Can Site Specific parameters help to identify the Seismically Induced Damage Pattern: An Assessment

doi:10.21203/rs.3.rs-2721236/v1

Indo-Gangetic plains are vulnerable to damage during earthquakes due to the presence of soft soil cover. Local geology and topography play a significant role in amplifying the damage and controlling the pattern of damage. In this study, an effort has been made to investigate the pattern of damage by using the different parameters which significantly varies spatially such as lithology, shear wave, soil texture, basement depth and distance to the fault. The logistic regression was performed on the training data was selected from the common damage zones provided in the catalogue a generalized damage pattern map was prepared for North Bihar without considering the earthquake source effect. The model achieved 55% accuracy and, further the prepared map was validated with the isoseismal map of 1934, 1988, 2015 Mainshock, and 2015 major aftershock, the area under the curve (AUC) determined as 0.54, 0.57, 0.56, and 0.60 respectively. The generalized map can be used with the ground motion parameter to estimate the extent of hazard after an earthquake.

AUC

Logistic regression

Indo-gangetic plain

Mainshock

Major aftershock

The Indo-Gangetic plains (IGP) observed severe damage intensity due to the frequent earthquakes in the adjacent highly tectonic Himalayan belt. Aside from its proximity to the Himalayas, it is comprised of the thick alluvium of the Ganga river (Keshri, Mohanty and Ranjan, 2020). The accumulated stress as a result of India's convergence into Tibet is periodically released in the form of large earthquakes (Wesnousky, 2020). The IGP and the Himalayas both suffered significant damage and a high death toll as a result of the Himalayan earthquakes Gorkha (2015) Mw 7.8, Kashmir (2005) Mw7.6, Assam (1950) Mw 8.5–8.7, Bihar–Nepal (1934) Mw 8.4, Kangra (1905) Mw7.8, and Shilong (1897) Mw 8.2–8.3 Assam) (Dal Zilio et al., 2021). In this trail, the 6.6 Mw earthquake in 1988 and the 7.3 Mw earthquake in 2015 both caused catastrophic disasters in the Bihar plains in the IGP.

The Probabilistic seismic hazard assessment (PSHA) for the whole Indian subcontinent was performed by (Ndma, 2011; Nath and Thingbaijam, 2012), the PSHA was also performed for the IGP, the peninsular Indian region and specific cities in India by other researchers (Das, Gupta and Gupta, 2006; Mahajan et al., 2010; Anbazhagan et al., 2019; Keshri, Mohanty and Ranjan, 2020; Sinha and Sarkar, 2021). As a result of the development and deployment of multi-criteria decision-making (MCDM) approaches in recent years, it is now possible to estimate the seismic hazard while taking into account a variety of site-specific factors that can increase the severity of damage. An MCDM approach using machine learning applied by (Jena et al., 2020; Pancholi et al., 2022) for earthquake disaster mitigation utilizing the factors such as shear wave, faults, lithology, amplification factor, soil type and , PGA. Site effects is a phenomenon that occurs during an earthquake and results in extra ground shaking, liquefaction, and landslides. It can alter the amplitude, frequency, and duration of bedrock motion when it reaches the surface (Anbazhagan, Kumar and Sitharam, 2013).

In present study an MCDA used to investigate the seismic hazard assessment using the Logistic Regression considering the site specific factors such as Lithology, Proximity to fault, Soil texture, Groundwater, Amplification factor, Shear Wave and PGA. The groundmotion model developed by (Bajaj and Anbazhagan, 2019) was used to evaluate the PGA for Northern Bihar for 1934, 1988, 2015A, 2015B earthquake.

Study Area

Study area located in the North Bihar region from 27.49N, 83.81E to 25.24N, 88.28E, the study area is a part of world’s largest alluvial plain Indo-gangetic plain (IGP). The study area has various structures such as West Patna Fault (WPF), East Patna Fault (EPF), Munger Saharsa ridge, Gandak fault, all these faults are spread over the whole North-Bihar and with any earthquake hazard these structures has a significant role in enhancing the damage intensity.

