Degradation of fracture porosity in sandstone by carbonate cement, Piceance Basin, Colorado, USA

Natural fractures can significantly control flow of formation and production fluids, especially in low-permeability rock units. They provide pathways between organic-rich source layers and matrix pores during hydrocarbon charge, and between matrix pores, hydraulic fractures and the wellbore during production (Laubach 2003; Cumella & Scheevel 2008; Warpinski & Lorenz 2008). Whether fractures are cemented or cement-lined, and the topology of fracture cement and porosity can significantly influence fracture permeability, with consequences for well producibility and reservoir performance (Laubach 2003; Nollet et al. 2005, 2009; Philip et al. 2005; Olson et al. 2009, 2010; Witte et al. 2012; Tokan-Lawal et al. 2015, 2017; Landry et al. 2016; Wennberg et al. 2016). On a larger scale, the diagenetic state of fractures may influence basin-scale fluid flow (Fischer et al. 2009; Bjørlykke & Jahren 2012). The type and amount of fracture cement can influence fracture spacing (Hooker & Katz 2015) and, among other factors, cement in pre-existing fractures can affect subsequent fracture growth (Virgo et al. 2014; Lee et al. 2018) and interfere with engineering operations such as hydraulic fracture stimulation. Understanding how and why cement accumulates in fractures is therefore essential for predicting the flow and storage behaviour of these rocks.

In tight gas sandstone reservoirs, quartz cement is widespread in fractures. Fractures are commonly either lined with quartz and thus partially open for fluid flow, or completely cemented and thus hydraulically inactive relative to the host sandstone. Understanding of quartz cement growth has provided insight into fracture development and permeability evolution of fracture networks (Beach 1977; Fisher & Byrne 1990; Laubach et al. 2004; Becker et al. 2010; Lander & Laubach 2015). However, as important as these reactions are, other mineral phases, notable carbonate minerals, are also widespread (Finley & Lorenz 1988; Laubach 2003). Carbonate cement in sandstone is commonly heterogeneous in composition, spatial distribution and time of precipitation within the burial diagenetic evolution (Boles & Ramseyer 1987; Bjørkum & Walderhaug 1990). Carbonate cement in many tight gas sandstone reservoirs completely occlude fractures, whereas nearby quartz-lined fractures of similar width preserve residual fracture porosity and thus hydraulic conductivity (Laubach 2003). Yet, controls on the occurrence of these cements are largely unexplored. Prediction of fracture cementation or openness is thus a primary objective in the characterization of tight gas sandstone and shale gas reservoirs that is of equal importance as the prediction of fracture abundance. Where occurrence of sealed fractures is heterogeneous, knowledge of the pattern of sealed and open fractures can be used in exploration and development decisions.

Here we use core observations in a tight gas sandstone from the Piceance Basin to investigate the controls on the distribution of the carbonate minerals – primarily calcite, ferroan calcite and ankerite – that commonly seal large fractures (apertures >1 mm). Petrography, micro-imaging and geochemistry from a wide range of wells across the basin (Fig. 1) document evidence for the distribution and origins of the carbonate cements that selectively destroy fracture porosity in these rocks. We hypothesize that the destruction of fracture porosity by calcite is fundamentally linked to diagenetic processes occurring in the host rock. Under this premise, we test the prediction of Laubach (2003) that host-rock petrographical observations can be used to infer the degree of cement occlusion or degradation of nearby fractures.

We explore these ideas by characterizing mineral phases, chemical reactions and conditions associated with calcite precipitation in host rock and fractures, and by comparing host-rock diagenesis to observed fracture occlusion by calcite cement. We studied carbonate cement in core samples of sandstone in the Upper Cretaceous Mesaverde Group collected from three well sites in the Piceance Basin, two wells in the Piceance Creek Field (MF31-19G and Piceance Creek Well 1) and one well in the Mamm Creek Field (Fig. 1). We show that the distribution of carbonate phases is heterogeneous, in some cases varying from abundant to absent from bed to bed or laterally over distances as short as interwell spacings (possibly 1000 m or less). The results demonstrate the interplay between host-rock diagenetic processes and the preservation of fracture porosity in sedimentary basins. Results can be used to make predictions of the distribution of open, porous and sealed fractures, and may have economic implications for exploration and development decisions (Almansour 2017).

During the Cretaceous Period, the Piceance Basin region was located in a foreland basin setting along the western margin of the Western Interior Seaway, receiving clastic sediment from the Sevier thrust belt to the west (Johnson & Nuccio 1986; Patterson et al. 2003) (Fig. 1). During the Late Cretaceous–Eocene Laramide Orogeny, the basin began to take its current asymmetrical shape with gently dipping flanks on the west and a steep flank on the east (Fig. 1). Following burial reconstructions of Yurewicz et al. (2003, 2008), Zhang et al. (2008), Fall et al. (2012, 2015) and Tong et al. (2017), maximum burial was during late Oligocene–early Miocene time, followed by uplift associated with top-to-the-west high-angle reverse faults along the White River uplift (Lorenz & Finley 1991).

