23: Lithological and Petrophysical Controls on Production of the Mississippian Limestone, Northeastern Woods County, Oklahoma
Published:January 01, 2019
- PDF LinkChapter PDF
Katherine M. Lindzey, Matthew J. Pranter, Kurt J. Marfurt, 2019. "Lithological and Petrophysical Controls on Production of the Mississippian Limestone, Northeastern Woods County, Oklahoma", Mississippian Reservoirs of the Midcontinent, G. Michael Grammer, Jay M. Gregg, James Puckette, Priyank Jaiswal, S. J. Mazzullo, Matthew J. Pranter, Robert H. Goldstein
Download citation file:
Mississippian carbonates of northern Oklahoma were deposited on the Anadarko shelf (ramp) as several shallowing-upward sequences. In Woods County, Oklahoma, the Mississippian ranges in thickness from 350 ft (105 m) to the south to as little as 100 ft (30 m) to the north due to uplift and erosion. Lithologies observed in core are chert conglomerate, tripolitic chert (tripolite), dense chert, chert-rich limestone, dense limestone, and shale-rich limestone.
To evaluate the spatial distribution of Mississippian lithologies and petrophysical properties, and to explore the controls on production, this study integrates 3-D seismic with core and well-log data. As a constraint for 3-D lithology modeling, lithology logs were estimated using a neural-network approach with core and log data resulting in 65.1% accuracy. A P-impedance volume from seismic inversion was used to constrain the spatial distribution of tripolite in the model, the main reservoir lithology. Lithology-constrained 3-D porosity and water saturation models show that tripolite is the most porous and heterogeneous lithology. Comparing lithology, porosity, and water saturation models to production data illustrates that production from vertical wells is primarily controlled by porous tripolite distribution, whereas horizontal wells produce from both tripolite and chert-rich limestones and are most sensitive to water saturation variations.
This study explores the lithological and petrophysical controls on production from the Mississippian limestone in northeastern Woods County, Oklahoma (Figure 1). As described herein, references to the Mississippian limestone include the Mississippian-age deposits that are primarily carbonate and silica-rich and that exist above the Woodford Shale (the Mississippian-age portion of the upper Woodford Shale is not included). The Mississippian carbonates and silica-rich deposits of this interval have also previously been referred to informally as the “Mississippi lime.” Despite the name, much of the production is from silicate-rich tripolitic chert zones that commonly consist of diagenetically altered silicate sponge spicules and silica from other sources, rather than from the carbonates themselves. Rogers et al. (1995) defined “chat” as the Mississippian Osage chert interval. It was coined “chat” by the drillers because of the chattering noise and how the bit bounced during the drilling process. In the study area, less tripolite as compared to other Mississippian limestone fields exists, and the tripolite that is present is generally less porous.
Detailed field studies throughout Kansas and Oklahoma have examined the primary controls on production and discuss the major obstacles in drilling (Camargo and Fons, 1966; Rogers et al., 1995; Montgomery et al., 1998; Watney et al., 2001; Turnini, 2015). Rogers et al. (1995) and Watney et al. (2001) observed that fractures had limited impact on production as hydrocarbons were derived from the chert pore spaces and waterflooding provided limited recovery enhancement while Montgomery et al. (1998) established that fractures had a large impact on production in lower permeability fields. In general, Camargo and Fons (1966) note that fracturing is significant for reservoirs in which average porosity is below 5%.
There has been extensive research proposing depositional models for the Mississippian of the midcontinent from outcrop, core, and log data. This includes paleoreconstruction of the Osagean (Gutschick and Sandberg, 1983) and a proposal by Rogers et al. (1995) that the locations of sponge bioherms controlled the present-day distribution of reservoir facies. In the Mississippian, diagenesis is at least as important as deposition in shaping the lithologies observed today. Knauth (1979) presented a model for the method of chert replacement of the original carbonate fabric where the mixing of meteoric and sea water in the pore spaces creates pore fluids undersaturated with respect to calcite and supersaturated with respect to silica such that the original carbonate fabric is dissolved and replaced by chert precipitate. Rogers et al. (1995) expanded on this model as it relates specifically to the silicate sponges of the Mississippian.
Studies have also focused on the regional distribution of Mississippian facies. Rogers (2001) studied well logs and completion records from nearly 7000 wells in north-central Oklahoma. Rogers (2001) observed that the Mississippian interval south of far-northern Oklahoma lacked much of the chert that was of interest for development. Rogers (2001) hypothesizes that this lack of silica is due to a gradual regional dip toward the south so that regions farther south were in deep enough water to avoid the erosion and weathering that led to silica replacement of the carbonates.
Due to the cyclicity of the lithofacies in repeated shallowing-upward sequences, Watney et al. (2001) noted that it is impossible to determine which specific interval a well-log signature pertains to without correlating all cycles upward from the underlying Woodford Shale. Through studying production data, well logs, core, and conversations with those experienced in drilling the region, Rogers (2001) observed that the Mississippian limestone has an abrupt negative deflection on the SP log and lower gamma-ray response where the rock changes from Pennsylvanian shales to limestone or chert. Traditionally, in the silicate-rich intervals, a minimum of 25–30% porosity and water saturation less than 80% is necessary for an economic reservoir (Rogers, 2001).
