Fragility-based seismic assessment of traditional masonry buildings on Azores (Portugal) using simulated ground-motion records

Research Paper Earthquake Spectra 2024, Vol. 40(4) 2836–2861 ÓThe Author(s) 2024 Article reuse guidelines: sagepub.com/journals-permissions DOI: 10.1177/87552930241256287 journals.sagepub.com/home/eqs Fragility-based seismic assessment of traditional masonry buildings on Azores (Portugal) using simulated ground-motion records Vasco Bernardo , Shaghayegh Karimzadeh , Daniel Caicedo, Sayed Mohammad Sajad Hussaini , and Paulo B Lourencxo Abstract Traditional unreinforced masonry structures are extremely vulnerable to seismic events, featuring large losses in many countries worldwide. This study focuses on the seismic assessment of a traditional masonry structures located in Faial Island— Azores (Portugal), which was hit by an earthquake of M w = 6.2 on 9 July 1998. A set of analyses was conducted through a probabilistic performance-based seismic approach, employing a stochastic finite-fault ground-motion simulation method to derive region-specific records and nonlinear numerical models to compute the capacity of the building. Subsequently, analytical fragility curves were derived considering different seismic scenarios. The results show a significant probability of the structural typology to reach moderate to extensive damage in the case of the specific earthquake, as observed in reality. The study holds significant importance in regions with limited recorded seismic activity, as they offer potential benefits for seismic risk management and mitigation strategies. Keywords Masonry structures, Faial Island (Azores), ground-motion records, numerical fragility curves, seismic assessment Date received: 18 May 2023; accepted: 21 April 2024 Introduction The Azores archipelago is composed of nine volcanic islands located in the mid-Atlantic (see Figure 1a), approximately 1500 km west of Portugal. This region is situated on the Department of Civil Engineering, ISISE, ARISE, University of Minho, Guimara ˜es, Portugal Corresponding author: Vasco Bernardo, Department of Civil Engineering, ISISE, ARISE, University of Minho, Guimara ˜es 4800-058, Portugal. Email: [email protected] boundary between the North American and Eurasian tectonic plates characterized by frequent seismic activity. The seismicity of the Azores plateau can be divided into three main areas: (1) the eastern group of islands composed of Sa ˜o Miguel and Santa Maria; (2) the central group, which includes Terceira, Graciosa, Sa ˜o Jorge, Pico and Faial; (3) the western group, which comprises Flores and Corvo islands. The eastern group (1) is the most seismically active region, with frequent small earthquakes and occasional large events. The central group (2) is also seismically active but experiences fewer earthquakes than the eastern group, while the western group (3) experiences the least seismic activity. Despite some regions of the Azores archipelago being currently considered in the range of low-to-moderate seismic intensity levels, they have been affected in the past by large earthquakes. Faial Island was hit by an earthquake with M w = 6.2 in 1998 (Morais et al., 2021). In this regard, the nature of the observed damage was mainly associated with the traditional and predominant building typology in that region, which is mostly characterized by stone masonry walls and flexible timber floors, making these structures highly vulnerable to seismic actions (Costa and Areˆde, 2006). Following the earthquake, several inspections were conducted in the buildings, which determined that roughly 50% required minor repairs, 30% required extensive repairs, and 20% needed to be demolished (Neves et al., 2012). Some examples of damage suffered by buildings are depicted in Figure 1b and c. The past major event in Faial Island raised the interest of researchers in evaluating the seismic damage of different structures in the region and in predicting their behavior for future earthquakes (Bernardo et al., 2022a). In this sense, several studies can be found in the literature regarding the seismic performance of different structures. For instance, the research carried out in the work by Guerreiro et al. (2000) identified and described the damage modes and possible collapse mechanisms of 30 masonry churches based on in situ observation. The observed damage was correlated with the structural typology, seismic action level, quality of masonry, and the existence of previous retrofit or structural interventions. Likewise, the work of Neves et al. (2012) was conducted in three stages, including a comprehensive description of the building stock; a damage grade classification based on the observed damage mechanism; and finally, the seismic vulnerability assessment of the building stock. A probability of collapse of about 30% was found for moderate-tohigh intensities (VII and VIII within the European macroseismic scale (EMS) 98 scale) (Gru ¨nthal, 1998). Ferreira et al. (2017) addressed the seismic vulnerability of the old city center of Horta in Faial Island through the analysis of 192 buildings using a simplified vulnerability assessment method. The authors estimated the number of collapsed and unusable buildings considering different intensities and associated their results with the low Figure 1. (a) Azores archipelago. (b) and (c) Examples of damage suffered by buildings in the 1998 Faial earthquake with M w = 6.2 (Costa et al., n.d.; Neves et al., 2012). Bernardo et al. 2837 resistance and high vulnerability of the building stock. More recently, Maio et al. (2017) assessed the seismic vulnerability of two stone masonry buildings in the Faial Island using the N2 Method (Code, 2005) and the Capacity Spectrum Method (CSM) (Council, 1996), for a set of ground-motion records representative of the 1998 Faial earthquake. The authors highlighted the potential application of retrofitting solutions considering the inplane seismic response. It should be noted that most of these investigations have been limited to the availability of recorded accelerograms and post-damage in situ observations. In this regard, reliable seismic risk assessment demands region-specific hazard definition and proper characterization of earthquake ground motions. Given the lack of recorded accelerograms characteristic of large-magnitude events in regions of moderate levels of seismic hazard, the use of simulated records is a suitable alternative. Ground-motion simulations can be employed as input for vulnerability models to predict damage scenarios at urban scale (Antonietti et al., 2021). The seismic performance assessment of structures requires region-specific groundmotion definition. Given the lack of recorded accelerograms characteristic of largemagnitude events in regions of moderate levels of seismic hazard, the use of simulated records is a suitable alternative. Recent studies have implemented simulations and validated their use in earthquake engineering practice. Among them, Zonno et al. (2010) implemented the stochastic finite-fault method (Motazedian and Atkinson, 2005) to reproduce synthetic time series representative of the islands affected by the 1998 Faial Earthquake in the bedrock by validating the simulations at one near-fault station. The simulated data were used for damage estimation by comparing the mean damage index (Dolce et al., 1999) and macroseismic method (Lagomarsino and Giovinazzi, 2006). Burks and Baker (2014) focused on the validation of ground-motion simulations by proposing a list of proxy parameters for the response of engineering systems, including the correlation of spectral acceleration (Sa) across periods, the ratio of maximum-to-median Sa across all horizontal orientations, and the ratio of inelastic-to-elastic displacement. They emphasized that the interperiod correlation structure of ground-motion simulations should align with that of an empirical model, such as the work by Baker and Jayaram (2008). Similarly, Dreger et al. (2015) compared simulated and empirical residuals of RotD50 peak spectral accelerations (PSAs) for 12 earthquake events. Later, Bijelic ´et al. (2018) assessed the performance of two tall buildings (20 and 42 stories) under comparable sets of simulated and recorded motions, leading to the identification of bias induced by differences in correlations of spectral values in some of the stochastic ground-motion simulations. Bayless and Abrahamson (2018) directly investigated the impact of interperiod correlation on groundmotion simulations in structural assessment, underscoring the inadequacy of tested stochastic finite-fault simulations in addressing such correlations, which could lead to undesired variability in structural response or underestimation of dispersion and nonconservative estimations in risk assessment, prompting modifications to the stochastic point-source approach by Boore (1983) to enhance its ability to account for them. Wang et al. (2019) suggested an alternative method to enhance interperiod correlations in simulations through postprocessing techniques. Alternatively, Altindal and Askan (2023) demonstrated that accounting for uncertainty in the primary input-model parameters of scenario-based simulations implicitly addresses variability similar to empirical models. However, they observed that their simulations tend to overestimate interperiod correlation compared to existing literature (Baker and Jayaram, 2008), potentially due to the flattened spectral amplitudes of simulated motions when contrasted with real records. The authors suggested that this phenomenon could potentially lead to increased seismic demands, prompting a deeper investigation into using these ground motions as input for structural 2838 Earthquake Spectra 40(4) response history analysis. Finally, it is important to note that, as highlighted in the work by Burks and Baker (2014), the interperiod correlation gains particular significance in the context of seismic demands on multiple-degree-of-freedom systems, especially those subjected to excitation at multiple periods. Literature review reveals that limited studies have investigated the use of simulated motions for seismic assessment of masonry structures. Hoveidae et al. (2021) investigated seismic damage to a historic masonry monument in Tabriz, Iran, for different scenario earthquakes using stochastically simulated ground-motion records, while Karimzadeh et al. (2021) addressed the influence of ASCE/ SEI 7–16 (Loads, 2017) and Eurocode 8 (EC8) (Code, 2005) ground-motion selection criteria for estimating the seismic demand of masonry structures considering both real and simulated suites of records. Karimzadeh et al. (2024a) conducted a recent investigation into the efficacy of simulations of the 1998 Faial earthquake through site-based (Rezaeian and Der Kiureghian, 2010) and finite-fault (Atkinson and Assatourians, 2015) stochastic approaches against real records for seismic assessment of masonry prototypes. Despite differences in engineering demand parameters (EDPs) estimated through real and simulated records, the authors noted that simulated records show promise for estimating seismic demands on these types of engineering structures. However, they emphasized that the validation of simulations is highly dependent on the specific application and that the accuracy of simulations may vary depending on the parameters under analysis. The scope of this work is deriving fragility curves of traditional masonry buildings located on Faial Island through structural numerical analysis and simulated ground motions. This study represents the first comprehensive investigation of its kind in this region, providing essential information to support future risk studies. The obtained results in terms of expected damage are compared with the post-earthquake damage observations in 1998 to validate the methodology employed. Due to the scarcity of real records corresponding to the instrumental area in the region, simulated records are alternatively employed in this study. The simulateddatasetsusedinthisstudyarederivedfromtheworkbyKarimzadehetal.