Summary Reliable detection of weak signals is potentially a fundamental limitation for microseismic event detection and location, especially for sensors in the near-surface region, where large source distances and high noise levels affect the signal-to-noise ratio (SNR). Previously, a nonlinear stacking method was introduced that could increase the SNR for weak signals and included methods that are insensitive to changes in signal polarity. This paper introduces nonlinear filtering methods that are a generalization of the nonlinear stacking method where the stacking operation is replaced by a filtering operation. This method keeps the SNR and polarity benefits of the stacking methods, but we can now choose to modify the pass and reject bands. For instance, we can improve the SNR for weak signals, with polarity changes, and allow additional signal misalignment criteria to be included in the passband. We will show synthetic-signal and real-signal examples where the new method has clear benefits compared to conventional stacking or other nonlinear stacking methods.

Summary In the context of surface microseismic processing, a nonlinear stack method, the phase weighted nth root stack, was evaluated and benchmarked against linear stack. From a synthetic analysis, the parameter choice was evaluated to improve the detectability of small amplitude microseismic events. The choice and impact of small exponent values for the nonlinear stack method were confirmed on one stage of a multiwell, multistage hydraulic fracturing in the Marcellus shale formation. Compared to the linear stack, up to 30 % more events were detected and located for this data example.

Summary Noise attenuation is a key challenge for surface-acquired microseismic processing. A number of data conditioning tools have been proposed and applied with various degrees of success to improve the signal-to-noise ratio prior to detection and location of microseismic events. Random noise attenuation, trace-by-trace correlation with a large magnitude event, and nonlinear stacking techniques have all been shown individually to improve microseismic event detectability in surface-acquired microseismic datasets. This paper demonstrates how the combination of these approaches significantly increases the number of detected microseismic events while keeping the number of false triggers to a minimum. In particular, random noise attenuation and trace-by-trace correlation with a large magnitude event followed by nonlinear stacking at the stage of substack generation provide a data conditioning workflow that significantly attenuates the effects of statics, anisotropy, and, to some extent, 3D velocity variations. This work is a step towards an optimized data conditioning workflow for surface-acquired microseismic data.

The perforation of the borehole casing and cement is the final stage in a well-completion procedure to establish a connection with the reservoir for hydraulic fracturing purposes. Although they have been neglected to date, the generated seismic expressions of these explosions in the stimulation well display a very characteristic signature. A Fayetteville Shale stimulation concurrently monitored using a downhole and surface seismic array was analyzed to develop a better understanding of the perforation arrivals and to identify the influence of reservoir properties, using a combination of data processing and forward modeling. The perforation shots typically give rise to multiple P- and S-wave arrivals. Comparison of observed and modeled arrivals reveals the importance of incorporating two crucial geologic parameters: a 1.6° subsurface dip and a strong anisotropy in the Fayetteville Shale, typically between 40% and 50%. Of equal importance are shot-generated tube waves traveling through the treatment well, which convert into body waves after interaction with the perforation plugs that seal off the borehole and give rise to secondary high-amplitude, low-frequency seismic events. Hence, the waveforms related to source, propagation, and treatment operation effects become strongly interlinked and need careful separation and interpretation to allow for the correct identification of primary and secondary perforation arrivals and for the reliable derivation of velocity models for surface microseismic monitoring.

Summary A Fayetteville shale stimulation concurrently monitored using a surface seismic array and a wide aperture downhole array is analysed in order to develop a better understanding of the factors influencing the surface expression of perforation shots using a combination of data processing and forward modelling. The perforation shots typically give rise to multiple compressional and shear wave arrivals. A comparison of observed and modelled arrivals reveals the importance of two geological parameters: a 1.6 degree subsurface dip, which had been neglected in previous studies, and a strong Thomsen anisotropy in the Fayetteville shale, typically between 40% and 50%. Of equal importance are shot generated waves (tube waves) travelling through the treatment well which interact with plugs remaining in the well and give rise to secondary high amplitude, low frequency, seismic events. These arrivals, which are related to source, propagation and treatment operation effects, need careful separation and interpretation in order to allow the reliable derivation of velocity models for surface microseismic monitoring.