The stratigraphy of the Mississippian limestone poses a challenge for seismic interpreters, as the seismic facies are also linked to diagenetic and structural features, in addition to the stratigraphic architecture. Montgomery et al. (1998) identified the bases of silica-rich layers in south–central Kansas as strong positive reflectors on seismic data due to their increased impedance relative to the underlying less-altered limestone. Dowdell et al. (2013b) explored the use of seismic attributes in predicting lithologies within the Mississippian limestone. They observed that tripolite was associated with low impedance and high porosity and predicted that tripolite would be most abundant at the intersection of low impedance and high curvature. Curvature is frequently associated with fractures in the rock, which would present a path for pore fluids to diagenetically alter the strata. Dowdell et al. (2013a) used a neural-network approach to generate a 3-D density volume from the seismic and well-log data and observed that the 3-D density volume was the best method to predict locations of tripolitic chert.
Costello et al. (2014) generated 3-D lithology, porosity, and permeability models for the Mississippian limestone in Woods and Alfalfa counties, Oklahoma, that were constrained to core and well-log data. To expand upon their work, this study addresses the lithological and petrophysical controls on Mississippian production by integrating 3-D seismic data and attributes (e.g., acoustic impedance) with core, well-logs and neural-network-derived lithology logs.
The study area is in northeastern Woods County, Oklahoma (Figure 1), just south of Hardtner field in Barber County, Kansas—one of the most prolific gas fields in Kansas. Data for this study include a full suite of open-hole logs from 32 vertical wells, gamma-ray MWD logs from 50 horizontal wells, and a 70 mi2 (181 km2) 3-D time-migrated seismic volume. Core from one well, the Chesapeake Energy 1–14 Bann, is within the seismic survey area and core from a second well, the Chesapeake Energy 1–34H Albus, is outside the survey area. The Chesapeake Energy 1–14 Bann also has 111 thin sections and petrophysical measurements from core plugs (Figure 2) that were analyzed.
Key lithologies were identified from core descriptions and estimated in noncored wells using an artificial neural network (ANN). A stratigraphic and structural framework for the model was produced by dividing the Mississippian limestone into stratigraphic intervals based the correlation of lithology logs, well-log signatures, and seismic data. Three-dimensional lithology models were generated using the well control and seismic-derived probability maps for lithology. Petrophysical properties were populated within the lithologies to create porosity and water saturation models. Finally, the model results and cumulative production data were analyzed to evaluate the lithological and petrophysical controls on Mississippian limestone production in the area.
The deposition of the Mississippian limestone of the midcontinent occurred from approximately 359 to 318 Ma. During this time, the midcontinent region that is now northern Oklahoma was about 20° south of the equator and moving farther south as Laurasia and Gondwana collided. The modern-day midcontinent was covered by a warm, shallow sea (Elebiju et al., 2011) and a shelf margin (ramp) gradually sloped seaward from these relatively shallow waters (Mazzullo, 2011) to a deeper seaway to the south that paralleled the converging plate boundaries (Scotese, 1999).
Paralleling these shelf margins, sponge-microbe bioherms formed elongated mounds below storm wavebase and produced abundant SiO2 spicules, which led to the formation of spicule-rich wackestones and packstones (Watney et al., 2001). Surrounding these mounds, the shelf region was a highly productive carbonate factory, producing carbonate mud as well as skeletal fragments (especially crinoids and bryozoans). Therefore, deposition produced oval-shaped mounds of sponge spicules that grade laterally in all directions into skeletal wackestones and lime mudstones (Rogers et al., 1995). It is theorized that sediment accumulation was made possible by basement block movements that led to the development of the shelf slope (Watney et al., 2001). Farther south, the Mississippian limestone lacks the productive tripolitic chert beds, most likely because the deeper waters lacked both the sponge-microbe bioherms and the structurally down-dip facies underwent less secondary diagenesis.
Due to the presence of graded beds and sponge spicules that were transported down-dip from their bioherms as well as fossiliferous limestone clasts that are associated with the spiculite, it is apparent that gravity-flow processes were locally important with regards to sediment distribution and structures (Montgomery et al., 1998). Similarly, fossils within the spiculite beds are likely to have been transported from up-dip after erosion in a high-energy environment such as above wavebase (Rogers and Longman, 2001).
