Damage Probability Assessment of Hospital Buildings in Yogyakarta, Indonesia as Essential Facility due to an Earthquake Scenario

Indonesia is a country located in an earthquake-prone region, and is characterized by significantly increased peak ground acceleration value. The seismic hazard map of Indonesia stated in SNI 1726-2012 and the current statistics published by PUSGEN in 2017 emphasized on the significance of assessing building damage probabilities, especially for essential structures in Yogyakarta. However, immediate action is required to handle response and recovery operations during and after a disaster. The aim of this study, therefore, is to ascertain the vulnerability and damage probability of hospital buildings in Yogyakarta by employing the 2006 earthquake scenario, where reports showed the destruction of over 156,000 houses and other structures. Furthermore, a Hazard-US (HAZUS) method was used for structural analysis, while a ground motion prediction equation was adopted to produce the building response spectra, following the characteristics of the earthquake incidence. The vital step in this assessment involves building type classification and identification of seismic design levels. However, the damage tendency of buildings is determined using the peak building response, which ensures the generation of capacity curves. The most significant findings on building damage probability value were less than 15% in each damage state (slight, moderate, extensive, complete). In addition, the optimum value was achieved at the minimum level of damage (minor), while the least values were recorded at the highest damage level (complete).


INTRODUCTION
Based on numerous evidence, Indonesia is known to frequently experience major earthquake damages. Specifically, Yogyakarta is one of the heavily populated cities on Java island observed to have suffered a severe devastating earthquake on May 27, 2006 in a magnitude of 6.3 Mw. The casualty rate and economic loss on buildings and infrastructures were high, and a report by Bappenas (2006) showed a death toll of over 5,700, while those injured exceeded 60,000. Furthermore, total damages and losses in the residential buildings reached 52.4%, total estimated economic loss was 29.1 trillion rupiahs. Therefore, the government applied scientific progress to understanding the earthquake sources by analyzing the seismic map of Indonesia. This data served as a basis for geomorphology and the designs for earthquakeresistant buildings. The National Center for Earthquake Studies (PuSGen) and the Indonesia seismic hazard published maps displaying Earthquake Resistant Building code (SNI 1726(SNI -2012. Figure 1 shows a significantly higher peak ground acceleration (PGA) value in Yogyakarta region compared to the values in SNI 1726-2002, from 0.15g to 0.2-0.4g. This increase provided a distintitive damage impact, especially for the buildings designed using outdated standards.
Previous studies showed the application of seismicity to residential buildings in Indonesia, including a research by Saputra (2012), Kurniawandy, (2015), and Bawono (2016). Faizah et al. (2017) conducted a rapid assessment on the vulnerability of the Muhammadiyah school buildings in Kasihan and Bantul districts of Yogyakarta using the Rapid Visual Screening (RVS) method from FEMA 154-2002. Therefore, the analysis was performed through visual observation, using the assessment form on RVS-FEMA-2002. A similar research was recently conducted by Gentile et al. (2019), where a rapid visual survey was carried out on multi-hazard risk prioritization and numerical fragility was implemented in school buildings at Banda Aceh, Indonesia.  (2003), hospitals are classified as essential facilities charged with the responsibility to provide health community care services, and are expected to function in disaster emergency occasions.
However, previous literature reviews showed the damage inflicted on innumerable hospital buildings, with devastating impact levels. This current study is a follow up research of the report by Muntafi, (2016), which focused on the mitigation efforts related to catastrophic incidence in densely populated earthquakeprone areas. Therefore, this study aims to determine the vulnerability of hospital buildings, alongside the structural quality and earthquake risk perception for all damage states. The research population used include hospital buildings in Yogyakarta city, Indonesia.

METHODS
This study adopts the Hazards USA (HAZUS) Earthquake Model, known to deal with all aspects of the big cities. However, there is no specific technique to evaluate the vulnerability of numerous building types. Therefore, this method is designed to produce economic loss estimates for local governments and policymakers to apply in the determination of contingency plans (NIBS, 2002).
The procedure was preceded with building identification, based on HAZUS model type and categorization, as well as seismic design level classification, analysis of response spectra using a moderate to strong earthquake of 6.3Mw by the 2006 Yogyakarta earthquake scenario, and also by generating the capacity curve to obtain the peak building response from the HAZUS fragility curve method. The result, therefore, show the damage probability of each building obtained based on the value of the peak response.

