Use of a Petrophysical-based Reservoir Zonation and Multicomponent Seismic Attributes for Improved Geological Modeling, Vacuum Field, New Mexico.

INTRODUCTION

Improving reservoir performance and increasing ultimate recovery from petroleum reservoirs in an efficient, economic, and environmentally safe manner are among the primary goals of reservoir characterization, modeling, and reservoir management. Integrated reservoir characterization that incorporates independent data sets including three-dimensional seismic, well logs, core, and production data is essential to provide an accurate description of the physical properties of a reservoir.

Multicomponent and time-lapse seismic data are also important to characterize both static and dynamic reservoir parameters, including porosity, anisotropy, and fluid property variations within structurally and stratigraphically complex reservoirs. An accurate geologic model based on integrated reservoir characterization can be valuable when used to evaluate vertical and lateral variability of reservoir properties, estimate hydrocarbon pore volumes, or to simulate fluid flow for performance prediction. The geologic model can be used for reservoir simulation to predict reservoir performance under different development scenarios, including supplemental recovery, and to aid in designing an appropriate reservoir management plan. This integrated approach to reservoir characterization and geologic modeling was implemented to evaluate the San Andres reservoir at Vacuum Field, New Mexico.

Vacuum Field is located in southeast New Mexico, approximately 50 miles northeast of Carlsbad, New Mexico (Figure 1.1). Stratigraphic, structural, and diagenetic variability within the platform and shelf margin carbonates of the Permian San Andres and Grayburg Formations at Vacuum Field have created a very heterogeneous and compartmentalized reservoir. Detailed characterization and modeling of the vertical and lateral heterogeneities within the reservoir are necessary so that areas of potential bypassed pay can be targeted using supplemental recovery techniques or infill development, including horizontal wells.

The portion of Vacuum Field being evaluated is under waterflood operations (secondary recovery) and has been converted to a partial-field, CO₂ flood (enhanced recovery) and monitored to evaluate the effect of gas injection on reservoir performance and recovery. The CO₂ injection within the study area represents the third phase of a multi-phase injection program within the Central Vacuum Unit at Vacuum Field. The goal of CO₂ injection is to improve sweep efficiency beyond secondary recovery, reduce residual oil saturation within the reservoir, and therefore increase recovery.

The central goal of this research involves demonstration of an alternative technique to construct a geologic model using a petrophysical-based reservoir zonation to reduce vertical upscaling problems and use of multicomponent seismic attributes to improve the estimation of reservoir properties between wells. Inherent in the model building process is the development of a detailed reservoir description utilizing all available data. The primary research objectives are:

1. Develop an updated sequence-stratigraphic and structural framework for the San Andres Formation at Vacuum Field using three-dimensional seismic, well logs (including horizontal wells), core data, and information from existing outcrop studies.

2. Define petrophysical-based flow units (reservoir zones) within the gross sequence-stratigraphic framework using flow capacity, storage capacity, and pore-throat radius values (R35) derived from core and log data. Vertical variability or compartmentalization within the reservoir is identified using this approach.

3. Analyze and interpret multicomponent seismic data and establish relationships among reservoir parameters from well data and multicomponent seismic attributes. Lateral variability within the reservoir is identified and incorporated using these data.

4. Construct a three-dimensional geologic model based on the sequence-stratigraphic and petrophysical-based zonation and use the established relationships among multicomponent seismic attributes and reservoir properties to estimate reservoir properties between wells. Populate the geologic model with reservoir properties within individual flow units that are below seismic resolution.

5. Evaluate the spatial distribution of mineral components (eg. anhydrite cement) within the San Andres Formation to identify their contribution to reservoir compartmentalization and seismic response.

1.2 Methods

This section lists the available data within the study area and outlines the various methods employed to characterize the San Andres reservoir and construct a representative three-dimensional geologic model. Unless specifically stated, software from Landmark Corporation was used to construct cross sections, generate maps, interpret seismic, and construct geologic models.