The Mississippian strata span a lower-order transgressive-regressive cycle bounded by unconformity surfaces above and below (Manger, 2011). Within this cycle, four stacked higher-order transgressive-regressive cycles are identified as shallowing-upward cycles (parasequences) with prograding facies (Figure 3). Each of these sequences may be capped by porous, spiculitic rocks (Watney et al., 2001). Sufficiently large drops in relative sea level would have led to subaerial exposure of the sponge mounds as small islands of weathered rock (Rogers, 2001), thus allowing for weathering and diagenetic alteration that left an unconformity topping each parasequence above crinoidal, lime-grainstone lithofacies. Within each shallowing-upward cycle, carbonate rocks sometimes display increasing sponge-spicule content with decreasing depth. This steadily increasing spicule content eventually grades into the bioclastic wackestones and grainstones of the shelf (ramp). In addition to these higher-frequency trends, in areas the overall sponge spicule content has been observed to increase upward throughout the Mississippian, accompanied by increasing cycle thickness (Watney et al., 2001). The Mississippian strata thin abruptly to the north in Kansas because of significant uplift and erosion rather than subsidence that created accommodation space to the south (Watney et al., 2001).
After deposition, the Mississippian limestone underwent extensive diagenetic alteration. First, pore waters redistributed the silica. Siliceous volcanic ash and some macrofossils were dissolved leaving extensive microscale porosity and thus tripolite (Montgomery et al., 1998). The dissolved silica precipitated in pore spaces and partially or completely replaced some carbonate fossils. The preservation of fossil microscopic structure suggests that this replacement occurred molecule by molecule (Rogers, 2001). This initial diagenesis mostly occurred just below the sediment-water interface before complete lithification (Manger, 2011). Additional sources of silica may have resulted from volcanic ash related to the converging plates to the south of the midcontinent as well as from hydrothermal emanations (Rogers, 2001). The structural deformation to the south is hypothesized to have driven three pulses of hot brines out of the Anadarko basin that led to chert precipitation, baroque dolomite precipitation, and oil migration followed by calcite cementation (Goldstein and King, 2014).
After lithification was completed, Late Mississippian–Early Pennsylvanian uplift led to a decrease in relative sea level that allowed significant erosion. This created what is known as the sub-Pennsylvanian unconformity. The Osagean section has been completely eroded in the highest topographic regions of the Mississippian at the top of features such as the Nemaha uplift. This uplift exposed the strata to a second round of diagenesis through vadose zone migration of meteoric waters. In some sections, sponge spicules were selectively dissolved to create moldic porosity (Montgomery et al., 1998). In many places where the carbonates had been partially replaced by silica, this second stage of diagenesis dissolved the rest of the carbonate leaving behind extensive secondary porosity in the form of vugs (Rogers, 2001). Tripolitic chert with the most extensive porosity, and thus the best reservoir characteristics, commonly resides 10–20 ft (3–6 m) below the sub-Pennsylvanian unconformity where it was exposed to the most extensive secondary diagenesis (Montgomery et al., 1998). The two stages of diagenesis created two distinct porosity types: primary microporosity and secondary vuggy to moldic porosity. This leads to a dual-porosity system and a two-stage depletion of reservoirs (Rogers, 2001). For the most part, the carbonates that did not undergo significant replacement by silica also lacked the secondary diagenesis that created additional porosity. Therefore, sections of highly porous and permeable tripolite and spiculite grade into impermeable limestones and dolomites creating an environment ideal for trapping hydrocarbons (Montgomery et al., 1998).
PETROPHYSICAL ANALYSIS OF LITHOLOGY
The major lithologies of the Mississippian limestone were identified through detailed descriptions of cores (208.5 ft [63.6 m], total footage) from the Chesapeake 1–14 Bann (1–14 Bann) and Chesapeake 1–34H Albus (1–34H Albus) wells. Using a suite of open-hole logs, the lithologies were estimated for the Mississippian interval in 32 noncored vertical wells within the study area through the application of an ANN. The core descriptions include observations of lithology, sedimentary structures, fractures, and diagenetic alteration. Thin sections and photomicrographs of the 1–14 Bann core provided detailed information about grain composition, pore types, and fractures. In addition, petrophysical information for 111 core plugs from the 1–14 Bann well included porosity, permeability (ambient and at reservoir conditions), fluid saturation, grain density, total organic carbon, and X-ray diffraction-based mineral percentages.
The major Mississippian lithologies include (1) shale, (2) chert conglomerate, (3) tripolitic chert, (4) dense chert, (5) altered chert-rich limestone, (6) dense limestone, and (7) shale-rich limestone (Figure 4). Shale-rich limestone was interpreted to be deeper in the Mississippian interval based on well-log signatures that indicate an intermediate lithology between pure carbonate and shale end members.
At the top of both cores is red and green Pennsylvanian shales. In the 1–14 Bann well, the shales are red, poorly lithified, and have an abundance of root traces. In the 1–34H Albus well, the shales range in color from red to green and are more consolidated than in the 1–14 Bann well, and exhibit numerous root traces. The green shales within the 1–34H Albus well are likely glauconitic. The shale exhibits relatively high gamma-ray values (VShale >70). The shale is believed to have been deposited during the early Pennsylvanian following the uplift and subaerial exposure that led to the pre-Pennsylvanian unconformity.
