forecasting reservoir performance by mapping seismic …...forecasting reservoir performance by...

9
Forecasting reservoir performance by mapping seismic emissions Charles Sicking 1 , Jan Vermilye 1 , and Ashley Yaner 1 Abstract Streaming depth imaging (SDI) is a modified version of Kirchhoff migration that images the intensity and distribution of weak seismic waves emitted from rocks at depth. These images reveal the locations of the frac- tures and fracture networks in the reservoir. SDI allows for more informed forecasts for drilling, hydraulic fracturing, and reservoir management than is provided by traditional microearthquake mapping methods. Using passive data from surface and near-surface geophone grids, SDI integrates the seismic emissions over time to form the fracture activity volume. The fracture systems and the active production volume (APV) of the reservoir are calculated from this activity volume. In situ wellbore measurements indicate that the preexisting fracture systems in the reservoir rocks have substantial impact on the placement of the fluids during the hydraulic frac- ture treatment. They also strongly influence the locations of maximum oil and gas production and the decline rates of resource production. Mapping the fracture systems in the reservoir before drilling provides a strong forecasting value for optimal production sites for well placement. SDI can forecast hydraulic fracturing per- formance and improve the estimates of resource production volumes. Mapping the activity volumes during hydraulic fracturing shows the placement of the fluids during the treatment. SDI helps forecast the locations along the well that will have the best production. Time lapse mapping of the APV periodically during production shows the zones that are producing fluids and how they change over time. Our case histories indicate that this new seismic method has great promise for improved management of unconventional resources. Introduction Traditional microseismic methods record passive emissions only during hydraulic fracturing and do not provide for long-term reservoir management. Using streaming depth imaging (SDI), fracture systems and their changes in the reservoir can be mapped over the lifetime of their development and production. Peri- odic SDI monitoring of a reservoir provides information on the interaction between adjacent wells over time. SDI volumes show the actively producing volumes of rock and provide the information required for location of new wells. Today, the most widely used method for passive monitoring of hydraulic fracturing uses geophones at reservoir depth in vertical wells that are located near the hydraulic fracturing. Maxwell and Urbancic (2003) describe this downhole method for detecting micro- earthquakes (MEQ) generated during stimulation oper- ations and for imaging deformation associated with the injections. Another method for mapping MEQ during the hy- draulic fracturing uses surface or buried grid recordings. The basis of this method is Kirchhoff migration, and it is normally referred to as seismic emission tomography. Duncan and Eisner (2010) describe the surface geophone method for detecting and mapping MEQ. The focus of these hydraulic fracture monitoring methods is to use the MEQs to infer the creation of fracture permeability. The SDI alternative to MEQ mapping is the use of am- bient seismic emissions background emissions with- out controlled sources to directly map the fracture permeability in the rocks. Kochnev et al. (2007) describe a non-MEQ ambient imaging method for mapping the pro- gression of hydraulic fracturing. Their published paper is very brief, but our discussions with the authors have pro- vided insights into their methods and processing. Using data recorded on a surface array, they begin by identify- ing time windows with low-energy seismic waves that originate in a target zone. They focus the data in the se- lected time windows using traveltimes computed with an accurate velocity model, and then they compute coher- ency for each of the selected time windows. The coher- ency of each of the time windows is stacked to obtain a sequence of images that show the developing fractures during the pumping time. In contrast, the SDI method uses all of the recorded data to focus any waveforms that originate in the reservoir or near the target. The time win- dows with low-energy seismic waves are not identified prior to streaming into SDI, but rather, the presence of the waves are revealed by the SDI process. 1 Global Geophysical Services Inc., Plano, Texas, USA. E-mail: [email protected]; [email protected]; [email protected]. Manuscript received by the Editor 1 December 2015; revised manuscript received 19 April 2017; published ahead of production 22 June 2017; published online 24 July 2017. This paper appears in Interpretation, Vol. 5, No. 4 (November 2017); p. T437T445, 8 FIGS. http://dx.doi.org/10.1190/INT-2015-0198.1. © 2017 Society of Exploration Geophysicists and American Association of Petroleum Geologists. All rights reserved. t Technical papers Interpretation / November 2017 T437 Interpretation / November 2017 T437 Downloaded 07/26/17 to 47.186.10.220. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Upload: others

Post on 21-Sep-2020

12 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Forecasting reservoir performance by mapping seismic …...Forecasting reservoir performance by mapping seismic emissions Charles Sicking 1, Jan Vermilye , and Ashley Yaner Abstract

Forecasting reservoir performance by mapping seismic emissionsCharles Sicking1, Jan Vermilye1, and Ashley Yaner1

Abstract

Streaming depth imaging (SDI) is a modified version of Kirchhoff migration that images the intensity anddistribution of weak seismic waves emitted from rocks at depth. These images reveal the locations of the frac-tures and fracture networks in the reservoir. SDI allows for more informed forecasts for drilling, hydraulicfracturing, and reservoir management than is provided by traditional microearthquake mapping methods. Usingpassive data from surface and near-surface geophone grids, SDI integrates the seismic emissions over time toform the fracture activity volume. The fracture systems and the active production volume (APV) of the reservoirare calculated from this activity volume. In situ wellbore measurements indicate that the preexisting fracturesystems in the reservoir rocks have substantial impact on the placement of the fluids during the hydraulic frac-ture treatment. They also strongly influence the locations of maximum oil and gas production and the declinerates of resource production. Mapping the fracture systems in the reservoir before drilling provides a strongforecasting value for optimal production sites for well placement. SDI can forecast hydraulic fracturing per-formance and improve the estimates of resource production volumes. Mapping the activity volumes duringhydraulic fracturing shows the placement of the fluids during the treatment. SDI helps forecast the locationsalong the well that will have the best production. Time lapse mapping of the APV periodically during productionshows the zones that are producing fluids and how they change over time. Our case histories indicate that thisnew seismic method has great promise for improved management of unconventional resources.

