A wide variety of autopicking algorithms have been developed and are being used to detect phase arrivals from microseismic data. All of the available methods are effective in high SNR (signal-to-noise ratio) environments. However, with the reduction in SNR, there is always a possibility of the arrivals not being detected and classified by standard methods. Different autopicking workflows in use within the academia and the industry show different degrees of performance drop under different scenarios. This is particularly the case in very noisy environments associated with most hydrofrac projects. The quality of the first arrival detection is related to the near- and sub- surface structure, source type, and prevalent SNR conditions. Any of the available autopicking methods could be applicable in some specific scenarios and may fail in others based on the acquisition conditions. Therefore, finding a robust method to work under most circumstances is a major challenge. Moreover, due to large volumes of seismic data common in acquisitions involving passive seismic arrays and the complexity of the autopicking approach in use, the detection of the arrivals can be time consuming. Since detection of the first arrival in a fast and accurate fashion is the key step for additional processing, the aim is to work with data that is characterized by low SNR and poor overall data quality to develop an effective workflow to obtain P-wave and S-wave first arrivals with high accuracy by using combination of attributes in an ANN (artificial neural networks) framework. In this paper, we provide a comparative analysis of the efficacy of different picking algorithms on a synthetic dataset. We benchmark our novel ANN based autopicker with two contemporary autopicking algorithms and validate its applicability for the test case.

This content is only available via PDF.
You can access this article if you purchase or spend a download.