The c. 1400 m (c. 4600 ft) thick Upper Cretaceous Mesaverde Group includes marine and shoreline sandstones of the Iles Formation (Corcoran, Cozzette and Rollins sandstones). These mainly shallow-marine shoreface sandstones interfinger with transgressional tongues of Mancos Shale (Payne et al. 2000). The Iles Formation is overlain by the Williams Fork Formation, a sequence of non-marine floodplain shales, discontinuous fluvial sandstones and coal beds (Johnson & Flores 2003; Patterson et al. 2003; Pranter et al. 2007; Pranter & Sommer 2011). Mesaverde sandstones were deposited throughout several transgressive and regressive cycles of the Western Interior Seaway (Dutton et al. 1993). All samples used in this study were collected from the fluvial to paludal Williams Fork Formation, except for three samples (MF31-19G: 7321.1, 7324.4 and 7349 ft) collected in the fluvial Ohio Creek Conglomerate of the uppermost Mesaverde Group.

In the subsurface, fractures strike predominantly WNW–ESE and apparently form a single set. Previous fracture studies, including those at the Department of Energy's Multiwell Experiment (MWX) site and other wells in the Piceance Basin, have shown that natural opening-mode fractures occur in sandstone throughout the Mesaverde Group (Bredehoeft et al. 1976; Verbeek & Grout 1984; Grout & Verbeek 1985, 1992; Pitman & Sprunt 1985, 1986; Finley & Lorenz 1988; Lorenz & Finley 1991; Lorenz 2003; Cumella & Scheevel 2008; Hooker et al. 2009, 2014; Fall et al. 2012, 2015; Baytok & Pranter 2013).

Core analysis shows that fractures are cemented by variable amounts of quartz and calcite, and locally by a lesser amount of barite and clay minerals (dickite) (e.g. Pitman & Sprunt 1985; Finley & Lorenz 1988). Fractures cemented solely by quartz and with kinematic apertures (opening displacement) larger c. 0.1 mm frequently preserve fracture porosity (Hooker et al. 2009). Fractures cemented with calcite are typically occluded regardless of aperture (Pitman & Dickinson 1989). The distribution of these cements in a given locality or lithology is commonly highly heterogeneous (Pitman & Dickinson 1989; Taylor et al. 2000).

Quartz cement in fractures typically contains crack-seal textures that allow the sequence of fluid-inclusion assemblages to be ordered into a relative time sequence (Laubach et al. 2004; Becker et al. 2010). Fluid-inclusion thermometry indicates that quartz precipitated during an increase in temperature from c. 140 to 185°C, followed by a decrease to c. 158°C (Fall et al. 2012, 2015). Assuming that temperatures reflect burial depth and an accurate burial history, these results imply that fracture opening occurred over a time span of 35 myr during maximum burial and incipient exhumation concurrent with gas charge of the reservoir (Fall et al. 2012, 2015). Based on textural relations between fracture calcite and fracture quartz, observations and inferences about fracture quartz formation constrain the timing of fracture calcite formation as demonstrated below.

Optical microscopy in combination with high-resolution scanning electron microscopy (SEM) was performed on samples of fracture cement and host-rock grains and cement to identify minerals and their textural relations. For Piceance Creek Well 1, our petrographical observations on textural relations in the host rock complemented commercial point-count analyses performed on 170 samples (250 counts each) collected on 1–5 ft depth intervals over the cored well sections. In addition, we interpreted cement patterns in nine sidewall core sections collected for well intervals without conventional core, and on nine samples of conventional core containing fractures. Bulk rock powder X-ray diffraction analyses, and inductively coupled plasma atomic emission spectroscopy (ICP-AES) for major elements and inductively coupled plasma mass spectrometry (ICP-MS) for trace elements were performed in a commercial laboratory following standard procedures.

Optical microscopy was combined with high-resolution SEM-based imaging and electron probe microanalysis (EPMA) to constrain the chemical composition and variation of mineral phases. Quantitative analysis of carbonate composition was performed using a JEOL JXA-8200 electron microprobe (EPMA). Wavelength-dispersive spectrometry (WDS) analysis of carbonate phases employed a 10 µm spot, a 12 nA sample current (measured on brass) and 20 s count times. Dolomite, siderite and Sr-glass standards in the collection of the University of Texas at Austin Electron Microprobe Laboratory were used for calibrations. A FEI Nova Nano scanning electron microscope model 430 equipped with an EDAX energy-dispersion spectrometer (EDS) was used to obtain semi-quantitative elemental maps and backscattered electron images to assess compositional heterogeneity.