IntroductionTraditional microseismic methods record passive

emissions only during hydraulic fracturing and do notprovide for long-term reservoir management. Usingstreaming depth imaging (SDI), fracture systems andtheir changes in the reservoir can be mapped overthe lifetime of their development and production. Peri-odic SDI monitoring of a reservoir provides informationon the interaction between adjacent wells over time.SDI volumes show the actively producing volumes ofrock and provide the information required for locationof new wells.

Today, the most widely used method for passivemonitoring of hydraulic fracturing uses geophones atreservoir depth in vertical wells that are located nearthe hydraulic fracturing. Maxwell and Urbancic (2003)describe this downhole method for detecting micro-earthquakes (MEQ) generated during stimulation oper-ations and for imaging deformation associated with theinjections.

Another method for mapping MEQ during the hy-draulic fracturing uses surface or buried grid recordings.The basis of this method is Kirchhoff migration, and it isnormally referred to as seismic emission tomography.Duncan and Eisner (2010) describe the surface geophone

method for detecting and mapping MEQ. The focus ofthese hydraulic fracture monitoring methods is to usethe MEQs to infer the creation of fracture permeability.

The SDI alternative to MEQ mapping is the use of am-bient seismic emissions — background emissions with-out controlled sources — to directly map the fracturepermeability in the rocks. Kochnev et al. (2007) describea non-MEQ ambient imaging method for mapping the pro-gression of hydraulic fracturing. Their published paper isvery brief, but our discussions with the authors have pro-vided insights into their methods and processing. Usingdata recorded on a surface array, they begin by identify-ing time windows with low-energy seismic waves thatoriginate in a target zone. They focus the data in the se-lected time windows using traveltimes computed with anaccurate velocity model, and then they compute coher-ency for each of the selected time windows. The coher-ency of each of the time windows is stacked to obtain asequence of images that show the developing fracturesduring the pumping time. In contrast, the SDI methoduses all of the recorded data to focus any waveforms thatoriginate in the reservoir or near the target. The timewin-dows with low-energy seismic waves are not identifiedprior to streaming into SDI, but rather, the presence ofthe waves are revealed by the SDI process.

1Global Geophysical Services Inc., Plano, Texas, USA. E-mail: [email protected]; [email protected]; [email protected] received by the Editor 1 December 2015; revised manuscript received 19 April 2017; published ahead of production 22 June 2017;

published online 24 July 2017. This paper appears in Interpretation, Vol. 5, No. 4 (November 2017); p. T437–T445, 8 FIGS.http://dx.doi.org/10.1190/INT-2015-0198.1. © 2017 Society of Exploration Geophysicists and American Association of Petroleum Geologists. All rights reserved.

t

Technical papers

Interpretation / November 2017 T437Interpretation / November 2017 T437

Dow

nloa

ded

07/2

6/17

to 4

7.18

6.10

.220

. Red

istri

butio

n su

bjec

t to

SEG

lice

nse

or c

opyr

ight

; see

Ter

ms o

f Use

at h

ttp://

libra

ry.se

g.or

g/

Page 2: Forecasting reservoir performance by mapping seismic …...Forecasting reservoir performance by mapping seismic emissions Charles Sicking 1, Jan Vermilye , and Ashley Yaner Abstract

In related work, Tary and Van der Baan (2012) com-pute continuous time-frequency transforms that highlightsignals that have time-varying resonance frequencies.They conclude that these signals are the result of reso-nance in fluid-filled fractures or, alternatively, succes-sions of very small repetitive seismic events along thefractures. Tary and Van der Baan (2012) also observe cor-relations between the variations in the frequency contentof their recordings, the hydraulic fracturing conditions,and the occurrence of microseismic events. They notethat there is a direct correspondence between variationsin the slurry injection rate and the combined energy emit-ted by concurring events.

Seeking to better identify these ambient emissions asopposed to MEQ events, Chorney et al. (2012) presentresults on seismic energy sources that are associatedwith deformations such as tensile fracturing or slowslips. Furthermore, Bame and Fehler (1986) note thatthe ambient signals they observe are unlikely to be de-tected by searching with seismic event triggering meth-ods because these require sharp signal onsets.

Additional support for the origins of these episodicsignals that occur over long time intervals can be foundin the fracture mechanics literature. Vermilye and Scholz(1998) and Shipton and Cowie (2001) investigate the vari-ous release mechanics of stored elastic strain energyfrom rocks through field studies of fractures. This storedstrain energy is not evenly distributed in the earth’s crust,but it is preferentially released on fracture/fault surfacesand in the damage zones surrounding these fractures.

