Introduction

The development of unconventional reservoirs relies on hydraulic fracturing, a technique that stimulates fracture network and enhances permeability by injecting high-pressure fluids into production formations (Montgomery and Smith, 2010). Microseismic monitoring is currently the only technique that can provide real-time information about the geometry of stimulated fracture network (Maxwell, 2014). The key task in microseismic monitoring is locating microseismic events. The traditional method for microseismic imaging is arrival-time inversion with traveltime picking adopted from earthquake seismology (Gibowicz and Kijko, 2013). However, the microseismic data often contain unidentifiable P or S-wave signals emerging from strong background noise, e.g., surface microseismic data (Duncan and Eisner, 2010). This makes traveltime picking challenging.

Recently progress has been made in locating seismic sources with minimum or without traveltime picking. Rentsch et al. (2006) and Rentsch et al. (2010) developed a source location method inspired by Gaussian-beam migration, which has low sensitivity to picking precision. Kao and Shan (2004) introduced a source-scanning algorithm in which data are aligned and stacked based on the traveltime from speculative image points in a manner similar to diffraction-stack (Kirchhoff) migration. Instead of using traveltimes, Gajewski and Tessmer (2005) back propagated full-waveform data by reverse-time modeling (McMechan, 1982). Artman et al. (2010) and Witten and Artman (2011) generalized this approach using P- and S-waves. Zhu (2014) improved the source properties by applying attenuation compensation. Time-reversal imaging is capable of inferring both the location and the start time of a point source, however, it may fail to locate rupture propagation (Kremers et al., 2011), i.e., multiple sources clustered along the time axis. Detecting and discerning each hypocenter therefore requires an evolving microseismic image that has high-resolution in both space and time.

Distributed sensor networks are designed to perform computation in-situ and in real-time. Recently, they have been used for seismic tomography based on the advanced wireless sensor network technology and distributed computing algorithms (Shi et al., 2013; Song et al., 2015). Instead of collecting data to a central place for processing, the distributed seismic data processing and computing can be performed on a single sensor or locally on a few sensors. The results are then gathered for real-time visualization. Employing distributed sensor networks for microseismic monitoring requires a distributed microseismic location algorithm.

In this paper, we develop a distributed microseismic imaging algorithm that is inspired by both passive-source time-reversal imaging and active-source migration imaging. The key innovation is that the imaging principle requires that the imaged hypocenters must correspond to locations where all the wavefields individually backward-propagated from each receiver coincide in both space and time. Our method should be capable of producing high-resolution images of multiple source locations, even when the signal-to-noise ratio (SNR) is low. The processing framework detailed in this paper is naturally suited for distributed sensor networks.

We first introduce the cross-correlation imaging condition for locating microseismic sources and establish a connection with time-reversal imaging. Synthetic examples are used to test the ability of the proposed approach to image multiple microseismic locations at once. We discuss a possible application of the proposed framework to real-time in-situ microseismic monitoring on distributed sensor networks.