The SGC, in collaboration with researchers from the Geological and Mining Institute of Spain and the Global Earthquake Model (GEM) Foundation, developed a national seismic hazard model (Arcila et al., 2020). Overall, this national seismic hazard model comprises a set of tectonic environments and seismogenic sources. In that study, the seismicity of the Colombian territory was classified into four tectonic environments. Superficial events (cortical) correspond to events in the national territory down to depths limited by the upper crust–mantle boundary. Interplate earthquakes of the Colombian Pacific subduction zone correspond to earthquakes that occur in the area of contact between the Nazca and South American plates along the country's Pacific coast. Earthquakes in the Benioff area correspond to earthquakes inside the plate, which is subducting towards the east from the Colombian Pacific towards the country's interior. Bucaramanga's seismic nest corresponds to an area where earthquakes with moment magnitudes between Mw 4.0 and 5.0 usually occur at depths between 140 and 200 km (Prieto et al., 2012).

The building exposure model for the region has information about the building classes, the number of buildings, inhabitants, and the buildings' replacement costs. To develop this model for Sabana Centro, the methodology used by the South America Risk Assessment (SARA) project to develop exposure models in South America (Yepes-Estrada et al., 2017) was taken as a basis. The source of information to assign the number of buildings was the 2018 national census (DANE, 2018). The census allowed for having information on the number of dwellings and typical wall and roof materials, which were used to infer the different classes of buildings by municipality. A total of 156 628 dwellings were calculated; this number differs from that reported by the national census by 2.8 %, since it did not consider dwellings whose wall material is poured concrete. This material was not included, since there was no information available to relate it to any type of building class. The set of dwellings were related to the same building classes and same relationships (“mapping schemes”) used in Yepes-Estrada et al. (2017). As the census information is reported in terms of dwellings, the procedure used in Yepes-Estrada et al. (2017) to calculate the number of buildings was also followed. Then, this data were complemented using information collected during remote surveys carried out by students from the Universidad de La Sabana in the municipality of Chía. The building replacement cost refers to the cost of structural and non-structural components of a building, and it is a value associated with the building's rehabilitation. This study has only considered the structural cost per building calculated based on cadastral information available in the Territorial Statistics System (TerriData, https://terridata.dnp.gov.co/index-app.html#/, last access: 28 December 2022) of the country. This replacement cost was computer per building, expressed in USD. As the currency in Colombia is in Colombian pesos, the exchange to US dollars was made for an average exchange rate of USD 4080. Figure 7 shows the results of inhabitants, buildings, and their total replacement cost for the region. The bold numbers indicate the percentages for each municipality.

A total of 75 778 residential buildings in the region were classified into 33 building classes. Table 4 provides a description of these typologies along with the number of buildings within each category and their percentages. Among them, 6249 randomly distributed buildings in Chía were inspected in 2020 by civil engineering students of the Universidad de La Sabana. Their attributes were collected, making use of the rapid remote visual screening (RRVS) web platform (Haas et al., 2016), which allowed the use of the GEM v.2.0 taxonomy (Brzev et al., 2013) as a checklist while observing the buildings' façades through Google Street View. The resulting dataset is available in Arroyo et al. (2022). Four attributes of the GEM v.2.0 taxonomy were used for classifying these inspected buildings: the main construction material type, material technology, lateral load-resisting system, the expected level of ductility, and the number of stories. During the elaboration of the surveys, instead of “labeling” buildings as certain typologies, the collected attributes were used to classify them in a probabilistic manner. For such a purpose, the method proposed in Pittore et al. (2018) was used to evaluate the level of compatibility between the observed building attributes and each predefined building typology. Details of this process can be consulted in Arroyo et al. (2022). This procedure allowed us to compare the percentages of the building classes calculated based on the SARA methodology and discretize the buildings by height.