Data available to characterize the San Andres reservoir included both regional and site-specific data. Figure 1.2 illustrates those data available within the specific study area. Regional data, that also covered the study area, consisted of a 115 square-mile, compressional-wave, three-dimensional seismic survey (Maljamar 3-D) and several two-dimensional compressional-wave seismic lines. Within the study area, twelve three-dimensional surface seismic volumes (Vacuum 3-D surveys) that consisted of four compressional-wave (p-wave) volumes, four "fast" shear-wave (S1-wave) volumes, and four "slow" shear-wave (S2-wave) volumes were acquired (Figure 1.2). These data were acquired through the Colorado School of Mines Reservoir Characterization Project (Phase VI and Phase VII) as part of a time-lapse seismic (seismic monitoring) project associated with a CO₂ injection program at Vacuum Field. Borehole geophysical data acquired during Phase VI included a 9-component, single geophone, vertical seismic profile (VSP), walkaway vertical seismic profile (WAW), and downhole three-dimensional seismic survey, each acquired in well CVU-200. Borehole geophysical data acquired during Phase VII included two 9-component, multi-geophone VSPs, WAWs, and downhole three-dimensional seismic surveys, each acquired in well WS-15 (one set prior to CO₂ injection and one set post-CO₂ injection).

Log data consisted of conventional data from 120 wells, one FMIâ (Formation MicroImager) in well WS-2-26, and log data from a medium-radius, dual-lateral horizontal well (CVU-110). Six cores were available (Figure 1.2) and neural-network estimated permeability curves were provided by Texaco, the unit operator, for the majority of the wells within the study area.

Injection and production data were also provided by Texaco, which primarily consisted of monthly cumulative volumes of fluids injected or produced. In addition, seventeen wells had single or multiple injectivity profiles.

Various scales of cyclicity within the San Andres Formation were identified using 3-D seismic and core data (Appendix A). Lower frequency, third-order composite sequences were identified from a regional compressional-wave 3-D seismic volume and through identification of type-1 sequence boundaries in core. High frequency sequences (fourth-order) and individual shallowing-upward cycles (fifth-order) were also identified using core data.

Structural trends and significant faults within the study area were interpreted using compressional-wave 3-D seismic (amplitude and continuity volumes), structure contour and isopach maps generated from log data, and interpretation of log data from a medium-radius, dual-lateral horizontal well. Log data from the deviated sections of the horizontal well were corrected to true vertical depth and correlated to adjacent vertical wells to identify small-displacement faults (< 25 feet).

Cumulative flow capacity (kh) and storage capacity (f h) plots were used to define vertical variability within the reservoir and to further subdivide high-frequency sequences into flow units. To support the flow unit interpretation, capillary pressure measurements were obtained from 18 core plugs taken from various rock types (Appendix B). Multi-point capillary pressure tests were conducted on each sample by Marathon’s Petroleum Technology Center using a centrifuge with air displacing brine (wetting phase). Pore-throat radius values were computed for the most effective porosity using the capillary pressure data. Computed pore-throat radius values were related to core porosity and permeability and used to modify the coefficients of the empirically-derived Winland equation. The modified Winland equation was then used to estimate pore-throat radius (i.e. flow units ) using log data in non-cored wells. A grid of cross sections was then constructed using StratWorksâ and used to correlate zones across the study area. Mean porosity, mean permeability, and interval isopach maps were generated for each flow unit using well data (Appendix H).

Surfaces representing the top of each flow unit and two faults were generated using Z-MAP Plusä . Mapped surfaces were used in Stratamodelä to build the framework for the three-dimensional geologic model. The geologic model was spatially discretized into 37 rows, 37 columns, and 18 layers, resulting in approximately 26,000 cells. As a result, upscaling prior to flow simulation was not necessary. Each cell was 110 x 110 feet in areal size (with the exception of faulted cells) and each of the 18 layers represented a reservoir flow unit defined using the petrophysical-based method.

