Characterizing hydraulic fracture growth using distributed acoustic sensing-recorded microseismic reflections

Characterizing the geometry and orientation of hydraulic fractures can provide insights into subsurface reservoir behavior and contribute to more effective operational decisions and production strategies. Established techniques to evaluate hydraulic fracture effectiveness include postfracture pressure-decay analysis (Sullivan et al., 2019), fluid-tracer analysis (Kumar and Sharma, 2020), interwellbore strain using low-frequency distributed acoustic sensing (LF-DAS) observations (Jin and Roy, 2017), and microseismic monitoring (van der Baan et al., 2013; Maxwell, 2014; Eaton, 2018). Of these methods, microseismic monitoring provides the most direct approach for the surveillance of fracture extent and geometry. Common products of microseismic monitoring include the origin time, hypocentral location, magnitude, and source mechanism of detected microseismic events. Microseismic clouds can help to infer fracture geometry and orientation, reveal fracture-system connectivity, and estimate stimulated volume and permeability (Maxwell, 2014).

In recent years, distributed acoustic sensing (DAS) has been increasingly used for multiple purposes during hydraulic fracturing well completion. The DAS system operates by emitting coherent laser pulses into an optical fiber (Hartog, 2017). Part of the signal is reflected due to backscattering, and the phase change of the back-scattered signals can be converted to strain or strain rate using interferometry. Fiber cables can be permanently deployed (cemented) behind a casing (Daley et al., 2013; Naldrett et al., 2020) or temporarily deployed on a wireline (Eaton et al., 2022; Wang et al., 2022). Once a fiber-optic cable is deployed downhole, it can be used for many purposes, such as microseismic monitoring (Cole et al., 2018; Karrenbach et al., 2019; Verdon et al., 2020), measurement of strain changes (Jin and Roy, 2017; Zhang et al., 2020; Bourne et al., 2021), and integrated fracture characterization (Zhang et al., 2021).

DAS offers many advantages over traditional arrays. Fibers can be kilometers long and monitor an entire well with broadband frequency (from mHz to kHz) and high spatial resolution (spatial sampling <1 m), which allows the detection of a large number of microseismic events (Karrenbach et al., 2019; Lellouch et al., 2020a). Traditional arrays usually provide 3C point measurements; however, DAS can only measure a single (axial) component of strain or strain-rate changes along the fiber, averaged over a gauge length, which creates a challenge for determining event locations (Lellouch and Biondi, 2021). Fortunately, recording DAS data simultaneously in multiple wells can effectively mitigate this limitation (Cole et al., 2018; Verdon et al., 2020). Multiphases recorded by DAS, such as interface-reflected waves and guided waves (Lellouch et al., 2022), can further help to resolve the inherent ambiguity in event location. The gauge length effect can be described as a function of the velocity and frequencies of the seismic waves being measured, which may vary considerably (Dean et al., 2016; Hartog, 2017). In addition to the typical frequency of DAS data for seismic monitoring, the LF-DAS (<1 Hz) measurement is sensitive to mechanical strains induced by dynamic fracture propagation and is therefore suitable for monitoring the growth and geometry of hydraulic fractures (Becker et al., 2017). LF-DAS can provide a direct observation of fracture opening (extension) and closing (compression) and fracture-driven interactions (FDIs), also known as frac hits, at the monitoring fiber.

