ (j = n, s) are the direction cosines of the principal stresses (${{\sigma }_{i}}$) with the normal to the fault plane (n) and the direction of action of the shear stresses (s) on it.

It is clear from equation (1) that in order to determine Coulomb stresses, it is also necessary to be able to estimate the fluid pressure ${{p}_{{fl}}}$ acting in the fracture pore space of rocks.

There is a wide range of tectonophysical methods for determining the directions of principal stresses (Gzovsky, 1954; 1956; Nikolaev, 1977; 1991; Parfenov, 1981; 1984; Rastsvetaev, 1982; 1987; Gintov and Isai, 1984; Sim, 1996; Seminsky, 2003; Anderson, 1951; Arthaud, 1969; Alexandrowsky, 1985; etc.). There is also a large group of methods (Gushchenko, 1975; 1979; Nikitin and Young, 1976; Carey and Bruneier, 1974; Angelier, 1975; 1990; Yunga, 1979; 1990; Gephart and Forsyth, 1984; Michael, 1984; Lisle, 1987; 1992; etc.) in which, in addition to the direction of action of the principal stress axes, the shape of the stress ellipsoid characterized by the values of the Lode–Nadai coefficient is determined. Obviously, the data from these methods are not enough to estimate the normal and shear stresses on a fault.

There are only three tectonophysical methods created in the papers by (Reches, 1978; 1983; 1987; Angelier, 1989; Rebetsky, 1999; 2003), which developed algorithms for the second stage of stress reconstruction, allowing us to calculate the values of the maximum shear stresses and the confining pressure. These algorithms are based on the analysis of a homogeneous sample of geological or seismological indicators of fault deformations (fractures in the form of slickensides and earthquake source mechanisms) on Mohr’s diagram (Fig. 5). In the method of cataclastic analysis of discontinuous displacements (Rebetsky, 2003), the normalized maximum shear stresses and the effective pressure are determined under the assumption that the limit state has been reached in seismogenic zones (tangency of Mohr’s large circle of the massif strength limit line),

$$\begin{gathered} {{\tau }_{n}} + {{k}_{f}}\sigma _{{nn}}^{*} = {{\tau }_{f}}\,\,{\text{at}}\,\,{{\tau }_{n}} = \tau ~\tilde {\tau }_{n}^{F}\,\,\,{\text{and}} \\ \sigma _{{nn}}^{*} = - p{\text{*}} + \tau ~\tilde {\sigma }_{{nn}}^{F}, \\ \end{gathered} $$

(5)

Fig. 5.

Mohr’s diagram with data on normalized stresses in the sources of the earthquakes (white circles) whose focal mechanisms were used to determine the parameters of the stress ellipsoid. Point K corresponds to the earthquake with the minimum strength on the fault, determined only by the frictional stresses (${{k}_{f}}\sigma _{{nn}}^{*}$). The pentagons are the stress states of the faults for which Coulomb stresses are negative.

and determination of zero bonding strength for the earthquake source with minimal resistance of frictional forces

$$\begin{gathered} {{\tau }_{n}} + {{k}_{f}}\sigma _{{nn}}^{*} = 0\,\,\,\,{\text{at}}\,\,\,\,{{\tau }_{n}} = \tau ~\tilde {\tau }_{n}^{K}\,\,\,\,{\text{and }} \\ \sigma _{{nn}}^{*} = - p{\text{*}} + \tau ~\tilde {\sigma }_{{nn}}^{K}. \\ \end{gathered} $$

(6)

From formulas (5) and (6), using (4) for points B and K in Mohr’s diagram (Fig. 5), we can obtain:

$$\begin{gathered} \left\langle {\frac{\tau }{{{{\tau }_{f}}}}} \right\rangle = \frac{1}{{{\text{cosec}}2{{\alpha }_{f}} - \left( {\tilde {\tau }_{n}^{K} + {{k}_{f}}\tilde {\sigma }_{{nn}}^{K}} \right)}}; \\ \left\langle {\frac{{p{\text{*}}}}{{{{\tau }_{f}}}}} \right\rangle = \frac{{\left( {\tilde {\tau }_{n}^{K} + {{k}_{f}}\tilde {\sigma }_{{nn}}^{K}} \right) - {{{{k}_{f}}{{\mu }_{\sigma }}} \mathord{\left/ {\vphantom {{{{k}_{f}}{{\mu }_{\sigma }}} 3}} \right. \kern-0em} 3}}}{{{{k}_{f}}\left[ {{\text{cosec}}2{{\alpha }_{f}} - \left( {\tilde {\tau }_{n}^{K} + {{k}_{f}}\tilde {\sigma }_{{nn}}^{K}} \right)} \right]}}, \\ \end{gathered} $$

(7)

which determine two very important components of the stress tensor—the maximum shear stress and the effective pressure—with an accuracy up to normalization to the value of the massif bonding strength ${{\tau }_{f}}~$.

Since the last parameter should correspond to the strength of the massif on the scale of stress averaging in the process of tectonophysical reconstruction (for geological data—hundreds of meters, for seismological data with earthquakes in the magnitude range from 3 to 6—10–100 km), it is impossible to determine it in a laboratory experiment. With the type of data in (7), the limit relation (1) can be rewritten for normalized stresses:

$$\left\langle {\frac{{{{\tau }_{C}}}}{{{{\tau }_{f}}}}} \right\rangle = \left\langle {\frac{\tau }{{{{\tau }_{f}}}}} \right\rangle {{\tilde {\tau }}_{n}} + {{k}_{f}}\left( { - \left\langle {\frac{{p{\text{*}}}}{{{{\tau }_{f}}}}} \right\rangle + \left\langle {\frac{\tau }{{{{\tau }_{f}}}}} \right\rangle ~{{{\tilde {\sigma }}}_{{nn}}}} \right) \leqslant 1,$$

(8)

where ${{\tilde {\tau }}_{n}}$ and $~{{\tilde {\sigma }}_{{nn}}}$ are the reduced stresses containing data on the axes of the principal stresses and the shape of the stress ellipsoid (${{\mu }_{\sigma }}$) determined from the results of the first stage of stress reconstruction. Thus, the parameters in the angle brackets in the right part of equation (8) are determined from the results of the second stage of reconstruction.

