A Comprehensive Geophysical Analysis to Determine Induced Fracture Distribution from a Hydraulic Fracturing Operation in the Marcellus Shale Formation

 Reservoir monitoring of hydraulic fracturing plays an important role in both assessing risks and enhancing production from tight hydrocarbon reservoirs that otherwise would not produce economically. However, propagation and distribution of induced fractures in the subsurface can only be indirectly inferred since the effects of pumping large volumes of fluid at significant depths cannot be observed directly. Since hydraulic fracturing is a destructive industrial process that often occurs in close proximity to communities, understanding whether the created fractures are contained is important for safety, efficiency, and long-term viability. This research advances the understanding of induced fracture networks and subsurface changes resulting from hydraulic fracturing by analyzing a unique dataset that includes timelapse crosswell seismic and passive microseismic data. Each of these technologies are monitored at depth as close as possible to the features to be analyzed, and the repeat crosswell survey was recorded immediately after hydraulic fracturing to capture subsurface measurements before flowback and fluid diffusion reduced transient pressure effects. Passive microseismic data was recorded during hydraulic fracturing to relate the progression of subsurface changes to hydraulic pumping parameters in real time. 

The dataset offered unique processing possibilities because of the source used and the intense effort applied in acquisition. Crosswell acquisition applied shots per level with two perpendicular source settings that produced superior signal-to-noise and enabled application of a four-component vector rotation to isolate fast and slow shear modes as a measure of fracture intensity and azimuth. The seismic source generated directed compressional and shear wave energy, and recording with multicomponent geophones allowed wave mode separation and clear first arrivals for the travel-time inversion. Crosswell processing included vertically transverse isotropy in raytracing and an IRLS regularization algorithm with a compaction constraint developed by the Lawrence Berkeley National Laboratory that quickly converged to a solution, and resulted in improved imaging results compared with the contractor results. The timelapse results consistently showed that extreme shallow fracture creation occurred with breaching of the bounding formation. Differences between the compressional and shear response suggested that trapped residual fluid at high pressure remained in the fractures at the time of the repeat crosswell acquisition.  

 The analysis of the shear-wave data was extended by applying the Alford rotation and linear transform technique using the perpendicular source orientations at each shot level to isolate fast and slow shear modes. Estimates of fast shear azimuth and slow shear time delay before and after hydraulic fracturing indicated fracture creation throughout the overburden above the Marcellus shale and a vertical breach of the bounding Tully limestone formation. The magnitude of the timelapse change is consistent between travel-time and slow shear time lag inversion, but since the absolute change is small, it is likely that the increase in fracture intensity was not significant. Fast shear azimuth changes indicate a prevalence of natural fracture activation

.Analysis of the passive microseismic data focused on the single parameter of determining height of the microseismic event as a proxy for hydraulic fracture height. As this study used vertical seismic arrays where the vertical velocity profile is symmetric relative to the receiver array, the receiver depth of the earliest event arrival time is equivalent to the fracture height. This estimate can bemade without the need for precise arrival picking, polarizations, rotations or other processing aspects that introduce error. This approach allowed the use of the entire dataset, more than four times the number of events from standard processing,for comparison with hydraulic fracture pumping parameters. Simplifying the analysis runs counter to trends in microseismic processing that involve increased automation, and advanced analysis that can only be applied toa subset of suitable data. These approaches may miss many small events and suffer as a result of noisy data. Relating progression of hydraulic fracture depth in time with hydraulic fluid pumping parameters demonstrated once again the dominance of shallow fracture creation. Importantly, this study was able to relate the increase in shallow fracture activity to changes in pumping pressure and proppet concentration and concludes that the hydraulic fracturing operation likely used excessive pressureand was pumped too hard and too long.

 These results point to operational inefficiencies during hydraulic fracturing that could negatively impact production and pose risks to future development. An example is given for this dataset where a possible fracture encountered the monitor well at a shallow depth above top cement, a potential vertical pathway to the shallow geology. The study demonstrates the value of crosswell seismic surveys for reservoir monitoring and highlights observations that might be missed by a monitoring program that does not include timelapse seismic data and analysis of shear wave changes due to hydraulic fracturing