A reliable strategy for improving automatic first-arrival picking of high-noise three-component microseismic data

We present a reliable first-arrival picking method using wavelet multilevel analysis (WMA) and polarization analysis for high-noise three-component (3C) microseismic data. The proposed method is based on the autoregressive (AR) model using the Akaike information criterion (AIC) algorithm, named PWT-AIC (in which PWT is polarization wavelet transform). This study aims to address the problem that the AR-AIC picker selects the global least point as the first arrival; it will cause a questionable arrival identification while being applied to the 3C microseismic recordings with a low signal-to-noise ratio. Initially, we employ the WMA to extract the dominant signal for high-noise 3C microseismic data; we then use the reconstructed approximation data to perform the AR-AIC calculation to identify the first arrival. Furthermore, we adapt the polarization information of 3C microseismic data to determine the proper calculation section for the AR-AIC algorithm. We conduct a polarization analysis by adopting the eigenanalysis of the sliding covariance matrix. Compared with the popular short-term average/long-term average ratio and existing AR-AIC pickers, the presented algorithm can significantly reduce the picking error. A test using synthetic 3C seismic data with high-noise indicates that the onset time can be accurately identified, and the improved method has an error of between 1 and 2 sample intervals. Results using field microseismic recordings also confirmed that the proposed strategy can improve the accuracy of arrival-time estimation for noisy 3C microseismic datasets. In addition, we perform a time consumption comparison of the different approaches. Although the proposed picker requires more computation time than the other pickers, the results are acceptable, given the capabilities of modern computers.