The study area consists of six drainage basins Gandak, Burhi Gandak, Bagmati, Kamla Kosi, and Mahananda from which the according to the Morphotectonic analysis performed by Rawat et al., (2022) , the Gandak and Mahananda basins are very highly active while Kamla basin is less active and remaining basins are moderately active. The North Bihar has recorded significant liquefaction events in 1934, 1988 earthquake Rawat et al., (2022) have reconstructed the liquefaction map of North Bihar for the same events.

Training data selection

In this study an effort was made to identify the pattern of damage during historical earthquakes and prediction for the future. The earthquake groundmotion parameters are greatly influenced by factors such as seismic source, propagation path of seismic wave, and site specific factors. Different region specific ground motion models have been developed to estimate the seismic source and propagation of seismic wave effect, while most essential and complex factor is site effects. Site specific variables significantly changes locally and influences the intensity of damage. In this work the sites frequently observed higher damage intensity during previous earthquakes (1934, 1988, 2015 maishock, and 2015 major aftershock) was considered as reference data. The concept is different site specific factors present at these locations has some affinity to enhance the ground motion during earthquake. Considering this as the base of the model, first step towards the preparation of seismic hazard map was the selection of damage sites. In light of this the Martin and Szeliga, (2010) prepared a detailed catalogue of MMI observed at IGP during historical earthquakes from there the MMI data during 1934 and 1988 earthquakes were collected. For recent earthquakes the isoseismal map provided by USGS was used to extract the Intensity value for 2015 Nepal earthquake mainshock and aftershock. The damage sites were put altogether shown in Figure 2 and the locations which are common in minimum two earthquakes have been recorded and used as training data for training the Logistic Regression (LR) model.

In this study the site specific parameters considered are time averaged shear wave velocity V_S30, Soil texture, Distance from fault, Bed rock depth, and lithology.

Table 1: The correlation of site specific parameters with damage intensity

Site Specific Parameter	Contribution in hazard	Data Source
Shear Wave Velocity (Vs30)	Shear wave velocity is proportional to soil stiffness, thus site amplification is stronger where shear wave velocity is low. (Boore, 2004; Sundararajan and Seshunarayana, 2011; Sundararajan et al., 2019).	USGS
Lithology	The lithology of the region contributes in the amplifying the damage intensity of hazard, the local geology has a strong correlation with the damage (Spudich, Hellweg and Lee, 1996; Ayele, Woldearegay and Meten, 2021; Pancholi et al., 2022).	BHUKOSH
Basement depth	The structural damage is limited in zones where bedrock outcrops (Bol, 2012), and the sediment thickness above the bedrock amplifies the damage intensity.	BHUKOSH
Fault proximity	Structures located on faults or in the proximity of faults has increased risk of earthquake damage, countries like USA has prohibited the new construction near the fault lines (Jena et al., 2020).	Fault data BHUKOSH
Soil texture	Soil texture plays an important role in liquefaction hazard, the study area is an alluvial plain where liquefaction is common secondary effect of the earthquake.	NBSS, Nagpur

For all the earthquake events considered in this study the PGA layer was prepared using the Groundmotion model by Bajaj and Anbazhagan, (2019) shown in Equation 1. The individual PGA layers were integrated with the hazard map prepared using the LR model. The PGA values determined by Nath and Thingbaijam,(2012) for 2% and 10% probability of exceedance in 2475 yrs and 475 yrs respectively was also integrated with the hazard map to evaluate the futuristic scenario of hazard during earthquake.

Multicollinearity analysis

The multicollinearity analysis has been performed for the data before applying LR model. The primary assumption for LR model is that the independent factors should not be correlated or very little correlation should be among the factors. This analysis infers that all the factors considered for the LR model are not significantly correlated hence can be used for the model. The multicollinearity in the independent variable reduces the precision of the coefficients, which ultimately weaken the statistical power of the model.

