APPLICATION OF TIME-LAPSE SEISMIC SHEAR WAVE INVERSION TO CHARACTERIZE THE STIMULATED ROCK VOLUME IN THE
NIOBRARA AND CODELL RESERVOIRS, WATTENBERG FIELD, CO
by
c
Copyright by Staci K. Mueller, 2016 All Rights Reserved
A thesis submitted to the Faculty and the Board of Trustees of the Colorado School of Mines in partial fulfillment of the requirements for the degree of Master of Science (Geo-physics). Golden, Colorado Date Signed: Staci K. Mueller Signed: Dr. Thomas L. Davis Thesis Advisor Golden, Colorado Date Signed: Dr. Terence K. Young Professor and Head Department of Geophysics
ABSTRACT
Advances in horizontal drilling and completions in shale reservoirs have allowed operators to extract hydrocarbons within low permeability reservoirs that were once impossible to ac-cess. The integration of time-lapse multicomponent seismic data with engineering technology aids in the characterization of these reservoirs through monitoring. This thesis investigates the fast and slow shear wave components of a time-lapse, nine-component seismic survey to determine the stimulated volume in the Niobrara and Codell reservoir intervals. The time-lapse post-stack inversions of the shear wave datasets provide insight into how the shear impedance is affected by hydraulic fracturing through the work of cross-equalized seismic shear impedances and shear wave splitting. The study area is the Wishbone Section within Wattenberg Field, CO, which is owned and operated by Anadarko Petroleum Corporation and contains eleven horizontal wells that vary in spacing and completion methods. Shear seismic data sets were acquired over this section before and after hydraulic stimulation.
The time-lapse shear seismic inversions show an increase in fast shear wave velocity and a decrease in slow shear velocity after stimulation. The sensitivity of both the fast and slow shear seismic to stimulation correlates with the net pressure trends at each stage. Borehole image log interpretations are compared to the inversions to analyze the affect that a complex fracture network has on induced anisotropy.
The stimulated volume for the Niobrara and Codell reservoir intervals are now more accurately defined. Time-lapse shear seismic is the only technology that is able to define the stimulated rock volume and reveal areas that are not being accessed by the wells currently drilled. These areas are now detected within the Wishbone section, and may be candidates for future re-completion.
TABLE OF CONTENTS
ABSTRACT . . . iii
LIST OF FIGURES AND TABLES . . . vi
LIST OF ABBREVIATIONS . . . xii
ACKNOWLEDGMENTS . . . xiv
DEDICATION . . . xv
CHAPTER 1 INTRODUCTION . . . 1
1.1 Wattenberg Field Background . . . 2
1.2 Geology . . . 2
1.2.1 Stratigraphy and Depositional History . . . 4
1.2.2 Regional Faulting . . . 6
1.3 Petroleum System . . . 8
1.4 Study Area . . . 9
1.4.1 Seismic Data . . . 12
1.5 Research Objectives . . . 12
CHAPTER 2 SHEAR SEISMIC ANALYSIS . . . 14
2.1 Turkey Shoot Fracture Characterization . . . 17
2.2 Aquisition and Processing Overview . . . 18
2.3 Cross-Equalization . . . 19
2.4 Time and Amplitude Analysis . . . 23
2.6 Discussion . . . 28
CHAPTER 3 TIME-LAPSE POST-STACK INVERSION OF FAST AND SLOW SHEAR SEISMIC . . . 30
3.1 Model-Based Inversion Theory . . . 30
3.2 Inversion Parameters and Method . . . 34
3.3 Inversion Results . . . 38
3.4 Tuning Thickness . . . 44
CHAPTER 4 INTEGRATION AND INTERPRETATION OF SHEAR SEISMIC . . . 50
4.1 Pseudo Shear Wave Splitting . . . 55
4.2 Validation Data . . . 62
CHAPTER 5 CONCLUSIONS AND RECOMMENDATIONS . . . 73
5.1 Future Work and Recommendations . . . 75
LIST OF FIGURES AND TABLES
Figure 1.1 Structure contour map of the Denver Basin showing oil wells in red and gas wells in green . . . 3 Figure 1.2 Stratigraphic column modified from Higley et al., 2003 and Sonnenberg,
2013. . . 4 Figure 1.3 Study area is located between two major regional wrench faults,
modified from Higley and Cox. Black box outlines the time-lapse nine-component Turkey Shoot Survey. Highest seismic fold is over Wishbone section, where eleven horizontal wells were drilled and completed. Incoherence seismic attribute representing fault probability is shown within Wishbone section. Vertical wells containing synthetic shear sonic logs are shown. Green wells were used in inversions while
red wells were excluded. . . 7 Figure 1.4 Shear faults affecting fracturing . . . 8 Figure 1.5 Cross section of Wishbone with 7 Niobrara wells and 4 Codell wells
varying in spacing. . . 10 Figure 1.6 Type log for Wishbone section revealing Gamma Ray and resistivity
character between the top of the Niobrara to the Carlile. . . 11 Figure 1.7 Schedule of operations within the Wishbone section until just after
acquisition of the monitor seismic survey was complete (modified from White). Wells were completed from east to west. Wells 7N, 8C, and 9N were zipper frac’ed. . . 13 Figure 2.1 Compressional velocity is affected by the bulk modulus and shear
modulus, while the shear velocity is only affected by the shear modulus. . 14 Figure 2.2 Sketch of the transversely isotropic model with a horizontal symmetry
axis caused by vertical fractures . . . 16 Figure 2.3 Acquisition geometry for the Turkey Shoot survey. The source (green)
and receiver (dark red and yellow) line spacing are shown. The study
area is outlined in pink . . . 19 Figure 2.4 S1S1 and S2S2 NRMS values before and after cross-equalization. . . 22
Figure 2.5 S1S1 and S2S2 frequency spectrums before and after cross-equalization. . 24 Figure 2.6 S1S1 and S2S2 baseline and monitor RMS average amplitudes before
and after cross-equalization. . . 25 Figure 2.7 Time-lapse RMS amplitude pseudo shear wave splitting anomalies on
the original volumes. Positive anomalies are red and negative anomalies are blue. The positive anomalies represent a larger difference in fast and slow shear wave velocity after completion, and the negative anomalies represent a decrease in the difference between fast and slow shear velocity after completion. The largest anomalies occur near and within the major faults and on the western side of the Wishbone section. The incoherence seismic attribute representing probable fault
locations is shown in black. . . 27 Figure 2.8 Time-lapse RMS amplitude pseudo shear wave splitting anomalies on
the cross-equalized volumes. Positive anomalies are red and negative anomalies are blue. The positive anomalies represent a larger difference in fast and slow shear wave velocity after completion, and the negative anomalies represent a decrease in the difference between fast and slow shear velocity after completion. Both the positive and negative
anomalies increase regionally compared to the original volume, so it is more difficult to distinguish the eastern and western halves of the Wishbone section. The incoherence seismic attribute representing
probable fault locations is shown in black. . . 29 Figure 3.1 Iterative loop used in model-based inversion . . . 32 Figure 3.2 Outline of Turkey Shoot survey area with all wells used in wavelet
extraction. Green wells were used in the low frequency model, red wells were excluded. The eleven horizontal wells are shown in the center. . . 34 Figure 3.3 The S1S1 and S2S2 extracted wavelets and statistical wavelets used in
the inversions. . . 35 Figure 3.4 Well E shear log tied to the S1S1 and S2S2 baseline seismic.
