Probabilistic seismic hazard analysis of Peshawar District has been
performed for a grid size of 0.01

Peshawar is the capital city of the Khyber Pakhtunkhwa province of Pakistan
that has an important background in the history of the Indian subcontinent. The city
provides key access to the Central Asian states through Afghanistan along
the western borders of Pakistan. It is located at 710

Peshawar is characterized by high seismicity rates due to its proximity to
the active plate boundary between the Indian and Eurasian plates, which are
converging at the rate of 37–42 mm yr

Damage observed in Peshawar during the 2015 Hindu Kush earthquake.

The

Seismic hazard of Peshawar reported by various researchers.

The present study aims to recalculate the seismic hazard of Peshawar, based on the up-to-date earthquake catalogue and ground motion prediction equations, and compare it with that recommended by BCP (2007). The PGA value at bedrock level was calculated using the classical probabilistic seismic hazard analysis procedure. The area sources as suggested by BCP (2007) were those for which the earthquake catalogue was obtained from worldwide seismogram networks and historical records. The modified Gutenberg–Richter empirical model was used to calculate the seismic zone parameters for both shallow crustal and deep-subduction-zone earthquakes. The seismic hazard in terms of PGA at bedrock was calculated and plotted with a GIS tool. Different ground motion attenuation relationships compatible with the geology and seismicity of the local environment were used to quantify the model in terms of variability in seismic hazard of Peshawar District. Furthermore, the logic tree approach was used to take into consideration the epistemic uncertainty. The GIS-based seismic hazard map developed for a return period of 475 years was compared with that given in BCP (2007). Seismic hazard maps were prepared for various other return periods, i.e., 50, 100, 250, 475 and 2500 years.

The uncertainties in the location, size and rate of recurrence of earthquakes along with the variation in the ground motion intensity and spatial variability can be well considered in the probabilistic seismic hazard analysis procedures (Ornthammarath et al., 2011; Çağnan and Akkar, 2018; Rowshandel, 2018). The probabilistic seismic hazard analysis (PSHA) provides a framework in which these uncertainties can be identified, quantified, and combined in a rational manner to provide a holistic view of the seismic hazard.

According to the modified Gutenberg–Richter law the earthquake exceedance
rate

The mentioned attenuation relationships can be used for ground motion prediction of shallow crustal earthquakes. However, several researchers including Crouse et al. (1988), Crouse (1991), Molas and Yamazaki (1995), and Youngs et al. (1995) have pointed out different conditions of attenuation relationships for shallow and subduction zones. Lin and Lee (2008) and Kanno et al. (2006) have developed attenuation relationships for earthquake records of Taiwan and Japan, respectively. The study of Lin and Lee (2008) showed lower attenuation for subduction zones than for crustal shallow earthquakes. Therefore, the use of shallow-crustal-earthquake attenuation relationships may lead to underestimation of the seismic hazard for subduction earthquakes in probabilistic analysis.

In probabilistic seismic hazard analysis, the peak acceleration at a
location is a function of magnitude and distance that is lognormally
distributed with standard deviation. In the hazard analysis, the study area
is first divided into seismic sources based on tectonics and geotechnical
characteristics. The different seismic sources are assumed to occur
independently, and the seismic events are considered to occur uniformly over
the source. The acceleration exceedance rate

Location of study area.

The collision of the Eurasian and Indian plate has resulted in the formation of an active Himalayan orogenic system that is further classified into the Tethyan Himalayas, Higher Himalayas, Sub-Himalayas and Lesser Himalayas (Gansser, 1964). The divisions are based on the tectonic blocks formed and separated by major fault boundaries.

The

The seismic hazard software CRISIS (2007) was used to calculate the peak
acceleration at bedrock level for Peshawar District. Figure 2 shows the geographical location of Peshawar District within the geopolitical boundaries of the KP province of Pakistan. The hazard analysis requires seismic source geometry, the earthquake reoccurrence relationship and the selected ground motion attenuation relationship. In the present study the ground motion attenuation relationships of Boore and Atkinson (2008) and Akkar and Bommer (2010) were used for shallow crustal seismic earthquakes and those of Lin and Lee (2008) and Kanno et al. (2006) for deep-subduction-zone earthquakes. The earthquake events within a 50 km depth were considered shallow, while earthquake events occurring at depths larger than 50 km were considered deep earthquakes. The seismic hazard maps were prepared in a GIS environment based on a grid size of 0.01

Shallow seismic sources for Peshawar (BCP, 2007).

The

Seismic source identification with defined latitude and longitude. © Google Maps.

Empirical relationships for moment magnitude.

The homogenized catalogue was further subdivided into shallow (depth less than 50 km) and deep (depth more than 50 km) earthquake events. Figure 6 shows the shallow and deep earthquake records along with seismic zones as defined in BCP (2007). Furthermore, Table 2 reports the number of earthquakes in each seismic source along with the maximum and minimum magnitude of each source. Deep earthquakes were found primarily in seismic source 1 and seismic source 2 which included the Hindu Kush seismic region. The deep sources were selected in consultation with the National Centre of Excellence in Geology, Peshawar since deep sources had not been studied before for Peshawar.

