ASSESSMENT OF SEISMIC HAZARD AND PATTERN RECOGNITION FOR REGION AROUND KANGRA

Abstract

This research work was carried out with the following objectives: to identify seismically susceptible areas using Pattern Recognition (PR); segmentation of longer tectonic units using PR and assessment of seismic hazard by two methods: Deterministic Seismic Hazard Assessment (DSHA) and Probabilistic Seismic Hazard Assessment (PSHA) in a 7o by 7o area, defined by longitudes 73oE to 80oE and latitudes 29oN to 36oN, centered on the epicenter of the great Kangra earthquake of 1905. An earthquake catalogue compiled for the study area consisted of 1172 main events with moment magnitude (Mw) ranging from 3.5 – 8.0 for the period 1552-2012. This catalogue is referred to as the Merged, Homogenized and Declustured, MHD, catalogue for western Himalaya. Another catalogue, MHD catalogue-2, compiled for validation of results, contained 80 earthquakes with Mw 4.5 – 5.7 for the period January 2013 to September 2015. Tectonic data for the study area was compiled from Seismotectonic Atlas of India and its Environs (SEISAT, 2000). Initially 26 tectonic units were considered, in subsequent studies these were increased to 118. The PR technique consisted of six steps: identification and extraction of features, classification of data, discriminant analysis, i.e. training exercise, decision making based on results of training exercise and validation of results. This PR technique was used twice, once for identification of seismically susceptible areas and again for segmentation of MBT and MCT. A circle of 25 km radius was drawn around each epicenter of the MHD catalogue, which was a central earthquake. Twelve features were extracted from this circle. All epicenters in the MHD catalogue were classified into two training classes according to magnitude. 15 iterations of discriminant analyses were carried out by varying number of features and classification criteria. The best result in terms of percentage of classification of epicenters was retained for further analysis. A linear combination of extracted features and discriminant functions λ2, λ3,.., λ12 was used to compute seismic score for each earthquake, Ri. This transformed the original set of features of an earthquake into a single seismic score. This was used to maximize the difference between the two training classes Multivariate mean of the two classes was subsequently used for decision making. 196 grid points at half degree intervals were considered in the study area and the same 11 features were extracted from a circle of radius 25 km drawn around each site. Using λ2, λ3,.., λ12 values obtained ii in the training exercise a seismic score was computed for each grid point. These were compared to the multivariate means obtained in the training exercise and this was used to assign new classes to each site, henceforth named as classes, A′, B′ and C′. Envelopes were drawn around clusters of same class and after validation three areas were identified as: Area A″: most susceptible, Area B″: moderately susceptible and Area C″: least susceptible. The next step was segmentation of MBT and MCT using the PR technique. All epicenters in the MHD catalogue were again divided into two classes with respect to distance from MBT. Discriminant analysis was repeated with four features. Graphical representation of seismic scores indicated four prominent clusters of crossover epicenters which were located at same position in all iterations. These clusters of seismicity within a well defined tectonic environment together with presence of transverse tectonic units, change in seismicity pattern along the thrust and change in strike were used for segmentation of the MBT. The six segments identified are: Poonch segment, Udhampur segment, Kangra segment, Solan segment, Dehradun segment and Nainital segment. A similar exercise carried out for MCT yielded 5 segments: Mashko segment, Chenab segment, Kinnaur segment, Uttarkashi segment and Bageshwar segment. Tectonic units in the study area increased from 118 to 127 after segmentation and these were used for hazard assessment. Hazard maps were prepared for DSHA study with two different approaches: DSHA-1 and DSHA-2, for the entire study area. The map prepared for DSHA-1 showed that PGA varied between 0.012 - 0.470g. This study was repeated by refining several aspects such as 127 tectonic units, assigning maximum magnitude to each tectonic unit and calculating PGA using Boore and Atkinson, 2008 relationship. A hazard map was prepared which revealed that PGA in the study area was substantially higher than in the previous study and varied between 0.039 - 0.581g. For PSHA study