The Study of Seismic Hazard in Near-Fault Areas Using Probabilistic and Deterministic Approach

Earthquake is one of the most common natural disasters in Indonesia and usually destroys both high and low-rise buildings as well as triggers liquefaction and Tsunami. This means it is important to provide a robust building design with the ability to resist earthquake load and other induced phenomena. One of the methods commonly used to determine the relevant response spectrum of the bedrock is seismic hazard analysis which can be either Probabilistic Seismic Hazard Analysis (PSHA) or Deterministic Seismic Hazard Analysis (DSHA). The application of PSHA allows the representation of the response spectrum of an earthquake using the return period, thereby providing the engineers with the flexibility of selecting the appropriate natural period. In PSHA, all sources are considered when determining the response spectrum and all uncertainties has been considered through probability approach. On the other hand, DSHA is based on geological observations and empirical data that can be easily understood. Earthquake source must be identifiable and uncertainties are considered by either taking median value or median plus one standard deviation value. This research discussed the greater influence of seismic hazard analysis on the bedrock response spectrum of near-fault areas including Bandung situated at a distance of 12.9 km from Lembang Fault, Palu at 3 km from Palu Fault, and Yogyakarta at 8.5 km from Opak Fault. Moreover, EZFRISK Program was used to generate a response spectrum at bedrock and the results showed that PSHA is consistently more conservative than DSHA. It was also noted that there are significant differences at shorter periods for Palu site but these differences were observed at the natural period between 1s and 2s for Bandung and


INTRODUCTION
Earthquake is a common natural disaster in Indonesia.For example, Yogyakarta experienced an earthquake induced by Opak Fault in 2006 and this caused severe damage to the densely populated area with 5716 reported to be dead, 37927 injured, and between 0.5 to 1 million homeless (Consultative Group on Indonesia, 2006).Dozens of earthquake occurrences have also been linked to Opak Fault in the last 200 years such as the big one that occurred in 1867 reaching VIII MMI (Adelayanti, 2020).Another example is Sulawesi earthquake and tsunami which was triggered by Palu-Koro Fault (VII-VIII MMI) in September 2018 and caused liquefaction and landslide in Palu (Cilia et al., 2021).It affected 2.4 million people including 2000 reported to be dead, more than 4600 injured, and over 210000 displaced from their houses.Recent findings also showed the existence of the Lembang Fault located approximately 10 km north of Bandung City (beneath Lembang City) and active with small earthquake activity (Rasmid, 2014).Daryono et al. (2019) reported that the fault has caused at least three earthquakes and its vertical displacement was 40 cm in 2300 -600 BCE with the magnitude estimated to be 6.5 Mw and likely to be reactivated to cause damage in near-fault areas.
Engineering design requires considering seismic hazard specifically when the site project is prone to earthquake (Vaziri et al., 2022).The incorporation of earthquake effect on structure commonly requires a response spectrum at the surface.This can be achieved using the following steps: 1. Determining the response spectrum of the bedrock at the site which is referred to as the target response spectrum 2. Selecting the seed motion based on the appropriate earthquake 3. Creating synthetic ground motion 4. Propagating ground motion from bedrock to the surface (with or without structural model) The response spectrum of the bedrock at the site can be determined through seismic hazard analysis which is commonly separated into deterministic seismic hazard analysis (DSHA) and probabilistic seismic hazard analysis (PSHA) as previously indicated (Cornell, 1972;Baker, 2008;SNI 1726SNI , 2019)).The two methods (Kramer, 1996) require the identification of earthquake sources including the distance and magnitude.The attenuation equation can also be used to determine the effect of distance including the geometric spreading and damping between the project site and earthquake sources as well as the reconstruction of the bedrock response spectrum.The difference between PSHA and DSHA is the consideration of uncertainties in the analysis.The following uncer-tainties are usually considered in PSHA (Kramer, 1996): 1. Probability of distributing potential rupture locations 2. Recurrence relationship 3. Uncertainties inherent in the predictive relationship 4. Uncertainties in earthquake location, size, and ground motion parameter prediction In PSHA, the target response spectrum depends on the desired return period and is normally referred to as the uniform hazard spectrum.The overall model for PSHA is presented in Figure 1.Meanwhile, the target spectrum in DSHA is known as the deterministic spectrum.
DSHA is based on geological knowledge or field evidence and observation to select a reasonable maximum potential for earthquake based on individual sources.The overall model for this method is presented in Figure 2.