(2024b), where the stochastic finite-fault method based on dynamic corner frequency algorithm (Atkinson and Assatourians, 2015; Motazedian and Atkinson, 2005) is implemented to generate region-specific records. Subsequently, ground-motion selection is conducted from these simulated record datasets for various scenario events of magnitudes ranging from 5.0 to 6.6 accounting for dispersion in terms of peak ground acceleration (PGA). Nonlinear numerical models are then constructed for different buildings to estimate their seismic capacity. Subsequent analyses are conducted through a probabilistic performance-based seismic approach to compute the response of the buildings for different seismic scenarios. The computation of uncertainty in the seismic demand is estimated using cloud analysis. Finally, numerical fragility curves are derived considering different seismic scenarios, demonstrating a large probability of these structures to reach moderate to extensive damage stages for large-magnitude events. The results show a satisfactory prediction from the point of view of numerical damage estimates and those observed in situ. Seismic action modeling This study focuses on the Azores plateau in Portugal as the study area for the modeling of seismic action. This region is dominated tectonically by the Azores triple junction between the North American plate, the Eurasian plate, and the African plate. The central and eastern parts of the plateau are characterized by a diffuse and complex deformation zone, which underwent shearing under a dextral trans-tensile regime. More in detail, the study Bernardo et al. 2839 area comprises only five islands, namely Faial, Pico, Sa ˜o Jorge, Terceira, and Graciosa. The following sections describe the ground-motion simulation and selection approaches with the corresponding results. Ground-motion simulation approach and results This study employs the simulated dataset generated in the work by Karimzadeh et al. (2024b) based on the stochastic finite-fault ground-motion simulation methodology (Atkinson and Assatourians, 2015). In this approach, the simulation algorithm, proposed in the work by Motazedian and Atkinson (2005) and based on the original FINSIM code by Beresnev and Atkinson (1998), is improved from the suggestions of Boore (2009). This approach considers multiple factors, such as earthquake magnitude, fault geometry, slip distribution, density, rupture velocity, and strike to identify the fault rupture. The method combines the source contribution with attenuation parameters and site effects to obtain the seismic signal in the time domain at any observation site. In the stochastic finite-fault method, the ruptured fault plane is divided into smaller sub-sources represented as a grid, with each sub-source assumed to be a point-source with a v 22 source spectrum as proposed in the work by Boore (1983). The sub-sources rupture with a time delay that depends on their distance from the hypocenter, and the time domain summation of the contributions from the delayed sub-sources is carried out to obtain the final ground-motion simulation (Atkinson et al., 2011). It should be emphasized that this method offers only one random horizontal component for the ground-motion records. The study by Karimzadeh et al. (2024b) performed the simulations within the EXSIM12 platform (Atkinson et al., 2011) to model the acceleration time series of the scenario events. In their study, 23 scenario events featuring varying magnitudes M w ranging from 5.0 to 6.8 with a bin size of 0.1 were simulated due to the rupture of different active fault planes in the region. Madeira et al. (2015) found that the fault length indicates a maximum expected magnitude of 6.8 within the region. It is important to note that these simulations were not limited to a single fault but instead involved various active faults in the Azores Plateau, as referred to in Figure 2, thus covering the aleatory region-specific uncertainty using the proposed framework. All simulations were accomplished in 359 nodes inside all islands with a grid size of 1 km (as shown with triangular symbols in Figure 2) and were performed in the bedrock. Region-specific input parameters provided in the work by Karimzadeh and Lourencxo (2022) and Karimzadeh et al. (2024a), based on simulation validations against observed motions from the 1998 Faial (M w = 6.2) event, were adopted to calibrate the simulation parameters for these events. However, to account for the uncertainty associated with the parameters representing source and attenuation effects, key parameters such as hypocenter location, stress drop, pulsing percentage, quality factor, and kappa were handled as random variables. Regional models with probability distribution functions (PDFs) and their ranges were implemented using the study by Carvalho et al. (2016). For every event, 30 Monte Carlo simulations were performed, each with distinct combinations of input parameters. Therefore, each scenario event was simulated using 30 different random input-model parameters, resulting in a total of 690 simulations performed at 359 sites. Further information can be found in the research conducted by Karimzadeh et al. (2024b). The results of simulations carried out for different scenario events in the region effectively demonstrate the inclusion of aleatory uncertainty through stochastic simulations. This aspect is further exemplified by computing the 5% damped pseudo-Sa values of the 2840 Earthquake Spectra 40(4) simulated dataset for randomly chosen stations situated at the center of each island. These computations are performed for a sample scenario event with the largest magnitude (M w = 6.8) due to a fault rupture in Faial Island, and the resulting Sa values are plotted Figure 2. The study area showcasing tectonic plates, with triangular symbols indicating the used stations for ground-motion simulations. The black dots in each fault show the faulting mechanism, which is normal. Figure 3. Damped (5%) Sa for stations in the center of Azores Islands. Bernardo et al. 2841 in Figure 3. A comprehensive representation of the Sa values at each station is given, including the individual mean, median, standard deviation (std), minimum (min), and