Significant formations or key stratigraphic surfaces were interpreted on multicomponent seismic volumes using SeisWorksâ . Interval and event seismic attributes were extracted from specific reservoir intervals or at selected stratigraphic horizons (events) and compared to reservoir properties from well data. Amplitude and complex trace attributes were extracted using PALä (Poststack Attribute Library). Attribute values and reservoir properties that coincided at each well location were crossplotted using RAVEä to identify potential relationships among the various parameters.

Interval isochron maps across the reservoir interval and shallower intervals from the multicomponent seismic volumes were used to estimate shear-wave anisotropy ( {D t_s2 - D t_s1} / D t_s1 ). This attribute was also related to reservoir and rock properties.

Once significant relationships were established among seismic attributes and reservoir parameters, a combination of 2-D and 3-D estimation techniques was utilized to populate the geologic model with reservoir properties (eg. porosity and permeability) within individual flow units that were below seismic resolution.

Estimates of anhydrite cementation were conducted using PetroWorksâ by generating r _maa-U_maa crossplots on 27 wells that had density, Pe (photoelectric index), and neutron porosity curves (Appendix I). This information was used to estimate anhydrite distribution in three dimensions within the geologic model to asses the impact of anhydrite cementation on reservoir quality and seismic response. Estimates of sandstone/siltstone content were also conducted.

Figure 1.3 summarizes the input data and methods used to characterize and construct the geologic model of the San Andres reservoir within the study area.

Given the nature of this research, this section begins with an overview of several significant geological studies conducted on the San Andres and associated formations from surface exposures in the Guadalupe Mountain area, and through subsurface studies. In addition, a brief discussion of previous modeling techniques and specific reservoir characterization studies at Vacuum Field is included.

The San Andres and associated Permian formations have been the focus of numerous geological studies as part of early railroad and mineral surveys in the western United States and due to the significant quantities of oil and gas discovered within the Permian Basin. Excellent surface exposures of Permian strata present within the Guadalupe Mountain region (Figure 1.4) have been studied extensively since the middle to late 1800’s.

The earliest recorded geological work on the Permian section within the Guadalupe Mountains was conducted by G. G. Shumard (1858) as part of a railroad survey by Captain John Pope. Shumard located a water well for the railroad and collected numerous fossils from the region that were identified by his brother (B. F. Shumard, 1858) to be of Permian age. Following Pope’s expedition, several geological reconnaissance trips through the Guadalupe Mountain area were conducted (Jenney, 1874; Tarr, 1892; Girty, 1902, 1908; Richardson, 1910; Beede, 1910) that focused on the identification of fossil assemblages, regional correlation of formations, and early stratigraphic nomenclature. The type section of the San Andres Formation was roughly established within the San Andres Mountains of south-central New Mexico by Lee and Girty (1909) and was later refined by Needham and Bates (1943). Baker (1920) was one of the first to recognize the Yeso and San Andres within the Guadalupe Mountains of New Mexico. Darton and Reeside (1926) described lateral relationships among the Capitan Limestone and equivalent shelfal and basinal deposits.

A second phase of geological work began in the middle to late 1920’s following the discovery of oil and gas in Permian age rocks northeast of the Guadalupe Mountains. In 1929, the year Vacuum Field was discovered, several studies were published that addressed conceptual models of deposition, facies relationships, regional structural interpretations, and shelf to basin correlations (Blanchard and Davis, 1929; Crandall, 1929; Lloyd, 1929; Lloyd and Thompson, 1929; Willis, 1929). In 1930, Cartwright published one of the first regional cross sections of the Permian Basin from the Guadalupe Mountains to Jones County in North-Central Texas. Dickey (1940) and Woods (1940) also presented regional well-log cross sections across New Mexico and Texas showing the characteristics of the San Andres and related shelf formations in the subsurface east of the Guadalupe Mountains. Formal series nomenclature for the Permian section was proposed in 1939 to include the Ochoa, Guadalupe, Leonard, and Wolfcamp series (Adams et al., 1939). Based on this new series classification, the physical history of the Permian Basin and southern midcontinent region was presented by Hills (1942). Hills produced seven paleogeographic maps of the region and several supporting cross sections and related this work to the standard Permian section for North America established by Adams et al. (1939). Skinner (1946) presented new evidence regarding the age of the San Andres and associated formations and their regional correlation.

