ALGORITHM FOR EARTHQUAKE EARLY WARNING SYSTEM

Abstract

Earthquake Early Warning (EEW) system is considered as one of the real-time earthquake damage mitigation measures, which detects, analyses and transmits information of the impending ground shaking prior to the arrival of seismic waves at the potential user sites. The warning time is used to minimize property damage, loss of lives and to aid emergency response. Such systems can be broadly classified as regional and onsite warning systems. While Regional warning approach is network based, the Onsite warning approach uses single station observations for parameter estimation to provide quick warning. The countries like Japan and Mexico have developed the real-time operating EEW systems and are capable of issuing public warnings. Such systems are in preliminary stage of planning in India albeit having comparable seismic hazard to those countries or regions where such systems have been successfully implemented and functioning. The past and the contemporary seismicity reported from Himalaya region and the risk for the cities falling in the vicinity of this seismically active region implicitly require EEW system for Northern Indian region as a mitigation measure. Hence, an attempt has been made in the present study, to understand EEW, develop new EEW parameters, develop a multi parameter based EEW algorithm for accurate and reliable EEW, size estimation during the issuance of warning and propose EEW system for disaster mitigation in seismically active Northern Indian region. The basic requirement of an EEW system is the development of a real-time algorithm for fast calculation of earthquake source parameters and the estimation of their reliability. The EEW algorithm automatically detects the P-onset followed by the estimation of location, magnitude and intensity of the event. The instrumentation density deployed in the seismic active area along with the network deign to be used in an EEW system is also an integral part of the system which affects performance of the system. Initially, status of EEW system in different countries with respect to their development, implementation, and social resilience has been studied followed by a detailed review of several EEW parameters, methodologies and systems present worldwide which has helped in marking the research gaps. In the present study, a new EEW parameter called Root Sum of Squares Cumulative Velocity (RSSCV) has been introduced, which is a function of velocity of the incoming time series and in turn represents the energy component in the time series. The ii robustness of this parameter has been shown by detecting the P-onset in real-time based on RSSCV and also using it as one of the parameters for issuing warning. The comparison with classical amplitude based (STA/LTA type) approaches RSSCV has better performed in case of higher magnitudes which supports its use for EEW systems. Therefore, in the developed EEW algorithm, RSSCV parameter has been used for auto P-picking and warning threshold estimation as well. The strong motion data set for the study compiled from Indian regions reveals the paucity to carry out the regression analysis for development of EEW algorithm. Therefore, data has been imported for the present study from regions like Japan, where K-NET has produced strong motion data from active seismic region having dense network and Pacific Earthquake Engineering Research Centre-Next Generation Attenuation Project Strong Motion Dataset (PEER-NGA). The dataset for development of multi-parameter algorithm taken from K-NET seismic array in Japan consists of 1726 records from 105 events having 5 ≤ M ≤ 7.2 with epicentral distance ≤ 60 km. To indigenize the developed EEW algorithm using Japanese dataset, Indian strong motion data has been taken from four different regions namely Region-1: North West Himalaya (Himachal Himalayas); Region-2: Uttarakhand region (Garhwal and Kumaoun Himalaya); Region-3: National Capital Region (consisting of Delhi-Haridwar Ridge region) and Region-4: North East Himalaya (North East Indian region). Indian dataset comprises of 51 digital records of 28 events within epicentral distance upto 60 km and magnitude range varying between 3.3 to 6.8. To further validate the algorithm on worldwide dataset, the data has been taken from countries such as Southern California, Taiwan and Turkey which consists of 219 earthquake records of 14 earthquakes having magnitude range of 4.27 ≤ M ≤ 7.62 within 60 km of epicentral distance. The reliable issuance of warning by EEW system depends upon the accuracy and reliability of predicted parameters used to define the size of the incoming event in real time. Such parameters are estimated using the analyses of initial portion of earthquake records. In the present study, not only the individual parameters such as Maximum Predominant Period (τp max), Average Period (τc), Peak Displacement (Pd), Cumulative Absolute Velocity (CAV) and RSSCV have been used to develop an algorithm but also various combinations have been attempted to issue alarm and estimate magnitude with reliable accuracy in minimal time window. The estimated parameters are empirically iii regressed with the apriori known catalogue magnitude of the event at variable time windows starting from 1 sec to 5 sec to determine threshold values for the considered parameters to issue warning for event having M ≥ 6. For example, for a time window of 4 sec the threshold values of parameters are found to be 1.1 sec for τp max, 1.42 sec for τc, 0.95 cm for Pd, 23 cm/sec for CAV and 5.2 cm/sec for RSSCV, respectively. The threshold values calculated for issuing warning at different time windows have been compared with the threshold values suggested by other researchers and a close match has been noticed. The criterion for issuing warning is based on the alarm status of nearest four stations within selected epicentral distance of the event. Out of these four when three stations cross the preset threshold value of an EEW parameter alarm is issued. Based on the correctness of the issued alarm, the efficiency of the system is evaluated by classifying the alarms into Correct Alarm (CA), Missed Alarm (MA), Correct All Clear (CAC) and False Alarm (FA). Further, the developed algorithm also includes the multi-parameter approach for issuing warning. Under this approach, the status of CA/MA/CAC/FA has been tested on various possible combinations of considered parameters. The combinations of EEW parameter preference based approaches include three, four, five and logic combination of EEW parameters preference based approach which has been tested at five different time windows. The three parameter preference based approach has been found to be most efficient at a time window of 4 sec. Under this approach if three out of five EEW parameters show similar alarm status, the event has been marked with the same status. After issuing the warning for high magnitude event, the algorithm further explores the data for more accurate estimation of magnitude using Brune’s model based approach. Brune’s model has been fitted on initial P-wave data after P-onset at different time windows starting from 1 sec to 10 sec for estimating spectral parameters such as low frequency spectral level, corner frequency and cutoff frequency for calculating the seismic moment and in turn moment magnitude (Mw). The residuals with respect to window length reveal the direct proportionality of the length with accuracy of magnitude estimation. In case of EEW system, the trade-off between accuracy and time plays an important role where a way has to be adopted that provides reliable magnitude estimation in acceptable limits of time. The study has concluded that a time window of 5 sec gives an uncertainty of ±0.3 in magnitude estimation which is considered to be an optimal time to confirm the magnitude of the issued warning. iv The status of seismic hazard in Northern India and the risk associated with the cities around Himalayas reveal that the region between MCT and MBT is the zone of large numbers of seismogenic sources which establish the need for installing an EEW system for Northern India. The instrumentation and network connectivity for the EEW system have also been proposed in the present study followed by the calculation of possible lead time for the cities in northern India such as Dehradun, Hardwar, Roorkee, Muzaffarnagar, Meerut and Delhi. It is found that for all the cities the time available for alarm varies from 5 sec to 90 sec which is substantial time to act for saving human lives and for activation of emergency response measures such as immediate shutdown of industrial units, nuclear power plants, gas lines, pipelines, computers and slow down high speed train. The use of EEW system as a real time risk reduction measures for Northern Indian region for disaster mitigation and management can never be over emphasized. It is envisaged that this work of multi-parameter EEW algorithm will be suitably utilized in EEW systems in India.