Below the Pennsylvanian shales is an extremely heterogeneous chert conglomerate. Clasts are subangular to rounded and sand to cobble sized. Some of the larger clasts display extensive alteration with high porosity and extensive oil staining, while much of the matrix is a green, glauconitic siltstone with no evidence of oil staining. Thin-section analysis (Figure 5A) reveals that much of the chert is merely replacing the original carbonate fabric of skeletal packstone. Porosity within the clasts appears as moldic, intercrystalline, and vuggy. Other clasts show almost complete silica replacement of the carbonate fabric but lack the secondary enhanced porosity and oil staining except as a thin rind around the rim. Petrophysical analysis from the core provides a porosity range of 1–19% with an average of 9.4%, permeability of 0–100 mD with an average of 18.9 mD, and oil saturation of 3–74% with an average of 18.9%. The conglomerate is interpreted to be a result of the uplift and subaerial exposure during the early Pennsylvanian, which weathered and transported these chert clasts.
The most economically significant Mississippian lithology is tripolitic chert consisting of highly altered beds of nodular chert within a porous, bioturbated limestone and silica matrix. The chert nodules appear as beds and replacement within the burrows and contain abundant high-angle fractures, which are mostly calcite cemented. This lithology shows the highest degree of oil staining. Thin-section analysis (Figure 5B) reveals that the highest porosity occurs at the chert–limestone interface where diagenesis has led to extensive molds (especially those from the dissolution of the original sponge spicules) and vugs. Petrophysical analysis from the core provides a porosity range of 4–23% with an average of 12.4%, permeability of 0–5 mD with an average of 0.99 mD, and oil saturation of 6–43% with an average of 20.8%. The tripolite is interpreted to be an in situ product of uplift and subaerial exposure, where meteoric pore fluids supersaturated with respect to silica dissolved the original calcite fabric to replace it with silica. The abundance of spicule molds also indicates the proximity of a sponge bioherm as a primary source for sediment. In the 1–34H Albus core, tripolite is concentrated directly below the chert conglomerate, while the 1–14 Bann core has tripolite present throughout the uppermost 100 ft (30 m) of core suggesting alteration influenced from other, higher-order sea-level declines.
Due to the limited and localized presence of the chert conglomerate and its similar reservoir properties to the tripolitic chert, tripolitic chert, and chert conglomerate were combined into the tripolite lithology for 3-D reservoir modeling that is discussed in a later section.
Dense chert frequently appears as bands between a few inches and a few feet thick and displays almost complete replacement of the original carbonate fabric with silica, but it lacks the secondary porosity observed in the tripolite and chert conglomerate. Original fabric, including bioturbation and fossils, is preserved with an abundance of sealed and open high-angle fractures throughout. Apart from the fractures, the dense chert shows minimal oil staining. A photomicrograph (Figure 5C) reveals scattered dolomite rhombs and minimal porosity. Petrophysical analysis from the core provides a porosity range of 6–10.9% porosity with an average of 8.6%, an average permeability of 0.0006 mD, and an oil saturation of 7–30% with an average of 17.62%. This relatively high porosity and extremely low permeability is likely the result of “microintercrystalline pores between micron-sized quartz spherules” as noted by Rogers et al. (1995).
The least altered lithology present in the cores is dense limestone which appears only in the 1–14 Bann core. The grains range from pure micritic calcite to beds of a skeletal packstone or wackestone where the skeletal fragments predominantly belong to crinoids, brachiopods, and bryozoans. There is some chert replacement of original limestone fabric, but this is accompanied by limited porosity and permeability enhancement. The altered chert-rich limestone reveals slight oil staining under UV light. Thin section analysis (Figure 5D) shows a range from micritic limestone, sometimes accompanied by abundant dolomite rhombs, to a skeletal packstone or wackestone. Where chert is present, it appears to replace scattered carbonate grains rather than forming nodules and beds. Petrophysical analysis from the core provides a porosity range of 1–11% with an average of 5.89%, permeability of 0–0.08 mD with an average of 0.009 mD, and oil saturation of 0–39% with an average of 23.5%. Most of the fabric shows significant bioturbation from horizontal burrows suggesting deposition below fair weather wave base.
The most abundant lithology observed in both cores is an altered chert-rich limestone. Dense chert replacement of the original limestone matrix occurs as nodules within the bioturbated intervals and as thin beds. Open and calcite-sealed high-angle fractures are present within the chert but rare within the more carbonate-rich sections. This lithology assignment represents a combination of limestone with uniform chert replacement of the bioturbated areas and alternating dense limestone and chert at too high a frequency to differentiate on well logs and both examples of this facies display moderate oil staining. Enhanced porosity can occur at the interface between the chert and limestone with moldic pores where the original carbonate grains have been dissolved but chert replacement has yet to occur (Figure 5E). Petrophysical analysis from the core provides a porosity range of 2–13% with an average of 6.69%, permeability of 0–0.534 mD with an average of 0.104 mD, and oil saturation of 0–36% with an average of 20.59%. While this chert-rich limestone facies has similar porosity to the dense limestone, the increased permeability due to enhanced diagenesis supports the assignment of separate lithologies. Furthermore, the increased abundance of silica will enhance susceptibility to fracturing during reservoir stimulation.