The stress states on the fault sections where Coulomb stresses are positive fall in the band between the line of maximum strength limit of the fractured massif and the line of dry friction (Fig. 5). If the local strength of the fault section $\tau _{f}^{i}$is less than the maximum strength of the massif ${{{{\tau }}}_{f}}$, the limit state for normalized Coulomb stresses ${{{{\tilde {\tau }}}}_{C}}$ < 1 may be reached on it. For those fault sections where normalized Coulomb stresses are negative, their stress states lie below the line of dry friction in Mohr’s diagram (Fig. 5).

Thus, those tectonophysical methods that have developed algorithms for finding the values or normalized values of the maximum shear stresses and the effective confining pressure make it possible to perform fault zoning by Coulomb stress level.

It should be noted that in the cataclastic method (Rebetsky, 2003a), as part of the third stage of reconstruction, algorithms have been developed to evaluate the calculation of the bonding strength ${{\tau }_{f}}$ by using additional data in the form of the values of shear stresses relieved in the source of the earthquake with the largest magnitude in the region under study. Based on this algorithm, it has been shown that the bonding strength of rocks is 0.5–1.5 MPa in the crust (50–100 km) of active continental margins (Rebetsky and Marinin, 2006; Rebetsky, 2009) and 5 MPa in the crust of intracontinental orogens (Rebetsky, 2015).

TECTONOPHYSICAL RECONSTRUCTION OF NATURAL STRESSES IN EASTERN ANATOLIA

Immediately after the catastrophic and strong earthquakes that occurred in Eastern Anatolia (literally already on February 9), a tectonophysical stress inversion was performed based on seismological data on crustal earthquake focal mechanisms. It was possible to do this work so quickly because the catalog of earthquake focal mechanisms for this area was up to date at the moment: we completed it by the end of 2022 as part of the Russian Science Foundation’s project 22-27-00591 “Development of Methods for Tectonophysical Zoning of Active Faults in the Earth’s Crust.” Another important factor that facilitated this work was the fact that researchers from the Geological Institute of the Russian Academy of Sciences uploaded the “Active Faults of Eurasia Database” online (http://neotec.ginras.ru/database.html) (Bachmanov et al., 2017) in 2020.

The catalog of earthquake focal mechanisms for the 33°–40° E and 33°–40° N (Eastern Anatolia) section consisted of 223 events with the magnitudes Mb ranging from 3.0 to 6.8 that occurred at depths down to 56 km between 1951 and January 2023 (Fig. 6). This catalog was a compilation of earthquake data from catalogs of international seismological databases—GLOBAL CMT (http://www.globalcmt.org), ISC (http://www.isc.ac.uk), ESMC (www.emsc.eu), as well as regional databases—KANDiLLi Observatory and Earthquake Research Institute (http:// www.koeri.boun.edu.tr) and Russian Unified Geophysical Service (http://www.admobninsk.ru).

Fig. 6.

Earthquake focal mechanisms of the compiled catalog: 1, ${{M}_{w}}$ ≥ 5; 2, ${{M}_{w}} <$ 5. Hereinafter, the location of faults is shown according to the active faults of Eurasia database (Bachmanov et al., 2017).

Stress reconstruction for the area significantly larger than the region of the two strongest earthquakes in Kahramanmaraş Province was performed from the data of the compiled catalog of earthquake focal mechanisms using the STRESSseism program, whose algorithm is based on the cataclastic method (Rebetsky, 1999; 2003). The node spacing of the computational grid was 0.2 degrees of latitude, which was several times smaller than the scale of stress averaging. In the final version of the calculations, the orientation of the principal axes of the stress tensor and the value of the Lode–Nadai coefficient, which determines the shape of the stress ellipsoid—the first stage of the reconstruction—were obtained for 715 nodes.

The algorithm for calculating stresses in the nodes of the computational grid involves the creation of homogeneous samples of earthquakes, the data on the mechanisms of which are not mutually contradictory. The criterion for homogeneity of sampled mechanisms is based on the requirements of elastic energy reduction as a result of the earthquake on the calculated parameters of the stress tensor (the acute angle between the displacement vector on the seismogenic fault and the shear stresses for the desired stress tensor), as well as on the order of the relieved elastic deformations in the directions of action of the principal stresses (Rebetsky, 2003a; 2003b). These criteria result in a system of inequalities that impose certain constraints on the values of relieved elastic deformations in the directions of action of the principal axes of the stress tensor. These criteria are more physical and impose “milder” constraints on the relationship between the parameters of fracture systems and tectonic stresses than the effects of the Wallace (Wallace, 1951) and Bott (Bott, 1959) hypothesis that the displacement vectors and the shear stress coincide on the fault plane.

The concept that elastic energy decreases after a displacement along the fault in the form of corresponding inequalities that determine the decrease of elastic shortening deformations in the direction of the highest compressive stresses and the increase of these deformations in the direction of the lowest compressive stresses is the basis of the methods by J. Angelier (Angelier, 1975) and Gushchenko (Gushchenko, 1975). The concept of the order of relieved elastic deformations is the basis of the cataclastic method (Rebetsky, 1999, 2003a; 2003b) and is an extension of the principles of plasticity theory (Chernykh, 1988) to the pseudoplastic—cataclastic behavior of rocks resulting from the fracture flow. Inequalities following from this concept include the inequalities of the methods by J. Angelier and O.I. Gushchenko. They also include an additional inequality derived from the constraint for relieved elastic deformations in the direction of the intermediate principal stress.

The practical application of the STRESSseism program has shown that the system of inequalities used in it, which limits the orientation of the principal stress axes for a homogeneous sample of earthquake focal mechanisms, allows the parameters of the stress ellipsoid to be determined quite accurately for 10–12 earthquakes. Since even in seismically active regions the density of distribution of earthquake epicenters is not the same, when samples of earthquakes are taken with a constant window of their selection from the catalog of focal mechanisms, a large part of the area remains uncovered by stress calculations. The current version of STRESSseism contains an algorithm for iterative generation of earthquake samples, where the window of data collection is small in regions of high epicenter density and gradually (with a given step) increases to some limit in regions of low density. This approach provided crustal stress data at 715 nodes. In Fig. 7a, the number of iterations performed to search for the earthquake data collection window is shown at the nodes of the computational grid.

Fig. 7.

The characteristics of the reconstruction at the nodes of stress calculation: (a) the number of iterations (It) when the initial sample was generated; (b) the number of stress calculations (L) with different stress time periods at one node; (c) the radius (R) of stress averaging along the lateral (km); (d) the time intervals (DT) of stress averaging corresponding to the last (in time) calculation in the given node (years). The epicenters of the two earthquakes in Kahramanmaraş Province are shown here and in the following figures by white triangles.