Fracture mechanics theory predicts that stress con-centrations are associated with fractures. Accordingly,Vermilye and Scholz (1998) and Moore and Lockner(1995) report field and laboratory studies with clear evi-dence that these stress concentrations are recorded inthe fracture damage zones. Vermilye and Scholz (1998)show that damage zones consist of rock volumes with ahigh density of small fractures and that the density offractures increases exponentially with their proximityto the main fracture surface.

Ziv and Rubin (2000) show that the brittle crust is in astate of unstable frictional equilibrium. Therefore, verysmall changes in stress (approximately 0.01 bar or ap-proximately 1 kPa) can cause slippage onweak fractures.Lawn and Wilshaw (1975) show that failure occurs pref-erentially on small, optimally oriented fractures and inthe zones surrounding the fractures in which crack-tipstress concentrations amplify the stress magnitudes.Hubbert and Rubey (1959) show that this unstable equi-librium is further disturbed as additional fluid pressuresreduce the normal stress on preexisting fractures. Theyalso show that, during production, subtle movement offluid produces similar effects.

Imaging seismic activity recorded in passive data canreveal the fracture density and stress conditions in areservoir before drilling. Such data can reveal the loca-tions of the highest density of natural fractures with ac-tive fluid flow. When mapped before drilling, they showthe locations where hydraulic fracture treatments can

contact the largest number of natural fractures, therebyforecasting the most productive rock volumes. Passiverecordings taken a few times per year over severalyears provide important reservoir management data.These data begin with mapping the rock volume acti-vated by the hydraulic fracturing and the rock volumearound the well that is active during flow back. Sub-sequent recordings provide the data for tracking therock volume that is producing the most resource.

In this paper, we first review the SDI processingworkflow beginning with the filters used for preparingthe field data. This data preparation increases the sig-nal-to-noise ratio (S/N) in the field data and allows theweak seismic waveforms from a reservoir to be imaged.We next describe the SDI method itself. Finally, weshow the results of the filtering and SDI processing asapplied to several examples that show the value of us-ing SDI over the life of a reservoir.

Data filtering and SDI methodsThe amplitude of the sought-after signal in ambient

field data is typically at or below the amplitude of thecultural and weather noise in the recorded data.Retrieval of these signals is achieved through the useof well-established, but now extended, methods:(1) high density, multistation 3D geophone spreads,either on the surface or buried below the surface,(2) continuous passive recording for many hours, and(3) reflection-seismic signal processing algorithms fornoise suppression and coherent signal stacking. Moredetailed descriptions of these steps than what is givenhere can be found in Sicking et al. (2016), Hall (2006),and Geiser et al. (2012). The focus in this study is on theinterpretation of the final processed SDI volumes andthe resulting mapping of fracture systems for severalcase histories.

The requirement for a 3D station geophone spread,laid out without oriented geophone arrays, is to capturethe full and unbiased cultural noise, ambient signal, andhydraulic fracturing generated seismic wavefields. Therequirement for long continuous ground motion record-ers is to allow for separation of the long duration, butepisodic, ambient, and hydraulic fracturing signals fromthe various types of cultural and weather noises. Sta-tionary noise is one type of cultural noise that is con-tinuous for very long time intervals. The integrationmethod of SDI can accumulate this noise in the finalimage volume and it competes in amplitude with thesought-after signal. This noise derives from seismicwaves that are generated at single locations as thoseseen in Figure 1a showing a plan view with three differ-ent sources of stationary noise, including a well head,compressors, and hydraulic fracturing pumps. Thesesources have fixed locations during the ambient record-ing. Figure 1b shows a vertical slice view that illustratesthe raypaths of pressure waves, surface waves, andmultiples that propagate from the well head to a singlereceiver. Stationary noise waveforms follow the same

T438 Interpretation / November 2017

Dow

nloa

ded

07/2

6/17

to 4

7.18

6.10

.220

. Red

istri

butio

n su

bjec

t to

SEG

lice

nse

or c

opyr

ight

; see

Ter

ms o

f Use

at h

ttp://

libra

ry.se

g.or

g/

Page 3: Forecasting reservoir performance by mapping seismic …...Forecasting reservoir performance by mapping seismic emissions Charles Sicking 1, Jan Vermilye , and Ashley Yaner Abstract

raypaths from the source to the receiver for the entiretime of the recording.

After the ambient data for SDI are properly recorded,a variety of tools used in reflection seismic signalprocessing can be applied to suppress noise. One filternot generally used for reflection seismic data is thecepstral filter. This filter suppresses the stationary noiseand is a very important algorithm for processing ambi-ent data. Although most other noise and the episodicsignals recorded in passive data are observed to occurin time bursts of less than 1 s, stationary noise wave-forms are observed to be continuous for time intervalsof minutes. To be effective, the time window used incepstral filtering must be much longer than the tran-sient signals and other noise packets recorded in thedata. Suppressing the very long, continuous waveformsenhances the shorter, episodic signal waveforms arriv-ing from the fractures at depth. For this study, the timewindow used for cepstral filtering was 61 s. The station-ary noise that continues for the selected time windowwill be suppressed without attenuating the signals thathave shorter time lengths. For a detailed description ofthe cepstral filter and examples of its applications, seeSicking et al. (2016) and Hall (2006). Using a combina-tion of field data and synthetics, Sicking et al. (2016)demonstrate an S/N gain of 8–12 decibels for the ceps-tral filter alone.

