Carbonate petroleum reservoirs associated with faults and fractures with associated large-scale breccia porosity are common in many sedimentary basins. Typically, in these rocks, dolomitization and mineralogical alterations by epigenetic basinal fluids that migrate along the fault and fracture systems have been observed. Such systems have been popularized as “hydrothermal” dolomite plays (Davies and Smith, 2006). However, fault-fracture reservoirs may not necessarily be extensively dolomitized (in fact, dolomitization may degrade reservoir properties; Gregg, 2004). Fault–fracture reservoirs also frequently display evidence of sulfide mineralization and have been genetically associated with Mississippi Valley-type (MVT) base metal deposits (Gregg, 2004; Gregg and Shelton, 2012). A characteristic of many reservoirs similar to those mentioned above is normal faulting with a strike-slip component. The resulting structure is described as a negative flower structure (Harding, 1985), typically characterized by a normal fault-bounded graben with a central synclinal depression.

A number of fault–fracture hydrocarbon fields have been developed in North America. The most important of these, which is also considered as the “type example,” is the giant Albion–Scipio field in the southern Michigan Basin (Hurley and Budros, 1990; Grammer and Harrison, 2013). The Albion–Scipio reservoir consists of dolomitized breccias in the Middle Ordovician Trenton and Black River limestones associated with extensional faulting having a left-lateral strike-slip component (Hurley and Budros, 1990). Dolomitization is associated with sulfide mineralization similar to that observed in many MVT mineral deposits (Budai and Wilson, 1991). Original oil in place in the Albion–Scipio field was estimated to be 290 MMBO (Hurley and Budros, 1990). The Albion–Scipio model has been used to discover a number of other fault–fracture petroleum fields in the Michigan and Appalachian basins including the parallel trending Stony Point field, discovered in 1982 (Hurley and Budros, 1990; Grammer and Harrison, 2013), the Saybrook field, discovered in northeastern Ohio in 1997 (Sagan and Hart, 2006), and more recently, in 2008, the Napoleon field, east of and paralleling the Albion–Scipio trend in southern Michigan (Grammer and Harrison, 2013). Several other large gas fields discovered in North America that fall into this reservoir class include the Ladyfern field in British Columbia (Boreen et al., 2001; Boreen and Coquhoun, 2003) and the Deep Panuke field in offshore Nova Scotia (Wierzbicki et al., 2006).

Here, we discuss a structural depression in the Mississippian age limestones (hereafter referred to as “Miss Lime”) in Payne County, north-central Oklahoma (Figure 1A), that shares many of the characteristics of the above mentioned hydrocarbon fields. In northwestern Payne County, an east–west striking fault, downthrown to the north, has been documented in association with the Nemaha Ridge by Gay (2003a, b) (Figure 1B). The fault’s surface expression is probably hidden below the present day Lake Carl Blackwell (LCB; Figure 2A). A 3-D seismic survey (Survey 1; Figure 2A) covering the lake revealed the presence of a depression on the downthrown side of the LCB fault at the Miss Lime level (Figure 2B). We will hereafter refer to this structural depression as the LCB sag.

The study area is located on the east flank of the Nemaha uplift and is bounded to the east by the Cherokee Platform, to the southeast by the Arkoma Basin, to the south by the Arbuckle uplift, and to the west, and southwest respectively by the Anadarko Basin and Wichita uplift (Figure 1A). The principal structural feature affecting the study area is the Nemaha uplift to the west that is characterized by north to south trending reverse faults (Gay, 2003a) and a number of east to west trending extensional faults with a strike-slip component, including the LCB fault (Figure 1B). Faulting associated with the Nemaha ridge originates in the basement and is thought to have been active during the Late Mississippian through Early Pennsylvanian (Gay, 2003a).