It is important to realize that the stress data obtained in the tectonophysical reconstruction process correspond to the scale of averaging, which depends on the range of magnitudes of the earthquakes whose focal mechanisms are used, as well as on the density of earthquake epicenters. Since Figure 7a indirectly reflects the distribution density of earthquake epicenters, it therefore also characterizes the size of the stress averaging window (Fig. 7c). In the region under study, these parameters of the compiled catalog of earthquake focal mechanisms allow us to speak about possible scales of averaging at different points from 10 to 250 km laterally and over the entire crustal thickness. The scales of stress averaging from 35 to 95 km are the most representative (see the diagram in Fig. 7c).

Since the computational grid spacing was about 24 km in longitude and 18 km in latitude, good detail of stress changes was achieved only in regions with an averaging window size of less than 50 km (Fig. 7c). Such regions included practically the entire EAFZ and associated regions in the 100–150 km band range. Outside this zone, the detail of the reconstruction results is significantly poorer, which should be taken into account when analyzing the distribution of Coulomb stresses and predicting expected earthquakes.

In the STRESSseism program, the generation of a homogeneous earthquake sample is performed by successive (on a time scale) checks for the homogeneity of the focal mechanisms of earthquakes whose elastic relief regions include the calculation point. Therefore, with a larger number of earthquakes occurring near the calculation point, it is possible to obtain data on stresses corresponding to different time periods of the stress state.

A sufficient density of distribution of earthquake epicenters in a wide band of the EAFZ (about 200 km) made it possible to generate 6234 homogeneous samples of earthquake focal mechanisms for 715 nodes, and their data were used to calculate stresses. At the same time, less than four stress calculations were performed for more than 300 nodes (Fig. 7b). On average, there were about 9 such calculations per grid node. Thus, based on the results of stress reconstruction, it was possible to perform stress monitoring. The time scale stress averaging was about 4.5 years for nearly 35% of the calculations and more than 30 years for 30% of the calculations (see the diagram in Fig. 7d). It should be noted that the shorter time range of calculations is often related not to the density of earthquake epicenters, but to the period of deployment of seismic networks in the region.

Here, we will show only two boundary time options of the calculations at each of the nodes: the earliest, usually corresponding to 1990–2010, and the latest, close to our time, mainly corresponding to the 2010–2022 period. The comparison of these two options will reflect the general trend of stress changes in the crust of the region under study.

Figure 8 shows the geodynamic type of the stress state for the earliest and latest calculations at each node. This parameter is determined by analyzing the orientation of the principal stress axes with respect to the vector to the zenith (Rebetsky, 2003a). From the data corresponding to the earliest stage of reconstruction (Fig. 7a), it can be seen that most of Eastern Anatolia has a horizontal shear stress regime with a subvertical orientation of the intermediate principal stress axes (strike slip faults). Regions with the geodynamic type of the stress state in the form of horizontal extension (normal faults) and shear with extension (transtension) are found only in the southernmost part of the EAFZ. The combination of horizontal shear and compression (transpression) is observed in isolated nodes for the western part of the stress reconstruction region and as an extended zone in the southwest. The horizontal compression regime (thrust faults) is found in the north of the EAFZ near its junction with the Sürgü fault, where the second of the strong earthquakes occurred. To the south, in the Dead Sea fault zone, there is a large region with a stress state regime in the form of vertical shear, for which the axis of the intermediate principal stress is subhorizontal and the other two principal stresses have a plunge of 30°–60° (cuts (Yunga, 1990)).

Fig. 8.

The geodynamic type of the stress state for the reconstruction corresponding at each of the nodes to: (a) the first (1990–2010); (b) the last (2010–2022)—calculation on the time scale. The lower right corner shows the location of the principal stresses of the axis to the Zenith in the octant, which determines the type of the stress state: 1—hor. extension; 2—hor. extension with shear or transtension; 3—hor. shear; 4—hor. compression with shear or transpression; 5—hor. compression; 6—vertical shear.

Comparing the data in Fig. 8b (the most recent stress reconstruction at each node) with those in Fig. 8a shows a significant change in the stress state. For 520 nodes, the difference in the time period of two such samples ranged from 1 to 60 years (the most representative was 20–25 years), which allows us to see the trend of the stress state of the region under study before the catastrophic Pazarcık earthquake.

The main changes in the stress state occurred in the southern part of the seismogenic fault section of the Pazarcık earthquake and to the west of it. The epicenter of this earthquake (see the first seismogenic fault in Fig. 3a) was also located in the zone that experienced a change in the stress state. These stress changes resulted in a significant increase in the region with the geodynamic type of the horizontal extension stress state in the southern part of the EAFZ and its immediate surroundings. Such changes should lead to a decrease in horizontal compressive stresses, which also results in a decrease in frictional forces on the fault.

We should also note the change in the geodynamic type of the stress state from horizontal compression to horizontal shear northeast of the junction of the Sürgü fault with the main EAFZ branch, and from horizontal shear to shear with extension further northeast. This may indicate that horizontal compression decreased on the EAFZ section northeast of the Sürgü fault, i.e., the strength of the fault created by frictional forces decreased.

The fact that the crustal stress field changes with time in the region under study is also supported by the results of comparing the data on changes in the Lode–Nadai coefficient between the early and late stages of stress reconstruction (Fig. 9). It should be specifically noted that this coefficient is a term of mechanics. It determines the shape of the stress ellipsoid. The geodynamic type of the stress state (Fig. 8) determines one of the principal stress axes that is closest in orientation to the axis to the zenith. It is a term of geodynamics (Rebetsky, 2003a).

Fig. 9.

The Lode–Nadai coefficient for the reconstruction corresponding at each of the nodes to: (a) the first (1990–2010); (b) the last (2010–2022) calculation on the time scale.

While in the early stage of the calculation the shape of the stress ellipsoid corresponding to uniaxial compression and its combination with shear (mech.) was widely represented in the crust of the EAFZ surrounding (more than 35% of the determinations), in the later stage the shape of the ellipsoid changed to pure shear at most of these calculation nodes. Such changes occurred to a greater extent in the northeastern segment of the region under study. In the northwestern segment of the EAFZ, where pure shear and its combination with extension were most prevalent in the early stages of the calculation, changes in the Lode–Nadai coefficient also occurred, but to a much lesser extent. It can be seen that the number of nodes with the pure shear stress ellipsoid shape also increased here, but this increase was 10–15% and resulted from the change in different initial states.