Correlation coefficient is 0.867. . . 36 Figure 3.5 S1S1 low frequency model used in the inversion. Well G shear sonic log
is displayed on the left. . . 37 Figure 3.6 Seismic shear impedance compared to unfiltered shear impedance logs
Figure 3.7 Cross plots of seismic shear impedance versus the unfiltered shear
impedance logs from wells A-F. . . 40 Figure 3.8 S1S1 baseline (top) and monitor (bottom) inversion results showing the
RMS average over a 50ms window below the top of the Niobrara. The Codell wells are yellow and the Niobrara wells are black. The top
Niobrara incoherence attribute overlays the inversion results. . . 42 Figure 3.9 S2S2 baseline (top) and monitor (bottom) inversion results showing the
RMS average over a 50ms window below the top of the Niobrara. The Codell wells are yellow and the Niobrara wells are black. The top
Niobrara incoherence attribute overlays the inversion results. . . 43 Figure 3.10 Approximate thicknesses in time and depth of packages within the
reservoir. . . 44 Figure 3.11 Cross sections of the model-based inversions for the S1S1 baseline (top)
and monitor (bottom) surveys. A shear impedance log of well G is
shown, this well was not included in the inversion. . . 45 Figure 3.12 S1S1 inversion results from the baseline survey subtracted from the
monitor survey. The figure is the RMS average of the amplitude envelope over a 50ms window. The highest shear impedance anomalies are highlighted in red, green, and yellow, while the lower impedance changes are shown in blue and white. The time-lapse fast shear impedances show the largest anomalies on the western half of the Wishbone section as well as in the fault network in the southeast side of the section (shown as the incoherence seismic attribute in black). . . 47 Figure 3.13 S2S2 inversion results from the baseline survey subtracted from the
monitor survey. The figure is the RMS average of the amplitude envelope over a 50ms window. The highest shear impedance anomalies are highlighted in red, green, and yellow, while the lower impedance changes are shown in blue and white. The greatest slow shear impedance changes occurring south of the east-west central graben
(shown as the incoherence seismic attribute in black). . . 48 Figure 3.14 Model-based inversions of the S1S1 and S2S2 surveys were subtracted.
The slice is the RMS average of the amplitude envelope over a 50ms window. The incoherence seismic attribute shown was extracted from
the top of the Niobrara. . . 49 Figure 4.1 Illustration of the behavior of an up-going shear wave that splits at an
Figure 4.2 Pseudo shear wave splitting amplitude anomalies. The maximum pseudo shear wave splitting anomalies are shown in red/yellow and the minimum PSWS are shown in blue/black. This figure covers a window of 50ms below the top Niobrara horizon. The incoherence seismic
attribute is shown in black overlying the amplitude slice. . . 52 Figure 4.3 The monitor S1S1 inversion minus the S2S2 inversion over the whole
reservoir interval (50ms window below top Niobrara) shown on the left. Microseismic moment shown on the right (created by Isabel White). The area outlined in black represents the approximate stimulated volume as shown by the microseismic and is being compared to the
S1S1 minus S2S2 monitor slice. . . 53 Figure 4.4 Cross-section of fast shear seismic data from baseline survey. The
yellow box represents the overburden window used to calculate the noise threshold. The green box represents the window used to calculate the Niobrara slices and the pink box represents the window used to calculate the Codell slices. These are approximate windows, as the actual windows follow the interpreted horizons. The top of the
Niobrara is shown in blue, and the graben is interpreted in black. . . 54 Figure 4.5 S1S1 baseline (left) and monitor (right) inversion slices over a 40ms
window below the top Niobrara plus 20ms, centered on the C Chalk. . . . 55 Figure 4.6 S2S2 baseline (left) and monitor (right) inversion slices over a 40ms
window below the top Niobrara plus 20ms, centered on the C Chalk. . . . 56 Figure 4.7 Normalized production for eleven horizontal wells in Wishbone section
(created by Erdinc Eker). . . 57 Figure 4.8 Time-lapse pseudo shear wave splitting over the Niobrara interval,
centered around C Chalk, with normalized production rank at bottom of image. The anomalies represent time-lapse changes in fast and slow
shear impedance. . . 58 Figure 4.9 Time-lapse pseudo shear wave splitting over the Codell interval,
including Fort Hays and Carlile, with normalized production rank at bottom of image. The anomalies represent time-lapse changes in fast
and slow shear impedance. . . 59 Figure 4.10 Stimulated volume represented by translucent red polygons over the
Niobrara (left) and Codell (right). The calculated stimulated volume for the Niobrara is 37.95%, while the stimulated volume for the Codell
Figure 4.11 The time-lapse PSWS volume is shown on the left with only the largest anomalies shown. The 3D geobody over the Niobrara interval is shown
on the right; the Niobrara wells are in green. . . 61 Figure 4.12 Pseudo shear wave splitting in the Niobrara interval overlain by
borehole image log interpretations on wells 2N and 6N. The black arrows point out areas where the complex fractures correlate with
PSWS anomalies. . . 63 Figure 4.13 Net pressure modes as described by Nolte and Smith and interpreted by
Allie Grazulis. Mode 1 represents normal lateral fracture growth, mode 2 describes minimal fracture growth, mode 3 is when the fracture has stopped propagating and pressure is building up, and mode 4 depicts
rapid fracture height growth (growing out of zone). . . 64 Figure 4.14 The deviation of each wellbore is shown as different formations in the
upper figure. The net pressure trends were interpreted on the wells shown with a white box. Grazulis interpreted the net pressures, which are shown next to the wellbores with each stage labeled as the lithology it is located in. The incoherence attribute is shown to highlight the
relative location of each well. . . 65 Figure 4.15 The Niobrara pseudo shear wave splitting slice is overlain by mode 1
(top) and mode 2 (bottom) interpreted by Grazulis with the normalized production rank at the bottom. . . 66 Figure 4.16 The Niobrara pseudo shear wave splitting slice is overlain by mode 3
(top) and mode 4 (bottom) interpreted by Grazulis with the normalized production rank at the bottom. . . 68 Figure 4.17 Bar graph representing the number of mode 2 and mode 4 picks within
each range of percent differences in shear impedance. . . 69 Figure 4.18 Four Niobrara wells (top) and four Codell wells (bottom) shown with
microseismic events overlying time-lapse PSWS anomalies within the Niobrara and Codell intervals. Niobrara microseismic events are
clustered in the eastern wells and linear in the western wells, signifying two types of stimulation being achieved. The microseismic within the four Codell wells are located within the highest concentration of large
Figure 4.19 Niobrara time-lapse PSWS slice shown with fracture azimuths
calculated by Motamedi. The yellow lines represent baseline fractures and the red lines represent monitor fractures. The white circles show where the baseline fracture azimuths are different from the monitor
fracture azimuths. The incoherence seismic attribute is shown in black. . 71 Table 2.1 Borehole image log interpretation of 2N, 6N, and Codell well . . . 17 Table 2.2 Turkey Shoot shear seismic time-lapse processing flow performed by
LIST OF ABBREVIATIONS
Anadarko Petroleum Corporation . . . APC Barrels of Oil Per Day . . . BOPD Borehole Image Log . . . BIL Colorado School of Mines . . . CSM Common Depth Point . . . CDP Common Offset Vector . . . COV Fast Shear Mode . . . S1S1 Gas Oil Ratio . . . GOR Giant Wattenberg Field Area . . . GWA Horizontal Transverse Isotropy . . . HTI Million Barrels of Oil . . . MMBO Normalized Root Mean Square . . . NRMS Pre-Stack Time Migration . . . PSTM Pseudo Shear Wave Splitting . . . PSWS Pure Mode Compressional Wave . . . PP Pure Mode Shear Wave . . . SS Quality Control . . . QC Reservoir Characterization Project . . . RCP Root Mean Square . . . RMS Shear Wave Splitting . . . SWS
Slow Shear Mode . . . S2S2 Stimulated Reservoir Volume . . . SRV Three Component . . . 3C Time-Lapse Nine-Component . . . 9C/4D Total Organic Content . . . TOC Trillion Cubic Feet of Gas . . . TCFG
ACKNOWLEDGMENTS
I would like to sincerely thank my advisor, Dr. Tom Davis, who not only welcomed me into his research group, but gave me unending support and guidance over the past two years. His insight into my research is invaluable and I wish him the best as he retires this summer. My committee members, Dr. Paul Sava and Dr. Bruce Trudgill, have taught me many classes over the years and although they have helped me most recently on my thesis, I will remember them as two of the most intelligent, insightful professors that I’ve encountered.