The dense spatial sampling of DAS recordings provides rich wavefields with a large aperture, enabling the detailed characterization of subsurface wave propagation and the application of advanced processing and imaging algorithms. In addition to P- and S-wave direct arrivals, DAS-recorded wavefields potentially contain a near-field strain response (Luo et al., 2021), guided waves (Huff et al., 2020; Lellouch et al., 2020b), and reflected waves (Staněk and Jin, 2021). Thus, DAS-recorded waves contain information not only about the microseismic event location and source mechanism but also about the subsurface structure and how the velocity and medium properties change around the monitoring fiber. In particular, some recorded reflected waves can reveal the location of nearby induced hydraulic fractures or small faults and can thus be used to image fractures (Ma et al., 2022; Staněk et al., 2022). Treating microseismic events as active sources, local media between event hypocenter and monitoring fiber can be inverted by making use of traditional ray tracing or migration theory. With a sufficient number of events containing reflections, it is possible, in principle, to perform time-lapse monitoring of the changes in subsurface properties resulting from stimulation. Several case studies have illustrated the observation of reflected waves using 3C geophone arrays and the simultaneous inversion of the event location and velocity model (Lin and Zhang, 2016; Grechka et al., 2017). However, because reflected waves can be ambiguous to interpret using significantly sparser geophone arrays, fracture imaging with reflected waves is not widely applied in traditional microseismic monitoring.

In this study, we apply a recently proposed imaging method to monitor hydraulic fracturing growth using DAS-recorded microseismic reflections. First, we document DAS-recorded microseismic reflections acquired during a multiwell project and analyze reflected waves using a synthetic model. We then apply a ray-tracing-based workflow proposed by Staněk et al. (2022) to map nearby fractures using the extracted reflected S waves. Imaging results are integrated with LF-DAS strain data to confirm that DAS-recorded reflected waves are generated by newly created hydraulic fractures. Fracture imaging based on microseismic reflections reveals the fracture geometry between the treatment and monitoring wells, as well as the fractures that do not reach the fiber. Finally, we demonstrate the potential ability to use fracture images produced by multiple events during the same treatment stage to monitor dynamic fracture evolution.

The DAS data used in this study were acquired during a multiwell project in the Montney Formation, western Canada. The reservoir was stimulated in hundreds of stages along the horizontal sections of three horizontal wells T1, T2, and T3 (Figure 1). Three permanently installed fibers in horizontal (wells A and B) and vertical (well M) wells were used to monitor hydraulic fracturing (Figure 1a) with a 4 m gauge length (Hartog, 2017), which provides high-quality DAS microseismic data and LF-DAS data. Simultaneous stimulation of three wells (Figure 1b) generated complicated fracture geometry. Although the event hypocenter locations were calculated using fibers in multiple wells, the microseismic wavefields and LF-DAS data shown in this paper were recorded in well B.

The raw data were converted to strain and downsampled from 1 to 4 m channel spacing for microseismic data processing, with a 2000 Hz sampling rate. To analyze microseismic signals, we first remove constant bias, noise spikes, and the median value at each time sample to attenuate the system noise, and then we apply a band-pass filter between 10 and 150 Hz. To accelerate the processing of large DAS data volumes, a machine-learning-based workflow (Ma et al., 2020, 2023) is developed for two of the most time-consuming steps, event detection and arrival-time picking. Event detection uses a convolutional neural network to search for events separately on the two horizontal fibers. When events are detected on at least two horizontal fibers within a predefined time interval (we used 1 s for this data set), the events are labeled as being associated with the same microseismic source for the following arrival-time picking and hypocenter location. This relatively long-time threshold was chosen to mitigate the potential impact of false triggering and mismatches that could arise in more complex scenarios, such as when multiple events overlap. Using ray tracing and grid search, the hypocenters of all microseismic events from three stages are mapped, as shown in Figure 2a and 2b. The layered isotropic velocity model (Figure 2c) used for the traveltime calculation is based on well logs from vertical well C. The P- and S-wave first arrival times from all three fibers were used to calculate the residual traveltime. Most of the events are close to the treatment well. A shallower cluster of events 2 km away from the treatment well is also observed, which may reveal local structure reactivation that is known to occur in this area (Wozniakowska and Eaton, 2020). The solid dots represent the events with reflected waves. Although events with fracture-induced reflections are visible in most stages of this project, this study is limited to the three stages (21–23) that provide representative examples of this phenomenon.