From Table 4 it is noticeable that 58.60 % of the buildings are constructed of non-ductile unreinforced masonry walls, including adobe blocks and dressed and semi-dressed stone. In addition, 20.92 % of the buildings are from non-ductile confined masonry and 3.32 % are non-ductile reinforced concrete frames, for a total of 82.84 % non-ductile buildings. Figure 8 shows the number of buildings for the three types of construction materials identified in the exposure model: concrete, masonry, and wood. This figure shows that the predominant construction material is masonry (65 272 buildings), mainly in Chía and Zipaquirá, with more than 10 000 buildings for each one. Then, there are those buildings made of concrete (6745), with more than 1000 in Chía, Zipaquirá, and Cajicá. Last, there are those structures made of wood (3762), especially in Chía, with more than 990 units. The number of masonry buildings represents the 86.14 % of the total buildings in Sabana Centro, whereas those of concrete and wood represent 8.90 % and 4.96 %, respectively. The number of buildings for each of the 33 typologies is depicted in Fig. 9. In the case of concrete, the predominant building class is one story non-ductile reinforced concrete moment frames, while for the masonry buildings, the non-ductile unreinforced masonry wall class is predominant.

A large part of the building inventory was constructed using unreinforced masonry and with characteristics that make them non-ductile or with low ductility. Therefore, it is necessary to make an appropriate assignation of the fragility curves to evaluate their physical vulnerability to ground shaking. In the absence of specific curves locally developed for the Sabana Centro province, fragility curves available in the literature were selected to represent these structures. Thereafter, a literature review was undertaken to select the fragility functions that most closely resemble the characteristics of the Sabana Centro building inventory. The Physical Vulnerability Suite of the GEM Foundation (OpenQuake Platform – Vulnerability, 2021) was considered for the review. The GEM database for the specific case of Colombia has the curves developed by Acevedo et al. (2017) for unreinforced masonry houses constructed in Antioquia, Colombia. There are some curves for reinforced concrete buildings with geographical applicability in Manizales, Colombia, by Bonett Díaz (2003) and the dataset of Villar-Vega (2017) for South America. Although the set of curves covers different types of buildings, they are calculated based on different methodologies and different damage states.

Another available dataset of fragility curves is that developed by Martins and Silva (2021), who covered nearly 500 building classes at a global level including Colombia. The fragility is calculated from nonlinear dynamic analyses performed on equivalent single-degree-of-freedom (SDOF) oscillators. They considered four damage states that are also intended to be studied in the present research: slight, moderate, extensive, and collapse. The corresponding damage thresholds were defined based on the spectral displacement of the structures. At a regional level is also the set of fragility curves for the residential building stock in South America (Villar-Vega et al., 2017), covering 54 common building classes. The methodology used for the derivation of the curves is similar to the one used in Martins and Silva (2021).

Based on the information collected, the fragility curves available in Martins and Silva (2021) were mainly used and complemented with those of Villar-Vega et al. (2017). These curves were selected in order to prevent a biased comparison of risk between the different municipalities in the region due to the different methodologies used to develop the fragility curves.

Therefore, a set of 33 fragility functions was used to represent the probability of exceeding a level of damage conditioned to ground-shaking intensity. These functions are comprised of 28 sets of curves reported by Martins and Silva (2021) and five sets developed by Villar-Vega et al. (2017). The last one is assigned to non-ductile confined masonry, one, two, and three stories, and ductile light wood members, one and two stories, since in the former these building classes were not included. These fragility functions are described by a cumulative probability curve with a lognormal distribution, and examples of some of them are presented in Fig. 10. This set of fragility curves was used to calculate the damage to the buildings included in the exposure model. Based on these curves, vulnerability functions were developed to evaluate the losses in the region. The loss ratios used in this study are 2 %, 10 %, 50 %, and 100 % for the slight, moderate, extensive, and collapse damage, respectively.

To determine the level of social vulnerability (SV) of the municipalities of the Sabana Centro province, this paper estimated a social vulnerability index (SVI) based on the methodology proposed by Cutter et al. (2003). The social equivalent to a quantitative physical risk assessment for earthquakes is an SVI. Social vulnerability is the reason for the different experiences of communities regarding the consequence of earthquakes (Burton and Silva, 2016). The construction of composite indicators based on the mathematical combination of a set of indicators, which consists of a group of variables, is one of the most common methods to objectively assess SV (Freudenberg, 2003). There are several methodological approaches for the construction of composite indicators, but in general, the steps include (1) the identification of pertinent variables, (2) the aggregation of variables into indicators and composite indicators, (3) multivariate analysis, (4) weighting, (5) convolution or link of variables, and (6) visualization and dissemination of results (Burton and Silva, 2016).

