1 Introduction

On February 6, 2023, two significant earthquakes struck southern Turkey on the same day, causing extensive loss of life and property damage across 11 cities. The two earthquakes and approximately 40,000 aftershocks resulted in a massive disaster in the region. Over 50,000 people in Turkey and around 5,000 people in Syria lost their lives in the earthquakes, which also affected the northern part of Syria. More than 230,000 buildings were destroyed or so severely damaged that they were uninhabitable (Büyüksaraç et al. 2024). It was proven that soil-structure interaction is the key parameter for the structure’s performance, especially in weak soils where the observed damages to the structures have been significant (Avcil, 2024).

After these catastrophic events, many scientists redirected their focus to this area in order to investigate potential “symptoms” of pre-earthquake processes. Some studied external phenomena, such as geospace weather (Ouzounov and Khachikyan 2024), but the primary efforts were concentrated on interpreting directly connected geological, geodetic, and seismological information.

Fault structures interact by redistributing accumulated energy within the medium during an earthquake generation. In other words, one seismic event can induce stress in an adjacent fault, leading to new ruptures. Such interactions are observed both over short-term time frames and longer periods. This situation occurred during the 2023 earthquakes (Alkan et al. 2024). Various factors play a crucial role in the interaction between faults and seismic activity, including dynamic stresses in the Earth’s crust, fluid diffusion in rock pores, and more. Understanding the complexity of the interaction between earthquakes and geodynamic processes, as well as predicting seismic activity, will be focus of future research.

The Eastern Anatolian Fault System undergoes deformations due to contemporary movements of the African and Arabian plates relative to the Anatolian Plate. As of February 6, 2023, no strong earthquakes with a magnitude exceeding M6 have occurred in the fault system’s segment engaged in seismic activity following the beginning of the 20th century. This period of relative calm indicates that significant deformations have been accumulated due to collisions between the tectonic plates.

In our study, we use an approach to calculate the regional spatial and time variations of the statistical parameters of seismicity (b-, Z-, β-values) using the extended software package ZMAP (Westerhaus et al. 2002). The software package is open-source, used to assess various seismic hazards (Westerhaus et al. 2002; Wiemer and Wyss 1994, 2000). ZMAP 7.1 contains many advanced seismological functions for earthquake catalog analysis, including estimating variations in b-, a-, and Z- values, time series analysis, temperature reduction, and more.

2 Methodology

As was mentioned above, our methodology is based on the resources of the ZMAP 7.1 software within the MATLAB environment. It is widely used in scientific research to analyze various earthquake catalogs and has proven to give valuable results (e.g., in Wiemer 2001; Ogata et al. 1991). Numerous researchers have examined the changes in seismic parameters (Ogata et al. 1991; Wiemer and Benoit 1996; Wiemer and Wyss 1997), showing that the value of parameter b- (from the Gutenberg and Richter equation, 1944) changes over time and has spatial variations. Some findings suggest that the average magnitude value increases significantly before a major earthquake, leading to a decrease in the value of b- (Smith 1986; Wyss and Lee 1973). A low b-value indicates high stress in the area, and varying b-values in different regions indicate different stress accumulation patterns (Kanamori 1981). The parameter a- (quantitive parameter of the seismic activity) exhibits significant variations from region to region due to its dependence on (1) seismic activity levels, (2) observation period, (3) size of the study area, and 3) earthquake magnitudes. The most well-known methods for calculating the parameters b-, Mc (Magnitude completeness) are the least squares and maximum likelihood methods (Wiemer and Wyss 1997). The present study applies the maximum likelihood method described by Wiemer and Wyss 1997; which is also incorporated in the used software.