Samples of fracture carbonate cement for Sr isotope, Sr concentration and stable isotope analyses were obtained by mechanical separation, followed by handpicking under a binocular microscope. Sr isotope analyses were conducted using an Isoprobe multicollector ICP mass spectrometer (MC-ICP-MS) with nine faraday collectors. Subsequent to ion chromatographic treatment of the samples, the purified Sr aliquots were diluted with 0.48 M HNO₃ to obtain a Sr concentration of c. 50 ppb. Strontium isotope ratio data were acquired in static, multi-collection mode using seven collectors to collect ion currents at mass 82, 83, 84, 85, 86, 87 and 88. The amplifier gain cross-calibration was performed daily before instrument optimization. A blank solution was first measured for 60 s for the background values, which would include and correct for interference from ions such as Kr. The half mass baselines for a sample or standard were then measured for 30 s. The Sr isotope data collection consisted of 10 cycles of 10 s integrations in each block, with four blocks being collected. The washout time between analyses was 10 min. The data from solution MC-ICP-MS analysis was processed online using the Isoprobe software. The ⁸⁴Kr/⁸²Kr ratio of 4.920030 and ⁸⁶Kr/⁸²Kr ratio of 1.4934206 (Rosman & Taylor 1998) were used for correction of the isobaric interference corrections of ⁸⁴Kr on ⁸⁴Sr and ⁸⁶Kr on ⁸⁶Sr. A ⁸⁷Rb/⁸⁵Rb ratio of 0.3857057 (Steiger & Jäger 1977) was used for correction of the isobaric interference of ⁸⁷Rb on ⁸⁷Sr. The online outlier analysis protocol from the Isoprobe software was used. Mass bias correction used the exponential law and a value of 0.1194 for ⁸⁶Sr/⁸⁸Sr.

Accuracy and reproducibility of the analytical protocol were measured by the repeated analysis of a 50 ppb solution of NIST SRM 987 strontium isotope standard during the course of analysis. The measured mean value for ⁸⁷Sr/⁸⁶Sr was 0.710257 ± 0.000027 (1 SD, n = 30) and for ⁸⁴Sr/⁸⁶Sr was 0.056491 ± 0.000092 (1 SD, n = 30). The ⁸⁷Sr/⁸⁶Sr ratio and the external precision were not significantly different from the value for ⁸⁷Sr/⁸⁶Sr of 0.710248 ± 0.000012 (1 SD, n = 427) for the NIST SRM 987 standard reported by high-quality thermal ionization mass spectrometry (TIMS) techniques (Thirlwall 1991). Our result is also in agreement with other reported solution Sr isotope data for the NIST SRM 987 standard using a Neptune MC-ICP-MS instruments with ⁸⁷Sr/⁸⁶Sr values such as 0.710245 ± 0.000018 (1 SD, n = 16: Ramos et al. 2004), 0.710273 ± 0.000017 (1 SD, n = 97: Copeland et al. 2008) and 0.710250 ± 0.000020 (1 SD, n = 41: Walther & Thorrold 2009), and ⁸⁷Sr/⁸⁶Sr of 0.710272 ± 0.000010 (1 SD, n = 14: Yang et al. 2011).

For measuring the Sr concentration, fracture samples were digested in 2% HNO₃ (trace metal grade) and the insoluble residue contents determined. The soluble weight and the weight of the 2% HNO₃ added determined the initial dilution factor, and established a basis for diluting to levels appropriate for ICP-MS (total dissolved solids (TDS) <0.2 wt%). Samples were analysed on an Agilent 7500ce quadrupole ICP-MS in no-gas mode (no collision or reaction gas), in addition to blanks (2% HNO₃) and three quality control standards. All samples were well above the detection limit. The Sr content of a procedural blank indicated no processing-related contamination.

The stable carbon and oxygen isotope values of carbonates were obtained at the Stable Isotope Biogeochemistry Laboratory, Stanford University, using a Thermo Finnigan Gasbench interfaced with a Thermo Finnigan Delta Plus XL mass spectrometer via a Thermo Finnigan ConFlo III unit. Between 0.11 and 7.34 mg of carbonate powder was weighed into sealed vials that were flushed with He gas and reacted with c. 0.25 ml of phosphoric acid (H₃PO₄) for 1 h at 72°C. External precision of oxygen and carbon isotope data is <0.1‰, based upon repeated measurements of two internal laboratory standards (calibrated against NBS 18, NBS 19 and LSVEC). The δ¹³C and δ¹⁸O values are reported relative to VPDB.

Stable isotopes for sample MF-31 were measured in the high-temperature stable Isotope Laboratory at UT Austin. Approximately 2 mg of carbonate-cemented sandstone was weighed into 20 ml Exetainer vials. The loaded Exetainer vials were heated at 80°C in an oven for 2 h, then sealed with a rubber pierceable septa and purged with dry helium to remove all the air from the vials. These vials were then heated to 40°C and 10 drops of concentrated phosphoric acid were added to each vial. Samples were allowed to react with phosphoric acid at 40°C overnight. The liberated CO₂ gas was sampled using a ThermoElectron MAT253 isotope ratio mass spectrometer coupled to a ThermoElectron GasBench with a CombiPal autosampler. Carbon isotope values are calibrated with respect to NBS-19 = 1.95‰ VPDB and oxygen isotope values are calibrated with respect to NBS-19 = −2.2‰ VPDB. An internal calcite standard was measured to assess precision and analytical drift. A precision of ±0.04 for δ¹³C and ±0.3 for δ¹⁸O was achieved.