The SVI index aims to identify those municipalities in Sabana Centro whose inhabitants are more vulnerable to an earthquake based on a selection of specific variables, indicators, and composite indicators. The indicators were aggregated into five composite indicators constructed for the SVI of the SARA project (https://sara.openquake.org/development_of_indicators_of_social_vulnerability, last access: 12 January 2023): population, economy, infrastructure, education, and health. The composite indicator of the population considers the indicators that capture the capacity of the population to mitigate their risk and recover from earthquakes. In the current research the composite indicator of the population accounted initially for the female and native indigenous population, age dependence, population density, number of households, and people per household. The composite indicator of the economy includes indicators to assess the economic health of the community (Burton and Silva, 2016). The single indicators considered for this composite indicator were population unemployed, looking for employment, unsatisfied basic needs (UBNs), and impoverished. Poverty is an important aspect to consider because of its direct association with access to resources, which affects coping with the impacts of disasters (Fatemi et al., 2017). The composite indicator of infrastructure considers the access to basic services (Contreras et al., 2020b). The composite indicator of education links the educational level and the socioeconomic status, mitigation, and recovery potential (Burton and Silva, 2016). It is assumed that a lower education level results in lower income, poor ability to understand emergencies, and a low capacity to recover after a disaster (Cutter et al., 2003). The composite indicator of health includes the indicators related to access to health facilities and healthcare (Contreras et al., 2020b). The lack of access to healthcare increases people's susceptibility to the potential impact of disasters (Fatemi et al., 2017). Considering the aforementioned composite indicators and the availability of information for the region, a total of 26 indicators were selected initially. However, to avoid problems with interpreting the model and overfitting, we checked the multicollinearity by looking at the variance inflation factor (VIF) of each variable and indicator (see Table 5). The VIF was identified in a linear regression that included collinearity diagnostics produced in SPSS statistical software (Field, 2005). We excluded the variables and indicators that were potentially correlated with others and those that did not add significant information according to the collinearity diagnostics (Table 6). Eventually, the model included 10 independent and relevant variables and indicators to estimate the SV in the case study area (Table 7).

Much of the information used for the indicators is from the national census (DANE, 2018) database; studies such as the multipurpose survey (EM2017), which examined the quality of life of households in Bogotá and surrounding areas (https://sdpbogota.maps.arcgis.com/apps/MapJournal/index.html?appid=c984e588b0764efbb424ffc2207b5cf6, last access: 23 December 2022); and the analysis of the characteristics of the population in Sabana Centro (Sabana Centro Cómo Vamos, 2019).

The min–max normalization was used to standardize the SV indicators from zero to one to estimate the SVI per municipality. Higher scores indicate more socially vulnerable municipalities, and lower scores reflect less vulnerable ones. Then, the indicators were integrated by summing them with equal weight, as followed in Contreras et al. (2020c). The resulting SVI index is therefore used to adjust the percentage of economic losses with respect to the costs presented by the building inventory, i.e., multiplying them by 1 + SVI (Carreño et al., 2007).

The predicted damage for the Mw 5.95 earthquake scenario in Chía considered for the region is presented in Table 8. The mean and standard deviation are presented for each of the damage states considered in the scenario. The respective distribution of the ground motion field for the 5.95 earthquake is presented in Fig. 11.

In the region, 42.95 % of the buildings considered in the exposure model are expected to suffer some degree of damage. This result represents 32 598 out of the 75 778 analyzed buildings. Table 8 shows that the type of damage with the highest occurrence is slight (21.31 %), followed by collapse (9.85 %), moderate (7.36 %), and extensive damage (4.44 %). Overall, 14.28 % might suffer extensive or collapse damage; hence they will not fulfill their life safety functionality. Nearly 10 % of collapse raises concerns from a decision-maker perspective, but two Colombian events put the results in perspective: the Mw 6.1 earthquake in Armenia (1999) and the Mw 5.5 earthquake in Popayán (1983). In the former, the records indicate that 17 551 buildings were destroyed, 18 421 had severe damage, and 43 474 had moderate damage. In the latter, which occurred at an estimated depth between 12 and 15 km, 12 % of buildings suffered complete damage. In both earthquakes, damage concentrated in unreinforced masonry buildings, constructed prior to the enactment of the Colombian seismic design code in 1998. More than 60 % of the building stock in Sabana Centro is comprised of that type of building, and furthermore, 35 % are two- and three-story houses (Table 4), which are more vulnerable than one-story houses (Heresi and Miranda, 2022a). These buildings are expected to withstand significant damage during an earthquake such as the Chía Mw 5.95 shown here, which is similar in magnitude and depth to the Armenia earthquake and for which the percentage of collapse herein presented is similar to that from the Popayán earthquake.