The Gardner-Knopoff (Gardner & Knopoff, 1974) is applied for declustering the earthquake catalog by linking events into clusters. This is being made according to adaptive spatial-time interaction zones and the assumption that the background field’s spatial and temporal components follow a Poissonian distribution. Considering the exponential distribution of earthquakes by magnitude (Gutenberg and Richter 1944), the value of the slope of the recurrence graph is estimated by the method of maximum likelihood (Utsu 1965).

$$\:b=\frac{\text{log}e}{{M}_{mean}-{M}_{min}}$$
(1)

,

where Mmean – is the average value of magnitude in the sample, Mmin - is the minimum magnitude of the sample and is calculated as: Mmin=Mc-∆M/2, where ΔM - takes into account the rounding of magnitude and here is selected ΔM = 0.1 (Schorlemmer et al. 2004), and Mc – is the magnitude of the completeness of the sample. The root mean square error of the estimate is:

$$\:{\sigma\:}_{b}=\frac{b}{\sqrt[2]{N}}$$
(2)

The standard normal deviation Z-test is one of these statistical methods frequently used for analyzing seismic quiescence. We applied the ZMAP method for imaging the areas exhibiting a seismic quiescence. In order to rank the significance of quiescence, we used the Z-test, generating the log term average (LTA(t)) function for the statistical evaluation of the confidence level in units of standard deviations. According to Öztürk and Bayrak (2012)

$$\:Z\left(t\right)=\frac{{R}_{all}-{R}_{wl}}{\sqrt{\frac{{\sigma\:}_{all}^{2}}{{n}_{all}}}-\frac{{\sigma\:}_{wl}^{2}}{{n}_{wl}}}$$
(3)

, where Rall is the mean rate in the overall period including TW (from t0 to te), Rwl is the mean rate in the considered time window (from t to t + TW); σall and σwl are the standard deviations in these periods and nall and nwl the number of samples; t is the “current time” (t0 < t < te).

The β-statistic (Matthews and Reasenberg 1988; Reasenberg & Matthews 1988) is frequently used to compare seismicity rates in two-time intervals and has been applied for the detection of dynamic triggering in several studies (e.g., Kilb et al. 2000; Gomberg et al. 2001). For n1 and n2 events during time periods 1 and 2 with time lengths t1 and t2, respectively, the β-statistic (Matthews and Reasenberg 1988; Reasenberg & Matthews 1988), which is used to determine whether the seismicity rate in period 2 is more significant than that in the period 1, is given by

$$\:\beta\:=\frac{{n}_{2}-\frac{{n}_{1}{t}_{2}}{{t}_{1}}}{\sqrt{\frac{{n}_{1}{t}_{2}}{{t}_{1}}}}$$
(4)

,if the investigated sequence is assumed to follow a Poisson process. The β-statistic is expected to follow the standard normal distribution when the seismicity rates in the two periods are the same. Thus, from the computed β-statistic, we can determine the observed significance level to reject the null hypothesis that the two seismicity rates are the same.

Reasenberg and Simpson (1992) regarded the difference between the two rates as significant if |β| > 2.

When the Z-test is applied to seismicity rate change, it tests the significance of the difference between the frequency Poisson distributions of two independent earthquake samples from the same sequences. It defines that |Z|=1.64 responds to a 90% confidence level, |Z|=1.96 responds to a 95% confidence level, |Z|=2.57 responds to a 99% confidence level (Habermann 1987).

2.1 Data

Understanding the spatio-temporal variations of the seismic activity of a region is crucial for analyzing the risk of natural disasters and developing seismotectonic environments. To quantify these variations in a particular area, we need an earthquake catalog (Burton 1990).

A continuously updated catalog existing for the period 1905 − 05.09.2023 was utilized from the Turkish Bogazici University KOERI1 to illustrate the very high modern seismic activity in a vast area around the Eastern Anatolian Fault System (Fig. 1).

Fig. 1
figure 1

Spatial distribution of earthquake epicenters after declustering of the catalog; the selected polygon is indicated by a black line

Full size image

The earthquake epicenters cover a spatial window of 34.0° N − 44.0°N and 25.0°E − 46.0°E; the total number of earthquakes is near 500 thousand (499,989). Depths range from 0 to 62 km, and for magnitude estimates, the magnitude scale Mw. The catalog consists of independent events, defined as residuals, after identifying and removing clustered and duplicated events using the Gardner and Knopoff algorithm (Gardner & Knopoff, 1974).