Advantages
Based on (Vaziri et al., 2022) (a) Widely used and known all around the world (b) Provides a forecast for seismicity and ground movement using seismogenic origin models (c) Considers all possible earthquake events (a) Provides reliable evaluation of seismic hazard from geology regardless of time.(b) Maximum credible earthquake (MCE) is commonly taken at median plus one standard deviation (84 percentile).However, it is possible to select a higher percentile for more conservative values or less for lower conservative values, indicating flexibility in meeting the project requirement.(c) Identifies individuals' faults with their estimated MCE and provides earthquake potential from each fault on hotspots and zones within the source area.

Disadvantages
(a) It is not based on valid physics because earthquake does not have regular occurrence through space and time (b) The probabilistic values for the operating basis earthquake (OBE) are produced by tinkering with the exceedance and this normally leads to arbitrary results with no basis in either science or engineering.The probability calculation also contains uncertainty.(c) Seismic probabilistic uses the mean value of ground motions obtained from the projection through time.The number can be added with 50% to be considered equivalent to 84 percentiles but the values do not represent the motion from the accelerograph.(d) PSHA assumes the possibility of combining earthquakes to project earthquake occurrences through time but this can make the individual earthquake sources lose their identities.(e) De-aggregation focuses on smearing earthquake in a region together into an amalgam in order to provide a value that represents no individual earthquake or a specific place in the region.This nonrepresentative value is normally projected through thousands of years into the future based on unverifiable calculation because there are no data for projection and later mathematically "deaggregated" into fragments to fill a grid with dubious interpretations.(f) It is difficult to understand and use.The results showed that DSHA map provided extremely high seismic hazard levels in some areas in Thailand and surrounding countries.
Meanwhile, PSHA map had the same trend but with a lower seismic hazard level.Eftekhari et al. (2021) also studied the effect of near-fault areas on PSHA in Iran and found large variations in the spectral accelerations among different locations at short oscillator periods and the absence of any significant difference at longer periods.However, the results obtained using PSHA and DSHA methods can complement each other in seismic hazard analysis (Moratto et al., 2007) and this is the reason they can be considered when constructing a target response spectrum (SNI 1726(SNI , 2019)).This study, therefore, compares the target response spectrum produced for the three sites of Yogyakarta, Palu, and Bandung using PSHA and DSHA in order to determine the method with a more critical target spectrum.

Project Location
The faults reviewed in this study were located at a distance of 0 -15 km from the site.For example, Project X is located approximately 12.86 km from Lembang Fault, Project Y is 3.05 km from Palu Fault, and Project Z is 8.48 km away from Opak Fault.The information from the 2017 Indonesian Earthquake Map showed that Lembang Fault has a maximum magnitude of 6.8 Mw with a slip rate of 2 mm/year, Opak Fault has 6.8 Mw with 0.75 mm/year, and Palu Fault has 6.8 Mw with 10 mm/year, respectively.This was the reason the three project locations are classified as near-fault areas (ASCE, 2017;SNI 1726SNI , 2019)).

Earthquake Data Collection
Earthquake catalog provided by Indonesian National Seismic Map (PUSGEN, 2018) was used to obtain Megathrust and Fault Earthquake Sources as indicated in Figures 3 and 4.
Earthquake catalog provided by the U.S. National Earthquake Information Service (USGS, 2022) was also used to obtain the background earthquake in- cluding shallow crustal and Benioff with a focus on those within a 500 km radius from the site.It is important to note that this study is only interested in the mainshock in relation to the background earthquake but the database can also contain foreshocks and aftershocks.This means there is a need to separate foreshock, mainshock, and aftershock earthquake.Gardner and Knopoff method (Gardner and Knopoff, 1974;Knopoff and Gardner, 1972) was adopted to filter out foreshock and aftershock based on the assumption that they are all related to each other and exhibit non-Poissoniandistribution.It is also important to note that the determination of non-Poissonian and Poissonian-  (Gardner and Knopoff, 1974) Table 2. Advantages and Disadvantages of PSHA and DSHA (Vaziri et al., 2022 ;Wang, 2011;Krinitzsky, 2003 ;Baker, 2008) Earthquake distributed shocks can be used to identify the mainshock.Therefore, the status of the background earthquake before the filtering process is presented in Figure 5 but only the mainshock is indicated in Figure 6 using Gardner and Knopoff method.

