1 Introduction

The report explores the potential impact of major earthquakes on the Büyükçekmece district in Istanbul, Turkey, by utilizing stochastic ground motion simulations. The region is situated near the North Anatolian Fault (NAF), which is one of the most seismically active regions globally, making it susceptible to earthquake damage. Due to these characteristics of the area, researchers worldwide have always been drawn to studying the seismicity of the region and assessing earthquake hazard through various methods (including but not limited to Pulido et al. 2004; Sørensen et al. 2007; Tanırcan 2012; Zengin and Çaktı 2012; Paolucci et al. 2018; Infantino et al. 2020; Stupazzini et al. 2020).

Unlike previous studies, this report concentrates on assessing earthquake hazard in a specific district of Istanbul, which is the final in a series of few studies. These studies are developed in collaboration with multiple academic institutions, Büyükçekmece Municipality and Istanbul Metropolitan Municipality. The findings from these studies are anticipated to inform future city-wide planning initiatives.

Initially, we carefully selected five earthquakes that occurred on NAF to validate the EXSIM simulation algorithm and we successfully validated the accuracy of this algorithm by simulating these earthquakes for the wider metropolitan region of Istanbul (Süleyman and Çaktı 2020, 2021, 2022) using data from the Istanbul Early Warning and Rapid Response System (IEWRRS).

To obtain a better insight into the potential impact of a major earthquake, we extensively analyzed the site response of this district in a recent study (Süleyman et al. 2023). We utilized the seismic ambient vibrations recorded in this region to calculate the depth-to-bedrock values, which were used to compute quarter-wavelength amplification factors in this study. These amplification factors were utilized in stochastic ground motion simulations, enabling us to prepare our estimations with expected improved accuracy.

In this study, we aimed to gain a better understanding of the impact of earthquakes with magnitudes ranging from 6.3 to 6.7 and 7.2 to 7.42 on the district. In the first stage, we conducted a total of 50 scenario events of these magnitudes, with a particular focus on the Büyükçekmece district. We thoroughly discussed these simulated scenario events, focusing on the impacts of different magnitude events and various period spectral accelerations on the district. In the study's second stage, we aimed to gain a deeper understanding of the uncertainty stemming from variations in stress parameter and hypocenter location. To achieve this, we significantly increased the number of scenarios examined compared to the first stage. As a result, our analysis shifted towards probabilistic investigations to better investigate the impact of these two key source parameters on the outcomes.

This study highlights the potential for earthquake hazards in this district in the event of medium to large earthquakes. The study emphasizes the importance of understanding the potential impact of earthquakes in Büyükçekmece and taking measures to minimize the impact.

2 Seismotectonics of the Marmara region

The NAF is a significant right-lateral strike-slip fault that stretches over approximately 1500 km and runs along the transform boundary between the Eurasian plate and the Aegean-Anatolian plate. The Marmara fault segments of NAF are located in a seismic gap, and their total length is approximately 100 km (Bohnhoff 2013; Ergintav et al. 2014). According to Ambraseys (2002), the recurrence interval for a major earthquake in this area is between 150 and 400 years, with the last event occurring in 1766. Drab et al. (2015) highlighted the importance of the Çınarcık segment, also known as Princes' Islands segment, which acts as a permanent seismic barrier. The authors identified four earthquake sequences in the Marmara Sea between C.E. 740 and 1999, with only one large earthquake propagating within a single event through the bend between the Çınarcık and the western portion of the İzmit segment.

Although major earthquakes occurred sequentially from east to west along the NAF in the twentieth century, culminating with the 1999 earthquakes, the seismic rupture has not yet broken the Marmara Sea segment. According to Ergintav et al. (2014), Diao et al. (2016), and Lange et al. (2019), this segment has the potential to trigger an MW 7.4 event. Furthermore, some past studies estimated a 35% to 70% probability of an MW > 7 earthquake occurring along the Marmara segment within a time frame of 30 years (Parsons 2004; Murru et al. 2016).