Earthquake records from homogenized catalogue and with defined seismic sources.

No. of earthquakes and minimum and maximum magnitude in shallow and deep seismic source.

In seismic hazard analysis the probability of earthquake occurrence is considered to follow a Poisson process, which considers the independent events to occur randomly in time and space. Only the main shocks are considered for hazard analysis. This is to avoid overestimation of the seismic hazard. The dependent events (foreshocks and aftershocks) are temporally and spatially dependent on the main shocks. For this purpose declustering was performed to remove the dependent events for the catalogue. The Gardner and Kenopoff (1974) declustering algorithm method was used for removing foreshocks and aftershocks. This performs a windowing procedure in time and space on the event magnitude to identify the dependent events. To perform this analysis ZMAP coding developed by ETH Zurich (freely available) was used. The homogenized catalogue was converted into ZMAP-specified format to perform the routine analysis. A total of 926 independent events remained after declustering.

The catalogue also report events from very far in the past, which cannot be considered
complete for all the magnitudes and the whole time span. The time window starts from
the year 1500; however, since then the catalogue has not been updated on a regular
basis. The instrumental observation of seismic data started after 1960, and it now observes and documents complete details of the earthquake events on a
regular basis. Due to these reasons the specified time window (1500–2015)
cannot be considered in obtaining the activity rate, as this would result in
the underestimation of the activity rate. For this purpose completeness analysis
was performed using the visual cumulative method (CUVI) proposed by Mulargia and
Tinti (1985). It is a simple procedure based on the observation that earthquakes follow a stationary occurrence process. It is used to find the completion point (CP) after which the catalogue is considered to be complete (Tinti and Mulargia, 1985). The procedure is to divide the magnitudes from 4 to 8 into various bands with a 0.5 step size. The selected bands are 4.00 to 4.50, 4.51 to 5.00, 5.01 to 5.50, 5.51 to 6.00, 6.01 to 6.50, 6.51 to 7.00 and 7.01 to 7.7. In each band the cumulative number of total earthquakes is plotted against the year of earthquakes; the period of completeness (

Completeness period for earthquake catalogue for specified band.

Completeness intervals and completion period of each magnitude band.

The modified Gutenberg–Richter (G–R) reoccurrence law, as mentioned earlier, was
used in the present seismic hazard analysis to characterize the G–R parameters. The seismic source parameters (i.e.,

Seismic source parameters for shallow and deep sources.

1–7 are shallow seismic sources and 1

The graph of

The attenuation relationships for a site are developed using substantial dataset information (Cotton et al., 2006); however, this is not available for Pakistan because of the scarcity of available strong-motion data. The alternative to this is to use the already-available attenuation relationships of other regions which have similar tectonic and geological conditions to Pakistan. In the case of shallow earthquakes, the candidate attenuation relationships for north Pakistan should be the ones developed for the active tectonic crustal earthquake region. Thus, the ground motion attenuation relationships of Akkar and Bommer (2010) and Boore and Atkinson (2008) were used to calculate the PGA for shallow seismic sources. However, the ground motion attenuation relationships of Lin and Lee (2008) and Kanno et al. (2006) developed for subduction zones were used for deep seismic sources. The seismic hazard in terms of PGA was then calculated at the bedrock site for different return periods, such as 50, 100, 250, 475 and 2500 years, as the cumulative seismic hazard due to both shallow and deep seismic sources. The various ground motion prediction equations (GMPEs) were combined through a logic tree approach and assigning equal weightings to each GMPE. The ground motions calculated were plotted in a GIS environment to obtain the seismic hazard maps for these different ground motion attenuation relationships.

The seismic hazard levels (Table 5), based on peak acceleration, defined in BCP (2007) were considered as a basis for the zoning of the seismic hazard at bedrock level.

Seismic hazard levels used for seismic zoning, obtained from BCP (2007).

The seismic hazard maps for a return period of 475 years in the case of shallow
crustal earthquakes on the one hand and deep earthquakes on the other for Peshawar District are reported
in Figs. 9 and 10, respectively. Figure 9 shows that for a return period of
475 years, the predictive relationship of Akkar and Bommer (2010)
overestimates the PGA value in comparison to that of Boore and Atkinson (2008), especially in the northern parts of the district. According to Arango et al. (2012), the distance scaling factor of the latter appears to be more adequate then the previous. Furthermore, Table 6 shows a slight
comparison of both ground motion prediction equations that suggests that in
terms of

Comparison of predictive equations used for shallow crustal earthquake (after Arango et al., 2012).

Seismic hazard maps for shallow crustal earthquake using different attenuation equations.

Seismic hazard maps for deep subduction earthquake using different attenuation equations and for a return period of 475 years.