Fracture-induced reflections can be identified based on their linear moveout characteristics, S-wave apparent velocity, and the intersection of the linear arrivals with the direct S wave. The wavefields of most events have near-symmetrical moveouts of direct P and S waves. P waves are often not visible around the apex and S-wave signals have stronger amplitude than P waves due to a combination of the fiber radiation and source radiation patterns (Karrenbach et al., 2019). In this study, we focus on complex wavefields as shown in Figure 3. Two representative waveform examples recorded by well B show clear direct P and S waves, S-P converted waves, near-field strain, and reflected S waves (Figure 3). As elaborated next, reflections may indicate nearby induced fractures from previous or current stages. In the case of event A, most reflected S waves intersect the direct S waves that are propagating toward the toe of the well (Figure 3a).

The linear moveout of these reflections is best explained by reflection from an interface that is approximately perpendicular to the fiber. The location of the intersection points with the direct wave (relative to the apex of the microseismic event), coupled with the lack of reflections with the opposite sense of dip (i.e., unstimulated side), suggests that these reflections likely originated from new fractures that opened in previous treatment stages closer to the toe. Conversely, event B contains reflections from the left and right sides of the apex (Figure 3b), suggesting that the reflection with the opposite dip (i.e., toward the heel) may be from a preexisting natural fracture or complex fracture connectivity; however, this needs further analysis with independent proof from LF-DAS and stimulation data. Some reflections do not appear to intersect with the direct S wave (arrows in event A); we interpret this geometry to indicate that the fracture does not extend as far as the monitoring well. Mapping of these short fractures could complement the fracture connection map from LF-DAS and contribute to understanding the sensitivity of LF-DAS to observe the strain change caused by proximal fractures. The reflections observed at the apex of event A (Figure 3a) are interpreted as fracture channel waves (Nihei et al., 1999) that were reflected at least twice between two fractures opened in the current stage.

Although reflected P waves are also visible in some events, as shown for event B, only reflected S waves are used to image fractures in this study because reflected P waves are less common and generally have a lower amplitude than the reflected S waves. Event B also contains arrivals after the direct S waves, which are interpreted to be from horizontal velocity interfaces (Figure 3b). Different from the fracture-induced reflected waves, reflections from horizontal interfaces usually follow the moveout of direct P and S waves within the coda wave and have a symmetric moveout around the apex in horizontal fiber recording (Appendix  A). Based on the computed hypocenter locations, the events with fracture-induced reflections are not likely spatially correlated and could occur at different times, thus providing a wide aperture for 3D fracture imaging. The computed hypocenter of the most distant event with reflected S waves is located more than 1 km away from the monitoring well (Ma et al., 2023).

To verify that the reflections are indeed generated by fractures, we first simulate synthetic seismograms using a finite-difference method (Petersson and Sjögreen, 2017). The basic objective is only to fit the traveltimes of field-observed direct P and S waves and reflected S waves. To simplify the task of fracture modeling, we treat each fracture as a thin low-velocity layer, assume all fractures are perpendicular to the fiber (Figure 4), and conduct 2D modeling within the plane of microseismic events and the monitoring fiber. In the case of event A, we only simulated reflections marked with the dashed lines in Figure 4, using the same layered velocity model (Figure 2c) used to process microseismic hypocenter locations. The model was developed using a trial-and-error approach and was set up based on event A. The synthetic data (Figure 4c) have been converted to axial strain based on the fiber geometry and exhibit similar arrival times for the direct and reflected S wave as field data. The dashed lines indicate the consistent first channels of synthetic and field reflections, which demonstrate that reflected microseismic signals observed by DAS could be created by nearby fractures.