In terms of municipalities, higher damage occurs in Chía and Cajicá, with 3522 and 2271 collapsed buildings. Compared to the total buildings of each municipality, collapses account for 19.0 % and 19.7 %, respectively. Overall, they account for 5793 out of the 7463 collapsed buildings for this scenario (77 %). In contrast, Cogua was the municipality with the lowest number of damaged buildings, as roughly 90 % of the inventory did not experience any type of damage, and only 0.33 % of them collapsed. Nemocón had the least damages after Cogua, with 2.4 % of collapses. These results are reasonable because Chía and Cajicá are closer to the epicenter in this scenario, and they have the highest building inventory of the region, together with Zipaquirá. Besides, a significant part of their building inventory is comprised of non-ductile unreinforced masonry. On the other hand, despite having a similar distribution of the building inventory, Cogua and Nemocón are the farthest municipalities from the epicenter. The main difference between these two is that Nemocón has softer soils, with roughly one-fourth of the municipality under 180 m s-1, thus the higher percentage of collapses.

The highest concentration of building collapses in the region is expected for houses constructed of unreinforced masonry (Table 9), mostly involving non-ductile unreinforced masonry walls of one and two stories (22.8 % and 30.5 %, respectively). Notably, these two types of buildings account for 53.30 % of collapses. Three-story unreinforced masonry houses account for 10.12 % of buildings, making the overall contribution of this structural system more than 6 out of 10 collapses. The percentage of three-story houses collapsed was smaller than the one from two-story houses (which are less vulnerable) because three-story houses are less frequent in the region.

The number of expected damaged buildings for each municipality due to the Chía Mw 5.95 scenario is presented in Fig. 12. It also shows their percentages for each damage state relative to the province's total. These results show that Chía, Cajicá, and Sopó have the highest percentage of collapsed buildings, with 47.20 %, 30.43 %, and 7.37 %, respectively. In this scenario, the least affected municipalities are Cogua and Nemocón, with less than 1 % of collapse percentages.

The damage calculations were conducted for all the seismic events presented in Table 2. Table 10 shows the resulting percentage of buildings that suffer damage for each of the 18 seismic risk scenarios. The results show that the worst scenario for the region is the Mw 6.95 in Cota, which has 21.37 % collapsed buildings, and the highest percentages of severe and moderate damage. Interestingly, the Mw 6.85 in Tocancipá had 8.12 % of collapsed buildings, roughly 2.5 times less than the Mw 6.95 in Cota. This difference is a consequence of the uneven distribution of the building stock in the region, as nearly 40 % is in Chía and Cajicá, which are close to Cota. A similar situation occurs for the earthquakes of Mw 6.25 in Sopó, Tabio, and Cajicá.

Figure 13 shows the variability of the results for each damage state of Table 10. The results illustrate the notable variability of the damage estimates between the different seismic scenarios. This variability is higher for the no-damage state, which ranges between 32.49 % and 76.99 %, corresponding to the Mw 6.95 in Cota and the Mw 5.35 in Tenjo scenarios. These two also had the highest and lowest percent of collapse, respectively. The median results of the 18 scenarios for the damage states are 53.6 %, 24.1 %, 8.1 %, 4.7 %, and 9.5 %, values that are close to those of the Mw 6.25 in Tabio.

According to the methodology presented in Sect. 3.4, the SVI is calculated and shown in Table 11.

Considering population, the most vulnerable municipality is Cota and the least vulnerable is Tenjo. Regarding economy, the most vulnerable is Gachancipá and the least vulnerable is Tabio. In the case of infrastructure, the most vulnerable municipality is Cajicá and the least vulnerable is Gachancipá. In terms of education, the municipality of Cogua is the most vulnerable, followed by Nemocón, unlike Chía. The health composite indicator shows that Zipaquirá is the most vulnerable municipality and the least vulnerable is Gachancipá. The economy composite indicator shows higher vulnerability indices than the other categories for most municipalities. Evaluating all of the categories, it was found that Zipaquirá is the municipality with the highest SVI, while Tabio is the least vulnerable.