During the Mississippian, carbonate rocks were deposited across the ancient Burlington Shelf including portions of Colorado, Nebraska, Kansas, Oklahoma, Arkansas, Missouri, Iowa, and Illinois (Lane, 1978; Gutschick and Sandberg, 1983). The depositional strike was roughly east–west, with shallower water deposition on a widespread carbonate shelf to the north and northeast on the Burlington Shelf. The shelf system was bounded by the transcontinental arch to the north and northwest and by the Ozark uplift to the east. The depositional system deepened to the south and southwestward into the Arkoma and Anadarko basins as distally steepened carbonate ramps (LeBlanc, 2014).

Stratigraphy of the study area comprises the following older to younger succession. The crystalline basement is Proterozoic age. It is overlain by dominantly carbonate Cambrian–Ordovician strata including the Arbuckle Group and Viola Limestone that are in turn overlain by the Devonian age Woodford Shale. The Woodford is overlain by the Mississippian age carbonate unit, which is overlain in turn by Pennsylvanian siliciclastics and limestones (Shelton and Gerken, 1995). The Pennsylvanian “Pink Lime” forms a key seismic marker in the broader study area, and the Woodford is the main source rock in this part of the southern midcontinent (Clare, 1963).

Two 3-D seismic surveys are of significance to this study. The first is a larger survey (Survey 1; Figure 2A), which covers the Carl Blackwell lake and contains seven oil wells of different vintages that are used for porosity estimation within the sag. The second is a smaller survey (Survey 2; Figure 2A) that contains a salt-water disposal well, Ruark, which is the only well within the survey boundaries that has both density and sonic logs required for well-to-seismic tie. Survey 1 has a composite terrain comprising both land and water, which required three types of sources for completion, vibroseis and dynamite on land and airgun in water. As a clarification, although different source and receiver types were used in Survey 1, it was not conducted in pieces, that is, hydrophones at the bottom of the lake were recording when a shot was being fired on the land. Further, the terrain in Survey 1 was not equally accessible, which led to a nonuniform source–receiver distribution, and consequently an acquisition footprint toward the east end of the survey (Jaiswal et al., 2019). The source–receiver distribution within the lake itself, which overlies the sag, was fairly uniform. In Survey 2, the terrain was less complex. Only one kind of source (vibroseis) was used and the source–receiver distribution was also fairly uniform.

Survey 2 was tied to the Ruark logs using the conventional method of preparing a synthetic seismogram at the well location and then progressively (and minimally) stretching and squeezing it such that best possible phase match is obtained with the seismic volume (see figure 2 in Jaiswal et al., 2019). Following this, key markers such as the Pink Lime, Miss Lime, Woodford, and Viola were interpreted on Survey 2. To extend the interpretation to Survey 1, both surveys were tied through a procedure comprising phase rotation, amplitude balancing and bulk shifting Survey 1 in time to properly correlate the key reflectors. After extending interpretations of the individual horizons into Survey 1, an inspection of the Miss Lime top time horizon (Figure 2B) as a quality control measure displays no kinks at the survey boundary, indicating that the interpretation process was effective and reliable.

The LCB fault appears in the Miss Lime top time horizon (Figure 2B) as an east–west trending feature. Adjacent to the LCB fault, on the footwall, toward the west end of the survey, a structural depression can be observed, which forms the core of a negative flower structure (Figure 2B). Since the volume being used for interpretation is in time (versus depth), whether the depression is structural or a velocity effect (e.g., from near-surface lake sediments or shallow gas) needs to be determined. We examined two mutually near-orthogonal cross sections, northeast–southwest AA′ and northwest–southeast BB′ (Figure 2B). In AA′ (Figure 3A), the sag extends from approximately CMP 3500 to 6000 and in BB′ (Figure 3B), it extends from approximately CMP 2500 to 5500. In both profiles, the depression appears to be increasing with depth, which cannot be due to velocity effects because, otherwise, the intensity of the depression would remain constant below the low velocity horizon. The depression appears to be structural and, although its true curvature can only be realized in a depth image, relationships between the various structural elements can be understood through the time image.