To map fractures using the previously mentioned reflections, we apply a ray-tracing-based fracture image workflow proposed by Staněk et al. (2022) (Figure 4a). This method assumes that the background medium is homogeneous and isotropic, represents the fractures as a planar interface, and assumes that the fractures are perpendicular to the fiber. The velocity in Figure 2c does not show abrupt changes within the Montney Formation, which supports the simplifying assumption used here of a homogeneous model for fracture imaging. The workflow treats every point on the 2D plane between the fiber and the microseismic event hypocenter $[xs,0]$ (the yellow area in Figure 4a) as a potential reflection point $[xf,yf]$⁠. Then, the DAS channel $yr$ along the fiber that records the reflected S wave and the corresponding traveltime of the reflected S wave $ts$ can be obtained based on the ray-tracing theory. In this way, every data point from the DAS data at channel $yr$ and time $ts$ can be mapped into the 2D plane that contains the monitoring fiber and the microseismic event hypocenter. After searching through all potential reflection points on the 2D plane, the observed amplitude is assigned to the corresponding reflection point, which completes the process of fracture imaging. For further details of the method, the reader is referred to Staněk et al. (2022).

To extract reflections from the raw DAS data, we applied a preprocessing workflow (Staněk et al., 2022) as shown in Figure 5, which is similar to conventional vertical seismic profile data processing. An f-k filter is first applied to separate the raw data into heel- and toe-ward wavefields (Figure 5b). Then, all signals above the red lines (Figure 5c) are muted to exclude the influence of the direct waves on the final fracture images. Next, we apply the fracture imaging workflow to heel- and toe-ward reflections using a constant S-wave velocity and then merge two images to form the final complete fracture image. We also apply postprocessing steps, including a low-pass filter below 250 Hz and a median filter using a window of seven samples for display purposes.

Figure 6 shows the fracture imaging results of two example microseismic events (events A and B shown in Figure 3) using the previously described workflow, originally developed by Staněk et al. (2022). Figure 6 shows the fracture images in space (left column), input DAS microseismic wavefields, and separated reflected S waves. Each red-blue image represents a 2D plane between the event hypocenter and the monitoring fiber in well B. The distance of event A and event B from microseismic sources to the fiber is 184 and 543 m, respectively. The red stars represent the microseismic location and the red arrows indicate the imaged and interpreted fractures. Because we mute signals around direct arrivals, including strong coda waves and part of reflections that contain near-fiber fracture information, the imaged fractures do not extend entirely to the fiber (⁠$x=0$⁠) even for long reflections that are connected with direct S waves. This limitation also explains the blank area with no data coverage around the apex of the microseismic event. The dashed red lines show the reflected S waves that arrive later than the direct S arrivals, which implies that the corresponding fracture does not intersect with the fiber.

The image results of both events indicate that newly created hydraulic fractures produce strong S-wave reflections. The imaged fractures are approximately perpendicular to the fiber for event A, which shows fractures that extend at least 50–125 m from the fiber. Fractures around the apex (the dashed red lines) are approximately 50 m away from the fiber, which can be further validated by integrating with LF-DAS data as discussed in the next section. The imaged fractures appear to fade out with distance from the fiber, sometimes abruptly, which highlights the observation limit of the monitoring fiber to fractures (i.e., the section of the fracture that is illuminated by the microseismic source). The actual fractures could be longer than the imaged results because at far distances the fractures possibly have insufficient width or impedance contrast to reflect waves emitted by nearby microseismic events. In addition to S-wave reflection/transmission coefficients that vary with the angle of incidence, the observed amplitude of reflections includes path effects caused by geometric spreading and DAS angular response. Correction for such effects would benefit from advanced waveform-based imaging theory, which could provide the basis for the quantitative interpretation of hydraulic fractures in future implementations of this approach.

In Figure 6b, most of the fractures appear to be oblique to the fiber. A possible reason for this is that event B may be shallower (by approximately 300 m) than the treatment depth and therefore affected by velocity heterogeneity that is not accounted for by our model. Diffractions from the upper and lower tips of the fracture also should be considered in complete fracture imaging. Event B includes a high-amplitude imaged event that appears to be curved. This event is an artifact from the S-P converted phase because we used the S wave to locate fractures. Furthermore, wavefield separation of the S-P converted wave and velocity calibration are recommended for future advanced imaging that can handle converted phases.