It should be noted that when the shape of the stress ellipsoid is close to uniaxial compression (${{\mu }_{\sigma }} \to + 1$) or uniaxial extension (${{\mu }_{\sigma }} \to - 1$), the variability of the location of shear fractures increases, which tend to be located around the principal stress axes of maximum and minimum compression, respectively, creating fracture cones (Rastsvetaev, 1987). When the shape of the stress ellipsoid tends to pure shear (${{\mu }_{\sigma }} \to 0$), the variability of fracture locations in the rock massif begins to approach the system of conjugate pairs. The geodynamic type of the stress state acting in the massif allows us to understand how the forming fracture cones or systems of conjugate fracture pairs will be oriented in space.

In Fig. 10, at the calculation points, the orientation of the projections on the horizontal plane of the axes of the three principal stresses is also shown for the two boundary stages of stress calculation at each of the nodes. It can be seen that along with the overall stable pattern of orientation of the principal stress axes for the early and late stages of the calculation, there are also variations that are more pronounced where there are changes in the geodynamic type of the stress state (Fig. 8) or in the values of the Lode–Nadai coefficient (Fig. 9).

Fig. 10.

Projections of the principal stress axes on the horizontal plane for: (a)–(c) the first (1990–2010); (d)–(f) the last (2010–2022) calculation on the time scale. (a), (d) Minimum compression (extension); (b), (e) intermediate principal stress; (c), (f) maximum compression. The axes are plotted at the calculation node (circle) in the plunge direction. When the plunge angle is less than 7.5 degrees, the length of the axes is maximum with a circle in the middle. For the lowest (a) and highest (c) compressive stresses, the color of the axes shows the geodynamic type of the stress state and the value of the Lode–Nadai coefficient, respectively (see the legends of Figs. 8 and 9).

The orientation of the principal stress axes is the basis for tectonophysical analysis of the stress state. In this case, it is quite common to use data on the orientation of the highest (${{\sigma }_{H}}$) and lowest (${{\sigma }_{h}}$) horizontal compressive stresses for this purpose as well (Zoback, 1992). It should be noted that in many cases the projection directions of the corresponding principal stresses are used for this purpose (Zoback, 1992; Heidbach et al., 2010; 2018), i.e., it is assumed that the azimuths of these principal axes determine both the directions of action of the highest and lowest horizontal compressive stresses.

In reality, this is not the case (Rebetsky, 2003a; Lund and Townend, 2007; Rebetsky and Polets, 2018). It should be noted that in spite of the above–mentioned papers, one of which was carried out as part of the WORLD STRESS MAP project (Lund and Townend, 2007), the data on the azimuths of the principal stresses of the highest and lowest horizontal compression are still listed in the database of this project as directions of the axes of the highest and lowest horizontal compressive stresses.

According to the formulas:

$$\begin{gathered} {{\sigma }_{H}} = - p \\ + \,\,\tau \left[ {\left( {1 - \frac{{{{\mu }_{\sigma }}}}{3}} \right)l_{{H1}}^{2} + \left( {\frac{{2{{\mu }_{\sigma }}}}{3}} \right)l_{{H2}}^{2} - \left( {1 + \frac{{{{\mu }_{\sigma }}}}{3}} \right)l_{{H3}}^{2}} \right] \\ {{\sigma }_{h}} = - p \\ + \,\,\tau \left[ {\left( {1 - \frac{{{{\mu }_{\sigma }}}}{3}} \right)l_{{h1}}^{2} + \left( {\frac{{2{{\mu }_{\sigma }}}}{3}} \right)l_{{h2}}^{2} - \left( {1 + \frac{{{{\mu }_{\sigma }}}}{3}} \right)l_{{h3}}^{2}} \right] \\ \end{gathered} $$

(9)

the location of the highest and lowest horizontal compressive stresses can be obtained by cross–sectioning the stress ellipsoid with a horizontal plane or by searching through all horizontal directions for stresses and selecting those that correspond to the highest and lowest horizontal compression, respectively.

Figure 11 shows the orientation of the highest and lowest horizontal compressive stresses in the crust of the region under study. It can be seen that the orientation of these stresses is much more stable in space than for the principal stress axes (Fig. 10). Changes in the stress state at the calculation point between two stages are often indicated by the geodynamic type of the stress state rather than by the orientation of horizontal stresses.

Fig. 11.

The orientation of the stress axes of the highest horizontal compression for the two stages of the stress state calculation: (a) the first one (1990–2010); (b) the last one (2010–2022). The geodynamic type of the stress state is: 1—horizontal extension; 2—horizontal extension with shear or transtension; 3—horizontal shear; 4—horizontal compression with shear or transpression; 5—horizontal compression; 6—vertical shear.

The data in Fig. 8a show that at the initial temporal stage of the reconstruction, the horizontal shear regime covered almost the entire EAFZ zone. The exception was its small southeastern section with the regimes of horizontal extension and extension with shear (transtension), which propagated into the zone of the Levant strike–slip fault, where they changed to the vertical shear regime. Prior to the Pazarcık earthquake (Fig. 8b), the zone with the horizontal extension regime expanded sharply up to the epicenter of the future catastrophic earthquake. An extensive region of horizontal extension with shear appeared in the northeastern segment of the EAFZ where it contacts the North Anatolian Fault. Another region that expanded and became more intense was the zone of horizontal compression in the southwestern sector of the EAFZ; in addition, a new zone of horizontal compression appeared to the south of the Pazarcık earthquake epicenter.

It should be specifically noted that the new regions of horizontal extension and compression are confined to the western side of the EAFZ. The formation of such pairs of regions with diametrically different stress state regimes according to the papers by D.N. Osokina (Osokina, 1987; Osokina and Friedman, 1987) may indicate that during the reconstruction period sinistral slip stresses occurred in the system of echelon faults west of the EAFZ. In this case, these faults include the Kozan, Göksun, and Andırın faults, as well as the Türkoğlu fault, which is continuous with the Pazarcık fault to the southwest (Fig. 3a). It is also possible that such movements also affected the Malatya and Ovacık faults branching off the EAFZ in the northeastern segment of the region under study. It should be noted that along the Pazarcık and Ergani Gunus faults, a series of strong earthquakes with magnitudes of 6.1–6.8 have occurred in the last 25 years, preceded and followed by a series of less strong events.