In both the AA′ and BB′ profiles, the LCB fault can be clearly interpreted based on the Pink Lime to Viola reflection terminations and offsets (Figure 3). Although the section stratigraphically deeper than the Viola is not very clearly imaged in either profile, the LCB fault appears to be continuing below the Ordovician into the basement. As a clarification, the base of the sedimentary column and top of the crystalline basement in the study area does not have tight well control. However, based on studies by Widess and Taylor (1959), the reflection at 900 ms in Figure 3 can be safely considered as a unit lying below the sedimentary column and within the crystalline basement. Thus, it is very likely that the origin of the LCB fault could be related to the formation of the Nemaha ridge (Gay, 2003a,b). The displacement along the LCB fault decrease upward and falls below seismic resolution at the stratigraphic level of the Pink Lime (Figure 3). Therefore, whether or not the same fault actually continues to the surface and creates the depression that holds the present day LCB cannot be determined from current seismic data.

Antithetics to LCB fault are more challenging to interpret. In AA′, the antithetic fault F1 (Figure 3A) can be confidently interpreted immediately to the northeast of the LCB fault based on the Miss Lime-to-Viola reflector termini and offsets. Unfortunately, no clear offsets in the Pink-Lime-to-Viola reflection package exist any further to the northeast. However, the undulations in the Miss-Lime-to-Viola reflection package are strongly indicative that the strata is highly fractured up to the beginning of the flexure at CMP 3500. A second antithetic fault, F2, marking an intermediate step and a third antithetic fault, F3, suggestive of the northeast limit of the sag, have been interpreted with dashed line (Figure 3A). In the BB′ profile at least two antithetic faults, F2 and F3 (Figure 3B) can be interpreted based on Mississippian-to-Viola reflection termini and offsets. Much like in the AA′ profile, undulations, suggesting subseismic resolution fractures, exist in the Miss Lime-to-Viola package. In both profiles, amplitudes weakening within the sag between the Pink Lime and the Miss Lime horizons, compared to the reflections outside the sag at the same stratigraphic level, may suggest that lithification within the sag is poor. In map view (Figure 4A), the shape of the antithetic faults (more tapered union with the LCB on the west side) weakly suggests that the strike-slip movement along the LCB fault could be right lateral (the downthrown block moves away from the Nemaha Ridge structure).

Note that in Figure 3, the interval between the Pink Lime and the Miss Lime grows dramatically to twice its thickness across fault LCB in the downthrown section, while the deeper units, Miss-Lime-to-Woodford and Woodford-to-Viola, remain fairly consistent in thickness across the same fault. The interpretation suggests that faulting occurred primarily during the Miss-Lime-to-Pink Lime time interval. The sections above the Pink Lime also seem to maintain their thicknesses across the LCB fault. This is consistent with the prevalent understanding of the broader study area (Shelton et al., 1979).

For porosity estimation, we have used the EMERGE toolbox within the commercial software package, Hampson-Russell. We are providing a summary of the workflow below and guide the reader to Hampson et al. (2001) for details. The modeling approach of EMERGE is guided by the basic philosophy of multiattribute analysis such as that proposed by Schultz et al. (1994a, b, c), which advocates combining two or more attributes, such as phase and frequency, in a manner such that a convincing correlation with a target rock property (such as porosity) is developed. Ideally, rock properties are to be generated from measurements on actual rock (core or outcrop). However, in practice they are usually inferred through petrophysical analysis of well logs, which in a sense is also a geophysical measurement. Due to difference in bulk volume over which the information is averaged, the same rock will manifest differently in sonic and ultrasonic datasets. Thus, the multiattribute analysis, particularly in this paper, is an attempt to reconcile log measurements (measured over ultrasonic frequencies) with surface seismic measurements (measured over sonic frequencies). The biggest conceptual challenge in this process is that the mathematical operators required for such a reconciliation changes from one data set to another; that is, the same attributes may not correlate in the same manner from one study area to another or from one seismic volume to another. Although the multiattribute rock property estimation appears an ad hoc process where an interpretation forces a model on data, an increasing number of publications are showing that it is in fact a valid technique (Nissen et al., 2009; Yenugu and Marfurt, 2011; Roy et al., 2013; Guo et al., 2014; Qi et al., 2015).