This research would not have been possible without the generosity of Anadarko Petroleum Corporation in providing us with the seismic and a myriad of other geologic and engineering data. It has been a privilege working with Janel Anderson, who has given so much of her time answering unending questions and reading sponsors meeting reports.
I am especially thankful to CSM, in particular the Geology and Geophysics Departments, for providing me with six years of challenging, inspiring courses. I have had supportive, intelligent classmates over the years that have encouraged me and, within RCP, have guided the integration in my research.
For not only technical, but personal guidance, I want to thank Sue Jackson for her dedication to RCP and the Wattenberg team. Her door is always open for the students, and I have gone to her office many times for advice.
Finally, I want to thank my family and friends for their patience and for always listening to my technical ideas. Throughout the past six years, they have been there through all of the challenges, and have helped me overcome many obstacles.
CHAPTER 1 INTRODUCTION
In collaboration with Anadarko Petroleum Corporation (APC), the Reservoir Charac-terization Project (RCP) at the Colorado School of Mines began Phase XV in 2013 with the acquisition of a time-lapse nine component (9C/4D) seismic survey. A myriad of other data were provided by APC, including: a regional PP seismic survey, a three component (3C) seismic survey, a microseismic survey, completions information, production data, tracer engineering results, well logs, and core samples. The objective of Phase XV is to integrate geophysical methods with engineering and geology to determine the optimal well spacing and completion strategies to maximize hydrocarbon recovery in Wattenberg Field. The study area is the Wishbone Section, which is owned and operated by APC and contains eleven horizontal wells that vary in spacing and completion method.
Advances in horizontal drilling and completions has allowed operators to access hydro-carbons that were once impossible to access. The integration of time-lapse multicomponent seismic data with engineering technology aids in the characterization of these reservoirs through monitoring. The four square mile 9C/4D seismic survey utilized in Phase XV was first acquired after APC drilled the eleven horizontal wells, and a monitor survey was ac-quired after the wells were hydraulically fractured. This thesis investigates the fast and slow shear wave components of the 9C/4D survey to determine the effectiveness of the completion strategies implemented in the Niobrara and Codell reservoir intervals. The time-lapse post-stack inversions of the shear wave datasets provide insight into how the shear impedance is affected by hydraulic fracturing, and the original and cross-equalized seismic amplitudes and shear impedances are analyzed to investigate pseudo shear wave splitting.
1.1 Wattenberg Field Background
Located northeast of Denver, CO, the giant Wattenberg Field area (GWA) spans approx-imately 3200 square miles. It produces oil, gas, and condensate from at least nine Cretaceous horizons (Figure 1.1). The field was first discovered in 1970 by Amoco Production Company with gas completions in the Lower Cretaceous J Sandstone, while the first significant Nio-brara production from vertical completions occurred in 1976 (Sonnenberg, 2013). Up until 1999, approximately 1.57 TCFG and 76.4 MMBO had been produced. Horizontal drilling in the Niobrara began in 2009, and currently the initial production from horizontal completions ranges from 100 to 700 BOPD with a GOR range of 500 to 10,000 cu feet per barrel. The estimated ultimate recovery per well is over 300,000 BOE. The Niobrara resource estimate in Wattenberg Field is 3-4 billion barrels equivalent (Sonnenberg, 2013).
1.2 Geology
The Denver Basin is a large asymmetric basin formed during the Laramide Orogeny. The deepest part of the basin lies close to and parallels the Front and Laramie ranges of Colorado and Wyoming, respectively. Wattenberg is considered to be a basin-centered petroleum accumulation and has a positive temperature anomaly. Both thermogenic and biogenic petroleum accumulations occur in the Niobrara. The accumulations in the deep part of the basin are thermogenic oil and gas, while the accumulations along the shallow east flank of the basin are biogenic gas, shown in Figure 1.1 (Sonnenberg, 2013).
The overpressured Niobrara contains four chalk units and three intervening marl intervals at drilling depths between 6200 and 7800 feet. The Fort Hays Limestone is located below the below the Niobrara, and the whole package is grouped as the Smoky Hill member. In descending order, the chalk units are known as the A, B, C, and Fort Hays. Operators in Wattenberg Field have focused primarily on production from the B and C chalks and the underlying Codell Sandstone/Fort Hays.
Figure 1.1: Structure contour map of the Denver Basin showing oil wells in red and gas wells in green (Higley and Cox, 2007).
1.2.1 Stratigraphy and Depositional History
The Upper Cretaceous Niobrara was deposited in a foreland basin within the Western Interior Seaway during a time of a major marine transgression. This Western Interior Cre-taceous Basin was asymmetric with the thickest strata deposited along the western margin. It is a complex basin that developed between Mid to Late Jurassic to Late Cretaceous time with sediment sourcing from the east and west and organic matter developing from pelagic sedimentation. The basin was bordered by mountains to the West and a broad stable cra-tonic zone to the East. It subsided in response to thrust and synorogenic sediment loading followed by pulses of rapid subduction and shallow mantle flow (Sonnenberg, 2013).
Figure 1.2: Stratigraphic column modified from Higley et al., 2003 and Sonnenberg, 2013.
The dominant lithologies of the Niobrara Formation are limestones, chalks, and interbed-ded marls (Figure 1.2). The chalk-marl cycles represent changes from normal to brackish
water salinities possibly related to sea level fluctuations or paleo-climatic factors. The chalk contains a higher percentage of carbonate, and was deposited due to the introduction of warm Gulfian currents into the Western Interior Cretaceous seaway during relatively high sea levels (Sonnenberg, 2013). The marls were deposited when the sea level lowered.
The Niobrara ranges in thickness from 100 to 300 feet in the eastern side of the basin to over 1500 feet in the West. Different rates of sedimentation may have caused thinning in the Niobrara as well as unconformities at the base, within, and at the top of the formation (Longman et al., 1998). The upper unconformity removes the top chalk bed in Wattenberg Field.
Deposition of the Niobrara in the Western Interior Basin was heavily influenced by the in-teraction between the warm north-flowing currents from the paleo-Gulf of Mexico and cooler southward-flowing currents from the Arctic region, plus the variation in sea level. The Gulf water brought in rich carbonate flora of coccoliths and promoted carbonate production and deposition. Siliciclastic input from the West and cooler Arctic currents inhibited carbonate production and deposition (Sonnenberg, 2013).
The Pierre Shale overlays the Niobrara on the east side of the Western Interior Basin, and the Mancos Shale resides in the western part of the basin (age-equivalent to Niobrara). Its Sharon Springs member is an excellent source rock with TOC’s ranging from 2 to 8 weight percent. The Niobrara overlies the Carlile Formation. The Niobrara itself contains two members: the Fort Hays Limestone Member and the Smoky Hill Shale Member. The Fort Hays is a calcareous limestone that has high resistivity and low gamma ray values in well logs. According to Sonnenberg (2013), it has low porosity, but its proximity to source rock makes it a potential drilling target. The Smoky Hill is divided into the A, B, and C benches as well as basal units. Within the Wishbone section, the A bench has eroded to a very thin marl interval to represent the top of the Niobrara. Wattenberg Field contains Niobrara thicknesses that range between 200 feet and 400 feet at 7200 feet to 8000 feet deep. The A, B, and C benches are divided into chalks and marls; however, the boundaries are
gradational so there are not many differences between the bench facies (White, 2015). 1.2.2 Regional Faulting
Within Wattenberg Field, the primary basement controlled right-lateral wrench fault zones extend to the northeast (Figure 1.3). A secondary fault network is oriented north-west and occurs in between the wrench fault zones. The maximum compressional stress is associated with the right-lateral wrench faults, none of which intersect the Wishbone section. The types of faulting in the Denver Basin include polygonal faulting, wrench faulting, listric normal faulting (Sonnenberg and Underwood, 2013), thrust faulting, and reactivated normal faulting. Listric, normal faults extend throughout the Cretaceous interval, which includes the Niobrara, Carlile, and Greenhorn Formations (Davis, 1985). Within the study area, the normal faulting is due to the mid-Tertiary secondary extensional tectonic regime.