Throughout the chert-rich sections of core, there is the development of vuggy porosity at the intersections of fractures. In addition, beds of millimeter-scale pyrite nodules are present within the shale, tripolite, and chert-rich limestone facies, while some stylolites appear within the carbonate and tripolite sections. Overall, the 1–34H Albus core shows significantly enhanced diagenetic overprint relative to the 1–14H Bann core despite being located just 8 mi (12.9 km) away clearly demonstrating the short distances over which alteration occurs.
In addition, the presence of a shale-rich limestone is inferred at the base of the Mississippian section in an obvious shallowing-upward sequence. Although there are no core data to confirm this lithology assignment, the intermediate nature of the gamma-ray log (values from 25–115 API) combined with the stratigraphic location of this interval make it clear that it is neither pure limestone nor shale and is instead an intermediate lithology.
Initial attempts to estimate lithologies in noncored wells using log-based cutoffs yielded inconclusive results. This is likely due to the limited range of variation observed in the gamma-ray and resistivity logs. Therefore, a modified ANN approach was used to classify lithologies in noncored wells. Both the 1–14 Bann and 1–34H Albus cores were used to train the ANN to identify tripolite, dense chert, limestone, and chert-rich limestone using the logs for bulk density, PE, deep resistivity, and neutron porosity. Cross validation was set at 50% with a 2% error limit. However, when shale was included in the model, the ANN had difficulty identifying this lithology likely due to the scarcity of shale in the cored interval. To account for this, a cutoff was applied to the neural-network-produced lithology logs whereby shale or shaly limestone (for the lower sections of the Mississippian interval) was defined as VShale (generated from the gamma ray) values greater than 0.7.
Of 208.5 ft (63.6 m) of core, the estimated lithology correctly matches the core lithology for 135.8 ft (41.4 m), which provides 65.1% accuracy (Figure 3). The ANN approach was then applied to the remaining 31 vertical wells in the study. However, lithology assignments in the lower two-thirds of the Mississippian interval should be considered with caution due to the lack of core data below the top 150 ft (45.7 m).
STRATIGRAPHIC AND STRUCTURAL FRAMEWORK
The stratigraphy and structural characteristics of the Mississippian were interpreted and a stratigraphic and structural framework was developed and represented as a 3-D reservoir model grid for lithology and petrophysical modeling. The Mississippian limestone was divided into four stratigraphic intervals, A–D, based on the gamma ray, deep-resistivity, and estimated lithology logs. The basal interval, or Mississippian A, is characterized by an overall coarsening-upward sequence with decreasing gamma-ray values and slightly increasing resistivity up to a clean limestone. The top of this interval is marked by a minor flooding surface that is visible in only some of the wells. Additionally, many of the wells show one or more clean limestone beds topped by a flooding surface for this interval, perhaps because of mass transport from the up-dip shelf margin. The interval thickness ranges from 120 to 250 ft (37–76 m) with an average of 175 ft (53 m).
The Mississippian B is typically a clean carbonate with gamma-ray values below 25 API and moderate but variable resistivity ranging from 2 to 20 ohm m (6.5–65.5 ohm ft). This interval is capped by a flooding surface in some, but not all, wells and it has been partially truncated by erosion to the north. The thickness ranges from 85 to 190 ft (26–58 m) with an average of 130 ft (40 m).
The Mississippian C is an interval of noticeably lower resistivity (typically <10 ohm m [32.8 ohm ft] in the lower half and <20 ohm m [65.5 ohm ft] in the upper half) and uniformly clean limestone bounded above and below by minor flooding surfaces observed in the gamma-ray log. This interval exhibits a majority of tripolite in the 1–14 Bann core consistent with the typically low resistivity of tripolite. This interval has the most variable thickness ranging from completely absent to the north to 78 ft (24 m) in the south.
The Mississippian D is the uppermost interval and is characterized by the cleanest gamma-ray readings (below 15 API) and highest resistivity (40–100 ohm m [131–328 ohm ft). This interval represents the lowest stage of sea level preserved in deposition before the erosion of the overlying pre-Pennsylvanian unconformity with significantly higher gamma-ray readings from chert conglomerate and Pennsylvanian shales. It is the thinnest interval ranging from completely absent to 47 ft (14 m).
From south to north, the entire Mississippian interval thins with the top two intervals (C and D) eventually being completely truncated in the northern half of the study area (Figure 6). Wells with sonic logs (N = 18) were tied to the time-migrated seismic volume and the seismic reflectors that correspond to the top and base of the Mississippian interval were identified (Figure 6). The top of the Mississippian exhibits a broad peak, while the top of the Woodford Shale manifests a very continuous and high amplitude, well-defined trough. The seismic data are not adequate to resolve the Mississippian A–D intervals. While the well-log-based stratigraphy indicates truncation of beds by the Mississippian–Pennsylvanian unconformity, there are no truncated reflectors at the top of the Mississippian on the seismic volume. This is likely a combination of limited seismic resolution and the alteration at the top of the Mississippian interval. Roy et al. (2013) suggest that the diagenetic overprint resulting from subaerial exposure masks the original depositional bedding and alters the seismic reflections.