The direct total economic loss arising from the considered Mw 5.95 Chía scenario is USD 900.49 million, which represents 14.41 % of the total replacement cost of the building inventory. The municipalities that contribute the most to these losses are Chía, with 52.60 %, and Cajicá, with 23.97 %, as shown in Table 12. This result was expected, since these municipalities' urban growth is high compared to the other municipalities. The smallest contribution comes from Cogua and Nemocón, with 0.06 % and 0.18 %, respectively. Overall, in the case of this Mw 5.95 earthquake scenario, the direct economic loss in terms of replacement costs would be approximately 21 % of the region's GDP. The economic losses for each municipality are presented in Table 12 and Fig. 14a. Cajicá is the one with the highest percentage of losses, with 22.66 % of the replacement costs. Other municipalities for which high economic losses are expected are Chía and Sopó, with 21.71 % and 17.22 %, respectively. The municipalities with the lowest losses are Zipaquirá (0.93 %) and Cogua (0.67 %).

Figure 14b shows the adjusted economic losses by municipality and includes SV. After considering the SV in the region, the economic losses increase by 27.42 %, from USD 900.49 to 1240.72 million. The municipality that would have the highest economic losses is Cajicá (22.66 % of the building replacement cost), with USD 215.80 million. When the SV is included, its potential economic losses increase to USD 306.61 million (32.19 % of the replacement cost of Cajicá buildings). In the case of Zipaquirá, the most socially vulnerable municipality, the losses were initially USD 5.08 million and increase to USD 7.85 million when the SVI is accounted for.

The percentage of economic losses concerning the building types relative to the total losses in the region is shown in Table 13. A total of 41 % of losses come from unreinforced masonry buildings (MUR/LWAL/DNO/H1 and MUR/LWAL/DNO/H2). This result is expected because this building type has a high seismic vulnerability, while the lowest contribution would come from CR/LDUAL/DUC/H:4,7​​​​​​​ houses, with less than 0.02 %. Three-story non-ductile houses constructed in reinforced concrete frames and confined masonry experienced the highest collapsed buildings per taxonomy, with nearly 20 % for both. In the case of the frames, 18.63 % of two-story houses collapsed, more than twice of those of one story. In the case of non-ductile confined masonry this ratio was 5, while for ductile confined masonry it was 4. These results agree with the findings by Heresi and Miranda (2022a) and add evidence to the need of avoiding lumping low-rise houses (one to three stories) into a single taxonomy for seismic risk calculations.

The seismic ground motion fields expected for each earthquake scenario listed in Table 2 were simulated 1000 times to account for their aleatoric uncertainty as advised by Silva (2016), making use of the OQ Engine. The physical vulnerability was calculated in a similar manner as for the Mw 5.95 earthquake scenario. Figure 15 shows the loss-exceedance curves that describe the probability of exceeding a given percent of economic losses for each earthquake scenario.

Among the simulated scenarios, the most significant economic losses might occur with the scenario that considers an earthquake of Mw 6.95 in Cota and the smallest with the simulation of the earthquake of Mw 5.35 in Tenjo. For example, the probability that economic losses exceed 20 % in the 6.95 Cota scenario is 61 %, while for the scenario 5.35 in Tenjo, the probability is 3 %. Besides magnitude, the uneven distribution of the building stock in the region is an important factor that exerts an influence on economic losses, as shown by the differences between the three Mw 6.25 events in Fig. 15. The highest economic losses are in the municipality of Cota with 24.29 % of losses (USD 1517.56 million). On average, economic losses when including social vulnerability increase by 37 % as shown in Table 14.

The findings presented in this work show that the Sabana Centro province is exposed to a considerable level of seismic risk. The simulations of 18 seismic scenarios with a return period of 475 years extracted from a probabilistic seismic hazard assessment (PSHA) show that the damaged buildings ranged between 23 % and 67.5 % of the building stock depending on the earthquake epicenter. Worryingly, the median value of buildings that would experience extensive damage or collapse in the 18 scenarios is 14.2 %. This situation stems from the fact that 83 % of the houses in the province are constructed using non-ductile structural systems. These houses accounted for more than 90 % of collapses in most of the scenarios. The damage results also highlight the importance of discretizing buildings with the same structural system by height, at least for houses between one and three stories as suggested by Heresi and Miranda (2022a) because two- and three-story houses had a significantly higher percentage of collapses compared to one-story houses.

In terms of economic losses, the median expected cost of the 18 earthquakes selected from the PSHA in the province is 19 % of its GDP, which accounts for USD 809.46 million, almost 13 % of the replacement cost of the building inventory. This result only accounts for the cost of physical damage of the building stock, however, and does not represent all of the potential impacts.