Description of The Study Area
Yogyakarta was selected as the study area due to the significant rise in PGA values. In addition, this highly populated city experienced a devastating earthquake in 2006, and destruction of over 156,000 was recorded (Elnashai et al., 2006). Preliminary surveys led to the selection of fifteen hospitals with locations presented in Figure 2.
The equation model used was based on data from Western North America type of shallow earthquake. In addition, the parameters utilized include magnitude of 5.5 -7.5 Mw, less than 80km distance as well as strike-slip, reverse-slip, and unspecified faulting styles. Subsequently, the spectral acceleration (g) of each period on Table 1 were obtained using Equation 1 and plotted as a function of spectral displacement by converting the units of g (the acceleration due to Earth's gravity, equivalent to g-force) to inches with the NIBS (2002) Equation (3) below: where Sd is spectral displacement (g), Sa is spectral acceleration (in), and T is time period (sec). The selected structures comprised a variety of reinforced concrete moment resisting frames Identified as C1 by the HAZUS system. According to the seismic resistance code design and prior to collapse during earthquakes, frame members of older buildings tend to undergo brittle failure. However, modern structures in zones of high seismicity exhibit ductile behavior and are more likely to undergo large deformation.

Seismic Design Level
During an earthquake, damages occur due to ground shaking and ground failure. This study, therefore, aims to determine the seismic design level from the classification by Muntafi (2018), and detailed descriptions from Eleftheriadou et al., (2014) and also the year of construction and building code of each structure as shown in Table  3.

Building Damage State and Cumulative Probability
The HAZUS system predicted and classified structural and nonstructural damage into slight, moderate, extensive, or complete state. Hence, slight structural damage is described as Flexural or shear type hairline cracks in some beams and columns near or within joints of reinforced concrete moment resisting (C1) frames. Moreover, moderate structural damage occurs when most beams and columns exhibit hairline cracks. Larger flexural cracks and concrete spalling Indicates yield capacity has been exceeded while larger shear cracks and spalling tend to be observed in the non-ductile variant. Extensive structural damage refers to a scenario where some frame elements have attained ultimate capacity. This is indicated in ductile frames by large flexural cracks, spalled concrete and buckled main reinforcement, while nonductile frames feature and shear or bond failures at splices as well as broken ties or buckled main reinforcement in columns possibly leading to partial collapse. Furthermore, a complete structural damage implies imminent danger or collapse due to brittle failure or instability in non-ductile frame elements. Approximately 20% (low-rise), 15% (mid-rise) and 10% (high-rise) of C1 buildings in the area are expected to experience this challenge.
The incidence of Ground failure and shaking are known to generally damage the functions of essential structures. These experiences are reported in the HAZUS document as lognormal fragility curves and are used to determine a building's tendency to reach or exceed damage for a specific potential earth science hazard (PESH) parameter, including response spectrum displacement. Therefore, the probability, ds, (P is the spectral displacement median value at the threshold of damage state (ds). In addition, β is the natural logarithm standard deviation of spectral displacement for ds, and is the function of standard normal cumulative distribution.  Figure 3 shows examples of fragility curve patterns for the four damage states.