The seismic interpretation within the Wishbone section reveals an east-west trending graben with about 150 feet of throw as well as a northeast-southwest trending graben on the northwestern side. Fault movement within this area, especially normal faulting with throws over 100 feet, suggests that stresses are compartmentalized (White, 2015). The anti-thetic faulting and complex damage zones most likely caused dense fracturing. Many of the natural fractures are calcite-filled, and likely healed during reactivation of a post-Laramide extensional regime (Vincelette and Foster, 1992). Along with differential compaction, the recurrent tectonism may have formed migration pathways, complex fracture systems, or fault seals through calcification (Figure 1.4).
Extensive polygonal faulting has been observed in the Niobrara; these are extensional dewatering features that formed because there was not an overriding stress orientation due to fault trends. These features can be seen by extracting the incoherence attribute on high resolution seismic data. The fact that the polygonal faults do not resemble hexagons indicates that both time and concurrent stress regimes have affected the system (Cartwright, 2011). The structural complexity suggests both a compartmentalized stress regime and an intricate fracture network.
Figure 1.3: Study area is located between two major regional wrench faults, modified from Higley and Cox (2007). Black box outlines the time-lapse nine-component Turkey Shoot Survey. Highest seismic fold is over Wishbone section, where eleven hori-zontal wells were drilled and completed. Incoherence seismic attribute representing fault probability is shown within Wishbone section. Vertical wells containing synthetic shear sonic logs are shown. Green wells were used in inversions while red wells were excluded.
Figure 1.4: Shear faults affecting fracturing (Davis, 1985).
1.3 Petroleum System
Hydrocarbon generation in the Denver Basin began in the Late Cretaceous, and source rocks include the Mowry, Huntsman, Graneros, Greenhorn, Carlile, Niobrara, and Sharon Springs member of the Pierre Shale. According to Sonnenberg (2013), other potential source rocks may be located in the Skull Creek and other areas in the Pierre Shale. Pyrolysis analysis done on the Niobrara indicates that organic-rich beds in the formation have organic carbon values that range from 0.5 to 8 weight percent and average 3.2 weight percent (Sonnenberg, 2013). The Kerogen present in the Niobrara is Type-II or oil-prone and the reservoir is self-sourcing with TOC values between 2% and 6% (Sonnenberg, 2013).
The lithology in the Denver Basin consists of interbedded calcareous shale, shaly lime-stones, and marls. These units become shalier and sandier towards the west side of the basin. Carbonates are still present in the West, however clastics begin to take over. The Niobrara reservoir rocks have gone through mechanical and chemical compaction and contain very low porosity and permeability. The factor most affecting the porosity is burial depth; chalks that have original porosities of 50% or greater only contain 6% porosity at 7000 feet. The initial average pore sizes are only a few tenths of a micron, and those are reduced even further through diagenesis (Sonnenberg, 2013). Therefore, fracturing is a key component for reservoir performance.
1.4 Study Area
Eleven horizontal wells were drilled in the Wishbone section; seven in the Niobrara and four in the Codell (Figure 1.5). The spacing between wells vary from 600 feet to 1200 feet, and the number of stages also vary. Some other parameters that vary across the section include the completion fluid volume and type, the proppant volume and type, pressure, and the hydraulic fracturing technique (wells 7N, 8C, and 9N were zipper frac’ed). Although the wells were planned for the Niobrara C Chalk and Codell Sandstone units, geosteering reports and gamma ray logs reveal that the structure in these intervals caused the lateral to land out of zone in many cases.
A type log for the Wishbone section is shown in Figure 1.6, highlighting the entire Nio-brara interval including the high resistivity C Chalk target as well as the Codell Sandstone. Since there are no compressional or shear velocity logs within the Turkey Shoot survey, the synthetic compressional sonic logs and shear sonic logs shown were calculated (for eleven wells) based on gamma ray, resistivity, neutron porosity, and density logs using a neural network. A synthetic log was tested against its real counterpart in a well just outside the study area, and was found to have approximately a 95% correlation (based on a conversation with Matthew Bray).
Figure 1.6: Type log for Wishbone section revealing Gamma Ray and resistivity character between the top of the Niobrara to the Carlile.
1.4.1 Seismic Data
The RCP seismic surveys include a 50 square mile merged 3D P-wave seismic survey, an 11 square mile 3C 3D survey, and a 4 square mile 9C time-lapse survey. All of these surveys contain the Wishbone section with the eleven horizontal wells; however, only the 9C Turkey Shoot acquisition was designed so that the Wishbone section has maximum fold. The eleven horizontal wells in the Wishbone section were drilled from south to north beginning with the eastern-most well and ending on the western side of the section. The first well was drilled (1N in Figure 1.5) in early summer in year 1, while the western-most well (11N) was drilled in the summer of year 2. The Anatoli survey was acquired over half a month in the summer of year 1 (also serves as the baseline Turkey Shoot), and image logs were also gathered at this time. The Wishbone wells underwent hydraulic fracturing in the fall of year 1, and the completion took place in the same order as drilling. The Turkey Shoot monitor survey was then acquired, finishing at the end of October in year 1. A second monitor seismic survey was acquired in January of year 4, after two years of production had taken place. Figure 1.7 shows the complete day-to-day well operations until just after the acquisition of the first monitor.
1.5 Research Objectives
This thesis investigates the fast and slow shear wave components of a 9C/4D survey to determine the effectiveness of hydraulically fracturing within the Niobrara and Codell reservoir intervals. Shear wave data are not sensitive to changes in fluid, therefore it can detect the presence of induced fractures. The time-lapse amplitude changes are analyzed to discriminate induced fracturing, and the fast and slow shear seismic time-lapse inversions are generated in order to investigate the sensitivity of shear impedance changes to hydraulic fracturing. The sensitivity of both the fast and slow shear seismic to stimulation is investi-gated and correlated with microseismic and net pressure trends. Time-lapse inversions have not been previously performed on the shear seismic component of the Turkey Shoot survey.
Figure 1.7: Schedule of operations within the Wishbone section until just after acquisition of the monitor seismic survey was complete (modified from White (2015)). Wells were completed from east to west. Wells 7N, 8C, and 9N were zipper frac’ed.
CHAPTER 2
SHEAR SEISMIC ANALYSIS
When a seismic source releases energy in the form of either an explosion or through a vibroseis source, its energy travels through the subsurface as an elastic wave. The simplest wave to analyze is the compressional wave, which is an acoustic disruption much like sound. A seismic reflection occurs with a change in acoustic impedance, which is the multiplication of rock density by the seismic wave velocity. These reflections can either be positive or negative depending on whether the impedance of the upper or lower rock layer is larger.