Fluid-inclusion microthermometry was carried out using a Fluid, Inc.-adapted USGS-type gas-flow heating–freezing stage mounted on a Olympus BX 51 microscope equipped with a ×40 objective (N.A. = 0.55). The stage was calibrated using the CO₂–ice melting temperature at −56.6°C of H₂O–CO₂ synthetic fluid inclusions, and the ice melting temperature at 0°C and critical homogenization temperature at 374.1°C of pure H₂O synthetic fluid-inclusion standards (Sterner & Bodnar 1984). Liquid–vapour homogenization temperatures (T_h) were determined to ±0.05°C by thermal cycling using temperature steps of 0.1°C (Goldstein & Reynolds 1994).

Mesaverde sandstones in our study wells are extensively cemented by quartz and clay minerals (chlorite, illite, illite–smectite), consistent with previous petrographical studies that documented that these diagenetic phases predate carbonate cement, with the exception of siderite pore cement that formed locally prior to quartz (Hansley & Johnson 1980; Pitman & Dickinson 1989; Pitman et al. 1989; Crossey & Larsen 1992; Klimentidis & Welton 2008; Ozkan et al. 2011). Quartz textures are compatible with protracted growth, compatible with models of quartz accumulation (Lander & Walderhaug 1999; Taylor et al. 2010).

Carbonate pore cement includes calcite, ferroan-dolomite, ankerite and siderite, whereas calcite is the only carbonate mineral phase in fractures (Figs 2–4). Absolute abundance of carbonate pore cement and the relative abundance of different carbonate species are highly variable within the three study wells (Fig. 5). Iron- and magnesium-bearing carbonate pore cement occurs only in the presence of detrital dolomite (Fig. 3a, b), with the proportions of ferroan-dolomite and ankerite being variable along the Piceance Creek Well 1. Ferroan-dolomite and ankerite preferentially occur with detrital dolomite as substrate (Fig. 2e, f), less commonly with feldspar (albite). In general, the iron content increases in these carbonate phases with growth (Fig. 4). Authigenic ferroan-dolomite and ankerite precipitated on detrital dolomite is observed to penetrate and to replace what appears to be a ‘pre-existing’ phase, preferentially albite (Fig. 3a) resembling similar replacement textures described by Milliken (2003).

Siderite cement is a minor component of the sandstone, only occurring in deeper samples as small (<40 µm) rhombs rimming detrital grains (Fig. 3g). This siderite cement is enveloped by quartz overgrowths and therefore predates quartz overgrowth cement, indicating that siderite formed early in the diagenetic history and prior to fracture opening and cementation.

Calcite pore cement occurs in both detrital dolomite-bearing and dolomite-free stratigraphic units. Based on point-count analyses, calcite pore cement content is higher in shallower samples and decreases with increasing depth for the Piceance Creek Well 1 (Fig. 5). A similar trend was observed by Pitman et al. (1989) in the MWX wells in the Rulison Field (Fig. 1).

Where present, ferroan-dolomite, ankerite or siderite cement are the dominant cement phases, and calcite is only a minor component in the Piceance Creek Well 1 (Fig. 5). Similar relations are observed in the Mamm Creek and MF31-19G wells. Where calcite pore cement occurs together with detrital dolomite and ferroan-dolomite/ankerite, calcite is the youngest pore-filling carbonate generation (Fig. 3b, h).

Calcite pore cement shows leaching textures and generates secondary porosity (Fig. 2g). Porous albite, K-feldspar and calcite appear in a patchy texture that is characteristic of the albitization of detrital feldspar (Fig. 3c). Furthermore, muscovite hosted in albite indicates sericitization of primary feldspar. Albitization of anorthitic plagioclase thus provides a source for Ca²⁺ available for carbonate precipitation. Although calcite is observed to be the dominant Ca-species that is associated with albite, ferroan-dolomite is also observed in textural association with albite (Fig. 3d). However, ferroan-dolomite cement occurs only in the presence of detrital dolomite. The presence of detrital dolomite thus controls which carbonate forms during albitization.

Fractures are cemented by earlier quartz and, where present, by later calcite (Fig. 2b). Quartz lines fractures and forms crack-seal cement bridges (Fig. 2c). Fracture calcite cement occurs both as banded crack-seal (Figs 2h and 6) and as sparry cement, with a crack-seal texture providing evidence for incremental fracture opening and cementation (synkinematic cement) (Becker et al. 2010). Blocky cement is interpreted to form after fracture opening (post-kinematic). Cross-cutting relations demonstrate that calcite-cemented fractures post-date ferroan-dolomite and ankerite pore cement (Fig. 3e, h). The timing relations between calcite pore cement and calcite fracture cement are ambiguous. Pore and adjacent fracture cement in optical continuity do not exclude that both generations formed contemporaneously. On the other hand, calcite pore cement cross-cut by quartz-lined fractures indicates that some calcite pore cement formed prior to quartz fracture cement, which predates fracture calcite.

