Method for real-time automatic identification of S-phase： Application to local seismicity

This paper proposes a real-time automatic method for S-phase identification in earthquake early warning. The proposed method directly analyzes three-component data without any filtering processing to the original signals. Firstly， the proper window length to identify the seismic phase is determined based on the predominant frequency of the first 0.5 s data prior to the P phase. Secondly， a characteristic function is applied to estimate the identification interval of S-wave. The function is defined as the product of the square of deflection angle and the square of ratio of horizontal seismic wave energy to total seismic wave energy. Trigger threshold of the characteristic function is taken as the sum of 5 times the average of the calculated characteristic function values and 5 times the variance of the calculated characteristic function values. Then VAR-AIC (variance-Akaike information criterion) method is applied to the two horizontal component data in the identification interval. The first break time of S-phase is determined after judgments on the two identification results. We apply the picking method to 118 three-component earthquake records with various magnitudes and compare the results from the automatic identification method and the interactive phase picking method. The results show that 89.39% of all tests have identification time errors less than 0.1 s from the proposed method for seismic recordings with signal to noise ratio of S waves to P wave coda higher than 5 dB.