By taking into account the findings of these studies and being aware of the major earthquakes that will occur in the near future with the potential of significantly affecting Istanbul, the authors of this study put their focus on the major seismic gaps, highlighted in these past studies–the Marmara Sea segment and the neighboring Princes’ Islands segment.

3 Seismic hazard assessment studies conducted for istanbul

Since the devastating 1999 earthquakes, Istanbul has been the focus of many academic studies. These studies used probabilistic and deterministic approaches to assess the seismic hazard of the region. One such recent study is Infantino et al. (2020), which aimed to improve ground-motion estimation in Istanbul by running 66 3D physics-based numerical simulations (PBSs) of earthquakes along the NAF with magnitudes ranging from MW 7 to 7.4 offshore Istanbul. The PBSs were compared with well-established ground-motion prediction equations. The researchers found that for MW 7 and 7.2, PBSs are in agreement with empirical prediction models, whereas, for MW 7.4, PBSs provide higher ground-motion estimates as a result of directivity effects amplified by the specific geometry of the portion of the NAF facing Istanbul. Another study, by Paolucci et al. (2018), also utilizing 3D PBSs, discusses the advantages and current limitations of this approach by focusing on Istanbul as a case study. They provide an overview of the advancements in the field of earthquake ground-motion prediction, highlighting the potential of 3D PBSs in improving the accuracy of ground-motion predictions.

Stupazzini et al. (2020) performed seismic hazard assessments in Istanbul using three approaches: physics-based (footprint & GAF), and standard GMPE methods. Results showed good agreement between physics-based approaches, while both diverged from GMPE, highlighting the significance of near-fault directivity effects. This was further confirmed by comparing it with a directivity-specific GMPE. The physics-based approaches provided a wider range of hazard estimates, covering both lower and higher values than the standard method. The study emphasizes the importance of considering directivity for accurate seismic hazard assessment in Istanbul.

Pulido et al. (2004) used a combination of deterministic and semi-stochastic procedures to simulate strong ground motion in the Sea of Marmara region, Turkey, based on fault rupture scenarios and a source asperity model. The earthquake scenarios involved the rupture of the closest segments of the NAF system to Istanbul, with three fault rupture scenarios defined based on the location of the hypocenter. Results showed that the location of the hypocenter is a critical parameter for determining the worst scenario earthquake at Istanbul, and anelasticity has a significant effect on the regional attenuation of peak ground accelerations. The simulated ground motions resulted in large acceleration response spectra at long periods, which could cause building damage during an actual earthquake.

Zengin and Çaktı (2012) conducted a stochastic finite fault simulation in Istanbul based on several fault rupture scenarios. The region-specific source, path, and site model parameters were calibrated by simulating the ML 5.1 Marmara Sea Earthquake, which occurred on July 25, 2011, on the Central Marmara Fault Segment. Synthetic site-specific ground motions were generated for each grid, and the simulation model parameters were verified by comparing the simulation results with ground motion prediction equations.

Sørensen et al. (2007) simulated strong ground motions caused by a scenario earthquake (M 7.5) in the Marmara Sea and investigated the effect of varying input parameters on the broadband frequency ground motion. They used a hybrid model combining a deterministic simulation of low frequencies (0.1–1.0 Hz) with a semi-stochastic simulation of high frequencies (1.0–10.0 Hz). The most significant parameters in terms of ground-shaking level were the rise time, rupture velocity, rupture initiation point, and stress drop.

Tanırcan (2012) simulated strong ground motions in Istanbul for a MW 7.2 scenario earthquake using a finite difference simulation technique and a three-dimensional velocity structure. Three different asperity models were used for ground motion simulations with hybrid Green's functions. The results showed that unilateral rupture propagation towards the northwest increased peak accelerations and velocities on the European side, while unilateral rupture propagation towards the southeast caused only localized high accelerations.