In contrast to quartz which lines or bridges fractures and rarely seals wide fractures (Fig. 2c), calcite tends to seal fractures completely (Figs 2a, h and 3e, f, h). Several partially cemented fractures lined with euhedral calcite crystals and kinematic apertures >8 mm were observed in well MF31-19G. The distribution of calcite-sealed fractures is heterogeneous. Partially cemented fractures are observed in core in close proximity to sealed fractures, as shown below in the subsection ‘Correlation between rock-mass calcite content and fracture fill’.

The different carbonate pore- and fracture-filling phases fall into distinct compositional groups (Fig. 4). Although of different colour in cathodoluminescence (CL) (Fig. 2h), the calcite pore and fracture cements cannot be distinguished based on their major and minor elemental composition. They are both of calcite end-member composition, with traces of magnesium and iron (Fig. 4). Pore-filling magnesium–iron-bearing carbonates tend to increase their iron content over time (i.e. ferroan-dolomite followed by ankerite). Iron content of carbonate in Piceance Creek Well 1 samples, for which microprobe analyses were performed, ranges from 12 mol% in the ferroan-dolomite cements to up to 28 mol% in the ankerite cements. The increase in iron goes along with a decrease in magnesium, whereas the calcium content does not change significantly. EDS analyses in samples from the other cores gave qualitatively similar trends.

Bulk-rock elemental and mineralogical (point count and X-ray diffraction (XRD)) analyses show a general decrease in Ca²⁺ and calcite abundance with depth in the Piceance Creek Well 1 (Fig. 5). Albite and Na⁺ content decrease with depth, consistent with albitization as the major source of Ca²⁺ for calcite precipitation. A positive spike in albite (and Ca²⁺) content at a depth of around 10 940 ft is accompanied by a higher concentration of ferroan-dolomite rather than calcite (Fig. 5).

However, no correlation is observed between albite (counted as plagioclase) and authigenic carbonate abundance if results are plotted for individual thin sections or core plug samples (shown in Fig. 7a based on XRD data for Piceance Creek Well 1). A good correlation is observed if results are averaged over the cored depth intervals (Fig. 7b). These trends between Ca²⁺ and total carbonate mineral content are unaffected by differences in abundance of carbonate species between XRD and petrographical point counts.

The δ¹⁸O_PDB isotope composition of calcite fracture cement and pore cement analysed for all three wells varies between −18.1 and −13.4‰; and δ¹³C_PDB values between −17.0 and −2.0‰ (Table 1; Fig. 8). In general, and particularly for well MF31-19G, the δ¹⁸O_PDB varies more for pore cement than for fracture cement. The larger spread in δ¹⁸O values for pore cement may reflect a wider range of temperatures over which pore cement formed compared to fracture cement. The δ¹³C composition of calcite cements generally becomes more positive with increasing core depth. This increase in δ¹³C with depth has also been reported for fracture cement by Pitman & Dickinson (1989). For analysed pairs of fracture and adjacent pore cement, both carbon and oxygen of fractures are heavier than pore cement for the Mamm Creek samples but lighter for the Piceance Creek Well 1 (Fig. 8). No systematic variation is observed for MF31-19G samples.

Strontium isotopic values, obtained for carbonate fracture cement for all three available wells, show a wide spread with highly radiogenic ratios (Table 1; Fig. 9). These Sr isotope ratios >0.7123 are distinctly higher than the seawater ratio during Late Cretaceous times (0.7078: McArthur et al. 2001). ⁸⁷Sr/⁸⁶Sr values for the Piceance Creek Well 1 range from 0.71493 to 0.71813, with shallower fracture samples having more radiogenic values. Sr isotope ratios for the M31-19G well shows a range from 0.71261 to 0.71739. Two shallow samples have a ratio of 0.7127, whereas three deeper samples sampled within 1 ft (0.3 m) show a range of 0.7126–0.7174 (Fig. 9; Table 1). ⁸⁷Sr/⁸⁶Sr values for the shallower Mamm Creek samples are comparable to the shallower fracture cements in the MF31-19G well.

Fluid-inclusion microthermometry was performed on fracture carbonate cements to determine the temperature conditions under which the cements formed. Calcite fracture cements containing fluid inclusions suitable for microthermometry were observed in two cores (MF31-19G and Piceance Creek Well 1). Both primary and secondary fluid-inclusion assemblages were observed. Primary inclusions appear in small 3D clusters and as isolated negative-crystal shaped inclusions (Fig. 6). The secondary inclusions occur as short, healed microfractures. The shape of the inclusions varies from irregular to negative crystal shapes; sizes range from c. 1 to c. 20 µm in diameter.