Figure 10 shows the seismic hazard maps for deep subduction earthquakes
using the Lin and Lee (2008) and Kanno et al. (2006) attenuation equations for a return period of
475 years. According to Fig. 10 both the attenuation equations resulted
in roughly similar seismic hazard for Peshawar District. Furthermore, it is
also evidenced from Fig. 10 that the inclusion of deep subduction zones in
the seismic hazard does not contribute significantly; i.e., it remains low
(0.08–0.16

In probabilistic seismic hazard analysis (PSHA), one of the major sources of uncertainty is the epistemic uncertainty arising from the selection of predictive relationships. Thus, the different ground motion attenuation relationships already discussed were further used to find out the epistemic uncertainty in the seismic hazard analysis. This was accomplished through the logic tree approach, assigning an equal weighting factor to each GMPE (Fig. 11); the seismic hazard was combined from all the GMPEs.

Logic tree for incorporating epistemic uncertainty.

Figure 12 shows the seismic hazard maps for shallow and deep events after incorporating the epistemic uncertainty. As can be seen in Fig. 12a, the seismic hazard of Peshawar District becomes balanced when the average of the seismic hazard calculated using the Akkar and Bommer (2010) and Boore and Atkinson (2008) predictive equations was taken. The reason for this is the provision of equal weighting to both the predictive relationships in hazard analysis. The seismic hazard in the case of deep-subduction-zone earthquakes remains roughly the same after incorporating epistemic uncertainty (Fig. 12b). It can also be further concluded that the earthquakes produced by deep subduction zones are not significant in terms of seismic hazard and may be reasonably ignored. Thus, the shallow seismic sources are sufficient for the seismic hazard assessment of Peshawar. The calculated seismic hazard map after incorporating epistemic uncertainty is compared with the hazard map from BCP (2007). For the return period of 475 years, a close agreement between the two seismic hazard maps can be noticed (Fig. 13). After this check the seismic hazard maps for other return periods, i.e., 50, 100, 250, 475 and 2500 years, were prepared (Fig. 14), which may be used for seismic risk assessment. Hazard maps for various cases are reported in Fig. A1 through Fig. A8.

Seismic hazard maps after incorporating epistemic uncertainty for 475-year return period.

Comparison of seismic hazard maps for a return period of 475 years.

Mean seismic hazard maps for various return periods, i.e., 50, 250, 475 and 2500 years, considering all GMPEs and both shallow and deep earthquake sources.

The following was concluded on the basis of a literature review of past
seismic hazard studies of Peshawar and classical PSHA conducted for Peshawar
in the present study:

The selection of an appropriate ground motion prediction equation is crucial in defining the seismic hazard of Peshawar District. In the case of shallow crustal earthquakes, the predictive relationship of Akkar and Bommer (2010) provides a higher estimate of the PGA value in comparison to that of Boore and Atkinson (2008). The distance-scaling factor of the latter appears to be the reason for this disparity between the two models.

The inclusion of deep subduction earthquakes does not add significantly to hazard and may be neglected in terms of seismic hazard. Therefore, only the shallow crustal earthquakes contribute to the seismic hazard of Peshawar District. However, recent earthquakes in Peshawar from deep sources have caused widespread destruction in various parts of the district. This raises concern for the existing GMPEs and the classical PSHA procedure to simulate such effects.

The epistemic uncertainty was used by providing equal weighting to the attenuation equation of Akkar and Bommer (2010) and Boore and Atkinson (2008). The mean seismic hazard map thus produced was balanced and was found to be in close agreement with the design base seismic hazard given in BCP (2007) for bedrock hazard. However, the BCP places Peshawar in Zone 2B, which is reasonable for most of the locations, but it underestimates ground motions especially in northern parts of the district.

The mean seismic hazard calculated for Peshawar was also compared with previous studies (Table 7). It can be observed that the seismic hazard obtained by independent researchers suggests an average PGA equal to about 0.24

Placement of Peshawar based on the present study: classical PSHA with areal sources, considering both shallow and deep earthquakes. Listed in ascending order of peak ground acceleration.

It is worth mentioning that the focus of the present study was to provide the base maps for seismic hazard in Peshawar. Site-specific soil properties were not known; therefore, they were not addressed in the present study. Alternatively, the code suggests amplification factors for various soils from Type C to Type E as per NEHRP soil classification. This may be considered to amplify or deamplify the seismic hazard provided in the present study.

All data, models and code generated or used during the study are available from the corresponding author by request (Naveed Ahmad, naveed.ahmad@uetpeshawar.edu.pk). Items which may be requested are the earthquake catalogue (raw and processed data), Excel sheets used for G–R parameters derivation, CRISIS input files, etc.

NA contributed as the advisor and supervisor of the research and in the analysis of the earthquake catalogue for the derivation of seismic parameters, selection of GMPEs, preparation of input files for CRISIS program and paper drafting. UK contributed in the compilation of the earthquake catalogue and processing of data, analysis of hazard through the CRISIS program and developing hazard maps. KM has contributed to the data organization, literature review, integration of work tasks and preparation of the initial paper draft. QI has contributed in the selecting and designing of seismic sources for both shallow and deep earthquakes, assignment of source parameters in the CRISIS program, output data compilation, and result plotting.

The authors declare that they have no conflict of interest.

This paper has been produced from MSc research work on the seismic microzonation of Peshawar, Pakistan. The authors are grateful to the reviewers for their constructive remarks that improved the quality of the manuscript.

This paper was edited by Maria Ana Baptista and reviewed by two anonymous referees.