The second stage of the MCA reconstruction provided data on the maximum shear stresses τ and the effective confining pressure $p{\text{*}}$, normalized to the massif bonding strength ${{{{\tau }}}_{f}}$ (see equation (7)). Figure 12 shows the stress state parameters corresponding to the earliest and latest earthquake samples at each node.

Fig. 12.

The values of the effective pressure (a), (c) and maximum shear stress (b), (d) normalized to the bonding strength for two calculation stages: (a), (b) the first stage (1990–2010); (c), (d) the last stage (2010–2022).

As can be seen from the data in Fig. 12, for both stages of the calculation, the EAFZ section in the zone of the future Pazarcık earthquake was mainly represented by the data on medium and low stress values. The regions of the lowest stress values corresponded to the second segment (Fig. 3a) of the seismogenic fault of this earthquake. The third segment had a slightly higher stress level. The points with the highest stress levels on the EAFZ were located outside the source of the Pazarcık earthquake.

The epicenter of the Pazarcık earthquake (first segment) was located on the fault section with the lowest level of the effective pressure and the medium level of the maximum shear stresses for both stages of the calculation. It should be noted that there are calculation points with high stress levels near the epicenter of the earthquake. Taking into account that the computational grid spacing of 40–50 km is 1.5–2 times smaller than the linear scale of stress averaging, we can say that the Pazarcık earthquake occurred on the fault section with the highest stress gradient. According to our studies of catastrophic earthquakes that have occurred since the beginning of this century in subduction zones (Rebetsky, 2018; Rebetsky and Guo, 2020), their epicenters gravitated to such sections.

ZONING OF HAZARDOUS FAULT SECTIONS IN EASTERN ANATOLIA

In our studies, the data on the stresses reconstructed in the crust of the region under study by the algorithms of the first two stages of the cataclastic method are the initial data for calculating the normalized Coulomb stresses ${{{{\tau }}}_{C}}$ according to Eq. (8). According to geomechanical criteria for a brittle fracture of rocks (Rebetsky, 2003a), these stresses are responsible for the proximity of faults to the critical state. Coulomb stresses can be positive if the shear stresses ${{\tau }_{n}}$ are greater than the frictional stresses (${{k}_{f}}\sigma _{{nn}}^{*}$). A brittle fracture will occur if Coulomb stresses reach a local bonding strength on the fault section $\tau _{f}^{m}$, which may be lower than the average bonding strength of the fault ($\tau _{f}^{m} < {{\tau }_{f}}$). Under the MCA, after the second stage of reconstruction, it is possible to calculate effective normal $\frac{{\sigma _{{nn}}^{*}}}{{{{\tau }_{f}}}}$ and shear stresses $\frac{{{{\tau }_{n}}}}{{{{\tau }_{f}}}}$, normalized to ${{\tau }_{f}}$, on the known planes of fault sections. Thus, according to (1), we can calculate the ratio of Coulomb stresses to bonding strength, which should be less than 1.

As follows from (1), in order to calculate Coulomb stresses on a fault, it is necessary for each fault section to have not only information about its location (strike), but also about the direction and angle of the plunge. For the faults of the region under study, such information was partially present in the active faults of Eurasia database (Bachmanov et al., 2017). For a large group of faults, it had an indication of their plunge direction, as well as the kinematics of the faults.

Since these data are not sufficient to solve our problems, we have created a program that determines their plunge angles based on the interpretation of data on the focal mechanisms of earthquakes that occurred at distances less than 50 km from the faults.

Thus, based on the data from the two stages of stress reconstruction using the cataclastic method and the data we obtained on the fault parameters, we calculated Coulomb stresses for those fault sections near which the nodes of the computational grid were located (at distances not exceeding 100 km).

Faults were zoned for hazard by dividing the positive Coulomb stress values into four intervals: (1) $0.8 < \frac{{{{\tau }_{C}}}}{{{{\tau }_{f}}}} \leqslant 1$; (2) $0.5 < \frac{{{{\tau }_{C}}}}{{{{\tau }_{f}}}} \leqslant 0.8$; (3) $0.2 < \frac{{{{\tau }_{C}}}}{{{{\tau }_{f}}}} \leqslant 0.4$; (4) $- 0.2 < \frac{{{{\tau }_{C}}}}{{{{\tau }_{f}}}} \leqslant 0.2$, which we refer to as: extremely hazardous, hazardous, alarming, and neutral. The last interval of changes in Coulomb stresses $\frac{{{{\tau }_{C}}}}{{{{\tau }_{{\text{f}}}}}} \leqslant - 0.2$ is considered to be safe. According to this gradation, the level of shear stress relief will change for fault sections in an earthquake, which also determines the specific density of elastic energy release in an earthquake.

As can be seen from Fig. 13, most of the faults (more than 75%) for which stress data are available are in a safe state. For the southwestern segment of the EAFZ and its branches in the northern coast of the Eastern Mediterranean and Syria, the ratio of hazardous to safe faults is close to 1 : 1.

Fig. 13.

Zoning for two stages of the stress state calculation: (a) the first stage (1990–2010); (b) the last stage (2010–2022)—of active fault sections from the database of (Bachmanov et al., 2017) by Coulomb stress level: (1) $0.8 < \frac{{{{\tau }_{C}}}}{{{{\tau }_{f}}}} \leqslant 1$ (dark red); (2) $0.5 < \frac{{{{\tau }_{C}}}}{{{{\tau }_{f}}}} \leqslant 0.8$ (red); (3) $0.2 < \frac{{{{\tau }_{C}}}}{{{{\tau }_{f}}}} \leqslant 0.5$ (light red); (4) $- 0.2 < \frac{{{{\tau }_{C}}}}{{{{\tau }_{f}}}} \leqslant 0.2$ (yellow); (5) $\frac{{{{\tau }_{C}}}}{{{{\tau }_{f}}}} \leqslant - 0.2$ (green); (6) no stress data on the fault (black). The stars show the epicenters of the Pazarcık and Elbistan earthquakes. See also the caption for Fig. 1.

The experience gained in the study of Coulomb stresses, and in particular the study of the preparation region of the 2008 Wenchuan M_w= 8.0 earthquake (Rebetsky et al., 2021), has shown that its source was located on the Longmenshan fault in the region, where, in addition to sections with high levels of normalized Coulomb stresses (${{\tau }_{C}} > 0.8{{\tau }_{f}}$), there were sections with medium levels of these stresses and even short sections with low positive values (${{\tau }_{C}} < 0.3{{\tau }_{f}}$). In this case, the epicenter of the Wenchuan earthquake source was located near a section about 20 km long that corresponded to a critically high level of Coulomb stresses (${{\tau }_{C}} > 0.6{{\tau }_{f}}$). This pattern of the stress state on the fault was consistent with the characteristics typical of the development of the seismogenic fault—the Wenchuan earthquake source, where there were extended sections of relatively slow seismic motions (Liu Jiao and Rogozhin, 2017), as well as sections where the greatest damage was observed and the released seismic energy was significantly higher (Yin et al., 2010).