Table 2: Shows the Multicollinearity analysis of different factors used for the hazard map preparation, analysis shows that factors are not correlated with each other or very lightly correlated.

	Bedrock depth	Distance to fault	Lithology	Shear wave velocity	Texture
Bedrock depth	1
Distance to fault	0.098	1
Lithology	-0.007	0.11	1
Shear wave velocity	0.042	-0.08	-0.26	1
Texture	0.15	0.04	-0.003	-0.08	1

Logistic regression

Logistic regression is a supervise machine learning multivariate statistical approach used for binary classification by predicting the probability of an outcome, the analysis result range falls between 0 to 1 (Althuwaynee et al., 2014). The model performs analysis over one or other independent parameters in order to establish a relationship to predict the output in a discrete classes. The technique is widely used for the evaluation of the mathematical probability of an event to occur or not. The logistic regression uses a sigmoid function to map predictions and their probabilities. The assumptions considered in the LR model is the dependable variable should be binary, there should be very low or no multicollinearity between the independent variable, LR model requires large sample size for prediction. The model prepared through the Logistic regression is equation 1

1.47 – 1.36 × 10^-4(bedrock depth) + 7.9 × 10^-6 (distance to fault) – 4.2 ×10^-6 (lithology) + 2.1 × 10^-5(Soil Texture) – 6.7 × 10^-3(Shear wave velocity) ……………………………………………………….…..1

The accuracy of the model was also tested using the ROC characteristics, it was observed that the model has received 55% accuracy in predicting the outcome. The curve for ROC is plotted in Fig 5

The damage pattern map shown in fig 4 depicts Gandak and Burhi Gandak basins shows the higher probability of damage intensity. Theses basins consist of paleochannels in abundance, the maximum depth to bedrock in this region is determined by GSI is to 6000 m and which is maximum in the whole study area. The soil texture is medium to fine infers medium permeability, and Gandak fault is present at this part. The combination of all these site conditions making this region vulnerable to damage during earthquake. Along the rivers Ganga, Gandak, Kosi, and Mahananda a prominent red or yellow zone is highlighted infers that in recent alluvium the shaking will be higher due to seismic amplification and damage will be more relative nearby region where compact soil is present. For illustration during Bhuj earthquake 2001, which occurred 330 km from Ahmedabad, and due to thick sedimentary cover of Sabarmati river a soil amplification was recorded results in strong shaking followed by heavy damage (Narayan and Sharma, 2004).

Along the West Patna fault probability of damage intensity varies from 0.49 - 0.55, while along the East Patna fault it is 0.4-0.49, the East Patna fault is considered to be responsible for 1934 and 1988 earthquake as a result the damage intensity observed will be higher (Verma, Pati and Sharma, 2017). The probability of damage intensity along the Ganga river the near Patna and Munger is 0.61 – 0.90 which is high to very high. In 1934 Bihar-Nepal earthquake the Munger region recorded heavy damage and the damage intensity was X in MMI classification in this region. Damage intensity from Munger to Patna was confined in a small strip along Ganga river and its flood plain (Kayal, 2014).

Validation

Validation performed using the ROC curve for the prepared hazard map from the isoseismal map of 1934, 1988, 2015 mainshock, 2015 major aftershock. The figure 7 depicts ROC curve and AUC value for all the events from the prepared hazard map. It was observed that minimum AUC 0.54 for 1934 event, and for 2015 major aftershock maximum AUC determined as 0.60. The accuracy of the results is nearby 0.5 which is moderate, the earthquake hazard also influences by the earthquake source and propagation path which is not considered in this study. The isoseismal map for 1934 and 1988 were prepared from the catalogue provided by (Martin and Szeliga, 2010) and isoseismal map provided in (Kayal, 2014). The isoseismal maps for the 2015 Nepal earthquake mainshock and aftershock was prepared from the USGS data of Did you feel it survey? Observing the resulted map and the determined AUC values it can be confirmed that the model is passed evaluating the hazard map.