maximum (max) values for all simulations. This information is useful in providing a clear understanding of the variability in the dataset and the potential range of Sa values that could be observed at each station. The range of probable spectral values demonstrated by the plotted Sa values serves as evidence of the stochastic behavior of the dataset. This observation is similarly observed across other scenario events due to ruptures occurring in other active faults within the region. Moreover, as the distance from the nearest island, such as Faial, to the farthest one, such as Terceira, from the ruptured fault plane increases, the peak value of Sa shifts toward longer periods, and there is also attenuation in terms of the amplitudes. This pattern aligns with the physical characteristics of distance-dependent damping in ground motions. Finally, it is crucial to emphasize that, because of the scarcity of empirical data and the lack of a suitable backbone ground-motion model, further validation of simulations (e.g. interperiod correlation) seems challenging. As indicated in the research by Karimzadeh et al. (2024b), the simulated dataset from the stochastic approach faces challenges in precisely capturing the coherent, long-period pulses that impact the features of ground motions near faults, particularly in the context of peak ground velocity (PGV) and PSA at extended periods. However, given that this investigation focuses on masonry structures, primarily characterized by short periods, it is believed that using these record sets poses minimal discrepancies for their seismic assessment, as supported by findings in the research conducted by Karimzadeh et al. (2024a), where an engineering validation of the simulated dataset was carried out for masonry prototypes in this region. Yet, it is important to note that before implementing these datasets in broader engineering applications, it is imperative to conduct additional validations, taking into account inherent limitations of stochastic finite-fault approach. Furthermore, it is essential to acknowledge the inherent case dependency in validation processes and emphasize the importance of employing a robust approach. In addition, recognizing the challenges in accurately capturing interperiod correlation of real records through stochastic simulations is crucial, as it can introduce uncertainty in seismic demand estimation, especially when multiple periods of the structure are excited. Ground-motion selection approach and results At this stage, sets of simulated records are selected for M w values of 5.0, 5.1, 5.2, 5.5, 5.6, 5.8, 6.0, 6.2, 6.4, and 6.6, respectively. Each suite corresponds to a specific value of M w in which 10 records are selected to cover the PGA dispersion for each scenario earthquake. On top of that, it has been extensively demonstrated how masonry constructions are highly sensitive to acceleration intensities, normally correlated with the PGA metric (Calderoni et al., 2016; Koc et al., 2023; Penna et al., 2014b; Tomic ´et al., 2021). Figure 4 shows the Sa of the 10 sets of the selected records, covering the distribution of PGA for each scenario earthquake. Changes in spectral amplitudes are clearly noted within all 10 scenarios selected for the analysis. Spectral amplitudes sequentially increase as with M w , reaching PGA values larger than 1000 cm/s 2 for some of the simulations with M w ø6.2. Furthermore, the distribution of spectral amplitudes is consistent with the observations after 9 July 1998, earthquake in the Faial Island (M w ø6.2; Catita et al., 2005; Matias et al., 2007). It should be noted that the accuracy of the simulation methodology, adopted in this research to characterize the seismic action, was formerly examined in the work by Karimzadeh et al. (2024a) from an engineering standpoint. This research showed the 2842 Earthquake Spectra 40(4) Figure 4. Sets of selected records for each scenario earthquake. Bernardo et al. 2843 consistency of the simulated dataset (Karimzadeh et al., 2024b) in terms of structural response predictions of monumental masonry against real records available from four stations near the site of the event (epicentral distances less than 150 km). However, broader applications in engineering practice require additional validations, considering inherent limitations. Structural capacity and seismic response This section defines the representative building typologies analyzed in the scope of the present work and the corresponding tridimensional numerical models. The seismic response of the buildings is computed for different scenarios presented in the previous section by considering a performance-based approach. Building typology description The traditional construction in Faial reflects the island’s history and cultural heritage, which is an important part of its identity. The building stock on the island has evolved over time to meet the practical needs of the local population and can be classified according to various construction system types, including the type of vertical resisting structure, type of floor, and roof structure. A clear characterization of the building stock in Faial Island is presented in the work by Neves et al. (2012). In the scope of this study, the construction type analyzed corresponds to the Unreinforced Masonry (URM) buildings composed essentially of stone masonry load-bearing walls and timber floors. This typology represents around 60% of the building stock in Faial Island and is considered one of the most vulnerable types of structures in case of seismic ground motions (Neves et al., 2012). The type of construction under interest involves buildings up to 3 stories high with regular geometry (see Figure 5). In rural areas, isolated buildings with a single story or 2story height are more typical, as the population density is generally lower and the need for housing is not as great. These types of buildings may serve as houses, agricultural buildings, or small commercial structures. In contrast, in urban areas, 2or 3-story buildings are more common and are typically enclosed in an aggregate. These buildings often serve as housing or mixed-use buildings that include both commercial and residential spaces. An important aesthetic aspect of these buildings is the size and arrangement of the openings in the facades that depend essentially on the location of the building (rural or urban), building purpose, and the social class that could also influence the architectural style. Usually, large windows are found in urban areas, while in rural regions, the windows are Figure 5. Example of the traditional building construction on Faial Island—Neves et al. (2012). 