Compressional velocity is affected by the bulk modulus and shear modulus, while the shear velocity is only affected by the shear modulus. This means that P-waves change the shape and volume of the rock matrix and fluid-filled pore spaces, while its velocity is dependent on the magnitude of fluid compressibility relative to that of the reservoir matrix. While the S-wave motion is perpendicular to the direction of its wave propagation; therefore, the shape of the rock may change but the volume does not (Figure 2.1). This is why shear waves are relatively insensitive to fluid changes (Johnston, 2013). Overall changes in the pressure of stress in the rock framework results in changes in the moduli and in P- and S-wave velocities.
Figure 2.1: Compressional velocity is affected by the bulk modulus and shear modulus, while the shear velocity is only affected by the shear modulus.
The objective of this thesis is to analyze the pure shear data (S1S1 and S2S2 volumes) to determine the effect of hydraulic completion methods on shear impedance and shear wave splitting. Many studies have characterized shear waves as being highly sensitive to fractures; especially in unconventional reservoirs where time-lapse analysis has been used to monitor changes in direction-dependent velocity due to hydraulic fracturing (Steinhoff, 2013).
Within these types of shale plays, shear waves are polarized in two orthogonal fast and slow orientations that travel with different velocities. A clearer explanation of this phenom-ena is when a shear wave propagates through a simple model of isotropic rock that contains vertically aligned fractures (Figure 2.2). This wave, with its particle motion perpendicular to its propagating direction and azimuthal propagation direction dependent on its source, travels vertically through this hypothetical horizontal transverse isotropy (HTI) media. Its particle motion is split into two orthogonal directions: the first follows the fracture strike and the second is perpendicular to it. The wave that has a particle motion perpendicular to the fracture strike can deform the rock easily because the fractures cause weakness, or a low effective rigidity (Thomsen, 1988). This wave travels at a low velocity and is labeled S2, or slow shear wave. The wave that propagates parallel to the vertical fractures travels at a faster velocity because the fractures do not affect the particle motion (Thomsen, 1988). This wave is known as the S1, or fast shear wave. According to Thomsen (1988), this po-larization occurs instantaneously at the boundary between the two media, therefore the two waves travel independently at different velocities (dependent on the source and how well it approximates a spike impulse).
In reality, this earth model will have multiple layers, and the fast and slow shear waves will have independent reflections at each boundary encountered after initially splitting. However, if the orientation of fractures remains constant, then only two waves will propagate. This shear wave splitting (SWS) is affected by the seismic source as well as medium properties such as the orientation of the maximum stress field and the density of fractures. If the maximum stress orientation is known during processing, the fast and slow components can
Figure 2.2: Sketch of the transversely isotropic model with a horizontal symmetry axis caused by vertical fractures (R¨uger, 1997).
be separated using Alford rotation.
The reflectivity of the vertical incident shear wave at each interface in which SWS occurs is equal to (Thomsen, 1988): RS1 = − ρ2VS1,2− ρ1VS1,1 ρ2VS1,2+ ρ1VS1,1 (2.1) and RS2 = − ρ2VS2,2− ρ1VS2,1 ρ2VS2,2+ ρ1VS2,1 (2.2) where:
RS1 = the fast shear wave reflection coefficient
RS2 = the slow shear wave reflection coefficient
ρ1 = density of the upper layer
ρ2 = density of the lower layer
VS1,1&VS1,2 = the fast shear wave velocities
VS2,1&VS2,2 = the slow shear wave velocities
In the case of an isotropic interface, Equations 2.1 and 2.2 are equal, therefore the S1 and S2 amplitudes are equal. This means that the only difference between the two shear
waves is the varying velocity of the rock (Reasnor, 2001). In fact, one of the most important uses for analyzing shear waves is the characterization of fractures. Fracture density can be approximated by calculating Thomsen’s SWS parameter, γ:
γ = VS1
2
− VS22
2VS22
(2.3) The ratio of the time difference between two reflections in both the fast and slow shear data or the difference in amplitude can be used to estimate the SWS.
2.1 Turkey Shoot Fracture Characterization
The following section summarizes a borehole image log (BIL) study by Dudley (2015) in which he characterized the fractures and maximum horizontal stress in wells 2N and 6N. As interpreted using the BIL, well 6N only contains some large faults, a maximum horizontal stress azimuth of N60W, and a primary natural fracture azimuth of N60E (nearly perpendicular to the wellbore). This well has 41 interpreted natural fractures, with a fracture intensity of 0.01 fractures/ft. Well 2N contained many faults, a maximum horizontal stress azimuth of N80W, and a natural fracture azimuth of both N50E and N80W. This well has 54 interpreted natural fractures, with a fracture intensity of 0.01 fractures/ft. A well that was targeted in the Codell Sandstone contained more natural fractures than the wells targeted in the Niobrara. A summary of the borehole image log interpretations are depicted in Table 2.1.
Table 2.1: Borehole image log interpretation of 2N, 6N, and Codell well (Dudley, 2015).
Category 2N 6N Codell Well
Target Formation Niobrara C Chalk Niobrara C Chalk Codell
Borehole Image Log Quality High High Poor
Number of Natural Fractures 54 41 96
Number of Faults 19 5 9
Natural Fracture Azimuth N50E, N80W N60W, N90W N65W
Max Horizontal Stress Azimuth N80W N60W N65W
Well 2N has 32% more natural fractures and four times as many faults as well 6N. Although these two wells pass through similar lithology, 53% of the fractures in well 6N are
located in the chalk intervals while only 26% of the fractures in well 2N are in the chalk. Based on cluster analysis and the BIL interpretations, the chalks do not contain more natural fractures than the marls. However, the Codell-targeted well showed a fracture intensity of 0.036 fractures/ft; significantly more than the Niobrara wells. The relevant conclusions of this study are as follows:
• The Codell Sandstone has more natural fractures than the Niobrara C Chalk
• The Niobrara contains sealed fractures, while the Codell does not have sealed fractures • The maximum horizontal stress azimuth does not significantly vary between wells and
has a general trend of northwest to southeast
The time-lapse shear seismic anomalies will respond to the anisotropy introduced by hydraulic fracturing. The number of fractures, their azimuths, and whether they are open or sealed play a part in the effectiveness of the completion designs. By studying the time-lapse pseudo shear wave splitting through an inversion analysis, the anomalies within the Niobrara and Codell intervals will reveal how the observed natural fractures are altered due to stimulation.
2.2 Aquisition and Processing Overview
As previously mentioned, the baseline Turkey Shoot survey was acquired before hydraulic stimulation and the monitor survey was acquired after stimulation and before production. The survey was designed to have relatively high fold centering over the Wishbone section where the eleven horizontal wells were completed in the Niobrara and Codell intervals. The fold is nearly identical between the baseline and monitor surveys do to the cropping per-formed to match geometries. The acquisition parameters are the following:
• 2ms sample interval
• Receivers used were single 3 component geophones, with a true north azimuth • 880 feet source line spacing and 660 feet receiver line spacing
• Source and receiver point intervals were 110 feet
Figure 2.3: Acquisition geometry for the Turkey Shoot survey. The source (green) and receiver (dark red and yellow) line spacing are shown. The study area is outlined in pink (White, 2015).
The acquisition geometry for the Turkey Shoot survey is shown in Figure 2.3. The source and receiver spacing are outlined in green and dark red/yellow, respectively. The eleven horizontal wells were drilled in the Wishbone section, outlined in pink. The processing of the shear seismic data (subject of this thesis) was completed by Sensor Geophysical, and the processing steps listed in Table 2.2 were applied in the fall of year 2.
2.3 Cross-Equalization
Cross-equalization is an important aspect of analyzing multicomponent seismic data. This process ensures that the differences in the reservoir between two datasets are related to variations of the signal; not acquisition differences or processing deficiencies. As a part of the processing, Sensor Geophysical designed a workflow to preserve amplitudes between the monitor and baseline surveys. The acquisition geometries were preserved and source-receiver
Table 2.2: Turkey Shoot shear seismic time-lapse processing flow performed by Sensor Geophysical.