Concluding remarks

In this work a damage pattern map has been prepared for the North Bihar considering the site effects of the studied region. It is concluded from the study that the damage intensity will be higher along the rivers, which is also confirmed in the previous earthquakes of (2001, 1934, 1988 earthquakes). The higher damage intensity observed in the regions where deeper bedrock depth is present, this fact could make an understanding to delineate the areas where bedrock depth is deeper, and the heavy construction should be avoided at those places. The settlement zones should not be close to faults. Study can help the urban planners to plan their strategies for the heavy construction, land distribution, utility management, and nuclear power plant site selection to avoid any future destruction.

Althuwaynee, O.F. et al. (2014) ‘A novel ensemble bivariate statistical evidential belief function with knowledge-based analytical hierarchy process and multivariate statistical logistic regression for landslide susceptibility mapping’, Catena, 114, pp. 21–36. doi:10.1016/j.catena.2013.10.011.
Anbazhagan, P. et al. (2019) ‘Probabilistic seismic hazard analysis using the logic tree approach-Patna district (India)’, Natural Hazards and Earth System Sciences, 19(10), pp. 2097–2115. doi:10.5194/nhess-19-2097-2019.
Anbazhagan, P., Kumar, A. and Sitharam, T.G. (2013) ‘Seismic Site Classification and Correlation between Standard Penetration Test N Value and Shear Wave Velocity for Lucknow City in Indo-Gangetic Basin’, Pure and Applied Geophysics, 170(3), pp. 299–318. doi:10.1007/s00024-012-0525-1.
Ayele, A., Woldearegay, K. and Meten, M. (2021) ‘A review on the multi-criteria seismic hazard analysis of Ethiopia: with implications of infrastructural development’, Geoenvironmental Disasters, 8(1). doi:10.1186/s40677-020-00175-7.
Bajaj, K. and Anbazhagan, P. (2019) ‘Regional stochastic GMPE with available recorded data for active region – Application to the Himalayan region’, Soil Dynamics and Earthquake Engineering, 126(August), p. 105825. doi:10.1016/j.soildyn.2019.105825.
Bol, E. (2012) ‘Determination of the relationship between soil properties and earthquake damage with the aid of neural networks: A case study in Adapazari, Turkey’, Natural Hazards and Earth System Science, 12(9), pp. 2965–2975. doi:10.5194/nhess-12-2965-2012.
Boore, D.M. (2004) ‘Estimating V̄s(30) (or NEHRP site classes) from shallow velocity models (depths < 30 m)’, Bulletin of the Seismological Society of America, 94(2), pp. 591–597. doi:10.1785/0120030105.
Dal Zilio, L. et al. (2021) ‘Building the Himalaya from tectonic to earthquake scales’, Nature Reviews Earth and Environment, 2(4), pp. 251–268. doi:10.1038/s43017-021-00143-1.
Das, S., Gupta, I.D. and Gupta, V.K. (2006) ‘A probabilistic seismic hazard analysis of Northeast India’, Earthquake Spectra, 22(1), pp. 1–27. doi:10.1193/1.2163914.
Jena, R. et al. (2020) ‘Earthquake hazard and risk assessment using machine learning approaches at Palu, Indonesia’, Science of the Total Environment, 749, p. 141582. doi:10.1016/j.scitotenv.2020.141582.
Kayal, J.R. (2014) ‘Seismotectonics of the great and large earthquakes in Himalaya’, Current Science, 106(2), pp. 188–197.
Keshri, C.K., Mohanty, W.K. and Ranjan, P. (2020) ‘Probabilistic seismic hazard assessment for some parts of the Indo-Gangetic plains, India’, Natural Hazards, 103(1), pp. 815–843. doi:10.1007/s11069-020-04014-8.