Each trace of the passive data is a 1D time sequence,and it is filtered independently from all other traces.Stationary noise has components of direct and scat-tered surface body waves, direct and scattered surfaceRayleigh waves, subsurface reflected P- and S-waves,and subsurface multiples of P- and S-waves. Becausethe waveforms in the stationary noise are a mix ofall types of waves arriving from many directions, thereis very little coherency in the noise across recordingchannels and thus cannot be removed using multichan-nel filtering methods. However, the stationary noisewaveforms at a single location are continuous and veryrepeatable for long time intervals making the cepstralfilter ideal for suppressing stationary noise.

Figure 2 shows the field data for a time window thatcontains the waveforms for a large-ampli-tude MEQ. Figure 2a shows the unproc-essed field trace data focused at thedepth of theMEQ. TheMEQ is not readilyevident in these data. The trace data forthe same data after the application of thecepstral filter are shown in Figure 2b. Theuplift from using the cepstral filter is evi-dent in the more coherent quality of thetrace data, and the S/N for the MEQ hasbeen substantially improved withoutchanging the phase of the waveforms.

After the trace processing for signalenhancement and noise suppression, thepassive data are processed with SDImethods to find the data volumes withfracture activity. The kernel of SDI is a

modified version of Kirchhoff migration (Yilmaz,2000). The 3D depth volume under inspection is first bro-ken into voxels, and the migration for each voxel is car-ried out in two steps. First, the traveltimes from the voxelto each receiver location is computed using a previouslyconstructed velocity model. Using these traveltimes,time shifts are applied to each data trace to focus thetraces to the voxel location. This focusing step alignsthe signals from that voxel such that they arrive on allreceivers at the same recording time. Second, the coher-ence is computed across all the traces and for the over-lapping time length of the shifted traces. The coherenceis a relative measure of how much signal is emitted fromthe voxel. It is shown by Sicking et al. (2016) that largeincreases in the length of the time window used to com-pute the coherency substantially improves the resolutionand location accuracy for the signal in the imagedvolume.

Figure 1. Stationary noise sources and raypaths to a receiver.

Figure 2. Field data before and after the application of a cepstral filter. Thestationary noise is suppressed and the MEQ buried in the noise in the upper plotis now clearly seen in the lower image (block arrows).

Interpretation / November 2017 T439

Dow

nloa

ded

07/2

6/17

to 4

7.18

6.10

.220

. Red

istri

butio

n su

bjec

t to

SEG

lice

nse

or c

opyr

ight

; see

Ter

ms o

f Use

at h

ttp://

libra

ry.se

g.or

g/

Page 4: Forecasting reservoir performance by mapping seismic …...Forecasting reservoir performance by mapping seismic emissions Charles Sicking 1, Jan Vermilye , and Ashley Yaner Abstract

In practice, the coherency for each voxel is computedfor each 1-min time interval. Each interval that containslarge amplitude MEQs or coherent noise bursts are iden-tified and deleted. The coherence for windows that passthis inspection is summed to reveal a 3D map of the rel-ative strengths of the emitted ambient and hydraulic-fracturing-generated seismic signals.

As an alternative, Nakata and Beroza (2016) describea wavefield-based imaging method for processing ambi-ent data. They use reverse time migration to extrapolatethe individual wavefields at each receiver to a selecteddepth voxel and use the geometric mean as the imagingcondition. Nakata and Beroza (2016) note that the geo-metric mean computation provides much better resolu-tion than arithmetic stacking but requires 100 timesmore computation. Nakata et al. (2016) provide thecomputational details of their method, which may besuperior to SDI the imaging method, but it is computa-tionally more expensive.

Mapping seismic emissions from the reservoirs:Case histories

Passive seismic monitoring and imaging of episodicseismic emissions has been used for (1) establishing op-timal well locations, direction of wells, and selection ofhydraulic fracture locations along the wells, prior to drill-ing, (2) evaluating the effectiveness of hydraulic fracturemonitoring and estimating the stimulated rock volume(SRV), and (3) recording time-lapse production monitor-ing for reservoir management. In some cases, SDI hasbeen used in all three phases of a reservoir’s life cycle.Examples of each of these phases are given here.

Predrill fracture mappingEffective reservoir management begins before wells

are drilled. In this phase, SDI is used to map the zones of

natural fracturing that are seismically active. These areinterpreted as the pretreatment permeability pathways.They can be used to plan the best drilling locationsand directions, forecast the performance of individualhydraulic fracturing stages, forecast interactions be-tween wells, and reveal the zones of the most produc-tive rock.

Two examples of ambient emission mapping prior todrilling are presented here. They both demonstrate thatthe zones of high ambient activity locate the zones ofbest production in the reservoir. The first example isa reservoir located in a large oblique-faulted structure,and the second is in the Eagle Ford Shale. In the firstexample, the passive data were recorded as a stand-alone project, with the geophone spread being laidout for this specific purpose. No hydraulic fracturingwas used during the well completions, so all the produc-tion came from the SDI identified natural fractures. Inthe second example, the passive data were recordedduring the acquisition of a 3D reflection survey. Thismode of “piggyback” SDI data acquisition reduces thecost of ambient to a small increment in the cost ofthe reflection survey.