To confirm that the imaged reflectors are newly created hydraulic fractures, rather than preexisting fractures, we integrated the fracture images with LF-DAS data. The LF-DAS data are sampled at 0.1 Hz and are recorded by the same fiber in well B as the DAS microseismic data. To make a detailed comparison, Figure 7 shows an enlargement of the imaging results of event A in the left column, which is the upper half-section in Figure 6a (left). LF-DAS data from multistages at the corresponding distance range from all three treatment wells are shown in the right column. The horizontal axis of the LF-DAS waterfall plot follows the treatment time stage-by-stage. The dashed red line and star mark the origin time of the microseismic event (well T3 stage 20). The left imaging result using one single microseismic event is a snapshot of the fracture or structure between the event hypocenter and the monitoring fiber at the moment when this event happened.

LF-DAS data show the evolution of measured strain from the initial treatment stage. Because three wells were simultaneously treated, we further imaged microseismic events before and after each stage to validate the correspondence between the imaged fractures and stages. Final interpretation results are shown by the solid black lines, which indicate that the imaged fractures using reflections are highly consistent with the features of frac hits or fracture opening observed by the right LF-DAS data. Figure 7b shows the imaging results of an earlier event before well T3 stage 12 (the dashed line shows the origin time) and the same LF-DAS waterfall plot. From the imaged fractures in the left column, no visible reflectors are observed between the distance from 3.6 to 4.0 km along the fiber, where all three wells were not stimulated at the moment when the event occurred, i.e., no hydraulic fracture generation in the corresponding area. The timing relationships expressed by the interpretation in Figure 7a and 7b provide compelling evidence that the observed reflected S waves between the distance from 3.6 to 4.0 km along the fiber in event A imaging results (Figure 7a) are indeed generated by the newly created hydraulic fractures from treatment stages between two events rather than the preexisting fractures. The reduced amplitude of the fracture response at 4.4 km in event A (Figure 7a) compared with the same reflector in event C (Figure 7b) suggests that the fracture may have closed eight days after the corresponding stimulation stage.

In addition to the agreement between long fractures that intersect the fiber and FDIs observed using LF-DAS, the imaging results can be used to determine the extent of fractures away from the monitoring fiber as well. In Figure 7a, one stimulation stage between days 12 and 13 and three stages between days 14 and 16 (the red double arrows) did not produce any visible frac hits; thus, FDIs observed using LF-DAS are not detected. However, two clusters of away fractures (marked with rectangles) can be imaged using reflections as shown in the left column.

The imaging results characterize fracture geometry between microseismic sources and the fiber at the time of microseismic events, even for cases in which the fractures do not intersect with the fiber. LF-DAS also can provide constraints on locations of hydraulic fractures after fractures generate strain changes near the fiber. This suggests that integrated interpretation could be used effectively to monitor dynamic fracture propagation during and after well completion. We observed 29 events with fracture-induced reflected S waves (the dots in Figure 8) out of 110 events in well T3 stage 13, most of which were detected after frac hits recorded in LF-DAS. The red stars represent the two visible frac hits in LF-DAS data. In addition to the microseismic events near the treatment well T3, the map view (Figure 8) of event hypocenters indicates a distant cloud in the northwest, which is associated with the stimulation of wells T1 and T2. Three representative events marked with the cyan dots near well T3 are chosen to characterize fracture growth from well T3 to fiber B, which is shown in Figure 9. For event 1 and event 2, only toe-ward wavefields were used for imaging due to the lower quality of heel-ward reflections (Figure 9a and 9b). By comparing the wavefields of three events, we observe an increase in the amplitude and visibility of reflected waves at the apex, which are generated by fractures in the current stage as the injection progresses. Fracture channel waves are also observed in the toe-ward wavefields of event 3 (Figure 9c) revealing that reflections may be attenuated by a newly opened fracture around the apex. The four columns in Figure 10 represent DAS-recorded microseismic wavefields, imaged fractures using the proposed workflow, sketches of fractures growth, and LF-DAS waterfall plots, respectively. Each fracture image can be considered as a snapshot of the near fractures at the origin time of the corresponding microseismic event. Two frac hits are interpreted from LF-DAS data and marked with solid black dots. Each row in Figure 10 shows the processing and interpretation of a selected event before or after the first frac hit. We applied amplitude processing to the imaging results of the first two events to exclude the influence of microseismic magnitude on the reflections and for display purposes as well. The solid red lines in waterfall plots represent the recorded time of microseismic events.