The results of the tectonophysical zoning of faults for the time period immediately preceding the two strongest earthquakes in the region of Eastern Anatolia under study identify several faults with the most extended (more than 50 km) hazardous sections with positive Coulomb stress values:

(1) The north–northeast–striking Göksun fault west of the EAFZ. It has two sections more than 50 km long with the Coulomb stress level $\left( {0.5{\kern 1pt} - {\kern 1pt} 0.8} \right){{\tau }_{f}}$, separated by a short zone (about 20 km long) with their negative values.

(2) The Andırın fault, sub-parallel to the Göksun fault to the east, had a section with positive Coulomb stress values about 65 km long and a zone with high values about 30 km long.

(3) The north–northeast–striking Misis fault north of the Göksun fault. It has two sections with positive Coulomb stress values about 30 and 60 km long, separated by a zone with their negative values (about 25 km long).

(4) The north–northeast–striking Latakia fault zone along the eastern coast of the Mediterranean Sea. It has a section with positive Coulomb stress values about 50 km long and a small zone with their high values (about 15 km long).

(5) The submeridionally striking Dead Sea fault system south of the EAFZ. It has a section with positive Coulomb stress values up to 100 km long and a zone with high values 40 km long. There are also several similarly striking faults with very high Coulomb stresses near the Dead Sea fault.

(6) The Kırıkhan fault, part of the southwestern EAFZ system. It has a section with positive Coulomb stress values about 70 km long and a zone with very high values about 30 km long.

(7) The Amanos fault southwest of EAFZ to the northeast of the Kırıkhan fault (the third segment of the Pazarcık earthquake source is located here). It has two sections with positive, high Coulomb stress values about 60 and 70 km long, separated by a region with negative values.

(8) The Pazarcık fault is part of the EAFZ system and lies northeast of the Amanos fault (the second segment of the Pazarcık earthquake source is located here). It has an extended (about 70 km long) section with positive Coulomb stress values of medium and low intensity.

(9) The north–northeast–striking Malatya fault north of the EAFZ, connecting it with the Deliler Teceger and Ovacık faults. It has a section with positive Coulomb stress values about 50 km long and a zone with high values about 20 km long.

(10) The northeast–striking Ovacık fault north of the EAFZ. It has a section with positive Coulomb stress values about 100 km long and a small zone with these medium–level stresses about 20 km long.

Other faults have sections with positive and even high Coulomb stress values less than 50 km long, consistent with the probability of magnitude 6.5–7.0 earthquakes occurring on them.

The conducted analysis shows that the EAFZ zone, which had highly detailed stress calculations, looks hazardous because the region of the future source of the Pazarcık earthquake had sections with very high Coulomb stress values (${{{{\tau }}}_{C}} > 0.8{{\tau }_{f}}$) for the Kırıkhan and Amanos faults (the third segment of the source in Fig. 3a) over a distance of about 300 km, although between them there was a zone (about 30 km long) with negative Coulomb stress values. In addition, there was an extended section with positive Coulomb stress values of medium and low intensity on the northeastern segment of the Pazarcık earthquake source (the second segment of the source in Fig. 3a) within the Türkoğlu fault. It was also separated from the hazardous section of the Amanos fault by an extended zone (about 70 km long) with negative Coulomb stress values.

The fault where the epicenter of the Pazarcık earthquake was located, which was a branch of the EAFZ system in its eastern wing, had positive Coulomb stress values along the entire length of the first segment of the source (Fig. 3a). The epicenter itself was located on the section with high Coulomb stress values. This fault lies north of the other two segments of the Pazarcık earthquake at a distance of 25 to 8 km and, according to the active faults of Eurasia database (Bachmanov et al., 2017), had no direct intersection with them.

It is important to note that the zone of the future source of the Pazarcık earthquake looked more hazardous for the initial period of stress reconstruction (Fig. 13a). In the time period close to the beginning of 2023 (Fig. 13b), the length and intensity of the hazardous sections for its northeastern part slightly decreased, while the hazard of the southwestern segment and the region near the earthquake epicenter increased. It can be assumed that these changes in Coulomb stresses were due to sinistral slips that occurred in the time interval between the two reconstruction periods presented.

For the sublatitudinal Sürgü, Çardak, and Sungun faults, where the second strong earthquake occurred, Coulomb stress values were all negative. Here, the stresses acting along the normal to the fault were so high that the frictional stresses (${{k}_{f}}\sigma _{{nn}}^{*}$) exceeded the shear stresses (${{\tau }_{n}}$). A similar condition was true for other sublatitudinal faults lying to the south.

DISCUSSION

Studies of the hazard distribution of active faults based on the results of the tectonophysical reconstruction of natural stresses have begun relatively recently (Rebetsky and Kuzikov, 2016; Rebetsky et al., 2021). In this respect, they are considerably less advanced than the studies of the patterns of Coulomb stress changes based on mathematical models (Stein et al., 1992; Okado et al., 1992; Harris et al., 1995; etc.). The use of the tectonophysical approach in the calculation of Coulomb stresses requires data not only on the parameters of the stress ellipsoid (orientation of the principal axes of the stress tensor and the shape of the ellipsoid), but also on the values of the maximum shear stresses and the effective pressure or their normalized values. This is much more information than is used in mathematical modeling of Coulomb stress changes.

Theoretical concepts of geomechanics suggest that a hazardous section of an active fault is a zone with positive Coulomb stress values, since the local strength of faults can vary greatly on their different sections. A particularly hazardous section is a relatively extended zone with a high level of positive Coulomb stress values (greater than 0.6–0.8${{\tau }_{f}}$). The expected minimum magnitude level of the predicted seismogenic fault section is determined by the scale of crustal stress averaging rather than by the spacing of the grid nodes where the stress data are obtained. This scale, in turn, is related to the magnitude range of the initial data on earthquake focal mechanisms used for stress reconstruction, as well as to the distribution density of earthquake epicenters. For the EAFZ, the minimum predicted magnitude was 6.5–7.0 because it had the highest density of earthquake epicenters for which data on their focal mechanisms were available (Fig. 4), and therefore had the smallest scale of stress averaging (Fig. 7b). At some distance from the EAFZ, for other regions, the scale of averaging increased to 100 km and more, which meant that the lowest magnitude level of the prediction was 7.5.