Example 1Figure 3 shows a predrill forecast of production from

a carbonate reservoir in which the production is fromnatural fractures. The structure is folded and cut byfaults with thrust and strike-slip displacements. Fig-ure 3a shows a cross section of the 3D velocity modelthat was computed using iterative prestack depth mi-gration for the 3D reflection data. It shows a strike-slipfault cutting an anticline formed by a preexisting thrustfault. Figure 3b is a depth varying horizon slice thatshows the intensity of seismic emissions along the res-ervoir horizon.

Ambient data recorded on a station-ary surface geophone grid were usedto compute the SDI emission intensityvolume for more than 20 hours of se-lected ambient data. The data wereprocessed using the noise suppressiondescribed in the previous section. Afterprocessing, much of the data had to beeliminated due to interferences by sur-face waves generated in a nearby zoneof earthquake activity. The final SDI in-tegration time for this image volume was15 h of recorded data.

The SDI map on Figure 3b shows thehighest emissions come from the hang-ing wall of the thrust fault. Because theproduction in this reservoir is known tocome from natural fracture porosity andpermeability, the SDI results were usedto target this location in the hangingwall. Subsequent drilling of severalwells confirmed this forecast (C. Sierra,personal communication, 2014).

Figure 3. Example from a complex structure with strike-slip and thrust fault.(a) A vertical cross section through the velocity model built using iterative pre-stack depth migration. (b) The passive seismic emissions for the reservoir layer.This example forecasts that the best production is in the hanging wall of the thrust.The red line in the velocity panel shows the path for the wells that are drilled downthe face of the thrust fault. The reservoir rock is dipping and lies in the hangingwall of the fault. The white arrows in panel (b) show the locations of the lateralwells cutting the zone of high activity.

T440 Interpretation / November 2017

Dow

nloa

ded

07/2

6/17

to 4

7.18

6.10

.220

. Red

istri

butio

n su

bjec

t to

SEG

lice

nse

or c

opyr

ight

; see

Ter

ms o

f Use

at h

ttp://

libra

ry.se

g.or

g/

Page 5: Forecasting reservoir performance by mapping seismic …...Forecasting reservoir performance by mapping seismic emissions Charles Sicking 1, Jan Vermilye , and Ashley Yaner Abstract

Example 2SDI provides predrill forecast of producing fracture

systems in the Eagle Ford Shale. The passive data forthis study were collected at night when the vibratorsof a 3D survey were shut down. Figure 4 shows the ac-tivity volume measured along this shale layer from theemission volume from approximately 9 hours of ambi-ent data. The faults shown as black lines were mappedusing the 3D reflection survey volume. High-emissionactivity was found in faulted sections of the shale, bothwhere the faulting either extended in to the overlyingAustin Chalk or was confined to the Eagle Ford. TheSDI results in Figure 4 show that areas where faultingdid not extend into the overlying Austin Chalk have thewells with highest production. The combined ambientand reflection data show that the east–west runningzone of high productivity and emissions does not followthe fault systems. The suggestion is that at these sitesthose faults that are confined to the shale have not al-lowed resource leakage into the Austin Chalk.

Hydraulic fracturing planning and SRV estimationUsing SDI for ambient recordings before a well

treatment can forecast the zones of optimal stimula-tion. However, it is the SDI volumes recorded duringhydraulic fracturing that reveal which zones of thewell are likely to have the most connected permeabil-ity, and hence production. The near-well featuresmapped during hydraulic fracturing are the basisfor computing the SRV. The SDI emission volumesare separate and independent of the hydraulic fractur-ing induced MEQ that are used for SRV by othermethods.

Example 3SDI provides ambient activity map-

ping for planning the hydraulic fractur-ing, mapping the activated emissionvolumes during hydraulic fracturing,and mapping permeability pathways.Figure 5a shows the ambient activityaround a well before hydraulic fractur-ing; Figure 5b shows the fracturesactivated during hydraulic fracturingoverlaid on the data in Figure 5a; and Fig-ure 5c shows a forecast of the permeabil-ity pathways from which the major partof the production will come. The latterplot suggests that the production willbe largest at the toe, small in the middle,and moderate at the heel of the well. Theimages in Figure 5 show that the rockvolume activated during hydraulicfracturing closely matches the naturalfracture systemsmapped before the treat-ment. The activity before hydraulic frac-turing (Figure 5a) was used for planningthe pumping pressures and flow times forthe various stages. The SRV is computed

using the rock volume of the activated fractures, thoseshown in Figure 5b. Imaging of the permeability path-ways, shown in Figure 5c, begins with computing sepa-rate SDI activity volumes for each stage of the treatment.The activity volume for every stage is computed for theentire volume surrounding the well. Each SDI volumethen contains data for only one stage. For each voxelin the volume, the number of times that voxel is activatedfor all stages is counted to compute the net pathway vol-ume. The final permeability pathway volume shows theconnectivity between the active voxels and the relativepermeability of the pathways.

Figure 4. Ambient seismic activity recorded before drilling inthe Eagle Ford. High-activity areas before drilling show thebest production areas. The green circles are sized accordingthe initial production from the well. This combined, inter-preted 3D seismic and highest SDI emissions map sectionshows that production is highest away from faults that pen-etrate up in to the Austin Chalk.