4 Methodology

4.1 Determination of site-specific parameters

This study was preceded by another study (Süleyman et al. 2023) that formed the foundations of the seismic hazard assessment of Büyükçekmece. The purpose of this preceding study was to estimate the site-specific parameters that affect the ground motion response at each selected spot in the study area defined within Büyükçekmece (Fig. 1). Seismic vibrations were recorded at 56 selected locations for approximately three hours each. The horizontal-to-vertical spectral ratio (HVSR) method was employed to observe site amplifications after processing the recordings. The HVSR curves were analyzed with consideration of their rotational dependency to assess the potential effects of site amplification at different orientations. Moreover, the fundamental frequency (f0) was determined at 42 of the 56 locations to calculate the depth-to-bedrock values. The lowest frequency most strongly represented in the ground motion response at each site was considered the fundamental frequency. This allowed for estimating the depth to bedrock values, crucial for evaluating the site response. Shear-wave velocities were also assigned to each observation location based on previous studies (IMM 2007; IMM 2009). The coordinates and the assigned VS,30 values are presented in “Appendix A”.

Fig. 1
figure 1

The location of the study area; adapted from City Scoping Report for Büyükçekmece District (2022)

Full size image

The results of this study provide valuable insights into the site-specific parameters of Büyükçekmece district. The estimated site amplification and depth-to-bedrock values can be utilized to improve the accuracy of seismic hazard assessments for the region. More detailed information on the methodology and results of this study can be found in Süleyman et al. (2023).

4.2 Validation of the ground motion simulation algorithm for Büyükçekmece

Ground motion simulations play an essential role in assessing seismic hazard and risk for a given region. In this study, we used the open-source algorithm EXSIM (Motazedian and Atkinson 2005; Boore 2009; EXSIM 2018), which has been validated in previous studies, to generate ground motion scenarios for Büyükçekmece, a region located in the western part of Istanbul, Turkey. The EXSIM algorithm accounts for finite-fault source models and stochastic simulation approaches, critical for accurately simulating realistic earthquake scenarios.

4.2.1 Source and path parameters used in the validation study

The validation exercise for the EXSIM algorithm, by utilizing five selected earthquakes was conducted in Süleyman and Çaktı (2021) and updated in later stages (Süleyman and Çaktı 2022). These five earthquakes considered in the validation process ranged in magnitude from MW 4.0–5.8 and occurred in the Marmara Sea region (Fig. 2). The selection of these events aimed to capture the diversity of the seismic sources in the region–such as, the diversity in epicentral locations, magnitudes, recording locations, and observed ground motion levels were a few of the selection criteria (Süleyman and Çaktı 2020).

Fig. 2
figure 2

The five earthquakes selected for the validation study, their magnitudes, and the dates of their occurrence (Süleyman and Çaktı 2020, 2022)

Full size image

For the largest earthquake (MW 5.8), a finite-fault source model developed by Karabulut et al. (2021) was used, while for the smaller events, point sources were utilized due to their relatively small rupture sizes. The validated set of parameters used in this study is similar to the ones used in Yenihayat et al. (2023) to validate the stochastic simulation algorithm for the region of Istanbul. Similarly, Infantino et al. 2020 used the same earthquake recordings to validate their 3D velocity model.

Table 1 summarizes the path, site, and source effect parameters used in the validation exercise. These parameters were globally and regionally developed, with particular attention paid to the Istanbul and Marmara region. The bottom lines of Table 1 relate to the source parameters of the five earthquakes included in the validation study.

Table 1 Summary of the parameters used in the validation exercise
Full size table

In order to accurately model the ground motion in the Büyükçekmece region, a site-specific approach was adopted, which involved the use of amplification factors generated for each of the 42 observation locations. In previous studies, a generic amplification model based on site classes had been used (Süleyman and Çaktı 2022). However, in this updated set of validated parameters, the Square-Root-Impedance (SRI) technique developed by Joyner et al. (1981) is employed for the amplification calculations. Further information on the application of this model can be found in the subsequent section.

Attenuation at high frequencies was measured based on the approach of Anderson and Hough (1984) in a previous study by Süleyman and Çaktı (2018). However, in the current study, to account for frequency-dependent anelastic attenuation, the model proposed by Horasan et al. (1998) was used. This model observes the spectral decay of S-waves at selected central frequencies over Fourier amplitude spectra. To describe geometrical spreading, the model of Akinci et al. (2013) was used. The duration of the simulated ground motions was scaled according to the model suggested by Boore and Thompson (2014). This validated model was then used to conduct scenario ground motion simulations for the metropolitan region of Istanbul in Süleyman and Çaktı (2022).