The input to EMERGE is a seismic volume and well log(s) that are tied to it. In EMERGE, logs are first calibrated to the seismic volume. The software then extracts a series of attributes from the seismic volume and for each attribute checks what kind of transform (e.g., inverse, cosine or log) regresses best with porosity, which is interpreted from log (minimizes the prediction error in a root-mean-square [RMS] sense). We extracted attributes at every 5 ms at the Miss Lime level to best match the log character. The software has capabilities to combine up to six transformed attributes to build a multiattribute regressive model but, in our case, only four transformed attributes were needed for minimizing the prediction errors: cosine of 30–35 Hz and 20–25 Hz band-limited data amplitude, cosine of average Miss Lime amplitude from full bandwidth (15–80 Hz) data, and average frequency of Miss Lime. The software then used all the four transformed attributes as an input set and the porosity as the output set to train a neural network. A neural network model is utilized because it provides lower prediction error (here, ~5% vs. 10%) over the regression model (e.g., Leiphart and Hart, 2001). The neural network model was then used to generate porosities at the Miss Lime level throughout the seismic volume. Results are shown in a cross-sectional view in Figure 3C, D and in a map view in Figure 4B. Both Figures 3, 4 show pockets of enhanced porosity zones existing within the LCB sag.

Vitrinite reflectance (R_o) data for the Woodford Shale (Cardott, 1989, 2012), which underlies the Mississippian, is shown in Figure 1 along with basement faults that have been identified in the region (Gay, 2003a; Darold and Holland, 2015). As shown in Figure 1A, R_o in the region falls largely in the range of 0.49–0.85, which correspond to sub-to-moderate organic maturity (moderate maturity suggests the oil window). Maturity increases to the southwest and to the southeast into the Anadarko and the Arkoma basins, respectively, as is expected to happen with increasing burial depth of the Woodford. Within the sub-to-moderate maturity zone, there are several anomalously high R_o values, most of which appear to be associated with basement faults. One of these anomalous R_o values (1.07) lies near the LCB structure. The reason for these anomalous increases in maturity is not clear although high-temperature (>100°C) fluid influx along faults could be the cause.

To examine temperature regimes of the fluids in the Mississippian strata, fluid inclusion studies have been performed by Mohammadi et al. (2019a,b). They observed two-phase (fluid and vapor bubble) aqueous and liquid petroleum-rich inclusions in fracture, breccia, and vug-filling calcite and quartz cements in the study area. Homogenization temperatures (T_h) for the aqueous fluid inclusions range from 47°C to 173°C with most of the T_h values clustered between 80°C and 150°C. These temperatures, without pressure corrections, represent minimum temperatures at which the calcite and quartz cements that contain them precipitated. Salinities of the included aqueous fluids range from 0.0 to 25 Wt.% NaCl equivalent. These data indicate that hot and saline hydrothermal fluids, most likely emanating from deeper Ordovician or crystalline basement, invaded the Miss Lime. The observation of petroleum-rich inclusions suggests that hydrocarbons, likely sourced in the Woodford Shale (Al-Atwah et al., 2019), followed the same pathways. The fluids likely moved upward along faults and fractures, such as LCB (Figure 3), which are connected with the underlying basement rocks, and mixed with less saline resident fluids in the Mississippian (Mohammadi et al., 2019a, b). If hydrothermal fluid invasion has indeed resulted in anomalously matured Woodford in the study area, the secondary-porosity pockets in the LCB sag have a high probability of having been being charged with petroleum.

Occurrence of high-porosity zones within the LCB sag along with possible fluid-invasion supports the hydrothermal reservoir concept. However, there are uncertainties associated with both the porosity and fluid inclusion datasets have to be considered. Although fluid inclusion studies are experimental and fairly direct, porosity models are mathematical. A detailed discussion on methods and principles underlying fluid inclusion studies can be found in Mohammadi et al. (2019a,b). Discussion of porosity estimation is presented herein.