The stratigraphic variability of lithologies for each zone is evident by the vertical lithology proportion curve (Figure 7) based on the lithology logs for all vertical wells. The base of the lowermost Mississippian A zone indicates a high but thin percentage of limestone topped by a majority of shale-rich limestone that transitions into limestone and chert-rich limestone stratigraphically upward. The Mississippian B zone features predominantly limestone and chert-rich limestone and is differentiated from the other zones by the relatively high abundance of dense chert. The Mississippian C interval shows the highest abundance of tripolite while the majority of the rocks are still limestone and chert-rich limestone. The uppermost Mississippian D zone shows a general trend of increasing shale and tripolite upward associated with a decrease in the abundance of limestone and chert-rich limestone. It is likely that most of the strata in the Mississippian B interval are located too far below the pre-Pennsylvanian unconformity to encounter the extreme diagenetic alteration responsible for generating the tripolite present in the overlying Mississippian C and D intervals.
At least two significant normal faults are observed in the area (Figure 8). One fault (Fault A) trends roughly northeast to southwest across the middle of the survey with the downthrown block to the south. The maximum offset is approximately 100 ft (30 m); therefore, this fault was incorporated into the 3-D model grid. A second, smaller, normal fault (Fault B) trends north–south in the far western part of the study area and is truncated to the south by the larger northeast–southwest fault. The smaller fault has a maximum vertical offset of less than 50 ft (15.3 m) and was not included in the 3-D grid framework. Both faults appear to be syndepositional with greater offsets observed at the Woodford Shale reflector than at the Mississippian reflector. The downthrown block of the northeast–southwest fault shows a broader peak in the seismic reflection at the top of the Mississippian immediately south of the fault compared to the upthrown block (Figure 6). This is most likely the result of greater abundance of chert conglomerate transported from the upthrown block down to the downthrown block or preserved thin-bedded units on the downthrown block that were not eroded.
The interpreted stratigraphic and structural framework based on wells and seismic data was used to create a 3-D model grid for the Mississippian (Figure 9). The Mississippian and Woodford seismic horizons were converted to depth using average velocities computed from well tops, and the surfaces define the top and base of the 3-D model grid (3-D stratigraphic and structural framework). Internal zones for the 3-D grid were generated from Mississippian A–D surfaces mapped using well tops. The 3-D grid consists of cells that are 200 × 200 ft (61 m × 61 m) aerially. The Mississippian A interval consists of 150 proportional layers. The Mississippian B–D intervals have layers that are 2 ft (0.6 m) thick and follow the base of their respective zones thus allowing for truncation by the upper zone boundary in a representation of the local unconformities that likely top each of these three upper zones. The resulting 3-D framework contains 31 million cells and shows truncation of Mississippian D and C zones with the Mississippian B subcropping toward the northern extent of the model. The resulting subcrop of the 3-D grid shows Meramecian-age rocks subcropping to the south and Osagean-age rocks to the north where the younger rocks have been removed (Figure 9).
SPATIAL DISTRIBUTION OF MISSISSIPPIAN LITHOLOGIES
3-D lithology models were generated to show the spatial distribution of tripolite, dense chert, chert-rich limestone, dense limestone, shale-rich limestone, and shale and were constrained to the lithology logs and lithology estimates based on 3-D seismic data. Dowdell et al. (2013b) observed a clear relationship between low impedance and the Osage A tripolitic chert zone whereby the higher porosity tripolite related to lower impedance relative to dense limestone. Turnini (2015) discovered a similar relationship between low impedance and higher curvature being associated with a relative increase in tripolite abundance for the Mississippian limestone of Kay and Noble counties, Oklahoma. Dowdell et al. (2013b) also noted a connection between reduced density and higher probability of tripolite and determined that a neural-network density estimation from the seismic data was most useful for providing a tripolite probability volume.
Like previous studies, an inverse relationship between low P-impedance (calculated from open-hole logs), high porosity, and the abundance of tripolitic chert based on the estimated lithology logs was observed for the 31 vertical wells (Figure 10A). However, shale and shale-rich limestone also demonstrate low impedance but somewhat lower porosity than tripolite. The low impedance of shale is likely due to the lower density as compared to the carbonates. Once the shale and shale-rich limestone intervals are removed (Figure 10B), the difference in impedance between tripolite and carbonates is clear. Tripolite is present between P-impedance values of 22,000 and 52,000 ft/s × g/cm3, with an abrupt drop in relative abundance for impedances greater than 40,000 ft/s × g/cm3 (Figure 10). Therefore, anomalously low impedance may be used to predict the likelihood of tripolite only in intervals of carbonates but should not be confused with the ubiquitously low impedance of the shale-rich areas. Fortuitously, tripolite is typically a result of subaerial exposure of sponge bioherms that grow in shallow water, while the shale-rich facies are expected in deeper water.