4.2.2 Ground motion simulation development using site-specific SRI method

The SRI method, commonly referred to as the Quarter-Wavelength Approximation, is a widely recognized technique used for estimating approximate linear site amplifications in simulations of ground motions and for comparing seismic velocity estimates at a single site (Boore and Joyner 1997; Boore 2003; Brown et al. 2002; Stephenson et al. 2005).

The SRI method offers distinct advantages over alternative approaches for calculating site amplifications. Notably, it is characterized by its efficiency, relying solely on a simple velocity model of the site. It assigns an amplification frequency to each depth and effectively smoothens multiple peaks. However, the SRI approach consistently provides lower estimates for the peaks observed in resonant systems and for gradient models lacking significant impedance differences, across a broad spectrum of frequencies. In the case of two gradient models commonly utilized in stochastic ground-motion simulations to derive crustal amplifications, this underestimation is limited to a maximum of approximately 30%. This margin is comparable to the uncertainties associated with several parameters utilized in simulations, and significantly smaller than the typical uncertainties in observed ground motions. Furthermore, Thompson et al. (2011) and Boore (2013) demonstrated that site responses computed using the SRI method exhibit closer alignment with observed site responses compared to results from the “full-resonant” method. Additionally, this study advocates for the adoption of the SRI method in constructing site response functions for average velocity profiles.

The method's core methodology entails estimating average seismic parameters, with a specific emphasis on the quarter-wavelength velocity (VsQWL), up to a depth denoted as z(f), equivalent to one-quarter of the wavelength associated with the frequency (f) of interest. Amplification factors for a given frequency are then computed by deriving the square root of the impedance ratio between the average depth and a reference point. To calculate the quarter-wavelength velocity corresponding to a particular frequency, a process of travel-time averaging across the input profile is employed by minimizing the expression as presented in Eq. 1 and 2.

$$\begin{array}{*{20}c} {\arg \min } \\ {z\left( f \right)} \\ \end{array} \left\|z\left( f \right) - \frac{{V_{s}^{QWL} \left( {z\left( f \right)} \right)}}{4f}\right\|$$
(1)

given that:

$$V_{s}^{QWL} \left( {z\left( f \right)} \right) = z\left( f \right)\left( {\mathop \int \limits_{0}^{z\left( f \right)} \frac{1}{{V_{s} \left( {z\left( f \right)} \right)}}dz\left( f \right)} \right)^{ - 1}$$
(2)

The depth-to-bedrock values and assigned shear-wave velocities were used in conjunction with the 1-D crustal shear-wave velocity model utilized by Karabulut et al. (2021) to generate unique amplification curves for each location. By adopting a site-specific approach and incorporating validated models for amplification, attenuation, and spreading, the simulations presented in this study are expected to offer a more precise representation of ground motion in the Büyükçekmece region.

For comparison, amplification factors computed for amplitudes exceeding 5 Hz at a site with a VS,30 value approximating 260 m/s equate to 4.32, whereas for a site with a VS,30 value approximating 340 m/s, the corresponding factor is calculated as 3.74. While it might be tempting to employ generic amplifications for each site class, this approach could introduce significant disparities in amplification factors, potentially resulting in overestimation or underestimation of stochastically simulated motions. Consequently, this study employs site-specific amplification factors to ensure a closer alignment between the simulated motions and the actual seismicity.

4.3 Scenario ground motion simulation development

As previously indicated, the authors of this study have emphasized seismic gaps along the North Anatolian Fault (NAF), specifically by developing earthquake scenarios for the eastern portion of the central Marmara segment (Source #1) and the Princes' Islands segment (Source #2) of the NAF. Source #1 and Source #2 have the potential for ruptures spanning approximately 100 km and 35 km, respectively. It is noteworthy that ruptures in either of these sources may trigger ruptures in the other, given that these segments intersect.