On the other hand, the experience gained from tectonophysical studies has shown that in the source of the 2008 Wenchuan earthquake (Rebetsky et al., 2021), the sections with high Coulomb stresses were interrupted by zones with their low positive values and even by short sections with negative Coulomb stress values. The results of the studies conducted for the Kahramanmaraş earthquakes confirmed these conclusions. Thus, the expected maximum magnitude level of the predicted seismogenic fault section, derived from the results of the Coulomb stress analysis, is determined by the entire intermittent zone with positive values. At the same time, at the ends of this hazardous zone, extended regions with negative Coulomb stresses should exist.

The fact that the Pazarcık earthquake source was not homogeneous and could contain both zones with high positive Coulomb stress values and sections with negative values is also confirmed by studies of changes in the intensity of seismic radiation directivity (Pavlenko, O.V. and Pavlenko, V.A., 2023).

Earthquake engineering specialists and seismologists believe that among the causes of building failures are the so–called pulse–like features (Baez and Miranda, 2000; Iervolino et al., 2012; Shahi and Baker, 2011; Pavlenko, O.V. and Pavlenko, V.A., 2023), which are characterized by the presence of acceleration pulses of long duration that correspond to velocity pulses of unusually large magnitude—ultrafast faults. Such pulses are observed at stations located near a seismogenic fault during the forward propagation of a rupture in the source. For standard velocities of the earthquake front propagation, these pulses are more prominent on horizontal components oriented perpendicular to the fault (Pavlenko, O.V. and Pavlenko, V.A., 2023). The predominance of parallel components is typical of ultrafast faults (Rosakis et al., 1999).

Pulse–like ground motions are observed when shear waves radiating from different points of the source propagating toward the seismic station arrive at the station almost simultaneously (Somerville et al., 1997). The resulting interference of elastic waves can increase the frequency of vibrations as their amplitude increases. Since the actual velocity of shear wave propagation along a seismogenic fault is the sum of the ground motion velocity for the fault front (about 5 km/s for the Pazarcık earthquake (Rosakis et al., 2023)) and the velocity of elastic shear waves (about 3.3 km/s), it is the first of the above factors that causes ultrahigh wave velocities and changes in the intensity of seismic radiation directivity. The change in the fault front propagation velocity and the ground motion velocity itself can be directly related to the change in the nature of the stress state along the seismogenic fault.

Thus, the data on the presence of ultrahigh wave velocities indicate that there are fault sections with a stress state favorable for increasing the fault front propagation velocity. While the appearance of pulse–like ground motions indicates that there are sections with significantly different stress states on the fault—both contributing and not contributing to fault formation. In this case, the section with a low velocity of the source front propagation is closer to the observation station than the section with a high velocity.

The EAFZ near the source of the Pazarcık earthquake is characterized by large–scale destruction. Thousands of buildings and dozens of cities and towns were severely damaged by the earthquake. In the paper by (Baltzopoulos et al., 2023), according to the data of the State Disaster and Emergency Management Authority (AFAD) under the Turkish Ministry of Interior, peak ground acceleration (PGA) was investigated from the records of seismic stations located at various and epicentral distances. It was shown that the spectral amplitudes of the recorded ground motions were as high as 2.8 g. These studies show that there were 6 stations where pulse–like ground motions were observed. These stations were on the north–northeastern part of the Narlı fault (the first segment of the source) and on the northeastern part of the Amanos fault (the third segment of the source).

As shown in the paper by (Abdelmeguid et al., 2023), the first station to record an almost 20% increase in the velocity of shear waves in the direction of the seismogenic fault compared to the velocity of these waves along the normal to it was the station located near the Narlı fault, north–northeast of the Pazarcık earthquake epicenter (about 30 km). According to the data in Fig. 13, there were two sections with low positive (from 0.2 to 0.5) and almost zero (from ‒0.2 to 0.2) values of normalized Coulomb stresses just in front of this station.

The data from both papers (Baltzopoulos et al., 2023; Abdelmeguid et al., 2023) agree well with the results of the tectonophysical zoning of the EAFZ by Coulomb stress level (Fig. 13). For the Pazarcık earthquake, its first and third segments looked the most hazardous according to the Coulomb stress level (Fig. 3). The highest Coulomb stress level was observed on the Narlı fault near the epicenter of this earthquake, as well as on two extended sections of the Amanos fault in its southern and central parts. These sections were separated by a zone with negative Coulomb stress values. An extended region with negative Coulomb stress values was also observed near the junction of the Pazarcık and Amanos faults, north of the first segment of the source (Narlı fault). To the north of it, along the Pazarcık fault, there was a section with medium to high Coulomb stresses.

According to data published on the U.S. Geological Survey website (https://earthquake.usgs.gov), the propagation of the source on the first segment lasted more than 10 s, with the fault front propagating mostly from south–southwest to north–northeast. Here, the fault initially propagated at a high velocity, opening up sections with high Coulomb stress levels. In the northern part of the Narlı fault, the fault propagation velocity slowed down as the fault reached the section with low and even negative Coulomb stresses. This was recorded by the stations located here with data on the presence of ultrahigh velocities of shear waves, on the one hand, and the shape of pulse–like motions, on the other hand.

After that, seismic motions began on the second segment of the source. Here, the fault front propagated both southwest and northeast along the Pazarcık fault. Thirty seconds after the beginning of the earthquake, seismic motions occurred on the third segment of the source, on the Kırıkhan and Amanos faults. Here, the fault front propagated mainly from north–northeast to south–southwest.

The presence of adjacent and practically subparallel segments in the source of the Pazarcık earthquake (the first and the third segments) should lead to changes in the stress state after coseismic motions occurred in the first segment. The issues related to the mutual influence of closely located faults were discussed in the paper by (Osokina and Tsvetkova, 1979; Osokina et al., 1979). The papers by (Lermontova and Rebetsky, 2012; Lermontova, 2021) show how they should be mutually arranged so that the second fault was activated after the first one, since it increases the Coulomb stress level. Such a mutual arrangement is also possible, when the change in the stress state caused by the activation of the first fault leads to a decrease in the Coulomb stress level and, consequently, makes the activation of the second fault impossible.