Figure 5. (a) The ambient seismic-activity volume and mapped fractures beforefracking. (b) The fractures activated during treatment overlaid on the ambientbackground. (c) A forecast of major permeability pathways from the combina-tion of pretreatment and treatment-time SDIs.

Interpretation / November 2017 T441

Dow

nloa

ded

07/2

6/17

to 4

7.18

6.10

.220

. Red

istri

butio

n su

bjec

t to

SEG

lice

nse

or c

opyr

ight

; see

Ter

ms o

f Use

at h

ttp://

libra

ry.se

g.or

g/

Page 6: Forecasting reservoir performance by mapping seismic …...Forecasting reservoir performance by mapping seismic emissions Charles Sicking 1, Jan Vermilye , and Ashley Yaner Abstract

Example 4SDI activity volumes around the well path can be

used in planning adjacent wells and their treatmentplans. These activity volumes help forecast hydraulicfracturing interference between adjacent wells andmaximize the utility of the treatments, including theelimination of production damaging stages.

Figure 6 shows an SDI map of the area surroundingtwo lateral wells. The fractures shown were mappedfrom the passive data recorded while well A was on pro-duction, after well B was drilled, but before the hydraulicfracturing of well B. This map indicates that several ac-tive features cross both wells: The sites of these are in-dicated by circles along well B. Well A was producingbefore, during, and after the treatment of well B. Thepressure at the head of well A was recorded continu-ously for well A during the treatment of well B. Signifi-cant changes in these pressures were observed duringthe treatment of the stages in well B shown by the circles(M. Caputi, personal communication, 2014). Evidently,these permeable pathways connect the two wells.

Time-lapse SDI: Tracking active producingvolumes

Using SDI to map the ambient emissions around awell while it is producing reveals the rock volumethrough which fluids are flowing into the well. Such ac-tive producing volumes (APV), and their changes overtime, can be used for reservoir management. This in-cludes planning retreatments and the drilling and stimu-lation of adjacent wells.

APV are best computed from ambient data recordedwith a permanent, shallow-buried grid of sensors thathas a density of between two and four stations persquare mile. The areal coverage of a target centered gridwith a radius of 1.5 times the depth for a reservoir depthd, is approximately 7 ! d2. Buried grids allow for re-peated observations with fixed receivers that are well in-sulated from surface noise sources. Ideally, the grid isinstalled before drilling and treatment and used to recordreservoir data two to three times per year thereafter. Foreach observation interval, a few hours of ambient pas-sive data are recorded and processed to compute a res-ervoir-wide activity volume.

The computation of the APV for a well starts by iden-tifying active voxels intersecting the well path. Thesevoxels comprise the amplitude limited activity volume.The voxels in the volume that are nonzero, and are con-nected to the well through other active voxels, are ex-tracted to form the APV. The APV volumes for any wellpath can be computed from the activity volume com-puted under the buried gird. This is true for testing wellpaths even before the well is drilled, allowing the APVto be forecast before the well is drilled.

Example 5APV changes due to well interference during hy-

draulic fracturing of an adjacent well are shown in Fig-ure 7. The SDI fracture maps in the figure suggest wellsA and B are connected by several permeable fracturesbefore the fracture treatment. This forecast that wassupported by the pressure change data recorded atthe well head in well A during treatment of well B. More-over, APV maps of well A before and after treatment ofwell B show that these pressure changes reduced theproducing rock volume of well A. The time-lapse APVsfor well A are shown in Figure 7. The APV in Figure 7awas taken before the treatment of well B, whereasFigure 7b was taken after the treatment of well B. Evi-dently, the treatment in well B reduced the APV inwell A.The percentage drop in production measured for well Ais on the same order as the observed volume reduction inthe APV.

Example 6APV changes as production declines over time. Time-

lapse ambient monitoring provides the data for comput-ing the APV over the years of production. The APV com-puted at two different times over three years for a wellin a typical nonconventional reservoir is shown in Fig-ure 8. The volume and locations of seismic emissions

Figure 6. Ambient seismic fracture systems before hydraulicfracturing in well B. The locations shown by the circles are thestages in well B that caused pressure changes in well A duringhydraulic fracturing in well B. These are the samewells shownin Figure 7.

T442 Interpretation / November 2017

Dow

nloa

ded

07/2

6/17

to 4

7.18

6.10

.220

. Red

istri

butio

n su

bjec

t to

SEG

lice

nse

or c

opyr

ight

; see

Ter

ms o

f Use

at h

ttp://

libra

ry.se

g.or

g/

Page 7: Forecasting reservoir performance by mapping seismic …...Forecasting reservoir performance by mapping seismic emissions Charles Sicking 1, Jan Vermilye , and Ashley Yaner Abstract

along the wellbore in this example wereobserved to significantly change overthis time interval. Figure 8 shows theSRV and two APVs. The SRV is com-puted for the data recorded during hy-draulic fracturing. The first APV wasmeasured after two years of productionand the second after three years of pro-duction. For this example, there was asubstantial decrease in the APV fromyear 2 to year 3.