Reformat Record length: 4 seconds
Sample interval: 2 milliseconds 3D geometry assignment 55 ft by 55 ft 3D CDP binning
Alford rotation From field coordinates to radial/transverse
Time-Lapse Match Match baseline to monitor
Amplitude recovery Spherical divergence correction + 4 dB/sec gain
Trace edits and mutes Singular value decomposition filter to remove surface generated noise Surface consistent spiking deconvolution 0.1% pre-whitening
Refraction static corrections Datum: 5200 feet Surface consistent residual statics
Surface consistent amplitude scaling
T-F adaptive noise suppression Offset consistent gain control Surface consistent residual statics
Mute 27 degrees
Alford rotation From radial/transverse to S1S1 and S2S2 Common offset vector binning COV Regularization
COV F-XY Deconvolution Time migration then sort to CDP,
relationships stayed constant between surveys, as the baseline geometries were cropped to mimic the monitor. A single statics solution was applied to both surveys and a single velocity model was applied to both surveys for PSTM. A time-lapse cross-equalization workflow was designed to enhance the signal-to-noise ratio, specifically in the reservoir.
Based on previous studies done by Paskvich (2016) and Motamedi (2015), the shear component of the Turkey Shoot dataset did not need to undergo cross-equalization. However in Figure 2.4, the NRMS values taken at the Lower Pierre (LP3, with a centered window of 200ms) ranged from 0-1.0; where <0.3 is considered a highly repeatable interval (White, 2015). Therefore, after completing model-based inversions on the original stacked shear data, a basic workflow was performed to lower the NRMS values in the Lower Pierre and subsequently adjust the reservoir character. The new values are highlighted in Figure 2.4.
Three cross-equalization steps were performed on a window centered around a Lower Pierre reflector residing approximately 500ms above the top of the Niobrara. First, the phase and time shifts were estimated on the S1S1 and S2S2 monitor surveys, and their respective baselines were used as the reference volumes. Those estimated shifts were then applied to the monitor surveys on a trace-by-trace scale rather than globally; this parameter significantly improved the NRMS outcome. Finally, a shaping filter was applied to the previous step’s output. The reference volumes were the baseline surveys, the window was the same as the first step (centered window of 200ms on Lower Pierre reflector), and it was performed trace-by-trace rather than globally. The shaping filter had the largest effect on the resulting NRMS values.
Figure 2.5 highlights the changes that occurred in the frequency spectrum due to cross-equalization. While the frequencies line up better between 0 and 18Hz, the higher frequencies deviate in the cross-equalized spectrum. Overall, there is not a significant improvement in matching frequencies between the baseline and monitor surveys. So to further check if this process is necessary, an amplitude comparison was performed.
2.4 Time and Amplitude Analysis
The S1S1 and S2S2 volumes were generated based on the assumption of a single fast orientation, N20W, throughout the Turkey Shoot survey. When characterizing fractures, the aim is to provide significant detail in the spacial variability of the splitting coefficient which calls for a residual Alford rotation rather than the assumption of a constant fast orientation. Motamedi (2015) found that both of the volumes contain mixed modes, therefore the analysis performed on these datasets may not reveal all of the areas in which shear wave splitting occurs. The shear wave splitting identified in this study is more accurately called pseudo shear wave splitting (PSWS).
The time-lapse changes in the original shear seismic amplitudes as well as the cross-equalized amplitudes are analyzed in this section. The normalized percent differences be-tween the S1S1 and S2S2 volumes were calculated within the reservoir interval to estimate the SRV. Since the datasets did not go through a residual Alford rotation and each of the volumes contain mixed fast and slow modes (evident from interval traveltimes in a previous study by Motamedi (2015)), the focus of analysis of stacked amplitudes is to identify regional time-lapse changes. Ideally, the largest changes should correlate with the stimulated areas.
First, the amplitudes of the S1S1 and S2S2 baseline and monitor surveys were compared to determine the effectiveness of cross-equalization and highlight the differences in amplitudes due to hydraulic stimulation. Figure 2.6 shows that a change in amplitudes does occur over the reservoir interval between the baseline and monitor surveys, and the amplitudes in the cross-equalized slices are slightly different than the original monitor amplitudes (most noticeable in between the eastern-most wells).
2.5 Results and Interpretation
In order to determine if the amplitude changes are significant enough to affect time-lapse pseudo shear wave splitting anomalies, the PSWS is observed for both the original volumes and the cross-equalized volumes.
Figure 2.7 highlights the time-lapse change in amplitudes between the S1S1 and S2S2 datasets based on the following workflow:
• RMS amplitudes were extracted over the reservoir interval from the baseline and mon-itor (original and cross-equalized) S1S1 and S2S2 volumes
• The percent differences in RMS amplitude were calculated between the S1S1 and S2S2 volumes and normalized over the S1S1 RMS amplitudes
• The normalized baseline percent difference in RMS amplitude was subtracted from the monitor differences SWS∆tAmp = [ S1S1,RMS− S2S2,RMS S1S1,RMS ∗ 100] Monitor− [S1S1,RMS− S2S2,RMS S1S1,RMS ∗ 100] Baseline (2.4)
Both the original volumes and the cross-equalized volumes reveal positive and negative variations of 50% after stimulation has occurred. The positive anomalies are red and nega-tive anomalies are blue. The posinega-tive anomalies represent when the slow shear wave velocity decreases, which is indicative of an increase in the fracture density; whereas the negative anomalies represent when the slow shear wave velocity increases, which is indicative of frac-tures sealing (an increase in pressure). In the original baseline and monitor volumes, the shear wave splitting indicates that the largest pressure anomalies occur near and within faults, while the largest fracture anomalies occur on the western portion of the Wishbone section.
Figure 2.8 shows a similar pattern of anomalies, however both the pressure and fracture time-lapse changes have intensified in comparison to the original volumes. The pressure anomalies in blue still correlate with faulted areas, and the fracture anomalies in red are still the highest on the western side of the wishbone section. However, the anomalies have increased regionally so it is more difficult to distinguish the eastern and western halves of the Wishbone section which vary in both well spacing and completion method, and
there-Figure 2.7: Time-lapse RMS amplitude pseudo shear wave splitting anomalies on the original volumes. Positive anomalies are red and negative anomalies are blue. The positive anomalies represent a larger difference in fast and slow shear wave velocity after completion, and the negative anomalies represent a decrease in the difference between fast and slow shear velocity after completion. The largest anomalies occur near and within the major faults and on the western side of the Wishbone section. The incoherence seismic attribute representing probable fault locations is shown in black.
fore should display different PSWS trends. For that reason, the post-stack inversions are performed only on the original volumes.
2.6 Discussion
The post-stack amplitude analysis discussed in this chapter has limitations in its appli-cation. Since the amplitude anomalies highlight time-lapse changes that occur between the fast and slow shear waves, stimulated areas with complex fracture sets may not be anoma-lous (Motamedi, 2015). Additionally, when observing the mixed mode data, the anomalies can represent either the fast or slow component which may not accurately depict time-lapse changes that occur in areas where the predominant fracture direction remains con-stant. Therefore, the interpretation of these time-lapse changes should be performed on a regional-scale, and the character of individual anomalies may not represent the presence and effectiveness of hydraulic fracturing.
A residual Alford rotation would allow for the correct separation of the fast and slow components, therefore the stimulated volume could be more accurately estimated. In this case, the individual anomalies could be analyzed to determine the extent and directionality of the induced fractures.