Similar to Dowdell et al. (2013b), the relationship between low-density and tripolitic chert presence was also examined as an additional tripolite predictor, but the relationship was not as well defined between carbonate and tripolite. Therefore, only an impedance volume was used as a constraint for tripolitic chert in 3-D lithology modeling. The impedance volume was generated from seismic inversion and was used to generate a tripolite probability volume where impedance values greater than 46,000 units were assigned to a probability of 0, impedance values less than 34,000 were assigned to a probability of 0.9, and values in between were scaled between the two end members. The volume was then used to generate maps of average tripolite probability for the Mississippian B–D intervals to be used as horizontal trends for the 3-D lithology modeling. Due to the abundance of shale-rich limestone and generally low impedance of the Mississippian A, this zone was likely deposited in deeper water and thus unlikely to contain significant tripolite, so a probability map was not used for tripolite prediction in this zone.
To constrain the distribution of shale in the 3-D lithology model, a neural-network estimation was used to generate a VShale volume from the gamma-ray logs and inverted seismic volume. From this, average VShale maps were produced for the Mississippian A–D zones. The maps show extremely low average VShale in the Mississippian B and C zones where the logs show clean carbonate and an overall higher average VShale in the Mississippian A zone where the logs show a shale-rich limestone. This suggests that the VShale volume is able to accurately predict locations of higher shale probability. However, the average VShale map for the Mississippian D zone shows higher than expected VShale, possibly a result of the inability to resolve the overlying Pennsylvanian shale from the Upper Mississippian limestone on seismic data. Additionally, the low impedance of the tripolite in this upper interval could also be contributing to the questionable estimation results in this zone. Therefore, the VShale maps for the Mississippian A–C zones were used to guide the distribution of shale in the 3-D lithology model.
The 3-D lithology model for the Mississippian limestone was generated using sequential-indicator simulation (SIS), which honored the upscaled lithology logs, the vertical lithology proportion curve (Figure 7), and variogram parameters. Vertical and horizontal variograms were generated and estimated to set the range for each lithology by zone. For all tripolite and chert lithologies, the major variogram range azimuth was set at 110° or roughly parallel to the present-day subcrop. The tripolite and shale lithologies were constrained to honor probability maps generated from the impedance inversion and neural-network-estimated Vshale, respectively to map the spatial lithology distribution in each zone.
The resulting model has 6.7% tripolite, 42.1% dense limestone, 19.6% chert-rich limestone, 12.5% chert, 17.6% shale-rich limestone, and 1.5% shale (Figure 11). A detailed examination of the lithology model (Figures 11, 12) reveals distinct differences in the lithology abundance of each zone that agrees with the lithology trend of the vertical proportion curve (Figure 7). The lowermost Mississippian A is dominated by shale-rich limestone and the transition to the middle Mississippian B zone is accompanied by an abrupt change to limestone- and chert-rich limestone-dominated strata with some relatively large bodies of chert. The very thin Mississippian C and D intervals have more tripolite relative to the underlying zones. The Mississippian B zone has the greatest thickness of tripolite, but the greatest proportion of tripolite is present in Mississippian C interval (24.5%) (Figure 12) followed by the B (9.4%) and D (8.9%) intervals. In the upper three intervals (B–D), dense limestone and chert-rich limestone are the dominant lithologies. Within these zones, there are overlapping bodies of tripolite and chert up to 5 mi (8 km) long encased within the limestone matrix. These tripolite and chert bodies are most abundant in the Mississippian C and D but are also present in the B. Tripolite does not exist in the lowermost zone, Mississippian A.
Some of the thickest intervals of tripolite are in patches along the major northeast–southwest fault and at the northeast where the seismic volume indicates low coherence (Figure 12). Near the fault and low coherence area, fracture density and fluid flow could be relatively higher compared to surrounding areas, thus enhancing the potential for diagenesis and tripolite formation. However, the area around the western north–south fault lacks significant tripolite presence perhaps due to limited sponge bioherms during the original deposition.
Models of total porosity and water saturation were generated using sequential-Gaussian simulation (SGS) and were constrained to the lithology model, upscaled well logs, and vertical and horizontal ranges estimated from variograms for each property by lithology (Figure 13). The ranges of each petrophysical property were set to half of the range used for the associated lithology, except for tripolite porosity. Due to its heterogeneous nature, the major and minor ranges for tripolite porosity and water saturation were set at 2500 ft (0.8 km) and 2000 ft (0.6 km), respectively. Tripolite exhibits an average porosity of 10.2% and significant porosity heterogeneity (Figure 13). Chert-rich limestone and shale-rich limestone have intermediate values of porosity. The diagenetic process of replacing the original limestone fabric with some chert left enhanced porosity in the chert-rich limestone, and the lower density of shale in the shale-rich limestone is likely contributing to higher total porosity values. Dense limestone and dense chert have the lowest values of porosity. Water saturation values show significantly less heterogeneity between lithologies than porosity, with an average of 44.2%. Tripolite has the greatest variability with a range of 7.2–100% and an average of 42.3% and a general trend of increasing water saturation from southwest to northeast. There are no systematic trends in water saturation for any of the other lithologies. In general, dense limestone has the greatest water saturation averaging 44.9%, while shale has the lowest saturation at 43.5% (Figure 13).