Fracture images of all three events in the second column show clear imaged reflections from two fractures at a measured depth of 3.86 and 3.90 km (the dashed red lines) along the fiber, with an especially evident evolution of a fracture at 3.90 km. The first event occurred 36 min before the first FDI, which is the first recorded event showing reflections associated with fractures from well T3 in this stage, and the corresponding imaged fracture at 3.9 km is approximately 80 m away from the fiber. The second and third events were recorded 4 and 36 min after the first frac hit, respectively, where the imaged reflectors at 3.90 km both intersect with the fiber. The imaged results using the third event reveal a more continuous fracture at 3.90 km with stronger reflected energy than the image of the second event, which implies that the visibility of fracture-induced reflections in DAS data increases with growing fracture aperture and impedance contrast as the treatment continues. The third event wavefield (Figure 9c) contains signals reflected by multiple fractures from various stages, which is similar to the examples shown in Figure 3a. Because fractures at distances further than 3.90 km are in close proximity to the apex, most of the reflected energy was removed during preprocessing by muting the direct waves, which caused the large blank area with no data coverage around the apex (marked with the gray lines). The dashed green lines in the first two events show a fracture at 3.97 km that is related to a previous treatment stage. By comparing fracture images captured by three distinct events in the same treatment stage, it appears that DAS microseismic reflection imaging has the potential to serve as a reliable monitoring tool for tracking the growth of dynamic fractures.

The focus of this work is to consider microseismic events as seismic sources and apply an imaging method on DAS-recorded microseismic data to monitor fracture growth during hydraulic fracturing. The results provide a high-resolution image of the fracture zone, including fractures that do not extend entirely to the fiber. Figures 7 and 10 show evidence for the effectiveness of reflection imaging, by showing fracture opening, growth, and closure consistent with strain evolution inferred from LF-DAS.

The current imaging method is straightforward and fast but based on several simplifying assumptions. As previously noted by Staněk et al. (2022), the workflow treats the fracture as a planar interface between events and the monitor fiber and assumes fractures are perpendicular to the fiber, which is similar in principle to a zero-aperture Kirchhoff migration. These assumptions should be satisfied if the horizontal section of the treatment well and monitoring well is nearly perpendicular to the maximum horizontal stress direction, which is often the case. In addition to reflections, diffractions from the upper or lower tips of a vertical fracture (Appendix  A) would be expected in the case of microseismic events above or below the vertical extent of the fracture, but these events are not considered here. Thus, in-zone microseismic events are most likely to produce reflections, resulting in fracture imaging that is primarily sensitive to the horizontal extent of the fractures. Event B (Figure 3b) may contain some edge diffractions because it is on the upper edge of the fracking zone. We use a homogeneous velocity model for fracture imaging because the velocity curves from well-log data in the field do not show abrupt changes within the Montney Formation. However, when the microseismic events are far away from the fiber (>500 m), this simplifying assumption may be not valid due to the effects of velocity anisotropy and heterogeneity. In cases where these assumptions are not satisfied, the fracture may be incorrectly located, or it may appear to be curved rather than planar (Figure 6b), which can be easily discriminated. To avoid such errors, it is important to carefully evaluate the applicability of all assumptions to the specific imaging context.

