Using beamforming to maximise the detection capability of small, sparse seismometer arrays deployed to monitor oil field activities

Abstract

Like most other industrial activities that affect the subsurface, hydraulic fracturing carries a risk of reactivating pre-existing faults and thereby causing induced seismicity. In some regions, regulators have responded to this risk by imposing Traffic Light Scheme-type regulations, where fracture stimulation programs must be amended or shut down if events larger than a given magnitude are induced. Some sites may be monitored with downhole arrays and/or dense near-surface arrays, capable of detecting very small microseismic events. However, such monitoring arrangements will not be logistically or economically feasible at all sites. Instead, operators are using small, sparse arrays of surface seismometers to meet their monitoring obligations. The challenge we address in this paper is to maximise the detection thresholds of such small, sparse, surface arrays, so that they are capable of robustly identifying small-magnitude events, whose signal-to-noise ratios may be close to 1. To do so we develop a beam-forming-and-stacking approach, computing running short-term/long-term average functions for each component of each recorded trace (P, SH and SV), time-shifting these functions by the expected travel-times for a given location, and performing a stack. We assess the effectiveness of this approach with a case study, using data from a small surface array that recorded a multi-well, multi-stage hydraulic fracture stimulation in Oklahoma over a period of 8 days. As a comparison, we initially used a conventional event-detection algorithm to identify events, finding a total of 17 events. In contrast, the beam-forming-and-stacking approach identified a total of 155 events during this period (including the 17 events detected by the conventional method). The events that were not detected by the conventional algorithm had low signal-to-noise ratios, to the extent that in some cases they would be unlikely to be identified even by manual analysis of the seismograms. We conclude that this approach is capable of improving the detection thresholds of small, sparse arrays, and so can be used to maximise the information generated when deployed to monitor industrial sites.