LITHOLOGICAL AND PETROPHYSICAL CONTROLS ON PRODUCTION
Scaled cumulative oil production from 28 vertical wells and 50 horizontal wells was compared to the lithology, porosity, and water saturation models to explore the dominant controls on production. All horizontal wells were of comparable lateral length and subjected to similar completion strategies. The best vertical producers are clustered in a thick tripolite interval in the northwestern area of the study, and most of the poorest producing wells are in areas with extremely low tripolite thickness (Figure 14). However, there are a few less successful wells in other areas with relatively thick tripolite indicating that tripolite thickness might be a significant control on production, but there are other factors as well. Average porosity appears to have a limited impact on oil production, and both the best and worst producing wells are in areas of particularly high water saturation. However, the examination of water saturation within just the tripolite lithology compared to production reveals that the poor producers that are in thick sections of tripolite are in areas of particularly high tripolite water saturation. Therefore, vertical wells appear to be dominantly producing from the tripolitic chert, and the lack of a relationship between porous tripolite and production is likely due to high porosity heterogeneity with no trends across the study area. However, water saturation associated with tripolite appears to play a role in production from the vertical wells, as the best wells are within a thick interval of tripolite that has low water saturation, while the less productive wells within thick tripolite intervals are areas with relatively high tripolite water saturation.
As the horizontal wells were not used in generating any of the models, they provide an independent control for determining modeling success away from the wellbores. Comparing the scaled cumulative production from horizontal wells (Figure 15), production appears to have no relation to tripolite or chert-rich limestone thickness. However, there appears to be some correlation between high chert-rich limestone porosity, low chert-rich limestone water saturation, and oil production, while a relationship between tripolite petrophysics and oil production appears to be nonexistent. This suggests that most production from horizontal wells is from chert-rich limestone rather than tripolite. It is likely that the hydraulic fracturing and other completion techniques used for the horizontal wells may be sufficient to overcome the permeability barriers in the chert-rich limestone to allow production from this lithology as well as from tripolite. Due to the much greater rock volume of chert-rich limestone than tripolite, production from the limestone would overwhelm any signals of production from tripolite, and it is likely that there is a sufficient thickness of chert-rich limestone in most of the study area that the thickness present does not present a limiting factor to horizontal wells drilled in the uppermost parts of the Mississippian limestone.
In north-central Oklahoma, Mississippian limestone deposition occurred on the Anadarko shelf below storm wavebase where a prolific carbonate factory also housed silicate-sponge bioherms. The major lithologies present within the Mississippian strata of northeastern Woods County, Oklahoma are (1) tripolitic chert, (2) chert conglomerate, (3) dense chert, (4) chert-rich limestone, (5) dense limestone, (6) shale, and (7) shale-rich limestone. A neural-network approach was used to predict the presence of tripolite (combined tripolitic chert and chert conglomerate), dense chert, chert-rich limestone, and dense limestone from a suite of open-hole well logs with 65.1% accuracy and a VShale cutoff allowed for the differentiation of shale and shale-rich limestones. The estimated lithology logs and other well-log data were used with 3-D seismic data and seismic attributes to interpret a stratigraphic and structural framework and 3-D model grid that consists of four reservoir zones and the dominant faults in the study area. In addition, the estimated lithology logs, lithology probability maps from P-impedance data, and other constraints were used to create a 3-D lithology model of the Mississippian limestone that represents the spatial distribution of the major lithologies. Lithology-constrained 3-D porosity and water saturation models show how tripolite is the most porous and heterogeneous lithology. Porous tripolite is most abundant along two faults within the study area as well as a region of particularly low seismic coherence, suggesting that fractures associated with these features may have allowed for enhanced fluid flow and diagenetic alteration.
Cumulative oil production from vertical wells appears to be most influenced by the tripolite thickness and water saturation. Cumulative oil production from horizontal wells appears to be most influenced by water saturation and porosity of chert-rich limestones from which the wells primarily produce. This difference in production contributors between vertical and horizontal wells is likely a result of the location of the horizontal wells (the very top of the Mississippian interval), and the use of hydraulic-fracture stimulation to enhance permeability of chert-rich limestone.
This research was funded through the Reservoir Characterization and Modeling Laboratory at the University of Oklahoma and the sponsors of the “Mississippi Lime” Consortium: Chesapeake Energy, Devon Energy, QEP Resources, and Sinopec (Tiptop Oil and Gas). We thank Chesapeake Energy for providing data. Additional funding was provided through the AAPG Foundation Grants-in-Aid (Jon R. Withrow Named Grant) and an Apache Corporation Scholarship. We are thankful to Schlumberger (Petrel) and CGG (Hampson Russell) for providing software.