A total of 50 simulations were conducted on the two seismic sources mentioned above–as separate rupture locations and jointly. Each of these scenarios represents a unique scenario with different combinations of magnitude, rupture, and hypocenter locations. “Appendix B” summarizes all the source parameters used in these simulations.

Simulations from scenario number 26 to 40 were conducted at Source #1 (Fig. 3), with magnitudes ranging from MW 7.2 to 7.4. Simulations from scenario number 1 to 25 took place at Source #2, (Fig. 3) involving MW 6.3 to MW 6.7 events. The simulations from number 41 to 45 involved combinations of MW 7.4 and MW 6.4 events, and the final five simulations involved combinations of MW 7.4 and MW 6.7 events with—the larger event triggering the latter one, forming larger magnitude events of MW 7.41 and MW 7.42, all taking place at two source locations consecutively. In all these scenarios, the hypocenter location was randomly assigned.

Fig. 3
figure 3

Source locations of the scenario ground motion simulations

Full size image

At all the ruptures considered slip weights are randomly generated at each sub-source of each simulation. The simulations only involved rectangular rupture plane scenarios due to the limitations of the EXSIM tool, which does not allow for rupture models with irregular shapes. The dip angles for all events were set at 90 degrees, given the predominantly horizontal nature of the faulting in the NAF. Stress values were set at 100 or 130 bars, as previously investigated in Süleyman and Çaktı (2022). The hypocenter depths of the simulated earthquakes range from 5.3 to 17 km, despite their expected shallow depths.

The simulations in this study were conducted for the 0.05–5 s (0.2–20 Hz) range. The simulated mean horizontal S-wave motions for each observation location considered in the Süleyman et al. (2023) study are presented as a result.

4.4 Accounting for variability in hypocenter location and stress drop values

The hypocenter location and stress drop values are recognized as two of the most uncertain parameters in the context of modeling ground motions. However, these parameters can significantly influence the ground motion values observed at near-source locations. The quantification of this uncertainty holds paramount importance in comprehending the range of ground motion levels, thereby enhancing the meaningfulness of outcomes produced through ground motion simulations. Despite the incorporation of diverse scenarios involving varying magnitudes and multiple hypocenter locations within this study, the researchers also undertook a probabilistic assessment of the aforementioned variability in hypocenter and stress drop.

In order to characterize the variability of ground motion at each observation point, a wide range of stress drop values was defined for each of the considered magnitudes. Consequently, for each of the selected stress drop values a hundred randomized scenarios were generated, each involving randomized hypocenter locations. This approach facilitated the quantitative analysis of this variability, leading to a more precise representation of the ground motion models.

5 Results

For each simulation, we obtained peak ground acceleration (PGA), peak ground velocity (PGV), and spectral accelerations (SA) at various periods, in addition to the mean horizontal S-wave motions. We compared our estimations with two ground motion prediction equations (GMPEs) that involve adjustment factors for Turkey, BSSA14 (Boore et al. 2014) and ASK14 (Abrahamson et al. 2014). The PGA and PGV estimations were compared against the GMPE curves derived for these simulated values, as presented in Fig. 4.

Fig. 4
figure 4

PGA and PGV from all simulations against RJB (red dots); mean (solid lines) ± 1 std (dashed lines) curves derived by utilizing ASK14 GMPE (left); mean (solid lines) ± 1 std (dashed lines) curves derived by utilizing BSSA14 GMPE (right). The left clouds in each graph belong to the MW 7.2–7.42 range, while the right clouds belong to the MW 6.3–6.7 range estimations. The black lines represent reference curves for VS,30 of 300 m/s and MW 7.3 case, while the red lines represent reference curves for VS,30 of 300 m/s and MW 6.5 case