Summary In 2011 Schlumberger and an independent operator jointly acquired a comprehensive dataset of hydraulic fracturing operations which stimulated the Fayetteville shale: signal and noise were tracked from reservoir to surface and then across the surface. The performance of alternative monitoring technologies such as surface and shallow borehole seismic arrays as well as a downhole seismic array are analyzed. The noise levels on the different seismic arrays are characterized and appropriate attenuation techniques are applied. Furthermore linear and non-linear stacking methods are utilized to increase the signal-to-noise ratio (SNR). The application of the Coalescense Microseismic Mapping algorithm to locate microseismic events concludes that an accurate source location requires an accurate velocity model together with consistent and aligned signal. Coherent noise, insufficient signal or noise discrimination are key factors influencing location uncertainty. Although surface line segments benefit from larger receiver apertures, SNRs of their stacks were usually only slightly higher than from stacks of surface patches using linear stacking methods. However, source localization is more accurate using surface lines. Nevertheless, surface patches are easily deployable so provided sufficient signal is recorded and noise attenuation methods are applied prior to stacking and source scanning, they may ultimately become the preferred surface monitoring configuration.

The recent expansion of surface and near-surface microseismic monitoring during hydraulic fracturing has led to the development of techniques that are directly applicable to more general permanent microseismic reservoir monitoring. Recent work associated with surface and near-surface microseismic monitoring of hydraulic fracturing for shale gas in the Fayetteville and Marcellus formations has helped to develop new methods to process low signal-to-noise ratio (SNR) data. Improvements to the SNR by stacking the data using novel non-linear stacking techniques and stacking of perforation shots allows the use of Coalescence Migration Mapping (CMM), an established technique used for downhole monitoring, to detect and locate microseismic events from surface data.

Coalescence microseismic mapping (CMM) is an established technique used for real-time detection and localization of microseismic events (Drew et al., 2005). It has historically been used for hydraulic fracture monitoring (HFM) where there is a downhole array located in a monitoring well. This paper presents a variation of CMM which is adapted for weak signals and its application to hydraulic fracture monitoring conducted using a surface seismic array. Data from a multi-well, multi-stage stimulation in the Marcellus Shale concurrently monitored using a surface array and a downhole array illustrate the effectiveness of this adapted CMM algorithm.

Locating microseismic events using a method such as coalescence microseismic mapping (CMM) is much more effective if the underlying detection function, such as STA/LTA, has reliably detected the events. This in turn depends on the SNR of the data that the detection function has been applied to. These issues are especially important for surface microseismic monitoring, where the SNR of the recorded data is significantly less than that of borehole data. In this paper, we introduce novel nonlinear stacking methods to improve automatic detection and location of weak seismic events that are typically below the background noise-level. We also introduce several novel methods that have applications for detection and location of signals with varying source radiation patterns that have a phase change. We show the outline of the methods on synthetic data and also demonstrate application to real data acquired during surface microseismic monitoring. In the real data case we compare one nonlinear stacking method with the conventional stack for a series of perforation shot arrivals and show how the signal-to-noise ratio is significantly improved for a real event, and how this improves the event detection and location sensitivity by a combination of signal-to-noise ratio improvement and noise discrimination.

In 2011, a comprehensive microseismic monitoring case study was acquired in the Fayetteville Shale using surface, shallow-well, and deep-borehole sensors to help evaluate and compare different hydraulic fracture monitoring techniques. We apply methods based on Coalesence Microseismic Mapping (Drew et al, SPE 2005) to the different arrays, introduce some new methods, and highlight some of the advantages and disadvantages of each data set.

Microseismic monitoring services were conducted during the stimulation of two horizontal wells that targeted the Fayetteville shale formation in Arkansas. Data acquisition was provided using a surface array, vertical arrays in a grid of five shallow wells, and a downhole array which extended from reservoir depth to the surface. Analysis of this data was conducted as part of a comprehensive test to assess and quantify the detection, accuracy and resolution of surface, near-surface and downhole acquisition systems.