Based on the studies of A.S. Lermontova (Lermontova and Rebetsky, 2012; Lermontova, 2021), it can be assumed that the change in stresses that occurred as a result of coseismic motions in the vicinity of the first segment of the Pazarcık earthquake source (Fig. 13) led to an increase in Coulomb stresses in the second segment of the source, which were previously positive but of low intensity. This contributed to the onset of seismic motions in it when the fault front propagating within the first segment was closest to the second segment of the source.

If the distance between the faults of the first and second segments of the Pazarcık earthquake source had been somewhat greater, the effect of the activation of the first segment on the second segment might not have been so significant. Then the earthquake would have ended by forming a seismogenic fault from the first segment only. In this case, the magnitude of the earthquake could have been only 6.5 or slightly higher. Investigations of Coulomb stress changes for the EAFZ performed in the paper by (Chen et al., 2023) using numerical modeling confirm our conclusions.

The transfer of seismogenic hazard from one fault to another in the case of their subparallelism with a strike slip is a fairly standard phenomenon studied in a tectonophysical experiment (Osokina et al., 1979). The occurrence of a magnitude 7.5 earthquake on the Sürgü fault, which has a significantly different strike from the eastern branch of the EAFZ, is a rare phenomenon. On this fault, tectonophysical zoning data show negative Coulomb stress values.

As mentioned in the INTRODUCTION, the source of the Elbistan earthquake was located in the western wing of the northern end section of the second seismogenerating segment of the Pazarcık earthquake source, and a huge downwarp 30 meters deep and several kilometers long also appeared near its end section (in the south), but already in the eastern wing, near the village of Tepehan. Thus, the second strong Elbistan earthquake and the “olive grove downwarp” relative to the first Pazarcık earthquake are located in the decussate sectors of additional extension formed in the vicinity of the seismogenic fault with a large component of sinistral slip (Fig. 14a) (Osokina and Friedman, 1987). The sectors that show the greatest additional extension and compression as a result of the slip also have the greatest rearrangement of the orientation of the principal stress axes (Fig. 14b).

Fig. 14.

The location of compression and extension sectors for the additional pressure (a) arising in the vicinity of a seismogenic fault with a sinistral slip, and the trajectory of principal stresses near the fault (b) according to (Osokina and Friedman, 1987).

Thus, the Pazarcık earthquake caused a rearrangement of the stress field over huge areas of the Earth’s crust, leading to the phenomena observed after the earthquake. Therefore, when zoning regions by seismic hazard, it is necessary to take into account the interaction factor between the activated fault and adjacent faults connected to it.

The conducted analysis has answered most of the questions raised in the INTRODUCTION. It can be concluded as follows. The main part of the seismogenerating faults of the two strongest earthquakes was not prepared to generate the magnitude that eventually occurred. Without the dangerously high level of Coulomb stresses on the echelon Narlı fault, the distribution of these stresses on the Pazarcık and Amanos faults could not have generated earthquakes with magnitudes greater than 6.5–7.0. In turn, without the Pazarcık earthquake, the Elbistan earthquake may not have occurred at all. It can be said that the Kahramanmaraş earthquakes had a domino effect, causing a cascading redistribution of stresses that followed the most unfavorable scenario.

CONCLUSIONS

According to the tectonophysical monitoring of the crust in Eastern Anatolia for the period of 50–70 years, there have been gradual, slow changes in its stress state. In the initial stage of monitoring, dating back to the early 2000s, the stress reconstruction showed that most of the crust had a horizontal shear setting, and only in the southern sector, several small sections with a linear size of 100–150 km were observed with horizontal extension and horizontal compression regimes. In the stress reconstruction performed just before the two February 6, 2023 Turkey earthquakes, the crustal regions with the horizontal extension setting expanded significantly, covering almost the entire southwestern segment of the EAFZ. In this case, the boundary of transition from the horizontal extension regime to the horizontal shear was located near the epicenter of the Pazarcık earthquake source.

The two strongest Turkey earthquakes that occurred in Kahramanmaraş Province progressed in different stress states that determined the hazard of their seismogenic activation. Most of the source of the first Pazarcık earthquake had positive Coulomb stresses ranging from low to high values. The highest level of these stresses corresponded to the first segment of the source (Fig. 3a). However, the level of maximum shear stresses and the confining effective pressure in the source was low and had a large variation in values. For the Elbistan earthquake, the Coulomb stress level on the fault prior to the first М_w = 7.8 earthquake was all negative.

If we consider each of the three segments of the Pazarcık earthquake source separately, magnitude 6.5–7.0 earthquakes could have occurred according to the data on the length of the fault sections with a high Coulomb stress level. However, a magnitude 7.8 earthquake occurred as a result.

This means that for systems of active faults, during the period when one of their sections becomes seismogenic—an earthquake occurs, there is a sudden restructuring of the stress state, which can change the entire course of the seismic process. This was true both for the first—Pazarcık earthquake, which became a 7.8 earthquake instead of an earthquake with magnitude not greater than 7.0 (for the first segment of the source), and for the second—Elbistan earthquake. In this case, for the first earthquake, the stress changes and the earthquake development occurred simultaneously, while for the second earthquake, the main changes in the stress state most likely occurred immediately after the Pazarcık earthquake, but the Elbistan earthquake occurred 9 hours after the first earthquake. It is possible that during this time, additional elastic deformations accumulated in the fault zone due to aftershocks from the first earthquake and slow aseismic motions. All of this contributed to an increase in Coulomb stresses on the Elbistan fault. According to the mutual location of the Elbistan fault relative to the strike of the Pazarcık earthquake source, the increase in Coulomb stresses was most likely caused by a decrease in compressive stresses normal to the fault, which also lowered the frictional stresses on the fault.

The conducted studies show a high degree of unpredictability in the occurrence of the strongest earthquakes in zones with dense and differently oriented fault systems. When an earthquake begins in such zones, it is difficult to predict which fault section it will be confined to. Here, it is possible to specify regions that are at risk for an earthquake of a certain minimum magnitude. However, it should be kept in mind that the interaction of active faults can increase this magnitude. Such conclusions show why prediction methods can produce both false alarms and omissions of strong earthquakes (Kossobokov et al., 2015).

Probably, for a better prediction of seismogenic activation of faults, it is necessary to combine the methods of tectonophysics, aimed at studying the stationary process of changes in deformations and stresses in a natural object, and the methods of geomechanics for numerical modeling of stresses in systems develo** in a dynamic—seismic setting.