Calcite cements contain fluid-inclusion assemblages of coexisting two-phase aqueous and single-phase hydrocarbon gas inclusions, similar to inclusions trapped in quartz cements (Fall et al. 2012, 2015). The coexistence of these two fluid phases indicates that the inclusions were trapped in an immiscible fluid system (aqueous fluid + free hydrocarbon gas), and the measured homogenization temperatures (T_h) of the aqueous inclusions represent true trapping temperatures (Goldstein & Reynolds 1994). The aqueous inclusions at room temperature contain 5–10 vol% vapour. Homogenization temperatures range from c. 138 to c. 163°C (Fig. 10), with a T_h variation of c. 1–15°C within a single fluid-inclusion assemblage. Liquid–vapour phases are consistent within single fluid-inclusion assemblages. Final ice melting temperatures provided salinities ranging from c. 2.0 to 3.5 wt% NaCl equivalent for inclusions in calcite, with no systematic trends in salinity with homogenization temperature.

A correlation between host-rock composition and cements found in fractures could arise in several ways (Laubach 2003). For example, fractures might be conduits for fluids that deposit cements in fractures and adjacent areas of the host rock. Alternatively, host-rock composition might influence the type of cement found in fractures. For example, in sandstone, the widespread occurrence of quartz deposited concurrently with fracture opening in fractures formed by various processes at a range of depths (e.g. Becker et al. 2010; Laubach & Ward 2006; Fall et al. 2012, 2015; Hooker et al. 2015; Ukar et al. 2016) is likely to reflect the silica-dominated geochemistry of the sandstones (Lander & Laubach 2015). Correlation between host-rock carbonate minerals and carbonate minerals (or their absence) in fractures can therefore provide some constraints on the geochemical processes governing calcite distribution.

The degradation index uses host-rock pore cement abundance as a proxy for fracture cement abundance to predict the degree of cement infill in large fractures in cases where such fractures are not directly observable because of a limited exposure of fractures in core samples. The index makes no prediction about the presence of, or size of, the fractures themselves. The degradation index also does not predict open fractures if those fractures are narrower than a predetermined threshold size (emergent threshold). Such narrow fractures are assumed to be filled with synkinematic quartz. A corollary of assuming that quartz tends not to fill large fractures is that for host-rock analysis, the volume of quartz cement in the rock mass can be neglected. The degradation index is a ratio of small quantities, since the volumetrically dominant cement in tight gas sandstones (or in deeply buried sandstones generally) is typically quartz, and host-rock porosity is typically low. As a ratio of small numbers, it is susceptible to a variability of the index due to small differences in the inputs. The practical value of the index arises from the low probability of encountering fractures with cores. The index can be measured even if no fractures are found in core.

Samples for routine point-count analyses were collected without consideration of the occurrence of natural fractures in the core (most are archival point counts). Degradation index values for routine point-count analyses for 145 samples of the Piceance Creek Well 1 are plotted in Figure 11 as green and red circles for samples with ≤50% and >50% degradation, respectively. No degradation value is defined for 25 samples without preserved porosity and no calcite pore cement, plotted as black circles in Figure 11. Based on these degradation values, we divided the core into intervals predicted to contain sealed (orange shading in Fig. 11) and open (grey shading) fractures depending on whether the calculated degradation values are high or low. Boundaries between intervals were somewhat arbitrarily, placed at the half distance between adjacent samples with degradation of ≤50% and >50%. Samples of undetermined degradation (black circles) were not considered in predicting fracture degradation.

Out of 44 fractures, degradation of 35 fractures (80%) in Piceance Creek Well 1 is predicted correctly by the degradation index. Degradation is variable (Fig. 11). Several high-degradation index intervals where sealed fractures are predicted have sealed fractures (e.g. 8881, 8925, 9520, 9526 and 9601 ft) and low-index intervals where open fractures are predicted have open fractures (e.g. 9581 ft, numerous fractures between 10 930 and 10 950 ft and 12 301 ft). At other depths, the correlation is weak or absent (e.g. 9478 ft and 10 872–10 930 ft). In the depth range of 10 872–10 930 ft, negligible porosity leads to indeterminable degradation index for many samples. Degradation values in this depth range are based on low calcite and porosity counts (1–4 out of 250 total counts) and are thus considered statistically not robust. Although the analysis merely returns ‘no prediction’, we counted the sealed fractures in this interval as failed predictions. Overall, open fractures, with average opening displacement <0.5 mm, tend to be narrower than sealed fractures, with average opening displacements of about 1 mm and as much as 3.3 mm.

The degradation index, as designed, predicts fracture attributes where no macroscopic fractures were encountered. Because our analysis relied primarily on archival point-count data, the majority of point-count analyses reported thus far did not sample the exact same depth where fractures were observed. Some apparent mismatches between predictions and fracture observations may thus reflect local heterogeneity in pore and fracture cement. To test if a higher density of point-count samples would improve the prediction, we point-counted and determined the degradation index for nine additional samples that contain fractures (triangles in Fig. 11). Eight of these nine samples predict calcite fracture degradation correctly.

Carbonate pore cement includes calcite, ferroan-dolomite and ankerite. Ferroan-dolomite and ankerite predate calcite cement. Their distribution and appearance correlates with the primary detrital composition, only occurring in the presence of detrital dolomite (Ozkan et al. 2011). The absence of dissolution features on the detrital dolomite grains indicates that detrital dolomite is not the source for ferroan-dolomite cement. Instead, the limited occurrence of ferroan-dolomite and ankerite suggests that detrital dolomite forms a preferred substrate for ferroan-dolomite and ankerite. Stockmann et al. (2014) have shown that calcite favours a calcite substrate for nucleation and growth; a similar preference can be inferred for the precipitation of ferroan-dolomite on detrital dolomite (e.g. Warren 2000).

