Sixth EAGE Workshop on Passive Seismic

The success of microseismic monitoring campaign strongly depends on using the correct sensor layout. Detection limit and location accuracy are affected by the capability of a network to uniformly capture the energy emitted by seismic events in the subsurface. We here highlight the importance of using a accurate, 3D velocity model.

We compare performance of a hypothetical network design using 1D and 3D velocity models. Our results show that the complexity of the 3D model translates into the key network performance indicators, such as distribution of magnitude of completeness, location accuracy and ground motion amplitudes.

Recent studies have identified and characterized a type of seismic event, known as an LPLD event, which have been detected in microseismic data sets acquired during hydraulic fracturing operations (e.g. Das and Zoback, 2011 ; 2013a ; 2013b ; Mitchell et al., 2013 ; Kwietniak, 2015 ). These events have been interpreted to be manifestations of slow-slip along preexisting fractures which are presumed to either be misaligned with respect to the current day principal stress directions or have high clay content ( Das & Zoback, 2013a ; 2013b ). A study by Caffagni et al. (2015) advise that care must be undertaken when analyzing and interpreting such events as regional earthquakes could be misinterpreted as LPLD events in vertical downhole seismic monitoring array data sets.

We here show that signals associated in time with such LPLD events could be observed on many Earthscope USArray stations, even at distances up to 350 km from the injection well. The spatial coverage of the USArray enabled all of the LPLD events to be relocated in the North Texas-Oklahoma region, outside of the stimulated reservoir volume. We conclude that these LPLD events are not directly related to the hydraulic fracture stimulation process or the induced reservoir deformation process.

We examine a case study in a North American shale play where a number of wells were stimulated. Because the completions were monitored with two whip arrays, we were able to perform seismic moment tensor inversion on many of the detected events. For the first well, we say many sub-vertical failures following two dominant orientations while the second well showed a predominance of sub-horizontal fractures, with a lower overall event count. The initial predictions indicated that the first well would have better production due to the larger volumes of more complicated stimulated volumes. However, when the production data was acquired, it was the second well that had the higher production. We resolved this paradox through examining the apparent stress of the events: the first well had higher apparent stresses more characteristic of fault activation; the second well had lower apparent stress values more characteristic of fluid induced events. Our interpretation is that the fluids allowed for bedding planes to slip and stimulate a larger volume whereas the first well did not succeed in transferring fluids to the seismically active volumes.

In certain shale or carbonated reservoirs, seismicity associated with Hydraulic Fracturing (HF) or Enhanced Oil Recovery (EOR) techniques can be detected and characterized by sparse surface networks of high-quality three-component instruments. Here we provide pilot project examples of surface seismic networks, initially deployed for induced seismic monitoring, generating rich data sets in the Duvernay and Montney shale plays in western Canada.

In particular, we show that in environments with favourable surface noise levels, high in-situ stress regime, low anelastic attenuation and shallow depth reservoirs, such networks can detect seismic events down to magnitude -1.0 within the stimulated volume along with induced seismicity on proximate faults.

The authors present and contrast the results of localizations within various velocity models: a unified 1D isotropic model, a unified 1D elliptically anisotropic VTI model; a unified 3D elliptically anisotropic VTI model; and an arbitrarily anisotropic 3D model. This use case highlights the value of implementing the appropriate level of physics to solve the problem effectively. The authors provide quantitative data to support the journey from the most simplistic isotopic result to the most advanced arbitrarily anisotropic result. The case for directly linking location accuracy to velocity model efficacy is evidenced on a West Texas microseismic project. Further, the authors briefly discuss the implications from an interpretative standpoint as we evidence the transformation of data into decisions.

Microseismic mapping provides essential information for monitoring effective fluid flow and stimulated reservoir volume, as well as for decision making on fracture operations and well completions. This work presents the microseismic monitoring results of a multiwell completion of a field located in the Sultanate of Oman. Target depth is about 1300 m from ground level. The treatments of three wells were monitored by both a surface array and a downhole array. The surface monitoring array was built using a 50-m inline spacing and 100 m crossline spacing geometry and comprised 1000, 974, and 975 3C geophones when monitoring the TW1, TW2, and TW3 wells, respectively. The downhole monitoring was performed using a 12-level 3C-geophone accelerometer array using 30-m-long interconnects deployed in the nearby well, MW. Both surface and downhole seismic data cover the complete stimulation campaign of all three treatment wells. Surface- and downhole-acquired data benefit from a combined inversion. Surface-acquired horizontal locations reveal the same event geometry as downhole results but with a more focused linear cloud. Mapped events indicate a well-bound stimulation along the vertical dimension while exhibiting predominantly strike-slip mechanisms.

Microseismic monitoring performed during hydraulic fracturing records rock failure that occurs during or after stimulation. The processed microseismic event catalog can be used to study the dynamic fracturing process and estimate the background fracture and stimulation efficiency. However, in microseismic applications the event catalogs are commonly used as static datasets, where the temporal and spatial evolution information is often ignored or not used to its full potential. Consequently, the natural fractures in the resevoir and the dynamic fracture growth information are not fully revealed. This paper describes how coupling microseismic data with well pressure measurements and image logs presents an enhanced workflow for analyzing microseismic events in a temporal and spatial manner. By using a dataset obtained in a Devonian shale gas reservoir, we illustrate the added value of using microseismic as a dynamic dataset to understand the fracture network growth due to injection.

After almost two years of monitoring of the Groningen reservoir in the Zeerijp area, Netherlands, the set of microseismic locations now enables to confirm at the local scale several properties of the seismic activity in the Groningen gas field. Events occur (i) mainly within the reservoir, (ii) along fault segments identified at the base Zechstein formation and (iii) have magnitude distributed according to a Gutenberg-Richter with a b-value of 0.82 ± 0.03. Our catalogue contains more than 500 events with M > −2.5 with a completeness level of 0.1. The period covered also underwent a significant rise in seismicity in relation to past activity levels, with the notable occurrence of four M > 2 events. Extrapolation of the Gutenberg-Richter law provides a 10 to 17 years return period for a M ≥ 3.9 in this area. The processing of such events proved to be particularly challenging due to the deployment of sensors within a low-velocity layer, leading to very complex waveform signatures. We had to develop a multi-phase location approach, in combination with comparison with synthetic waveforms to overcome this difficulty.

Nowadays, in a passive seismic processing STA/LTA based on the cross-correlation is computed. This processing is done in order to remove the noise (Earthquake) from the natural noise.

Detecting the Earthquake play a key role in passive seismic studies.

However, in a passive seismic monitoring system record files contain months and years of data and manually analyzing the data can be expensive and inaccurate.

In addition, in the presence of noise, detecting the arrival of weak microseismic events becomes troublemaker.

In this passive seismic study, we manually determined the known earthquake. As well as, we computed STA/LTA based on the cross-correlation of the energy ratios method.