2844 Earthquake Spectra 40(4) corresponding regression coefficients. The values of bD depicted in Figure 13 can be easily obtained by the standard deviation of the logarithmic error between the analytical powerlaw function fitted and the empirical data. Note that, the range considered for the best analytical approximation to the set of data considered only the limit state DS4 (near collapse) to avoid the contribution of PPs for levels of deformation of the structure beyond this final limit state. Through analyzing the response of the structures in Figure 12 for different seismic scenarios with M w = 6.0, it can be seen that as the height of the building increases, the performance level tends to reach the maximum capacity defined by the near-collapse limit state (LS) more quickly, as the maximum strength of the buildings is also lower. The same trend is also found for buildings with lower deformation capacity/less ductility. This relies essentially on the ratio of masonry walls in the seismic action direction, as buildings with a higher ratio can exploit greater values of deformation and consequently have a better performance. Naturally, this observation is restricted to buildings governed by in-plane mechanisms, in which out-of-plane mechanisms are prevented from occurring. The dispersion values in the seismic bDdemand increase as the behavior of the structure seems to exploit large deformation values in the nonlinear response, which results, for example, in higher bDvalues with the increase in the number of stories or typologies with lower ductility. In other words, bDvalues also tend to increase as the performance of buildings approaches collapse, which can be explained by the high nonlinear response of structures and the greater variability of demand in this range. Seismic analytical fragility curves The seismic fragility curves derived in this section describe the probability of reaching or exceeding a given DSi (see Table 3) for a certain seismic intensity level, expressed in terms of Sdas the EDP, and considering a standard log-normal cumulative distribution function F, as shown in Equation 1: Figure 11. Damage pattern for the damage states considered: Archetype A3 and different stories. Bernardo et al. 2851 P(DSijSd)=F1 bDSi ln Sd Sd,DSi  ð1Þ where bDSiis the standard deviation of the natural logarithm of variable DSi. The values considered for bDSiincluded both uncertainty in the capacity bCand demand bD. The uncertainty in capacity was computed through the recommendations of FEMA (ATC, 2018) by combining the uncertainty associated with the level of building definition and construction quality (e.g. material properties, geometry, and constructive details) with the uncertainty in the completeness and quality of the analytical model. The knowledge/information about the buildings was considered limited, and the uncertainty regarding the quality of the numerical model was considered moderate. Thus, a dispersion for the capacity bC= 0.47 was adopted. The values of bDadopted were obtained in the previous section. A total dispersion bDSiwas obtained by combining both uncertainties (capacity and demand) through the square root of the sum of squares. Note that, the large values for the Figure 12. Seismic performance for the archetypes adopted and different scenarios of M w = 6.0. 2852 Earthquake Spectra 40(4) dispersion considered in the capacity, namely the dispersion value for the quality of the analytical models—limited quality (ATC, 2018), attempt to indirectly cover that some failure modes are not incorporated in the model, such as out-of-plane mechanisms. Figure 14 shows the fragility curves for the analyzed typologies and the median Sd (EDP) for the seismic scenarios of M w = 5.5, 6.0, 6.2, and 6.6. Note that M w = 6.2 was the magnitude of the Faial earthquake in 1998. In addition, fragility curves are also computed in terms of IM (PGA), see Figure 15, as a function of the limit states defined in the capacity curves and through the relationship between PGA-Sd(IM vs EDP) obtained in Figure 13. Therefore, the median PGA values are linked to the limit states of each archetype. These results can be used for loss estimation in the region under investigation. Table 4 summarizes the probability of occurrence of a given limit state in each typology for a certain M w considering the results of Figure 15. As can been seen, a significant Figure 13. PPs (Sd) for different seismic intensity levels corresponding to the different scenarios considered and seismic demand computation. Bernardo et al. 2853 probability of occurrence of moderate to extensive damage for seismic scenarios larger than M w = 6.0 is achieved, independently of the building typology. Notwithstanding, the typology that seems to have a better seismic performance, namely for higher seismicity levels, corresponds to the Archetype A2. In contrast, typologies with Archetypes A1 and A3 achieved a similar damage level. This mostly reflects the ratio between the area of walls and floor area, as previously discussed. Furthermore, the probability of occurring a given damage also seems to increase with the number of stories. For M w = 5.5, most of the damage is slight, with a low probability of occurrence of moderate damage (less than 15%). Note that the current standard for Portugal (EC8) considers a seismic action with a 10% probability of exceedance in 50 years (475-years return period) which roughly corresponds of an event with M w = 5.0 for the region of Faial, and therefore, the expected damage is lower than the one for an event with M w =5.5.For M w = 5.5, the average probability of occurrence of a given SD (1-DS0) between archetypes seems to increase with height; however, this conclusion is not so evident for higher Mw. In the case of a scenario with M w = 6.0, the likelihood of buildings not being damaged is low (34% maximum); the sum of the probability of attaining moderate-to-severe Figure 14. Seismic fragility curves in terms of EDP (Sd) for the archetypes analyzed. 