DiscussionThe case histories described in the

previous section were recorded eitheras part of a 3D exploration-seismic sur-vey or as a standalone surface or buriednetwork. Subsequent drilling and hy-draulic fracturing in an established res-ervoir can be very effectively guided byperiodic SDI campaigns. The first ambi-ent data should be recorded beforedrilling to map the preexisting fracturesystems in the reservoir. The predrillSDI volumes are used to forecast the pro-duction volumes for proposedwell paths.They allow planning for maximization ofthe SRV during the well treatment and toavoid interference with existing produc-tion wells.

Ambient data should also be recordedduring hydraulic fracturing. The record-ing time should be started the day beforehydraulic fracturing and continued

through the flow-back time. The SRV found from thetreatment can be compared with the maps observed be-fore drilling to confirm the performance of the treatment.The ambient data recorded before and after treatmentprovide baselines for tracking the changes in the produc-ing rock volumes over time.

Experience shows that the fracture systems in therocks before drilling impact the fractures that areopened during the treatment and the volume of rockthat produces after the treatment. By following the evo-lution of the APV volumes with periodic SDI campaigns,the planning of secondary treatments and infill drillingcan be more effectively managed.

ConclusionSDI images the intensity and distribution of weak

seismic energy emitted from the rocks at depth. Theseseismic emissions show the locations of the fracturesystems in the reservoir, which form the permeabilitypathways mapped by SDI.

The preexisting fracture systems in the reservoirrocks have substantial impact on (1) the placement ofthe fluids injected during treatment, (2) the interferencesbetween wells, (3) the oil volumes extracted from eachwell, and (4) the decline rates for the wells. Mapping the

Figure 7. The producing rock volume around well A was computed (a) frompassive data recorded before the fracking of well B and (b) from the passivedata recorded after the fracking of well B. During the fracking of well B, wellA experienced pressure hits (see Figure 6). The pressure hits reduced the pro-duction in well A. The well head monitored production decline for well A is onthe same order as the measured reduction in APV volume.

Figure 8. Changes in APV over the producing life of a well.(a) The activated fractures (SRV) during hydraulic fracturing.(b) The producing (APV) volume after two years of produc-tion. (c) The APV after three years of production.

Interpretation / November 2017 T443

Dow

nloa

ded

07/2

6/17

to 4

7.18

6.10

.220

. Red

istri

butio

n su

bjec

t to

SEG

lice

nse

or c

opyr

ight

; see

Ter

ms o

f Use

at h

ttp://

libra

ry.se

g.or

g/

Page 8: Forecasting reservoir performance by mapping seismic …...Forecasting reservoir performance by mapping seismic emissions Charles Sicking 1, Jan Vermilye , and Ashley Yaner Abstract

fracture systems in the reservoir before drilling providesa strong forecasting value for reservoir management inthe placement of wells, hydraulic fracturing perfor-mance, and well production. Such predrill forecastingis not possible using traditional MEQ methods.

Mapping the activity volumes during hydraulic frac-turing shows the placement of the fluids during thetreatment. It allows for the computation of the SRV,and helps forecast the locations along the well for fu-ture production.

Time lapse mapping of the activity volumes periodi-cally during production shows the zones that are pro-ducing fluids and how they change over time. UsingSDI to map the seismic emissions over the life of thereservoir allows for efficient management of unconven-tional reservoirs by helping to plan repeat hydraulicfracturing and infill drilling.

The efficient use of SDI for reservoir management ismost efficient with the use of permanently installedburied geophone grids that have sufficient areal cover-age to allow imaging of the seismic emissions emittedfrom the reservoir.

AcknowledgmentsP. Malin has spent many hours editing this paper. His

experience and insights have substantially improvedthe quality. S. Simms and L. Bjerke provided the dataintegration in the field examples that helped demon-strate the impact of ambient imaging.

ReferencesBame, D., and M. Fehler, 1986, Observations of long period

earthquakes accompanying hydraulic fracturing: Geo-physical Research Letters, 13, 149–152, doi: 10.1029/GL013i002p00149.

Chorney, D., P. Jain, M. Grob, and M. van der Baan, 2012,Geomechanical modeling of rock fracturing and associ-ated microseismicity: The Leading Edge, 31, 1348–1354,doi: 10.1190/tle31111348.1.

Duncan, P. M., and L. Eisner, 2010, Reservoir characteriza-tion using surface microseismic monitoring: Geophys-ics, 75, no. 5, 75A139–75A146, doi: 10.1190/1.3467760.

Geiser, P., A. Lacazette, and J. Vermilye, 2012, Beyond ‘dotsin a box’: An empirical view of reservoir permeabilitywith tomographic fracture imaging: First Break, 30,63–69.

Hall, M., 2006, Predicting bed thickness with cepstral de-composition: The Leading Edge, 25, 199–204, doi: 10.1190/1.2172313.

Hubbert, M. K., and W. W. Rubey, 1959, Role of fluid pres-sure in mechanics of overthrusting: Bulletin of the Geo-logical Society of America, 70, 115–166, doi: 10.1130/0016-7606(1959)70[115:ROFPIM]2.0.CO;2.

Kochnev, V. A., I. V. Goz, V. S. Polyakov, I. S. Murtayev, V.G. Savin, B. K. Zommer, and I. V. Bryksin, 2007, Imaginghydraulic fracture zones from surface passive micro-seismic data: First Break, 25, 77–80.