Detailed and comprehensive studies of the Permian stratigraphy of west Texas and southeast New Mexico were conducted by King (1942, 1948). This work produced detailed geological maps of the area and introduced formal stratigraphic nomenclature for most of the Permian section. His publications described the stratigraphic and structural relationships from the Delaware Basin to the reef complex in the Guadalupe Mountains. King (1951, 1977) also presented a thorough review of the tectonic history of the area.

Adams and Frenzel (1950) and Newell et al. (1953) investigated the sedimentology and paleoecology of the Capitan reef complex and associated rocks. Their publications addressed the stratigraphy, petrology, and paleontology of formations within the Delaware Basin, basin margin, and shelf areas based on surface and subsurface data.

King et al. (1955) addressed correlation problems and aspects of Guadalupian and Leonardian sedimentation associated with platform to basinal facies within the Guadalupe, Brokeoff, and Delaware Mountains. The Brokeoff and Delaware Mountains are located immediately southwest and south of the Guadalupe Mountains, respectively.

The origin and distribution of the Delaware Mountain Group sandstones was interpreted by Hull (1957) to be related to cyclicity controlled by rates of relative subsidence. Hull proposed that sand was transported to the basin as turbidites when relative subsidence ceased.

Galley (1958) discussed the geologic history of the Permian Basin and the relationship among the occurrence of oil, sedimentary environments, hydrology, and tectonic setting. Galley also speculated about the origin and migration of oil and gas, and suggested that highly organic marine argillaceous rocks, rather than pure carbonate rocks, were the principal sources of hydrocarbons.

Boyd (1958) determined the facies relationships and established accurate correlations between Permian basin and shelf formations of pre-Capitan age in the central Guadalupe Mountains region. He interpreted the shelf and shelf-margin sedimentary environments, analyzed the diagenesis of the shelf sediments, and produced detailed geologic maps.

Pray (1959, 1961) presented a detailed analysis of Paleozoic rocks of the Sacramento Mountains Escarpment where he described the geomorphology, stratigraphy, and structural features and provided type sections for the different formations. Pray also discussed the tectonic evolution of the area.

Through detailed geologic maps of the Guadalupe Mountain area, Hayes (1964) attempted to clarify the stratigraphic relationships among platform, shelf margin (reef complex), and basinal formations and to address stratigraphic nomenclature issues.

Models for shelf deposition and cyclicity of Guadalupian strata in the Permian Basin were proposed by Silver and Todd (1969), Meissner (1967, 1972), and Dunham (1972). Dunham (1972) presented three alternative hypotheses for the morphology of the Capitan shelf margin which included barrier reef, uninterrupted slope, and the marginal mound hypotheses. Based on his work, Dunham favored the marginal mound environment. Silver and Todd (1969) and Meissner (1967, 1972) introduced the cyclic sedimentation concept for Middle Permian strata by relating sequences of carbonate and clastic units to variations in relative sea-level. These workers suggested that eustasy was responsible for basinwide depositional cycles and interpreted the lower San Andres Formation to be the result of a long-term transgression. The middle and upper San Andres were considered to be the product of a long-term regression. Meissner (1972), like Hull (1957), also interpreted the Brushy Canyon and lower Cherry Canyon sandstones in the Delaware Basin to be the result of multiple episodes of siliciclastic bypass across the upper San Andres shelf.

Kinney (1969) and Kottlowski (1969) presented a regional perspective of both surface and subsurface characteristics and hydrocarbon potential of the San Andres Formation within southern New Mexico. Kottlowski (1969) presented a detailed description of the San Andres type section and its relationship to the regional setting.