2854 Earthquake Spectra 40(4) damage ranges from 44% to 69%, being the moderate damage the most expected. Finally, for the scenarios of M w = 6.2 and M w = 6.6, the probability of occurrence of severe and extensive damage is much higher: for instance, in the case of an event with M w = 6.2 (estimated M w of the Faial earthquake in 1998), a mean probability of 41% is achieved for extensive damage (DS4—near-collapse LS) by aggregating all the typologies considered. Note that, according to post-earthquake damage assessment (Neves et al., 2012), 30% and 20% of the buildings required major repairs or even needed to be demolished, respectively. Therefore, the results achieved for the M w = 6.2 are also in line with in situ observations. Conclusion This work derived fragility curves of traditional masonry buildings on Faial Island (Azores, Portugal) using structural numerical analysis and simulated ground motions for several scenarios, providing important information for future risk studies. The results were validated with the post-earthquake damage observations of 1998 Faial earthquake (M w = 6.2). The ground-motion records used in this study are sourced from the work by Karimzadeh et al. (2024b), encompassing a total of 23 region-specific scenarios with magnitudes ranging from 5.0 to 6.6, generated in bedrock through the stochastic finite-fault Figure 15. Seismic fragility curves in terms of IM (PGA) for the archetypes analyzed. Bernardo et al. 2855 methodology based on a dynamic corner frequency concept (Atkinson and Assatourians, 2015). Afterward, suites composed of 10 simulated records covering the PGA dispersion for each scenario defined by M w values (5.0, 5.1, 5.2, 5.5, 5.6, 5.8, 6.0, 6.2, 6.4, and 6.6) were selected. The seismic input was used to estimate the seismic performance of traditional buildings represented by different archetypes. The main assumptions and procedures for the probabilistic assessment of seismic performance considered are summarized below: (1) definition of the geometry and material properties of the representative buildings based on the synthetic database generated by Bernardo et al. (2022a) and the survey of Faial’s building stock carried out by Neves et al. (2012) and Costa et al. (2008); (2) development of numerical models where only in-plane behavior is considered; (3) estimation of seismic performance using the improved CSM method; it is worth mentioning that in the estimation of the structure’s performance through the CSM, some improvements can be made to lead with multiple solutions resulting from non-smooth single records (e.g. Nettis et al., 2021); (4) calculation of the relationship between PGA (IM) and Sd (EDP) to assess the dispersion of seismic demand (cloud analysis); and (5) derivation of numerical fragility curves to estimate the probability of occurrence of a given limit state. Table 4. Probability of occurrence of different limit states for different typologies and scenarios 2856 Earthquake Spectra 40(4) The results of the simulations have revealed that the stochastic behavior of seismic actions can be captured through simulations. The variability and unpredictability of seismic actions have been demonstrated to be sufficiently accounted for in these simulations. As a result, the use of simulations facilitates the selection of an adequate number of seismic records, even in regions where recorded motions are scarce or limited. The ability to rely on simulations to generate synthetic seismic records has proven to be a valuable tool for earthquake engineering, where the availability of accurate data is essential for the design and analysis of structures. The previous ground-motion simulations were used on three archetypes of buildings representative of the building stock in the urban regions of Faial with 1 to 3 stories high. The nonlinear seismic performance of these structures subjected to different scenarios was computed to investigate the expected damage. Subsequently, analytical fragility curves were derived, considering the uncertainty in the capacity and response of the structures, allowing to evaluate the likelihood of reaching different damage states for multiple M w scenarios. The results from the fragility analysis led to conclude that, regardless of the building typology, there is a significant probability of achieving moderate to extensive damage for seismic scenarios with M w larger than 6.0. Furthermore, a large probability (greater than 60%) was found for buildings to reach near-collapse performance, specifically above M w = 6.4 scenarios. Thus, there is a large vulnerability of these buildings when subjected to moderate seismicity levels. Furthermore, the results achieved are in line with the observation of post-earthquake damage in the region, where it was reported that around 30% and 20% of the buildings needed extensive repairs or even demolished, respectively, due to extensive damage suffered. Further studies accounting the soil amplification at surface should be developed to take account of site conditions and evaluate its effects on the buildings performance. Finally, it is worth noting the importance of these studies in regions where recorded motions are limited, providing potential benefits for seismic risk management and mitigation strategies since it is possible to capture the random characteristics of earthquakes and to predict the damage for a certain seismic scenario. However, before using ground-motion simulations, it is crucial to consider the inherent case dependency in validation processes, enhancing the robustness of the approach. It is also crucial to recognize the challenges in capturing appropriate interperiod correlation of real records through simulations, which can introduce uncertainty in seismic demand estimation, particularly where multiple periods of the structure are excited. Declaration of conflicting interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/ or publication of this article. Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/ or publication of this article: This study was funded by the STAND4HERITAGE project that has received funding from the European Research Council (ERC) under the European Union’s Horizon 2020 research and innovation program (grant no. 833123), as an Advanced Grant. This work was also partly financed by FCT/MCTES through national funds (PIDDAC) under the R&D Unit ISISE (reference UIDB/04029/2020). This work is partly financed by national funds through FCT— Bernardo et al. 2857 Foundation for Science and Technology (grant no. UI/BD/153379/2022) attributed to the fourth author. ORCID iDs Vasco Bernardo https://orcid.org/0000-0002-5071-8155 Shaghayegh Karimzadeh https://orcid.org/0000-0003-3753-1676 Sayed Mohammad Sajad Hussaini https://orcid.org/0000-0001-9841-5498 Data and resources The simulated dataset generated in the work by Karimzadeh et al. (2024b) based on the stochastic finite-fault approach was adopted for the purposes of this research (https://doi.org/10.1002/ eqe.4040). The numerical models representative of the building stock in the Azores (Portugal) were taken from the research of Bernardo et al. (2022a, 2022b) (https://doi.org/10.1007/s10518-02201434-8). References Altindal A and Askan A (2023) Construction and validation of a simulated ground motion dataset for Turkey. Journal of Seismology 27: 1047–1065. Antonietti PF, Mazzieri I, Melas L, Paolucci R, Quarteroni A, Smerzini C, and Stupazzini M (2021) Three-dimensional physics-based earthquake ground motion simulations for seismic risk assessment in densely populated urban areas. Mathematics in Engineering 3(2): 1–31. DOI: 10.3934/mine.2021012. ATC (2018) FEMA P—58—1: Seismic Performance Assessment of Buildings. Volume 1— Methodology. Washington, DC: Federal Emergency Management Agency (FEMAP-58). Atkinson GM and Assatourians K (2015) Implementation and validation of EXSIM (a stochastic finite-fault ground-motion simulation algorithm) on the SCEC broadband platform. Seismological Research Letters 86(1): 48–60. Atkinson GM, Goda K and Assatourians K (2011) Comparison of nonlinear structural responses for accelerograms simulated from the stochastic finite-fault approach versus hybrid broadband approach. Bulletin of the Seismological Society of America 101(6): 2967–2980. Baker JW and Jayaram N (2008) Correlation of spectral acceleration values from NGA ground motion models. Earthquake Spectra 24(1): 299–317. Barbat AH, Pujades LG and Lantada N (2008) Seismic damage evaluation in urban areas using the capacity spectrum method: Application to Barcelona. Soil Dynamics and Earthquake Engineering 28: 851–865. Bayless J and Abrahamson NA (2018) Evaluation of the interperiod correlation of ground-motion simulations. Bulletin of the Seismological Society of America 108(6): 3413–3430. Beresnev IA and Atkinson GM (1998) FINSIM—A FORTRAN program for simulating stochastic acceleration time histories from finite faults. Seismological Research Letters 69(1): 27–32. Bernardo V, Campos Costa A, Candeias P and Costa A (2022a) Seismic vulnerability assessment and fragility analysis of pre-code masonry buildings in Portugal. Bulletin of Earthquake Engineering 20(11): 6229–6265. Bernardo V, Campos Costa A, Candeias P, Costa A, Marques A and Carvalho A (2022b) Ambient vibration testing and seismic fragility analysis of masonry building aggregates. Bulletin of Earthquake Engineering 20: 5047–5071. Bijelic ´N, Lin T and Deierlein GG (2018) Validation of the SCEC broadband platform simulations for tall building risk assessments considering spectral shape and duration of the ground motion. Earthquake Engineering & Structural Dynamics 47(11): 2233–2251. Boore DM (1983) Stochastic simulation of high-frequency ground motions based on seismological models of the radiated spectra. Bulletin of the Seismological Society of America 73(6): 1865–1894. 2858 Earthquake Spectra 40(4) Boore DM (2009) Comparing stochastic point-source and finite-source ground-motion simulations: SMSIM and EXSIM. Bulletin of the Seismological Society of America 99(6): 3202–3216. Burks LS and Baker JW (2014) Validation of ground-motion simulations through simple proxies for the response of engineered systems. Bulletin of the Seismological Society of America 104(4): 1930–1946. Calderoni B, Cordasco EA, Pacella G and Onotri V (2016) Critical issues in the assessment of seismic vulnerability of historical masonry buildings: A study case. In: Modena C, da Porto F and Valluzzi MR (eds) Brick and Block Masonry: Trends, Innovations and Challenges (Proceedings of the 16th International Brick and Block Masonry Conference, IBMAC 2016). London: Taylor & Francis, pp. 611–621. Carvalho A, Reis C and Vales D (2016) Source and high-frequency decay parameters for the Azores region for stochastic finite-fault ground motion simulations. Bulletin of Earthquake Engineering 14(7): 1885–1902. Catita C, Feigl KL, Catala ˜o J, Miranda JM and Victor LM (2005) InSAR time series analysis of the 9 July 1998 Azores earthquake. International Journal of Remote Sensing 26(13): 2715–2729. Code P (2005) Eurocode 8: Design of Structures for Earthquake Resistance-Part 1: General Rules, Seismic Actions and Rules for Buildings. Brussels: European Committee for Standardization. Costa A (2002) Determination of mechanical properties of traditional masonry walls in dwellings of Faial Island, Azores. Earthquake Engineering & Structural Dynamics 31: 1361–1382. Costa A and Areˆ de A (2006) Strengthening of structures damaged by the Azores earthquake of 1998. Construction and Building Materials 20: 252–268. Costa AA, Costa A, Areˆ de A and Oliveira CS (2008) Estudo experimental ‘‘in situ’’ de paredes em alvenaria do Faial, Acxores. Sismo 1998—Acxores. Uma de ´cada depois, Governo dos Acxores, SPRHI, S.A. https://www.researchgate.net/publication/274832376 Costa A, Oliveira CS and Neves F (n.d.) Caracterizacxa ˜o do Parque Habitacional das Ilhas do Faial e Pico. Elementos de Estudo do Comportamento Relativo a `Crise Sı ´smica de 9 de Julho de 1998 nos Acxores. Sismo 1998—Acxores. Uma de ´cada depois, Governo dos Acxores, SPRHI, S.A. https:// www.researchgate.net/publication/311993488 Council AT (1996) Seismic Evaluation and Retrofit of Concrete Buildings (2. Appendices). Redwood City, CA: ATC. Dolce M, Goretti A and Masi A (1999) Analisi dei danni causati dal sisma del settembre 1998 nel Pollino. Ingegneria Sismica 16(3): 5–15. Dreger DS, Beroza GC, Day SM, Goulet CA, Jordan TH, Spudich PA and Stewart JP (2015) Validation of the SCEC broadband platform v14. 