Capacity of Building
The building capacity of each hospital structure was evaluated based on the model type, seismic design level, and property age. Also, the curves were assumed to have a range of possible log properties and distributed as a function of the ultimate strength (Au) per curve. The Yield (Dy, Ay) and Ultimate Capacity Point (Du, Au) values per structure were determined using the parameters provided in Table 4 and Figure 4. Figure 4 showed the influence of seismic design level and structural height on spectral displacement value. Therefore, higher figures correspond to an older structure standard and a greater amount of building stories, as demonstrated in the graphs. Amongst the five graphics, the properties with mid-rise and low seismic design levels (C1M-Low), including Building A and O demonstrated the highest spectral displacement and possibly vulnerability value.   The vulnerability function in HAZUS method for this study was derived from two types of curves, including the capacity and demand or response spectrum. Furthermore, these parameters served as basis for the determination of peak building responses generated per structure, and calculated from Equation (1) and (3). The outcome was estimated as the intersection of both curves, converted into the Sa-Sd relationship using Equation (3), as shown in Figure 5. Meanwhile, Table 5 shows the values for spectral displacement. Figure 5 shows the highest peak spectral acceleration value in Building B and C, while the least were recorded in N and O. The coefficients b, h, and VA were determined in the selected GMPE, and determined to be involved in conjunction with several parameters, including magnitude, distance, and Vs value. Therefore, buildings evidently closest to the epicenter are relatively most affected with similar earthquake magnitude.

Fragility Curve
Furthermore, each fragility curve in this research was determined based on a median and lognormal standard deviation (β) value in relation to the model type and seismic design level per structure. These parameters were obtained from the cumulative probability calculation result using equation (4). Figure 6 shows the fragility curve for each typical hospital building. Figure 6 shows the close relationship between the variations of each damage level and the model type, height, as well as seismic design for the building. Furthermore, there were significant differences in the five typical model forms of structures at the same Sd value. Specifically, the shape of the fragility curve has a C1M building type fabricated with high seismic design code, and a tendency to be gentler. Moreover, damage probability in all states emerge only after Sd values above 0.2 inches. This phenomenon is depicted in the illustration with similar structure form of both low concrete moment frames (C1L) and medium level of stories (C1M). Therefore, using a greater code induces a smaller value, particularly in the complete state.

Probability of Building Damage
This parameter was obtained from the cumulative probability calculation using equation (4) for the respective chospital buildings and at each destruction level. Furthermore, all values depend on the median of spectral displacement, lognormal standard deviation, design code, and model-building type. Table 6 shows the computation result for the specific damage state based on the peak building response values of hospital structures, while Figure 7 illustrates the building damage probability.
The matrix depicts a destruction probability value below 15% per damage state in all hospital buildings. This phenomenon was affiliated with the 2006 Yogyakarta earthquake scenario. Also, structure G has the highest value in all levels, including slight, moderate, extensive, and complete, at 14.903%; 11.018%; 1.431%; and 0.157%, respectively. In addition, a low seismic design level was used despite the short stories (C1L) characteristics of the structures, and the distance to the disaster source was relatively small.   The lowest value of damage probability for moderate, extensive, and complete levels occurs in Building L, with values of 7.568%; 2.832%; 0.125%; and 0.001%, respectively. However, a high seismic design code level was used, despite being situated at the closest distance to the epicenter. This triggered the incidence of a wreckage due to lower earthquake scenario. The peak output was obtained at the lowest level (slight), while the least significant value was acquired at the highest level (complete). This findings indicate the higher propensity for minor damages to the hospital structures investigated rather than major.

CONCLUSION
The damage probability assessment of hospitals and other essential buildings is crucial as a disaster mitigation effort. Based on the evaluation using the 2006 Yogyakarta earthquake scenario for fifteen buildings, the closest distance to the epicenter for buildings with the same model type was determined to have the highest damage values. In addition, other major parameters involved in this appraisal include the height of the building story and the seismic or construction design standards used. Particularly, the fragility curves show a combination of low seismic design level, high building story, close proximity to the epicenter and high damage possibility. Also, all the structures investigated tend to have a greater chance for minor, compared to moderate or severe destruction. This study is expected to serve as an initial information source for further research, alongside other methods and earthquake scenarios with potentially high magnitude for future events. Furthermore, the research is particularly related to essential facilities, including primary communication institutes, fire, police and power stations, disaster or emergency operations centers, shelter, and other utilities required in a disastrous situation.

DISCLAIMER
The opinions and conclusions expressed in this manuscript are of the researchers. These authors do not have an associative benefit to prompt a conflict of interest in this study.