Pray and Esteban (1977) contributed a facies study of the shelf and shelf-edge deposits of the Permian Reef Complex of the Guadalupe Mountains. Their work focused on three major interpretive problems of the Guadalupain shelf and shelf edge that included genesis of the Capitan-massive, pisolite facies, and inner shelf carbonate-evaporite facies transition. Hurley (1978) verified that Dunham’s (1972) marginal mound hypothesis applied to the lower Capitan equivalent section.

Sarg and Lehmann (1986) refined the regional sequence-stratigraphic framework by identifying and correlating three third-order depositional sequences for the San Andres and overlying Grayburg Formations. They interpreted and correlated two San Andres sequences (composite sequences) in the subsurface from a regional 2-D seismic line with two San Andres sequences in Last Chance Canyon. Sarg and Lehmann (1986) interpreted the San Andres/Grayburg contact to be a regionally extensive subaerial unconformity (type-1 sequence boundary) caused by a relative sea level fall.

Sonnenfeld (1991a, b) and Sonnenfeld and Cross (1993) conducted a detailed sequence-stratigraphic analysis of the San Andres Formation and Cherry Canyon Sandstone tongue at Last Chance Canyon. They applied sequence-stratigraphic concepts to delineate fourth and fifth-order depositional sequences within third-order, seismic-scale sequences identified by Sarg and Lehmann (1986).

Senger et al. (1991), Harris et al. (1993), Hovorka et al. (1993), Kerans and Ruppel (1994), Kerans et al. (1994), and Kerans and Fitchen (1995) conducted detailed stratigraphic studies of the San Andres within the Guadalupe Mountains (especially along the Algerita escarpment) and also addressed the impact of lateral and vertical facies and diagenetic variations on permeability and porosity. Through detailed reservoir studies, Hild (1986), Ruppel and Cander (1988), Ruppel (1990), Major et al. (1990), Major and Holtz (1990), and Lucia et al. (1992a) also emphasized the control of cyclicity, facies variations, and diagenesis on porosity and permeability distributions within San Andres reservoirs. Tyler et al. (1990, 1992) observed that rapid lateral facies changes together with highly-cyclic shoaling sequences resulted in pronounced permeability variations both laterally and vertically, and that significant potential existed for additional recovery from San Andres and Grayburg reservoirs.

According to McNeal (1965), the Permian Basin is hydrodynamically active, and under hydrodynamic conditions the relative permeability to gas or oil may be less than under hydrostatic conditions, such that a greater percentage of water is produced with the oil or gas.

Fractures and karst-related features, including small caves, are characteristic of many San Andres reservoirs (Craig, 1988; Tinker et al., 1995), and add complexity to these heterogeneous carbonates.

Landes (1970) summarized early work and statistics on Vacuum Field. A detailed investigation was conducted by Purves (1990) wherein he described lithological and petrophysical aspects of the San Andres and Grayburg Formations.

Recently several studies have been carried out in Vacuum Field (Central Vacuum Unit) by students at the Colorado School of Mines. Capello de Passalacqua (1995) interpreted four fourth-order sequences within the San Andres reservoir interval, described depositional facies and effects of diagenesis, measured shear-wave and compressional-wave velocities of core samples, and concluded that Vp/Vs (ratio of compressional-wave and shear-wave velocities) was sensitive to the saturated bulk moduli. Swanson (1996) modeled the zero-offset seismic response of a 3-D model of Vacuum Field to show that changes in seismic velocities and reflection coefficients due to CO₂ injection could be detectable. Voorhies (1996) processed and interpreted a VSP (vertical seismic profile) from well CVU-200 in order to correlate the formations from that well with the three-dimensional surface seismic data. He also analyzed shear-wave polarization with depth (anisotropy), and estimated seismic attenuation with depth to detect changes in lithology and fluid saturations. Roche (1997) processed and interpreted two three-dimensional, multicomponent seismic surveys, acquired pre- and post-CO₂ injection, and established and applied a processing flow for repeatability when using time-lapse, multicomponent seismic data.