Very few research studies have tested the correlation between social vulnerability (SV) and losses. To the best of our knowledge, the relationship between SV and modeled losses has so far been informative rather than indicating that total losses (measured as dollar losses or debris generated) increase with SV (Schmidtlein et al., 2011). However, it was found that only relative losses (dollar losses per average family income) tend to increase with SV. Case study areas with a low SV tend to have more material goods with significant monetary value (dollar) exposed to risk than areas with high SV. Therefore, we should expect a negative correlation between property losses and SV (Cutter and Finch, 2008). It is important to understand that while the total loss (dollar) in case study areas with high SV is lower, the impact of those losses in their communities is high (Schmidtlein et al., 2011). Integrating the level of SV to the physical losses will not produce a significant increase in the last ones, considering that they are negatively correlated (Cutter and Finch, 2008), but the result will be a more holistic risk assessment, also useful to prioritize actions at the regional level. Sabana Centro is a province with a significant level of SV, representing many areas in Colombia and other countries in South America, with essential deficiencies in areas like health and education. The estimated integration of SV with the economic losses increases them to 26 % of the GDP, representing approximately USD 1.11 billion. The results of this study show that including SV is important in risk analysis, as it allows one to go beyond only considering economic loss assessments with respect to physical damage.

There are several limitations that should be addressed in future studies. One of them involves the selection of the building's seismic fragility functions. Recent research has demonstrated that assumptions about several input parameters used in physical seismic vulnerability significantly influence risk assessment in urban areas (Hoyos and Hernández, 2021). One concern held by the authors is that the field observations and surveys show that a big part of the building stock of Sabana Centro is the result of informal construction. Presently, these buildings are constructed using either confined masonry or infilled RC frames due to the influence of the Colombian design code (which has similarities to the ACI-318; Arroyo et al., 2019), which forbade unreinforced masonry. Research about fragility functions like these buildings in Puerto Rico (Murray et al., 2022) and Villavicencio (Feliciano et al., 2022) shows that the collapse probability may be even twice more than that of code-conforming buildings. The fragility functions by Martins and Silva (2021) and Villar-Vega (2017) used in this research do not account for the particularities of these buildings; thus the authors hold the hypothesis that the damage estimates should be considered a lower bound.

It is important to mention that the exposure model for residential buildings developed in this study considered two types of exposure modeling approaches. On the one hand, the census-based part is a top-down approach from aggregated data. On the other hand, the rapid remote surveys constitute a bottom-up approach. Although the latter allowed an assessment of the validity of the assumed building classes, both approaches were not fully integrated through a probabilistic approach (Pittore et al., 2020). Although this method requires more computational efforts, it is worth exploring in future studies.

Another aspect that was not explicitly addressed in this study was the consideration of spatial cross-correlation models in modeling the ground motion fields. Several studies (e.g., Weatherill et al., 2015; Heresi and Miranda, 2022b) have demonstrated the relevance of such models in seismic risk assessment for building portfolios when sets of fragility functions that consider several intensity measures (IMs) are implemented. Although when such models are accommodated, the loss outcomes typically show a greater dispersion (and are more likely to give extreme values), while when such models are disregarded, the mean loss values forecasted have been observed to be practically the same as for the cases when a spatial correlation model was used (e.g., Michel et al., 2017; Gomez-Zapata et al., 2021). Therefore, the results presented in this study are still informative, but once again, we remark they should be treated as lower bounds for the considered risk scenarios.

Furthermore, due to the proximity of the study area to the Usme fault, near-fault effects might be expected. Hence, the study of possible directivity effects might be relevant for future studies. The evaluation of this feature has been shown to be relevant in both seismic ground motion (Türker et al., 2022) and earthquake loss models (Gentile and Galasso, 2021). Therefore, a better understanding of their role in risk scenarios will benefit the outlined results.

Another point is that this study did not consider human casualties. These were not included due to the lack of accurate information about housing occupation, since in this province, it is expected that more than one family share one dwelling. Notwithstanding, the median results of physical damage assessment for the 18 scenarios suggest that 14.2 % of buildings will have a seismic performance below the life safety level. The limited number and quality condition of hospitals and healthcare facilities in the province would further exacerbate the potential impact of an earthquake on the population.