Estimating porosity from seismic data is an open-ended challenge and at the forefront of geophysical research. Seismic signals acquired at receivers are essentially source waveforms distorted primarily by geometrical spreading and attenuation (energy loss in exciting the rock and fluid particles). Attenuation, which modulates both amplitude and phases, is a complex function of rock and fluid properties such as mineralogy, viscosity, and pore architecture in addition to permeability and porosity. Therefore, in the absence of information about other rock and fluid properties, estimating porosity and permeability from the seismic waveforms is not possible. Studies on quantitative interpretation have shown that many attributes of the seismic waveform, such as amplitude, frequency and phase, contain vital information about rock and fluid properties (Doyen, 1988; Chen and Sidney, 1997; Alvarez and Bolivar, 2015). With an understanding that different attributes are effected to different extents by different rock and fluid properties, an interpreter searchers for models that map seismic attributes (as-is or transformed; individual or in combination) to rock properties such as porosity. In developing these models, it is expected, in principle, that the causative relation between a given rock (or fluid) property and a given data attribute is known. However, this is not usually the case. Most of the models are empirical, lacking rigorous physical explanations. This is sometimes attributed to the seismic noise and sometimes to lack of experimentation to uncover the physics governing the relations. In common practice, the models are developed in a trial-and-error manner heuristically exploring which attribute combination, if any, has predictive ability for a given rock property by testing at well locations. However, recently, as more and more sophisticated experimental and visualization facilities are becoming available, better physics-based explanations are also being formulated (Avseth et al., 2010).

In this application, in addition to the challenges described above, we also have data sparsity issues. First, porosity logs are mainly available in the Mississippian and upper Woodford intervals (Figure 5) and second, only seven wells (Survey 1; Figure 2A) are available in the entire 3-D volume for generating porosity within the LCB sag. Further, none of those seven wells have actually penetrated the sag, and therefore, we do not have a way of ground truthing our results at the target. This is the biggest weakness of our modeling. Moreover, seismic processing within the sag runs the risk of being least reliable in the area because of structural distortion. Therefore, at the outset, we acknowledge that the absolute porosity values generated by the neural network within the LCB sag may have large uncertainties and urge the reader to interpret Figures 3C, D; 4B as stating that there may be large porosity variation within the sag. After all, it is the porosity variation within the LCB sag that is relevant to the idea of fault–fractured reservoir rather than the absolute porosity values.

In evaluating the LCB sag as a potential petroleum reservoir, it is useful to consider a nearby analog (Figure 1). Located approximately 230 km (145 mi) northwest of the LCB structure, the Tri-State Mineral District covers more than 5000 km² (3100 mi²) in northeastern Oklahoma, southeastern Kansas, and southwestern Missouri (Figure 6A). The district produced more than $2 billion in zinc and lead prior to its closing in 1967 (Brockie et al., 1968). The ore deposits are primarily hosted by chert–dolomite breccias and are concentrated in the Mississippian Warsaw and Keokuk limestones (Brockie et al., 1968; Hagni, 1982). Ore distribution in the Tri-State Mineral District is believed to largely be controlled by patterns of faults and fractures that acted as conduits for saline, ore forming, basinal fluids moving upward from the underlying Ordovician strata (Brockie et al., 1968; Hagni, 1976; Wenz et al., 2012). Fluid inclusion and oxygen isotope data (δ¹⁸O) indicate that the ore forming fluids were saline brines with temperatures that ranged between 60°C and 175°C (Ragan, 1996; Wenz et al., 2012; and Mohammadi et al., 2019b). These hydrothermal brines also were accompanied by migration of petroleum as is evidenced by the presence of liquid petroleum inclusions in ore minerals and associated carbonate cements (Wei, 1975; Mohammadi et al., 2019a) and liquid petroleum in ore bodies (Fowler, 1933).