Lawn, B. R., and T. R. Wilshaw, 1975, Fracture of brittlesolids: Cambridge University Press.

Maxwell, S. C., and T. I. Urbancic, 2001, The role of passivemicroseismic monitoring in the instrumented oil field:The Leading Edge, 20, 636–639, doi: 10.1190/1.1439012.

Maxwell, S. C., and T. I. Urbancic, 2003, Passive imaging ofseismic deformation associated with injection for en-hanced recovery: 73rd Annual International Meeting,SEG, Expanded Abstracts, 458–461.

Moore, D. E., and D. A. Lockner, 1995, The role of micro-fracturing in shear fracture propagation in granite: Jour-nal of Structural Geology, 17, 95–114, doi: 10.1016/0191-8141(94)E0018-T.

Nakata, N., and G. Beroza, 2016, Reverse time migrationfor microseismic sources using the geometric mean asan imaging condition: Geophysics, 81, no. 2, KS51–KS60, doi: 10.1190/geo2015-0278.1.

Nakata, N., G. Beroza, J. Sun, and S. Fomel, 2016, Migra-tion-based passive source imaging for continuous data:86th Annual International Meeting, SEG, ExpandedAbstracts, 2607–2611.

Shipton, Z. K., and P. A. Cowie, 2001, Damage zone andslip-surface evolution over μm to km scales in high-porosity Navajo sandstone, Utah: Journal of StructuralGeology, 23, 1825–1844, doi: 10.1016/S0191-8141(01)00035-9.

Sicking, C., J. Vermilye, and A. Yaner, 2016, Pre-drill res-ervoir evaluation using passive seismic imaging, URTeC2460524: Proceedings of the Fourth Unconventional Re-sources Technology Conference.

Tary, J., and M. van der Baan, 2012, Potential use ofresonance frequencies in microseismic interpretation:The Leading Edge, 31, 1338–1346, doi: 10.1190/tle31111338.1.

Vermilye, J. M., and C. H. Scholz, 1998, The process zone:A microstructural view of fault growth: Journal ofGeophysical Research, 103, 12223–12237, doi: 10.1029/98JB00957.

Yilmaz, Ö., 2000, Migration, in S. M. Doherty, ed., Seismicdata analysis: Processing, inversion, and interpretationof seismic data, 2nd ed.: SEG, 463–654.

Ziv, A., and A. M. Rubin, 2000, Static stress transfer andearthquake triggering: No lower threshold in sight?:Journal of Geophysical Research, 105, 13631–13642,doi: 10.1029/2000JB900081.

Charles Sicking received a bache-lor’s in physics and a doctorate in geo-physics from the University of Texasat Austin. He was the vice presidentfor R&D for Global Geophysical Ser-vices and led the development ofthe company’s microseismic process-ing system. Before joining Global Geo-physical, he was chief geophysicist at

Weinman GeoSciences. He also served as a research geo-

T444 Interpretation / November 2017

Dow

nloa

ded

07/2

6/17

to 4

7.18

6.10

.220

. Red

istri

butio

n su

bjec

t to

SEG

lice

nse

or c

opyr

ight

; see

Ter

ms o

f Use

at h

ttp://

libra

ry.se

g.or

g/

Page 9: Forecasting reservoir performance by mapping seismic …...Forecasting reservoir performance by mapping seismic emissions Charles Sicking 1, Jan Vermilye , and Ashley Yaner Abstract

physicist at Atlantic Richfield Company, where he devel-oped multiple seismic processing algorithms andtechnologies for 2D and 3D velocity model building usingtime-to-depth and depth imaging applications. He also ledthe development of the seismic wavelet processing flowand methodology for ARCO’s seismic processing system.

Jan Vermilye received a bachelor’sin geology from the State UniversityCollege of New York, a master’s ingeology from Columbia University,and a doctorate in geology from Co-lumbia University. She was GlobalGeophysical’s manager for microseis-mic processing and interpretation.She has more than 20 years of experi-

ence working and teaching in the field of geology and hasexpertise in fracture mechanics and structural analysis ofnatural fracture systems from the microscopic to the fieldscale. She has been published in First Break, the Oil and

Gas Journal, the Journal of Structural Geology, theJournal of Geophysical Research, and Geology.

Ashley Yaner received a B.S. in geo-physical engineering from the Colo-rado School of Mines (CSM) and amaster’s degree in geophysics from theCSM’s Center for Wave Phenomena.She was the ambient seismic team leadat Global Geophysical Services, over-seeing and processing ambient seismicprojects. She was previously a micro-

seismic analyst and research geophysicist at Cimarex En-ergy, where she integrated well-log analyses with 3Dseismic data. Sheworked at the Center forWave Phenomenaat the Colorado School of Mines studying microseismic mi-gration, waveform tomography, and full-waveform velocityinversion. She is a senior data analyst on the R&D teamat Fracture ID, analyzing drill-bit geomechanics data and aid-ing in processing software development.

Interpretation / November 2017 T445

Dow

nloa

ded

07/2

6/17

to 4

7.18

6.10

.220

. Red

istri

butio

n su

bjec

t to

SEG

lice

nse

or c

opyr

ight

; see

Ter

ms o

f Use

at h

ttp://

libra

ry.se

g.or

g/