The observed iron increase from ferroan-dolomite to ankerite may either reflect an increase in Fe²⁺ activity relative to Ca²⁺ and Mg²⁺ or a decrease in temperature (Morad 1998). The formation of ferroan-dolomite and ankerite prior to fracture opening and prior to synkinematic quartz cementation of fractures under prograde burial conditions (Figs 2 and 3) (Fall et al. 2012, 2015) indicates that these carbonate phases formed during an increase in burial temperature. The observed increase in iron content of carbonate pore cement thus reflects a change in fluid chemistry rather than a temperature effect.

Calcite cement is the latest cement phase in both pores and fractures. Although pore and fracture calcite cement are indistinguishable under the transmitted microscope and in elemental composition as determined by EPMA, differences in isotopic composition between calcite fracture cement and adjacent pore calcite cement (Fig. 8), and the sharp contrast in CL colour between pore (bright red) and fracture calcite cement (darker red) (Fig. 2h) indicate that pore and fracture calcite cement are compositionally different, and thus precipitated under different pore fluid or temperature conditions and thus at different times.

Fluid inclusions in calcite fracture cement indicate precipitation from c. 138 to c. 163°C (Fig. 10) at lower temperatures than quartz precipitating at c. 140–188°C (Fall et al. 2012, 2015). This suggests that calcite fracture cement formed either prior to or after maximum burial. Calcite cement containing fluid inclusions lacked a clear cement stratigraphy, which prevented us from determining trends in temperature evolution trends for calcite. We thus cannot determine if fracture calcite precipitated during an increase or decrease in burial temperature. However, the sequence of the fracture cements, with early quartz cementation followed by calcite (Fig. 2b) (Fall et al. 2012, 2015) indicates that calcite fracture cement formed after peak burial and thus during exhumation. In addition, the crack-seal textures of some of the calcite fracture cements indicate that this calcite precipitated synkinematically with fracture opening. The overlap in the temperature range of calcite and quartz fracture cement suggests that calcite cementation partly overlapped with the precipitation of quartz within the sample set, and continued during partial exhumation after quartz cementation ceased. However, while quartz and carbonate fracture cementation overlaps for the entire sample set, no overlap is observed for any individual fracture through alternating calcite and quartz cement precipitation.

The good correlation between the albite and carbonate cements (calcite, ferroan-dolomite/ankerite) (Fig. 7b), observed when averaged over depth intervals of the entire core sections, suggests that albitization of plagioclase was the dominant source of Ca²⁺ for these carbonate minerals (with the exception of detrital dolomite, which is too sparse to affect the correlation). The strong correlation for albite against authigenic carbonate (Fig. 7b) averaged over core intervals, but the lack of correlation between individual core plug or thin section samples (Fig. 7a), suggests that Ca²⁺ was redistributed over the depth interval of the core segments (60–160 ft, 18–49 m).

A spike in the albite content in Piceance Creek Well 1 at a depth of around 10 940 ft (3335 m) is accompanied by an increased abundance of ferroan-dolomite and a lower abundance of calcite (Fig. 5). The low abundance of calcite correlates with the preserved fracture porosity (Fig. 11) as predicted by the fracture degradation index. We explain the low abundance of calcite in this layer by the high concentration of ferroan-dolomite, which relates to a higher abundance of detrital dolomite. The lower solubility of dolomite compared to calcite (Langmuir 1997) makes dolomite less conducive to dissolution and reprecipitation compared to calcite, thus allowing fractures in this interval to preferentially remain open.

The strontium isotopic composition of fracture calcite cement plotting distinctly above the Cretaceous seawater strontium ratio (Fig. 9) indicates that dissolution of marine carbonates does not constitute a significant source of strontium and, by proxy, calcium for fracture calcite. The strontium ratios are also significantly higher than lacustrine carbonates of the overlying Wasatch and Green River formations (Davis et al. 2009). Plagioclase dissolution and the albitization of plagioclase have been viewed as potentially significant sources of strontium in oilfield waters and for diagenetic calcite (Land 1987; Schultz et al. 1989). The ⁸⁷Sr/⁸⁶Sr ratios of plagioclase minerals of young plutonic and volcanic rocks are commonly lower than marine strontium isotopic ratios (Schultz et al. 1989; Faure & Mensing 2004). Likely Precambrian source rocks for Mesaverde Group sediment of the Laramide uplift are characterized by high radiogenic ⁸⁷Sr/⁸⁶Sr ratios of bulk rock (0.731–0.939: Rhodes et al. 2002) and plagioclase (0.720: Clow et al. 1997; 0.738: Patel et al. 1999). However, in the absence of unaltered feldspar available for strontium isotopic analysis in Mesaverde sandstone, we cannot exclude clay diagenesis as a contributor of radiogenic strontium. The wide range in strontium isotopic composition, including samples within close proximity (<1 ft, <30 cm) to each other, suggests a heterogeneous source of strontium and ineffective mixing of strontium by advective or diffusive processes on this scale (Fig. 9). This notion of inefficient mixing of strontium is supported by the wide spread in the distribution of strontium isotope ratios when plotted against strontium concentration (Fig. 9).