Much of the time-lapse analysis done in this study is based on percent differences. These differences do not represent a threshold above noise that would accurately represent the SRV. However, time-lapse changes were observed in the overburden both for the amplitudes and inversion results to estimate the noise level. In both cases, the time-lapse percent change decreased with a decrease in depth. The decrease in anomalies was more significant just above the reservoir in the cross-equalized volumes. A study done by Motamedi (2015) suggests that the small anomalies in the overburden are caused by the affects of stimulation rather than random noise caused by repeatability issues.
Figure 2.8: Time-lapse RMS amplitude pseudo shear wave splitting anomalies on the cross-equalized volumes. Positive anomalies are red and negative anomalies are blue. The positive anomalies represent a larger difference in fast and slow shear wave velocity after completion, and the negative anomalies represent a decrease in the difference between fast and slow shear velocity after completion. Both the positive and negative anomalies increase regionally compared to the original volume, so it is more difficult to distinguish the eastern and western halves of the Wishbone section. The incoherence seismic attribute representing probable fault locations is shown in black.
CHAPTER 3
TIME-LAPSE POST-STACK INVERSION OF FAST AND SLOW SHEAR SEISMIC
Post-stack seismic inversion is the process of reconstructing the velocity or impedance in the earth using stacked seismic traces. This is based on the 1D convolutional model (multiples are assumed to be negligible):
T(i) =X
j
r(j)W(i − j + 1) + n(i) (3.1)
where:
r(j) = the zero-offset reflectivity of the earth expressed as a time series W(i) = the seismic wavelet, assumed to be constant
n(i) = additive measurement noise
In other words, inversion is the process of determining the reflectivity, r(j) given the seismic trace, T(i), as the reflectivity directly relates to the impedance contrasts in the subsurface:
r(j) = I(j) − I(j − 1)
I(j) + I(j − 1) (3.2)
where:
I(j) = ρ(j)v(j) 3.1 Model-Based Inversion Theory
The model-based inversion in Hampson-Russell (1999) is based on a generalized inversion method formulated by Cooke (1981), and Cooke and Schneider (1983). It uses a forward model to calculate synthetic seismic data as a part of the inversion algorithm. The impedance profile is calculated based on the 1D convolutional model described in Equation 3.1 by updating model parameters, or the impedance, until it creates an output with the least
error. Some assumptions in this method are: the earth is 1D consisting of a certain number of layers, an earth model is represented by blocky impedance layers, and the wavelet is known (Herawati, 2002). Taylor series expansion is utilized in the generalized linear inversion algorithm: F(M) = F(M0) + ∂F(M0) ∂M ∆M + .... (3.3) where: M0 = initial model M = ”true” model
∆M = change in model parameters F(M) = observed seismic
F(M0) = synthetic seismic from initial model
The series is solved by truncating to the first order and introducing the matrix of deriva-tives, A:
∆F = ∂F(M0)
∂M ∆M = A∆M (3.4)
The solution to the above equation can be expressed as:
∆M = (ATA)−1AT∆F (3.5)
However, that assumes a stable inverse which does not include the effects of noise; there-fore a pre-whitening factor is introduced (Cooke, 1981). The generalized linear inversion method assigns blocks to an initial impedance value. This value changes with respect to thickness and time (Cooke and Schneider, 1983). An initial guess impedance is guided by both the interpolated impedance log and the horizon interpretation (low frequency model). It is the convolutional model (Equation 3.1) that relates the model to the observed seismic. When the reflection coefficient between two layers and the impedance of the upper layer are known, the impedance of a subsequent layer can be determined.
The inversion result is non-unique, therefore a model constraint must be introduced in the algorithm. Hampson-Russell software has an option to limit how far the resulting impedance may deviate from the initial guess. Figure 3.1 is a workflow generated (Cooke and Cant, 2010) for the model-based inversion.
Unfortunately, the derived reflectivity does not necessarily match the true earth reflectiv-ity. Simply changing the wavelet can change the inversion result. Also, because the seismic trace contains noise, the resulting model may be sensitive to the amount of noise. Even calculating the impedance of a certain layer is challenging because each impedance value depends on all of the reflection coefficients from the rock layers above. Therefore, insignifi-cant errors in those reflection coefficients can produce a large error in the derived impedance (low frequency trend error). This is where the non-uniqueness problem comes in. It comes from the fact that there can be many combinations of reflection coefficients to produce many models.
In Hampson-Russell, a constraint is used to distinguish between the set of possible models (Hampson-Russell, 1999). For each trace that goes through the inversion process, the soft-ware has an initial guess trace created during model building. This trace is generated using well logs scattered throughout the survey (in this case, six wells) to calculate one impedance trace for each seismic trace.
A constraint can either be applied as a soft constraint, where the initial guess impedance is a separate piece of information that is added to the seismic trace with a weighting of the two. A hard constraint sets boundaries on how far the final answer can deviate from the initial guess.
The model-based inversion summary is as follows (Hampson-Russell, 1999):
• The effects of the wavelet are removed from the seismic as it is known, and as long as it is the same phase as the seismic, the seismic does not have to be zero-phase
• The wavelet can introduce errors
• Model-based inversion has better resolution than other inversions • The inversion results may be too dependent on the initial model
3.2 Inversion Parameters and Method
As previously mentioned, the synthetic shear logs used to tie the vertical wells within Turkey Shoot were generated using a neural network, and had approximately 95% correlation with a test well. There are eleven vertical wells within the study area, and eight of those were used to extract a wavelet from the LP3 horizon to the J-Sand horizon. Two of the wells were not included in building the model because well H did not have a good tie and is very close to the edge of the survey, and well G is located within a graben (shown in red in Figure 3.2). The six wells that were included in the low frequency model are wells A-F, and they had a range of correlation coefficients between 0.785-0.905.
Figure 3.2: Outline of Turkey Shoot survey area with all wells used in wavelet extraction. Green wells were used in the low frequency model, red wells were excluded. The eleven horizontal wells are shown in the center.
A shear synthetic was created using the algorithm designed for a compressional synthetic by representing the shear sonic logs as compressional sonic logs with a wavelet rotated 180◦.
First, a wavelet was extracted using wells A-F, and a phase rotation was found to be 130◦
volume using the extracted wavelet, and then a statistical wavelet was generated for each volume with the appropriate phase rotations. The wells were then tied again to adjust to the new wavelet. Statistical wavelets were used in the inversions instead of extracted wavelets because they are smoother and contain more frequency; they provide a better output for deconvolution.
Figure 3.3: The S1S1 and S2S2 extracted wavelets and statistical wavelets used in the inversions.
The shear impedance logs were calculated by multiplying the shear sonic logs by the density. Figure 3.4 is an example of a good tie between well E and the S1S1 and S2S2 baseline volumes. The wells that had a correlation coefficient below 0.80 were either located within a graben or near the edge of the survey, and as previously mentioned, they were not included in the earth model.
The model used in the model-based inversion only includes low frequencies that the seismic does not contain. The model generated for the S1S1 baseline inversion was also used in the S1S1 monitor inversion, and the same method was used for the S2S2 inversions to keep consistency between the baseline and monitor workflows. Six horizons were input into the model, including the: Terry, LP3, LP4, Niobrara, Graneros, and J-Sand (shown in Figure 3.5). This was also used as a quality check for the well ties and to pick out problem wells before being used in the inversion analysis.
Figure 3.5: S1S1 low frequency model used in the inversion. Well G shear sonic log is displayed on the left.
The model-based inversion parameters are listed below:
• Soft constraint: 30% seismic driven • Pre-whitening of 1%
• 100 iterations
• Inversion performed over the interval LP4 to J-Sand • Kriging method of extrapolation
• Smoothing filter high cut and taper 3Hz/10Hz
These parameters were selected based on many iterations through the inversion work-flow (Figure 3.1). Figure 3.5 shows the final model used in the S1S1 baseline and monitor inversions.