Full size image

Each graph in Fig. 4 includes two mean GMPE curves derived for two sets of magnitudes: one for scenarios with magnitudes MW 7.2–7.42 and the other for scenarios with magnitudes MW 6.3–6.7. The VS,30 value used to generate these curves was selected as 300 m/s due to the close proximity of the observed VS,30 values at all sites used for these simulations (“Appendix A”). The Joyner-Boore distances of the MW 7.2–7.42 events range from 17 to 19 km, while for the magnitude 6.3–6.7 events, it ranges from 31 to 36 km. It is noteworthy to mention that the PGA and PGV values for the MW 7.2–7.42 events are higher than the mean curve, whereas those for the MW 6.3–6.7 events are lower than the mean curve generated for these events. Moreover, most of these values fall within ± 1 standard deviation of the means of the used GMPEs. The high decay of ground motion values for the lower magnitude range can be attributed to the high attenuation of seismic waves, as observed and discussed in the recent study by Malcioglu et al. (2022), particularly focusing on the MW 5.8 Silivri earthquake. However, as mentioned earlier, the area is predominantly characterized by soft soil conditions, consisting of thick layers of unconsolidated sediments. These sediments act as traps, intensifying seismic waves and significantly amplifying ground shaking during earthquakes. The depth of the bedrock also plays a crucial role, with shallower bedrock resulting in greater amplification. Overall, the PGA levels for the MW 7.2–7.42 and MW 6.3–6.7 scenario events range between 313–490 cm/s2 and 55–93 cm/s2, respectively, while the PGV levels range between 29–50 cm/s and 4–8 cm/s. In comparison, Infantino et al. (2020) provided PGV estimations via PBSs for MW 7.2 and MW 7.4 at a VS,30 of 650 m/s, which closely align with the PGV estimations in our study for equivalent distances. However, our region's VS,30 stands around 300 m/s. Notably, within our study area, Infantino et al.'s (2020) PBS approach notably overestimates PGVs in contrast to those estimated by Cauzzi et al. (2015). Stupazzini et al. (2020) present comprehensive maps delineating 975 years of return period PGV values for their footprint and GAF estimations across much of Istanbul. Remarkably, their estimations for this region are in strong concordance with ours for the MW 7.2–7.42 range. Zengin and Çaktı (2012) employ stochastic simulations for Istanbul, offering estimations for VS,30 of 270 m/s for the MW 7.4 case. Surprisingly, both their PGA and PGV estimations are lower than ours. Despite employing a rupture scenario akin to our MW 7.4 case, their set of input parameters diverges substantially from ours. Notably, a key factor contributing to this variance is likely the disparity in stress drop values; they utilize 100 bars, whereas our study incorporates a stress drop of 130 bars for the MW 7.2–7.4 range. This parameter holds significant sway over the ground motion values derived through the EXSIM algorithm.

Figure 5 discusses whether any geographical dependencies exist in the study area within PGA and within the SA value distributions on the map, considering the magnitude 7.4 simulation (scenario number 36 in “Appendix B”). Therefore, no significant trend is observed. This figure also depicts the inter-period acceleration change. It is evident from the figure that at 0.2 s SA values exceed 1.2, whereas SA values decrease substantially at 2 s.

Fig. 5
figure 5

The geospatial distribution of PGA and SA at 0.2, 1 and 2 s

Full size image

Specifically, we computed SA values for 49 logarithmically spaced periods for 50 simulations for each location. To facilitate meaningful comparisons, the computed SA values were plotted using 3D bar charts. An illustration of the SA spectra (0.5–5 s) for all the realized simulations at location #17 (“Appendix B”) is depicted in Fig. 6. The results indicate a systematic variation in the SA spectra as the magnitudes of the simulations decrease. As expected, the highest SA values were observed for MW 7.42 event simulations, while MW 7.3 and MW 7.2 scenario events also showed significantly high values. Our analysis revealed that the periods between the 0.06–0.9 s range are likely to exhibit the highest expected SA values.