The statistical parameters of seismicity vary with changes in the boundaries of the studied area, potentially affecting the interpretation and analysis of results. Therefore, selecting the region’s borders is one of the crucial tasks. The appropriate borderline, in this case, can be defined by following the region around a specific fault area associated with the two very strong earthquakes of February 6, 2023(T0 = 01:17:36; 37.170°N, 37.080°E, h = 20 km, Mw = 7.8 and T0 = 10:24:49; 38.110°N, 37.340°E, h = 10 km, Mw = 7.5) (Fig. 1). The chosen appropriate sector (indicated by a black dashed line on Fig. 1 has dimensions of approximate 1000 km in length and 250 km in width, with an area of 261,834 km². The total number of earthquakes for the period 1983 − 05.09.2983 with magnitude 0 < Mw ≤ 7.7 and depth 0 < h ≤ 115 km) before declustering is 87,511. After declustering and entering a minimum threshold of Mc ≥ 2.5 it is reduced to 29,346 events and depths ranging from 0 to 50 km for the same period (Fig. 1). This area encompasses a system of fault zones parallel to the border between the Arabian and Anatolian plates, as well as the contact zone of the Northern and Eastern Anatolian fault zones.

The Mc estimation using the maximum curvature method resulted in Mc = 2.5, with a depth limit of 50 km (Fig. 2). The b-value of the catalog of the selected region for investigation of the seismic parameters is also shown in Fig. 2. Changes to seismic parameters may occur when boundaries for these parameters are altered. The correct borderlines of the studied area can be determined using fault slip vectors and a migration model of seismic centers. To confirm the appropriate area selection, its evolution of seismicity is tracked for four 10-year lasting periods: (1) the period from 1983 to 1992; (2) the period from 1993 to 2002; (3) the period from 2003 to 2012; and (4) the period from 2013 to 2022 (Fig. 3a, b, c, d). For the entire period from 1983 until 2023, the catalog used includes 29,346 events in total.

Fig. 2
figure 2

Magnitude-frequency distribution for the period between 1983 and 2023 for the selected polygon

Full size image
Fig. 3
figure 3

Number of earthquakes in a 10 km by 10 km grid for the last four ten-year periods

Full size image

First period, there is an area with an increased number of earthquakes (100–120 events) observed in the northeastern part of the region, with a hypothetical center at approximately 41.50°E/39.50°N. During the second period, this area expands to the southwest while maintaining the same earthquake count (100–120 events). The hypothetical center shifts in the same direction to coordinates 39.90°E/38.90°N. Third period: the area with a high number of earthquakes maintains its hypothetical center at 39.90°E/38.90°N, but the earthquake count increases to 160–180 events. During the fourth period, this zone extends further southwest, reaching the epicenters of two earthquakes on February 6, 2023. For this fourth period, the hypothetical center of the zone has shifted to approximately 39.50°E/38.50°N, and the earthquake count around it is above 180. Close to the epicenters, the number of events increased from 60 to 80 in the earlier periods to 100–140 for the fourth period. This indicates that the seismic activity in the northeastern part of the fault and its movement southwestward caused stress accumulation in the hypocentral area, leading to two very strong earthquakes in February 2023.

Subsequently, the b-value (slope of the recurrence graph) is estimated, and the increase in seismicity rate is assessed by calculating the statistical values Z- and β- for the declustered catalog of events in the selected polygon (Habermann 1983; Matthews and Reasenberg 1988; Marsan and Nalbant 2005; Aron and Hardebeck 2009).

3 Results

The present study examines the values of the slope of the Gutenberg-Richter recurrence curve and other statistical parameters prior to the two seismic events on February 06, 2023, with epicentral parameters: 37.043°E/37.288°N, T0 = 01:17:13, Mw = 7.7, h = 8.6 km and the subsequent event with parameters 37.239°E/38.089°N, T0 = 10:24:48, Mw = 7.6, h = 7 km. The spatial variations in the recurrence curve (b-value) slope value characterize the seismicity in a given region (Tsukakoshi and Shimazaki 2008). Typically, when b ≥ 1, the Earth’s crust is considered heterogeneous with low stress, while b < 1.0 suggests a homogeneous crust with high stress (Bridges and Gao 2006). Studies show that the decrease in the b-value in a studied seismogenic region can be associated with an increase in stress before strong earthquakes (Nuannin et al. 2005; Stiphout et al. 2011a; Wiemer and Wyss 1997; Wyss and Stefansson 2006; Wu et al. 2008) Fig. 4.