Talley (1997) interpreted the structural framework within the Vacuum area (12 square miles) using a regional compressional-wave 3-D seismic survey (Maljamar 3-D). He also established relationships among multicomponent attributes from surface seismic and reservoir properties.

Adams (1997) interpreted the sequence stratigraphic framework, sedimentology, and diagenetic history of the San Andres Formation using data from two cores. Scuta (1997) identified bypassed pay within the San Andres using time-lapse logging with resistivity logs, time-lapse injectivity profiles, and porosity-resistivity overlays. He interpreted fractures and their orientations using data from a Formation MicroImager log, and built a three-dimensional geologic model for the reservoir interval using well data. DeVault (1997) developed and applied a technique for multicomponent AVO (amplitude variation with offset) analysis to estimate porosity, lithology, and fracture density. Mattocks (1998) processed and interpreted a VSP from well CVU-200 and concluded that the anisotropy model at Vacuum Field is orthorhombic with significant polar velocity variation and weak azimuthal variation.

Blaylock (1999a) identified relationships among multicomponent seismic attributes and reservoir parameters. He related lateral variations in porosity and mineralogy to changes in Vp/Vs. Galarraga (1999) interpreted the structural framework (115 square-mile area) for the Paleozoic section using a regional compressional-wave seismic survey (Maljamar 3-D). Bard (1999) conducted flow simulation using the geologic models of the Grayburg-San Andres reservoir to predict changes in fluid saturations and production under CO₂ injection. Duranti (1997, 1999) interpreted faults in the vicinity of the time-lapse seismic area using compressional-wave seismic volumes.

Cabrera-Garzón (1999) used two-dimensional and three-dimensional geostatistical techniques with multicomponent three-dimensional seismic attributes to estimate porosity and permeability between wells. Michaud (1999) processed and interpreted a multi-geophone VSP (vertical seismic profile) from well WS-15 to correlate the formations from that well with the three-dimensional surface seismic data. She also analyzed shear-wave polarization and seismic attenuation with depth. Lorenzen (1999) used one-dimensional shear-wave seismic inversion with VSP data to estimate acoustic impedance (product of velocity and density), and extended shear-wave inversion to three dimensions using the surface seismic data. Galikeev (1999) used prestack depth migration on the multicomponent surface seismic data to improve the seismic image. Mendez-Hernandez (1999) evaluated issues related to near-surface statics to improve the surface seismic processing. Maher (1999) attempted to locate significant fault planes and fractures within the reservoir interval using passive-seismic monitoring.

The major contributions of this research involve techniques for improving geologic models for use in flow simulation. Two areas that were addressed included the geologic model framework, and use of multicomponent seismic data (vs. compressional-wave seismic data alone) for reservoir characterization and for three-dimensional estimation of reservoir parameters. Since it was not possible to accurately correlate depositional cycles within the model area, a petrophysical-based method was used to define reservoir flow units within high-frequency sequences. This technique resulted in 18 flow units that became the "layers" for the three-dimensional geologic model. Each flow unit was roughly equivalent to a depositional cycle or cycle set. A significant advantage of this technique is that, in most cases, the geologic model does not need to be upscaled prior to flow simulation. In this case, the model honored the geologic details and consisted of a reasonable number of cells for direct use in flow simulation.

Faults were incorporated into the framework of the geologic model that were interpreted from compressional-wave seismic data and maps generated from well data. In addition, a technique for identifying subtle faults with small displacements (e.g. < 25 feet) was demonstrated using data from horizontal wells and adjacent vertical wells.

Relationships identified among multicomponent seismic attributes and reservoir parameters were used with two-dimensional and three-dimensional estimation techniques to populate the layers of the geologic model with porosity and permeability. In addition, these techniques allowed the estimation of reservoir properties in layers that were below seismic resolution. Although use of compressional-wave seismic data for parameter estimation is common, the use of multicomponent seismic attributes in geologic modeling is a significant contribution that has provided additional information about the reservoir that was not available with compressional-wave seismic data alone.