For further development of microseismic reflection imaging, the difference between microseismic and active sources and the unique features of DAS microseismic data should be considered. Microseismic reflection imaging offers the benefit of having sources located within the stimulated volume and in close proximity to sensors, resulting in higher frequency and resolution compared to the surface array. DAS microseismic data also benefit from dense spatial resolution, compared with traditional downhole microseismic array data. However, individual single microseismic sources only provide a limited aperture compared with wide-angle active seismic data, which may introduce interpretive ambiguities.

The illuminated area of each microseismic event is controlled by the geometry of microseismic sources and monitoring fiber (sensor array) and the subsurface medium properties. Stacking multiple microseismic sources may reduce such artifacts but various magnitude and focal mechanisms of different microseismic sources should be taken into account during the stacking, in addition to the use of a well-calibrated velocity model to reduce hypocentral location uncertainty. To image arbitrary fracture geometry in three dimensions, advanced imaging tools such as the Kirchhoff migration (Grechka et al., 2017) or the crosswell reverse time migration (Lin and Zhang, 2016) are being tested, with a more complex velocity model.

With high-quality fracture images, there is a potential that the amplitude of DAS microseismic reflections could be used to measure changes in fracture properties, fluid volumes, and proppant geometry. The visibility of fracture-reflected waves requires strong microseismic sources, proper geometry of the source hypocenter and monitoring fiber, and a sufficient impedance contrast. In our data set, reflected waves associated with the current stage usually become visible within a few hours after the start of the stage, when the fracture is sufficient to reflect microseismic energy. For this reason, the actual fracture may possibly be longer than the imaged reflectors.

Events with fracture-induced reflections are observed from most stages in this project, although the analysis presented here focuses on three representative stages. The duration of the reflections could last more than 10 days (Figure 7) after the corresponding treatment stage and, therefore, could be combined with production data to evaluate proppant performance and capture a propped fracture geometry (Reshetnikov et al., 2023). To analyze the complex relationship between the amplitude of reflections and the mentioned parameters, advanced wavefield separation methods, especially near-apex wavefield separation for DAS data, are required to obtain reflected waves with true amplitude for imaging. Automated detection of reflections and muting of direct waves also would accelerate the imaging process because microseismic events have different traveltime moveouts. Based on a robust machine-learning workflow for accurate location and wavefield separation, the fracture imaging workflow could be automated to be applied for the purpose of real-time hydraulic fracture mapping.

DAS microseismic data can capture rich microseismic wavefield that, in addition to direct P- and S-wave arrivals, includes reflections that are generated by nearby hydraulic fractures. The reflections can be used to image fractures using a straightforward mapping method based on ray tracing. In this case study, an induced fracture network is located between microseismic source events and the monitoring fiber, including fractures that do not reach the fiber. The resulting images provide insight into the evolution of hydraulic fracture, including its initiation, propagation, and eventual closure. By analyzing images from multiple microseismic events, our results show that time-lapse imaging is possible, which can be integrated with LF-DAS and microseismic clouds to monitor dynamic fracture propagation. In cases where there is an abundance of microseismic events that provide a distribution of sources to illuminate fractures, this method has good potential to provide a more complete understanding of the hydraulic fracturing process.

This research was funded through the Canada First Research Excellence Fund (CFREF) by the Global Research Initiative in Sustainable Low Carbon Unconventional Resources (GRI) at the University of Calgary. The authors are grateful to ConocoPhillips Canada for sharing the DAS data and for the permission to publish this work. We thank the sponsors of the Microseismic Industry Consortium for their ongoing support.

Data associated with this research are confidential and cannot be released.

This appendix contains two schematic diagrams that illustrate scenarios of (1) reflected waves from horizontal interfaces (Figure A-1) and (2) fracture-induced diffractions and reflections (Figure A-2). These signals were observed in the microseismic events discussed in this study and can provide additional insights into the DAS microseismic wavefields, with the potential to enhance our understanding of subsurface properties.

Biographies and photographs of the authors are not available.