3.3 Inversion Results
In order to quality control (QC) the parameters, an inversion analysis was performed on six wells before it was applied to the full seismic volume. Once applied, the resulting volume was compared to the unfiltered shear impedance logs from the wells in the initial model. Since the shear seismic data has much lower frequency content than the compressional seismic, and because a post-stack inversion was performed instead of pre-stack, it is expected that the inversions do not discriminate the individual Niobrara benches as well as inversions performed on the compressional volumes (Figure 3.6).
To quantitatively analyze the correlation between the inverted seismic and original shear impedance logs, a cross plot for each of the four inversions - S1S1 baseline, S1S1 monitor, S2S2 baseline, and S2S2 monitor - was generated (Figure 3.7). The correlations are all 0.93 with the exception of the S1S1 monitor inversion (0.87). The impedance logs were filtered to the seismic bandwidth as well, and the correlations are approximately 0.97. A blind well was also used for validation, and the filtered log for that well located in a graben has a correlation of about 93% to all four inversions.
Since the shear seismic has a median frequency of about 13Hz and a bandwidth 0/5/22/35 compared to the compressional seismic bandwidth of 0/5/60/65, the initial model has more of an influence on the inversion results. The resulting time-lapse S1S1 and S2S2 impedances can be used for input into a geomechanical model; however, this study focuses on the time-lapse changes to detect the stimulated reservoir rather than simply calculating shear impedance. Therefore, the low frequency model will have equal influence on both the baseline and monitor inversions and will be negated when subtracting the results.
Figure 3.6: Seismic shear impedance compared to unfiltered shear impedance logs from wells A-F.
The S1S1 inversion results are depicted in Figure 3.8 as the RMS average within a 50ms window below the top of the Niobrara Formation (entire reservoir interval including both the Niobrara and Codell). They show slightly lower impedance values in the northwest and higher impedance values in the southeast. The S1S1 monitor slice over the same interval as the baseline shows an increase in shear impedance within and on the northwestern side of the east-west central graben (shown as incoherence attribute) after stimulation has occurred; this correlates with the closer well spacing and zipper frac.
The S2S2 inversion results (Figure 3.9) are shown over the same interval as the S1S1 results. The baseline inversion results show a much lower impedance overall compared to the S1S1 inversions (the same color scales were used). However, a similar trend exists with lower impedance in the northwest and higher impedance in the southeast. The monitor slice highlights a decrease in impedance on both the northern and southern side of the east-west central graben. A discrimination between the east-west and east side of the Wishbone section cannot be quantitatively measured without subtracting the baseline volume from the monitor.
A qualitative analysis of the baseline and monitor inversion slices highlight an increase in the S1S1 impedance and a decrease in S2S2 impedance after stimulation. Since the impedance is a multiplication of shear wave velocity and density of the rock, these results indicate that the fast shear wave velocity increases after stimulation and the slow shear wave velocity decreases after stimulation. Therefore, the fast shear waves are able to prop-agate along more fracture planes and the slow shear waves are encountering more fractures perpendicular to their particle motion.
In a previous study done by Motamedi (2015), it was found that both the S1S1 and S2S2 volumes contain a significant number of mixed modes so it is difficult to interpret individual anomalies, but regionally the inversions are able to detect the induced fracturing. By subtracting the baseline from the monitor surveys, a quantitative analysis of the inversion differences can be made.
Figure 3.8: S1S1 baseline (top) and monitor (bottom) inversion results showing the RMS average over a 50ms window below the top of the Niobrara. The Codell wells are yellow and the Niobrara wells are black. The top Niobrara incoherence attribute overlays the inversion results.
Figure 3.9: S2S2 baseline (top) and monitor (bottom) inversion results showing the RMS average over a 50ms window below the top of the Niobrara. The Codell wells are yellow and the Niobrara wells are black. The top Niobrara incoherence attribute overlays the inversion results.
3.4 Tuning Thickness
The wavelengths before and after the model-based inversion were analyzed in order to cal-culate the tuning thickness of the original seismic amplitudes versus the resulting impedances. The resolution limit can be expressed in terms of the wavelength in the ground (Hill, 2015):
∆T = 1 2favg
= Pavg
2 (3.6)
This gives the resolution limit as: ∆Z = V ∗ ∆T 2 = v ∗ Pavg 4 = λavg 4 (3.7)
An interval velocity is assumed to be 7400ft/s (derived from the first part of Equation 3.7). The average wavelength of the original seismic is 60ms, while the inverted wavelength is 40ms. Therefore, using Equation 3.7 and converting λavg to feet, the tuning thickness
of the original seismic is slightly below 111ft, and the inverted seismic tuning thickness is slightly below 74ft.
Figure 3.10: Approximate thicknesses in time and depth of packages within the reservoir.
Figure 3.10 highlights packages within the reservoir (Niobrara and Codell) that are able to be resolved through seismic inversion. As previously stated, the individual benches are difficult to resolve with the low frequency shear seismic data.
Figure 3.11: Cross sections of the model-based inversions for the S1S1 baseline (top) and monitor (bottom) surveys. A shear impedance log of well G is shown, this well was not included in the inversion.
Figure 3.11 depicts the inversion results for the S1S1 baseline and monitor volumes. Individual packages within the Niobrara interval are resolved; however, based on the known tuning thickness, these are most likely grouped benches with a minimum resolution of 74ft. The impedance alternates throughout the Niobrara and Codell intervals, and dramatically decreases at the Greenhorn Lincoln Limestone.
To quantitatively analyze the time-lapse changes occurring in the fast and slow shear volumes before and after hydraulic fracturing, the baseline surveys were subtracted from the monitor surveys over the reservoir interval (50ms), and the RMS average of the amplitude envelope is shown in Figure 3.12 and Figure 3.13. The highest shear impedance anomalies are highlighted in red, green, and yellow, while the lower impedance changes are shown in blue and white. The time-lapse fast shear impedances show the largest anomalies on the western half of the Wishbone section as well as in the fault network in the southeast side of the section (Figure 3.12). Figure 3.13 depicts the largest slow shear impedance changes occurring south of the east-west central graben (shown as the incoherence attribute). The color scales on the two slices were kept the same for comparison.
Time-lapse shear wave splitting is another method of quantitatively characterizing the stimulated reservoir. The inverted S2S2 volume was subtracted from the inverted S1S1 volume in Figure 3.14 for both the baseline survey (left) and the monitor survey (right). The color scales were kept constant for comparison. This figure highlights the significant change in pseudo shear wave splitting that occurs before and after hydraulic fracturing. The monitor slice (right) shows significant splitting occurring - that does not appear in the baseline survey - within the east-west central graben and along the fault network in the northwest part of the section, which suggests that those areas were hydraulically fractured. Further analysis of this theory is performed when the engineering data and microseismic are correlated with the time-lapse inversion anomalies.
Figure 3.12: S1S1 inversion results from the baseline survey subtracted from the monitor survey. The figure is the RMS average of the amplitude envelope over a 50ms window. The highest shear impedance anomalies are highlighted in red, green, and yellow, while the lower impedance changes are shown in blue and white. The time-lapse fast shear impedances show the largest anomalies on the western half of the Wishbone section as well as in the fault network in the southeast side of the section (shown as the incoherence seismic attribute in black).
Figure 3.13: S2S2 inversion results from the baseline survey subtracted from the monitor survey. The figure is the RMS average of the amplitude envelope over a 50ms window. The highest shear impedance anomalies are highlighted in red, green, and yellow, while the lower impedance changes are shown in blue and white. The greatest slow shear impedance changes occurring south of the east-west central graben (shown as the incoherence seismic attribute in black).
Figure 3.14: Model-based inversions of the S1S1 and S2S2 surveys were subtracted. The slice is the RMS average of the amplitude envelope over a 50ms window. The incoherence seismic attribute shown was extracted from the top of the Niobrara.