Fig. 4
figure 4

Magnitude-frequency distribution for the two five-year periods

Full size image

Additionally, the spatial distribution of the b-value (declustered catalog with Mc ≥ 2.5) was analyzed and compared within four one-year long time intervals (2019, 2020, 2021, and 2022) before the earthquakes of 06.02.2023. The maps were created for a grid of points spaced 16.093 km apart. The b-value was calculated for each point based on at least 50 events occurring within a radius of about 100 km around that point (see Fig. 5).

Fig. 5
figure 5

Spatial distribution of b-values for the last one-year long time intervals - (a) 2019, (b) 2020, (c) 2021 and (d) 2022, before the February 06, 2023, very strong earthquakes

Full size image

The b-values for different periods near the epicenters of the two events are as follows:

  1. (i)

    2019: b = 1 ± 0.1;

  2. (ii)

    2020: b = 0.9 ± 0.1;

  3. (iii)

    2021: b = 1 ± 0.1.

For the most recent period, iv) 2022, the b-value falls within the range of 0.7–0.8 ± 0.055 for almost the entire area. Despite the inability to definitively determine an area with a low b-value for the last period, the b-value of 0.7–0.8 < 1 for almost the entire region indicates that stress has accumulated within the crust.

The spatial distribution of the Z-parameter prior to the earthquake on February 06, 2023 (Fig. 6a) has been computed for the same grid of points and a declustered catalog with Mc > 3.9, comparing two time periods: the first period from January 01, 2020, to June 30, 2021, and the second period from July 01, 2021, to December 31, 2022. Please take note of the following information:

Fig. 6
figure 6

(a) Spatial distribution of Z obtained in the comparison of two periods (01.01.2020–30.06.2021 and 01.07.2021–31.12.2022, i.e. 1.5 years each); (b) Variations of the cumulative number of earthquakes (blue line) and the parameter Z (red) for the period 1970–2023, in a sliding time window of 1 year and 6 months with a step of 14 days

Full size image

High (positive) Z-values on the maps indicate a decrease in the rate of seismic events (seismic quiescence) compared to the first period. Conversely, low (negative) Z-values indicate an increase in the rate. In Fig. 6, the epicenters are located at the boundary of a zone with relatively high Z-values (Z = 2.5–2.8), suggesting that the period chosen (July 01, 2021, to December 31, 2022) before the earthquake is a time of relative seismic quiescence. These relatively high Z-values of 2.5–2.8 indicate around 95% reliability of the result.

Figure 6b illustrates the variation in the cumulative number of earthquakes and the Z-parameter over the time interval from 1970 to 2023. The window for calculating the z-value is 1.5 years, shifted across the entire time interval with a step of 14 days. The highest positive Z-value (of + 10.51) after 1980 lasted approximately seven months (from September 28, 2020, to April 26, 2021) and is about one year and nine months prior to the earthquakes of February 06, 2023. Taking into consideration that reported seismic quiescence before major earthquakes varies from 1.5 to 5.5 years according to other authors, it can be concluded that regions where significant anomalies of seismic quietness have been observed can be interpreted as zones of future earthquakes with an average probability greater than 70% (Öztürk, 2011).

The definitions of Z- and β- values are similar, with Z-values being more symmetric than β-. Z- and β- values are based on the assumption that earthquake sequences follow a Poisson distribution. Under such assumptions, both follow approximately the standard normal distribution. The spatial distribution of the β- parameter prior to the earthquake on February 06, 2023 (Fig. 7a) has been computed for the same grid of points, catalog, and compared time periods. Low (negative) β- values on the maps can be interpreted as a decrease in the rate of two seismic events (seismic quiescence) compared to the first period. In Fig. 7a, the epicenters are located at the boundary of a zone with relatively low values of β- (β = -2.5-3.5), indicating that the chosen period (July 01, 2021, to December 31, 2022) before the earthquake is a period of relative seismic quiescence.