Attenuation Equation
The attenuation equation is normally used to estimate the level of ground shake caused by an earthquake with a certain magnitude, the distance between the source and a site, as well as the condition of the source.This means it represents the function related to the source information and wave propagation path of earthquake as well as the local condition of the site.The function was developed based on statistical regression analysis of the actual recording of ground motion or accelerograph on site (PUSGEN, 2018).The attenuation equation selected based on PUSGEN ( 2018) is presented in the following Table 2.

Seismic Hazard Analysis
Seismic Hazard Analysis using PSHA and DSHA methods was conducted through the EZ-FRISK program (Fugro, 2021).Earthquake sources used for PSHA were described in Section 3.1, including Megathrust, Benioff, and Shallow Crustal.Meanwhile, those used for DSHA included determined earthquakes such as Lembang, Opak, and Palu Faults.It is pertinent to note that the Uniform Hazard Spectra (UHS).

RESULTS
UHS and Deterministic Spectra in Figure 7 showed that the DHSA values in the three locations are consistently lower than PSHA values.Moreover, UHS is significantly higher than Deterministic Spectra and converges at the higher natural period at Palu.It was also discovered that UHS for both Bandung and Yogyakarta sites are higher than Deterministic Spectra.Meanwhile, the difference is not significant at the lower natural period but quite significant at the natural period between 1 and 2 seconds.

Palu
The results showed that the application of PSHA at Palu site is much more conservative than DSHA, specifically for low-rise buildings with a low natural period as indicated in Figure 7a.Meanwhile, there was no significant difference in UHS and Deterministic Spectra for high-rise buildings with a natural period of more than 2 seconds.

Bandung
Figure 7b shows that PSHA was more conservative than DSHA at Bandung site.This was indicated by the placement of UHS above Deterministic Spectra and this means Lembang Fault has the potential to provide more impact on the low-rise build-ings.It was also noted that PSHA is significantly more conservative on a longer period, specifically between 1 and 2 seconds.
Sari and Fakhrurrozi (2020) applied PSHA in Bandung Basin and PGA presented in Figure 8 showed that the sampling locations were far from Lembang Fault.However, it is interesting to note that PGA obtained from the spatial interpolation near Lembang Fault was estimated at 0.5 g and this is similar to the value obtained in this research.This means it is possible to adopt the PGA map from Sari and Fakhrurrozi (2020) to determine PGA located near the fault.

Yogyakarta
Figure 7c shows that PSHA was also more conservative than DSHA at Yogyakarta site, but the difference is not significant at the lower period.(Sari and Fakhrurrozi, 2020)

Figure 1
Figure 1 Overall model for PSHA modified from Vaziri et al. (2022) (a) It cannot be used on unidentifiable earthquake sources.(b) Error in determining earthquake sources.(c) Error in assigning earthquake potentials of fault.(d) There is no information on the recurrence of the controlling earthquake due to its neglect of time and frequency dimensions.(e) Difficult to determine the worst-case event because different earthquake sources can be the worst case at different period.(f) It is not a true worst-case event determinant because larger earthquake or ground motion can always occur in the future and the use of the 84 th percentile (mean plus one standard deviation) does not necessarily produce this scenario due to the absence of a theoretical upper bound on the amplitude of the ground motion.

Figure 7
Figure 7 UHS (Uniform Hazard Spectra) and Deterministic Spectra for Three Locations (a) Bandung, (b) Palu, and (c) Yogyakarta Figure7cshows that PSHA was also more conservative than DSHA at Yogyakarta site, but the difference is not significant at the lower period.Putra et al. (2018) also applied DSHA to construct a seismic hazard map.Their results showed PGA surface map around Opak Fault ranged from 1.1 to 1.2 while the amplification factors vary widely from 0.5 to 0.4.This indicated a large variation of bedrock PGA.The surface PGA and amplification factor obtained from the study area byPutra et al. (2018), respectively.This means the bedrock PGA is approximately 0.5 g which is very close to the value obtained using DSHA.According toPutra et al. (2018), the entire area near Opak Fault is at high risk of seismicity.