Mahajan, A.K. et al. (2010) ‘Probabilistic seismic hazard map of NW Himalaya and its adjoining area, India’, Natural Hazards, 53(3), pp. 443–457. doi:10.1007/s11069-009-9439-3.
Martin, S. and Szeliga, W. (2010) ‘A catalog of felt intensity data for 570 earthquakes in India from 1636 to 2009’, Bulletin of the Seismological Society of America, 100(2), pp. 562–569. doi:10.1785/0120080328.
Narayan, J.P. and Sharma, M.L. (2004) ‘Effects of Local Geology on Damage Severity during Bhuj, India Earthquake’, 13th World Conference on Earthquake Engineering, Vancouver, B.C., Canada, (2890), pp. 1–8.
Nath, S.K. and Thingbaijam, K.K.S. (2012) ‘Probabilistic seismic hazard assessment of India’, Seismological Research Letters, 83(1), pp. 135–149. doi:10.1785/gssrl.83.1.135.
Ndma (2011) ‘Development of Probabilistic Seismic Hazard Map of India Technical Report’, National Disaster Management Authority, p. 126.
Pancholi, V. et al. (2022) ‘Multi-criteria approach using GIS for macro-level seismic hazard assessment of Kachchh Rift Basin, Gujarat, western India – First step towards earthquake disaster mitigation’, Journal of Earth System Science, 131(1). doi:10.1007/s12040-021-01744-6.
Rawat, A., Kumar, D., Chatterjee, R S, et al. (2022) ‘A GIS-based liquefaction susceptibility mapping utilising the morphotectonic analysis to highlight potential hazard zones in the East Ganga plain’, Environmental Earth Sciences, 81(13), pp. 1–16.
Rawat, A., Kumar, D., Chatterjee, R. S., et al. (2022) ‘Reconstruction of liquefaction damage scenario in Northern Bihar during 1934 and 1988 earthquake using geospatial methods’, Geomatics, Natural Hazards and Risk, 13(1), pp. 2560–2578. doi:10.1080/19475705.2022.2122591.
Sinha, R. and Sarkar, R. (2021) ‘Deterministic Seismic Hazard Assessment of Dhanbad City, India’, Lecture Notes in Civil Engineering, 116 LNCE, pp. 1–14. doi:10.1007/978-981-15-9976-7_1.
Spudich, P., Hellweg, M. and Lee, W.H.K. (1996) ‘Directional topographic site response at Tarzana observed in aftershocks of the 1994 Northridge, California, earthquake: Implications for mainshock motions’, Bulletin of the Seismological Society of America, 86(1 SUPPL. B), pp. 193–208. doi:10.1785/bssa08601bs193.
Sundararajan, N. et al. (2019) ‘Shear Wave Velocity Characteristics in Parts of Muscat, Sultanate of Oman — A Measure of Earthquake Hazard Assesment’, Journal of the Geological Society of India, 93(5), pp. 515–522. doi:10.1007/s12594-019-1212-y.
Sundararajan, N. and Seshunarayana, T. (2011) ‘Liquefaction Hazard Assessment of Earth Quake Prone Area: A Study Based on Shear Wave Velocity by Multichannel Analysis of Surface Waves (MASW)’, Geotechnical and Geological Engineering, 29(3), pp. 267–275. doi:10.1007/s10706-010-9360-2.
Verma, A.K., Pati, P. and Sharma, V. (2017) ‘Soft sediment deformation associated with the East Patna Fault south of the Ganga River, northern India: Influence of the Himalayan tectonics on the southern Ganga plain’, Journal of Asian Earth Sciences, 143, pp. 109–121.
Wesnousky, S.G. (2020) ‘Great pending Himalaya earthquakes’, Seismological Research Letters, 91(6), pp. 3334–3342. doi:10.1785/0220200200.

No competing interests reported.

Can Site Specific parameters help to identify the Seismically Induced Damage Pattern: An Assessment

Status:

Version 1

Abstract

Figures

Introduction

Methodology & Materials used

Results & Discussion

References

Additional Declarations

Status:

Version 1