Fig. 6
figure 6

(top) SA at all periods, for all realized simulations at location #17; (bottom-left) The highest synthetic SA values for all periods; (bottom-right) The highest synthetic SA values based on each scenario

Full size image

The study conducted by Beauval et al. (2009) aims to develop a hybrid methodology for computing probabilistic seismic hazard, by employing an empirical Green’s function (EGF) simulation technique. The researchers analyze ground-motion variability at various stations, with a particular emphasis on understanding the sigma values. The researchers generated a large number of acceleration time histories based on the EGF method. By comparing these simulations with observed spectra and predictions from ground-motion models, the study highlights discrepancies and underscores the importance of sigma in hazard assessment. As the results of our study point out a significant uncertainty around ground motion values, we attempted to model this uncertainty by keeping the variability of stress parameter and the hypocenter location in the center of our attention. To deepen our analyses, we followed an approach similar to that of Beauval et al. (2009). We decided to increase the number of simulations drastically at this stage of the study. Such as, for the simulation scenario number 36 of MW 7.4 in “Appendix B”, we defined a wide stress drop range from 100 to 160 bars, with a 5-bar increment. For each of these increasing stress values a hundred simulations were realized to parametrize the uncertainty, in total yielding 1300 simulations for this exact location and magnitude. In each of these simulations, the hypocenter location was randomly generated within the defined rupture plane. This process was repeated for all the other scenarios and observation locations.

Figure 7 shows the PGA and PGV occurrence frequency distribution amongst 1300 realized simulations and their corresponding probability density (PDF) overlaid on the frequency distribution histogram. The PGA and PGV frequency distributions clearly show normal distribution observed in most of the other locations and scenarios, as well as in the case demonstrated in Fig. 7. Therefore, the PDF is modeled to observe the probability of occurrence of ground motion parameters for a pre-defined moment magnitude and rupture location, while stress drop and hypocenter location are the variables. It is also important to mention that sigma levels cover relatively wide PGA and PGV ranges, even when a single magnitude and the same observation location is considered. The phenomenon of a wider range of ground motions is caused due to high uncertainty of the stress value and hypocenter location, especially when such relatively large ruptures are considered.

Fig. 7
figure 7

(top) PGA occurrence frequency distribution amongst 1300 realized simulations with overlaid PDF distribution, for the simulation case #26, at location #1; (bottom) PGV occurrence frequency distribution amongst 1300 realized simulations with overlaid PDF distribution, for the simulation case #26, at location #1

Full size image

For this particular case shown in Fig. 7, the PDF shows a mean value of 422 cm/s2 for PGA and 44 cm/s for PGV. Stupazzini et al. (2020), follow a similar approach to check the ground motion distributions through histograms and lognormal distributions. As a comparison, for the MW 7.4 case, they develop 20 PBS scenarios. For the location of Hagia Sophia of Istanbul at 20 km distance from the NAF and VS,30 of 650 m/s, they observe mean PGA and PGV values of 309 m/s2 and 46 m/s, respectively through the PBS approach. As their mean estimation for PGA is lower than our estimation, the PGV estimation is nearly the same.

6 Conclusions

The NAF is an area with a high potential for seismic activity, and Büyükçekmece, a densely populated district in Istanbul, is particularly vulnerable to earthquake damage. This study aimed to use the EXSIM stochastic ground motion simulation approach to model potential earthquakes in the region and gain a better understanding of the impact that medium to large magnitude earthquakes could have on the district. The accuracy of the EXSIM algorithm was validated by simulating five past earthquakes in the region in a preceding study.

By using scenario ground motion simulations, we identified the potential earthquake hazard in an approximately 6 km2 area in Büyükçekmece. The site-specific approach adopted in this study provided unique amplification curves for each location, which were used to carefully model the ground motion in Büyükçekmece. The use of finite-fault source models and site-specific parameters allowed for the precise simulation of realistic earthquake scenarios. Our investigations revealed that the site-response in this part of Büyükçekmece is relatively similar, which resulted in similar estimations of ground motion levels at all locations. However, it is worth understanding the variations in ground motions, even though they are small, for a more accurate hazard assessment.

In this study, we presented practical ways to observe (1) the effect of magnitudes mostly for the magnitude increments of 0.1 and 0.01 on ground motion levels along a wide range of periods (0.05–5 s); (2) the importance of modeling the uncertainty caused by the stress parameter and hypocenter location. Involving scenarios differing in magnitudes, stress parameters and hypocenters allowed us to observe that the ground motions may significantly vary when different scenarios are applied. Moreover, difficulties in predicting the hypocenter location and stress values cause a significant range of uncertainty that the researchers of this study suggest to be carefully considered in such studies.