the study area was divided into nine seismogenic source zones (SSZ). These are named as SSZ1 to SSZ9 and are: the Kangra SSZ, Uttarakhand SSZ, Kashmir Syntaxis SSZ, Kaurik SSZ, Kargil Laddakh SSZ, Western Nepal SSZ, Karakoram SSZ, Jhelum SSZ, and Indo Gangetic SSZ. Hazard parameters were computed for each source zone. GMPE’s given by Abrahamson and Litehiser (1989) and Boore and Atkinson (2008) were used to estimate PGA in each SSZ. PGA in the entire study area was estimated to vary between 0.039 - 0.289g for 10% probability of exceedance in 50 years for a return period of 475 years. PGA varied between 0.038 - 0.723 g for 2% probability of exceedance in 50 years, for a return period of 2,475 years. By iii retaining the SSZ and hazard parameters of the earlier exercise the PSHA was repeated to study the effect of varying GMPE’s on PGA. PGA in the study area varied between 0.013-0.315g for 10% probability of exceedance in 50 years for a return period of 475 years and varied between 0.024 - 0.780g for 2% probability of exceedance in 50 years for a return period of 2,475 years. Four hazard maps were developed for PSHA studies. Validation of results of return periods for MW ≤ 5.7 was carried out for each of the nine seismogenic source zones. This could not be validated for higher magnitudes as no earthquake of magnitude MW > 5.7 occurred after 2012, as per the MHD catalogue-2 for validation. These earthquakes were not part of the data set used for computation of hazard parameters. Seismic hazard parameters computed for each SSZ showed that the most hazardous zone in the entire 7o by 7o study area is the Kangra seismogenic source zone, SSZ1. This zone is currently going through a rapid phase of techno–economic development and hydro electric potential is tremendous in this Himalayan zone due to the presence of many rivers and their tributaries. Return period computed for an earthquake of magnitude Mw= 8.0 revealed that it will occur somewhere between the years 2003 to 2109, earthquakes of magnitude Mw 7.0 and 6.0 are overdue since 1964 and 1966, respectively, and earthquakes of magnitude Mw 5.0 -5.7 are occurring more frequently than computed. This indicates that an impending earthquake disaster is overdue in the Kangra source zone and underlines the urgency of estimating future implications for the Kangra source zone. Susceptible areas superimposed on SSZ map showed that almost 73% area of Kangra SSZ comprised of susceptible area A′′, i.e. approximately 17,500 km2. Therefore the Kangra SSZ was narrowed down to this truncated area. Recent epicenters from MHD catalogue-2 overlaid on the Kangra SSZ revealed that 23 out of the 24 epicenters were within this truncated area. Therefore, the truncated area can be considered where most current events are located. PGA contours in the truncated area varied between 0.37g and 0.58g as per DSHA. Risk to loss of human life calculated for five districts located within the truncated area indicated that Kangra district would suffer maximum human casualties and injuries followed by Mandi, Hamirpur, Bilaspur and Chamba districts. It is pertinent to note that PGA computed was much higher than that assigned to zone V in the seismic zoning map of India as per, BIS 1893- 2002; therefore, actual casualty figures and injured may be much higher than computed. Isoseismals of four destructive earthquakes plotted on iv truncated area showed that maximum damage due to these four events is concentrated within the truncated area. Therefore, this area is under an enhanced threat perception. After all these studies the following can be deduced for the truncated area. It is an area where the following are concentrated: (1) Return period for an earthquake of magnitude MW= 8.0 is between the years 2003 to 2109, an earthquake of Mw 7.0 and 6.0 is overdue since 1964 and 1966, respectively, and earthquakes of magnitude Mw 5.0-5.7 are occurring more frequently than computed, (2) Computed PGA values are very high and are in the range 0.37g to 0.58g, (3) current seismic events are located, (4) meizo-seismal areas of four destructive earthquakes are located, and (5) risk to population is very high. This has tremendous implications in the Kangra SSZ for future. Starting with a 7o by 7o study area the results were narrowed down to hypothesize a predictive model for a smaller area, 17,500 km2, where frequent destructive earthquakes are expected in the near future. Therefore, urgent preparedness, emergency responses and disaster mitigation measures are required in this area. The results presented here for the Kangra SSZ can be obtained similarly for all the other eight SSZs and risk implications can be estimated.