Fig. 7
figure 7

(a) Spatial distribution of β- values obtained in the comparison of two periods (01.01.2020–30.06.2021 and 01.07.2021–31.12.2022, i.e. 1.5 years each) before the February 2023 strong earthquakes; (b) Variations of the cumulative number of earthquakes (blue line) and the parameter β- (red) for the period 1970–2023, in a sliding time window of 1 year and 6 months with a step of 14 days. It is used to confirm the result as long as Z and β- are very similar

Full size image

Figure 7b illustrates the variation in the cumulative number of earthquakes and the β- parameter over the time interval from 1970 to 2023. The window used for calculating the β value is 1.5 years, shifted throughout the entire time interval in steps of 14 days. The smallest negative β value (approximately − 1.51) after 1980 has the same duration as the Z- value (from September 28, 2020, to April 26, 2021) and occurs about one year and nine months before the earthquakes on February 06, 2023.

4 Conclusions

The present study analyzes seismic parameters within a selected area around the Eastern Anatolian Fault System using the ZMAP 7.1 software within the MATLAB environment. The entire examined earthquake catalog spans from 1905 to 2023, encompassing earthquake parameter information within a very wide region (more than 1.5 million km²) around the epicenters of the extremely strong earthquakes from February 6, 2023. The research focuses on a smaller period (1983–2023) and a smaller selected area (261 834 km²) around the Eastern Anatolian Fault Zone. The results show in detail the seismicity parameter’s changes across different periods from 1983 to 2023, exploring factors that could contribute to these changes. Statistical seismicity parameters are comparatively analyzed over the selected region’s last ten-, five- and one-year periods. This method enables the detection of temporal and spatial changes in seismic activity, investigating areas with increased earthquake occurrences that migrate over time. One possible explanation for this phenomenon is the build-up of stress within fault zones.

Analysis of the b-value slope of the Gutenberg-Richter recurrence plot reveals a 21.2% decrease during the second part of the last ten-year period (2018–2022). This reduction is interpreted as stress accumulation in the Earth’s crust before both significant seismic events of 2023 (see, for example, Tsukakoshi and Shimazaki 2008; Urbancic et al., 1992; Wu and Chiao 2006; Katsumata 2011; Maeda and Wiemer 1999; Stiphout et al. 2011a). The spatial distribution of b-values is further analyzed for a declustered catalog of seismic events within four annual time intervals. The spatial distribution of the b-value changes before the stronger earthquakes, indicating an abnormally low b-value covering the epicenters of the studied stronger earthquakes. These low values of b- can be interpreted as potentially locked areas or areas with accumulated increased stress in the Earth’s crust preceding the significant seismic events. Similar trends are observed for Z- and β-parameters, with positive values associated with seismic quiescence and negative values linked to increased seismic event rates.

Similar to previous studies in other regions, the analysis of b-, Z-, and β- seismic parameters indicates that seismic activity in the northeastern part of the Eastern Anatolian Fault Zone and its migration towards the southwest could be linked to the build-up of stress in the hypocentral region preceding the two very strong earthquakes in February 2023. The interpretation of all obtained results emphasizes the connection between the calculated parameters and impending earthquakes, identifying zones with anomalous values as high-risk areas for the very strong seismic events in the Eastern Anatolian Fault Zone region.

Based on the analysis, it can be concluded that the reduction of the b-value from 1.07 (for the period 2013–2017) to 0.843 (for the period 2018–2022) and the high Z-values defining the zones of relative seismic rest may indicate an impending release of significant stress accumulated from the interaction of the Arabian and Anatolian plates’ movements in the studied region.(Wiemer and Wyss 1997; Wu and Chiao 2006; Wyss and Stefansson 2006; Rudolf-Navarro et al., 2010; Stiphout et al. 2011a). Thus, the spatio-temporal variations of these three statistical parameters of seismicity could be interpreted as predictors of the strong seismic events that happened in the southern part of the Eastern Anatolian region of Turkey.