Ore bodies are controlled by large faults that trend northeast to southwest and other structures that trend normal to the larger northeast trending faults. Both the Miami trough and the Seneca trough to the east are normal fault systems with left lateral strike-slip components (Figure 6A). A section illustrated through a typical ore body on the Miami trough, the Blue Goose Mine (Figure 6B), is of comparable scale to the LCB structure in the study area.

We conceptualize the LCB play evolved in the following manner. First, the negative flower structure evolved in association with the Nemaha Ridge formation (Figure 7A). Movement of hydrothermal fluids occurred most likely in correspondence with the Ouachita orogeny (Appold and Garven, 1999; Leach et al., 2001; Gregg and Shelton, 2012), which likely further deformed the LCB fault zones and created breccia and dissolution porosity. At this time saline fluid moved upward from the underlying Ordovician carbonates and granitic basement along the fault–fracture system into overlying Mississippian limestones where it mixed with resident brackish fluids and precipitated authigenic minerals (calcite, quartz, dolomite, and possibly sulfides). The underlying Devonian shales sourced the hydrocarbons needed to charge the breccia porosity (Figure 7B). Overlying Pennsylvanian shales may have provided a seal confining hydrothermal fluids and petroleum in the Mississippian regionally (Goldstein and King, 2014) and locally in the LCB structure. Since these mechanisms are shared by fault–fracture petroleum reservoirs throughout the world (e.g., Davies and Smith, 2006), we speculate that if a core were to be cut through the LCB sag, the rock fabric would be comparable to similar oilfields (Figure 7C).

The concept of the LCB play evolution is fairly generic and should be applicable to the entire Mississippian play in the north–central Oklahoma. It is notable that several basement-related faults exist in the broader study area (Figure 1B). When striking at the correct orientation, several of these faults or their segments may have undergone the same process where a negative flower structure, with associated brecciation, has evolved. Fluid inclusion data support hydrothermal fluid migration at several locations in the broader study area, and the underlying Woodford Shale has reached the oil window throughout most of the study area. A foreseeable risk involved in this play concept is seal breaching because units overlying the sag that could have acted as seals may have been eroded, or would also have experienced similar mechanical and chemical alterations as the rocks inside the sag themselves. Therefore, whether the LCB sag is a working petroleum system requires further testing. Nonetheless, we judge the probability that LCB sag type structures in the study area may be holding significant petroleum reserves as high.

In this study we show the presence of a fault-bounded structural sag at the Miss Lime stratigraphic level in Payne County, Oklahoma, with key elements of a fault-fracture (hydrothermal) petroleum reservoir. The LCB sag encompasses the Pink-Lime-to-Viola horizons at least and appears to be highly fractured with the main LCB fault being basement involved. A multiattribute model suggests the presence of high-porosity zones within the LCB sag. Regional fluid inclusion studies indicate that the study area has experienced invasion of hydrothermal and saline fluid, most likely sourced from Ordovician or deeper units. Such fluid migration is associated with petroleum migration and a nearby analog, the Tri-State Mineral District, suggests that can create breccia porosity. Thus, not only does the LCB sag have several elements of a fault-fracture type play, the concept of fault–fracture hosted hydrothermally defined reservoir is also extendable to the entire north-central Oklahoma where basement faulting and anomalous fluid inclusions are evident. The play concept proposed in this chapter presents a testable hypothesis for any carbonate formation on the midcontinent.

This study was supported by the Oklahoma State University–Industry Mississippian Consortium and the Boone Pickens School of Geology. We thank all of the companies that supported this Consortium (American Energy Partners, Chaparral Energy, Chesapeake Energy, Devon Energy, Longfellow Energy, Marathon Oil, Maverick Brothers, Newfield Exploration, SM Energy, Samson Energy, Sinopec [Tiptop], Redfork Energy, Trey Resources, and Unit Petroleum).

We also would like to acknowledge CGG GeoSoftware for its donation of the Hampson Russell software products to complete the work on this paper. S.N. Bose program supported intern Satya who helped the lead author in drafting images. This is Boone Pickens School of Geology Contribution 2017-64.