3 simulation methods using pseudospectral acceleration data. Seismological Research Letters 86(1): 39–47. FEMA (2005) Improvement of Nonlinear Static Seismic Analysis Procedures. Washington DC: FEMA 440, Federal Emergency Management Agency. Ferreira TM, Maio R and Vicente R (2017) Seismic vulnerability assessment of the old city centre of Horta, Azores: Calibration and application of a seismic vulnerability index method. Bulletin of Earthquake Engineering 15(7): 2879–2899. Gencturk B and Elnashai AS (2008) Development and application of an advanced capacity spectrum method. Engineering Structures 30: 3345–3354. Gru ¨nthal G (1998) European Macroseismic Scale 1998. Geneva: European Seismological Commission (ESC). Guerreiro L, Azevedo J, Proencxa J, Bento R and Lopes M (2000) Damage in ancient churches during the 9th of July 1998 Azores earthquake. In: Proceedings of the 12th world conference on earthquake engineering, vol 12, Auckland, 30 January–4 February. Hoveidae N, Fathi A and Karimzadeh S (2021) Seismic damage assessment of a historic masonry building under simulated scenario earthquakes: A case study for Arge-Tabriz. Soil Dynamics and Earthquake Engineering 147: 106732. Karimzadeh S and Lourencxo PB (2022) Stochastic ground motion simulation of the 9th of July 1998 Faial Earthquake (Azores, North Atlantic). Authorea [preprints]. DOI: 10.1002/ essoar.10508875.3. Bernardo et al. 2859 Karimzadeh S, Funari MF, Szabo ´S, Hussaini SMS, Rezaeian S and Lourencxo PB (2024a) Stochastic simulation of earthquake ground motions for the seismic assessment of monumental masonry structures: Source-based vs site-based approaches. Earthquake Engineering & Structural Dynamics 53: 303–330. Karimzadeh S, Hussaini SMS, Funari MF and Lourencxo PB (2021) On the effect of different codebased ground motion selection approaches for the estimation of the seismic demand of masonry structures by using real ground motion data set. Authorea [preprint]. DOI: 10.1002/ essoar.10509375.1. Karimzadeh S, Mohammadi A, Salahuddin U, Carvalho A and Lourencxo PB (2024b) Backbone ground motion model through simulated records and XGBoost machine learning algorithm: An application for the Azores plateau (Portugal). Earthquake Engineering & Structural Dynamics 53: 668–693. Koc AB, Erberik MA, Askan A and Karimzadeh S (2023) The sensitivity of global structural parameters for unreinforced masonry buildings subjected to simulated ground motions. Buildings 13(8): 2060. Lagomarsino S and Giovinazzi S (2006) Macroseismic and mechanical models for the vulnerability and damage assessment of current buildings. Bulletin of Earthquake Engineering 4(4): 415–443. Loads MD (2017) Associated Criteria for Buildings and Other Structures. Reston, VA: American Society of Civil Engineers, pp. 6–15. Madeira J, Brum da Silveira A, Hipo ´lito A and Carmo R (2015) Chapter 3 active tectonics in the central and eastern Azores islands along the Eurasia–Nubia boundary: A review. Geological Society, London, Memoirs 44(1): 15–32. Maio R, Esteˆ va ˜o JMC, Ferreira TM and Vicente R (2017) The seismic performance of stone masonry buildings in Faial island and the relevance of implementing effective seismic strengthening policies. Engineering Structures 141: 41–58. Matias L, Dias NA, Morais I, Vales D, Carrilho F, Madeira J, Gaspar JL, Senos L and Silveira AB (2007) The 9th of July 1998 Faial Island (Azores, North Atlantic) seismic sequence. Journal of Seismology 11: 275–298. Mendes N and Lourenco PB (2010) Seismic assessment of masonry Gaioleiro buildings in Lisbon, Portugal. Journal of Earthquake Engineering 14(1): 80–101. Morais EC, Ferreira TM, Esteˆ va ˜o JMC and Oliveira CS (2021) Implementation and validation of an approach for the estimation of the magnitude of historical earthquakes in Azores Islands. International Journal of Disaster Risk Reduction 53: 102000. Motazedian D and Atkinson GM (2005) Stochastic finite-fault modeling based on a dynamic corner frequency. Bulletin of the Seismological Society of America 95(3): 995–1010. Nettis A, Gentile R, Raffaele D, Uva G and Galasso C (2021) Cloud capacity spectrum method: Accounting for record-to-record variability in fragility analysis using nonlinear static procedures. Soil Dynamics and Earthquake Engineering 150: 106829. https://doi.org/10.1016/j.soildyn. 2021.106829 Neves F, Costa A, Vicente R, Oliveira CS and Varum H (2012) Seismic vulnerability assessment and characterisation of the buildings on Faial Island, Azores. Bulletin of Earthquake Engineering 10(1): 27–44. NP EN 1998-3 (2017) Euroco ´digo 8—Projeto de estruturas para resiste ˆncia aos sismos. Parte 3: Avaliacxa ˜o e reabilitacxa ˜o de edifı´cios (Portuguese version of EN 1998-3). Costa da Caparica, Portugal: Instituto Portugueˆ s da Qualidade (in Portuguese). Penna A, Lagomarsino S and Galasco A (2014a) A nonlinear macroelement model for the seismic analysis of masonry buildings. Earthquake Engineering & Structural Dynamics 43(2): 159–179. Penna A, Morandi P, Rota M, Manzini CF, da Porto F and Magenes G (2014b) Performance of masonry buildings during the Emilia 2012 earthquake. Bulletin of Earthquake Engineering 12(5): 2255–2273. Rezaeian S and Der Kiureghian A (2010) Simulation of synthetic ground motions for specified earthquake and site characteristics. Earthquake Engineering & Structural Dynamics 39(10): 1155–1180. 2860 Earthquake Spectra 40(4)