The range of δ¹³C of pore and fracture calcite cement between −17 and −2‰ (Fig. 8b) is likely to reflect a mixed carbon source that includes carbon derived from the decarboxylation of organic matter (around −10 to −25‰: Bjørlykke 2010) and inorganic carbon derived from carbonate dissolution (around 0‰ for marine carbonate). Taylor et al. (2000) described the dissolution of detrital dolomite in stratigraphically lower shallow-marine to fluvial sandstone of the Mesaverde Group (Castlegate Sandstone) of the Book Cliffs. Evidence of detrital dolomite dissolution is not observed in our samples. The increase in δ¹³C with depth is consistent with an increasing component of inorganic carbon and carbonate dissolution in deeper sections of the Mesaverde Group but points away from a possible contribution of inorganic carbon due to the dissolution of carbonates in the overlying Wasatch and Green River formations. The variation in δ¹³C observed in fractures <2 ft (<60 cm) apart suggests a carbon pool that is heterogeneous in time and/or space. Fracture cements are generally too thin for the collection of multiple samples from the same fracture and lack growth zonation in CL that would allow cement-layer specific sampling for the determination of isotopic trends within fracture cements.

Using our fluid-inclusion trapping temperature of 135–165°C and the fractionation equation of Friedman & O'Neil (1977), fracture calcite δ¹⁸O values constrain the δ¹⁸O composition of the precipitating pore fluid to −2 to +5‰_SMOW. This range is consistent with either the diagenetic alteration of meteoric to brackish connate water or the mixing of diagenetically altered connate water with meteoric water entering the formation at the time of calcite precipitation. Modern spring water in the Piceance Creek Basin is characterized by a δ¹⁸O_SMOW composition of about −17‰ (Kimball 1984). Measurements of methane concentrations in fluid inclusions and equation-of-state calculations by Fall et al. (2012, 2015) indicate that the carbonate fracture cement precipitated under near-lithostatic pore-fluid pressure conditions. These fluid pressure conditions would prevent the influx of meteoric water into the formation at the time of fracture cementation. We therefore interpret the inferred porewater composition to represent the diagenetic modification of connate water of mixed meteoric to marine brackish composition, consistent with a paludal to fluvial depositional environment of the formation.

The differences in δ¹⁸O_PDB of adjacent fracture and pore cement of up to 2.2‰ amounts to a temperature difference of 19–36°C assuming a constant water δ¹⁸O_SMOW composition in the range of −2 to +5‰. For the Mamm Creek samples, fracture cement precipitated at apparent lower temperature than the pore cement, whereas the opposite difference is observed for Piceance Creek Well 1 (Fig. 8). These trends can be interpreted as the result of an increase in burial temperature between pore and fracture cementation at Mamm Creek and a decrease at Piceance Creek Well 1, consistent with our observation that fracture cement post-dates pore cement at both locations. This implies that fracture cementation at Mamm Creek occurred close to peak burial conditions and after peak burial conditions at Piceance Creek. Following Fall et al. (2012, 2015), we assume that fluid-inclusion temperatures reflect the ambient burial temperature at the time of fracture cementation without significant advective heat transport across stratigraphic boundaries. This interpretation also assumes that the water δ¹⁸O composition remained constant during the time span of calcite fracture cement precipitation, with diagenetic water–rock interaction largely completed at this late stage in the burial history. The observed isotopic trend towards lighter δ¹⁸O at Piceance Creek could be explained with isotopic fractionation occurring during fracture precipitation in a closed system. While it is conceivable that calcite fracture cementation results from the continued albitization of plagioclase, our observation of dissolution textures of calcite pore cement in MF31-19G samples (Fig. 2g) can be interpreted to suggest that calcite fracture cement results from the dissolution and reprecipitation of earlier formed pore calcite cement. This interpretation, rather than direct precipitation during albitization, is more consistent with the reported temperature range of albitization in Gulf Coast sandstones of 110–120°C (Boles 1982), which is distinctly lower than the c. 138–163°C temperature range for the fracture calcite formation (Fig. 10) reported by Fall et al. (2012, 2015).

The recognition that fracture degradation by calcite cement is not sensitive to the timing of calcite pore cement precipitation opens up the possibility that degradation can be calculated using a modal mineral calculation based on bulk-rock chemical analyses. Such data can be obtained from cuttings at the wellsite and combined with observations of sparry calcite in cuttings as an indicator of fracture cement. Continuous monitoring of carbonate content in the host rock based on cuttings while drilling could provide a measure of formation flow properties that could inform well completion strategies.

Compared to calcite, dolomite has a lesser tendency to dissolve and reprecipitate (Langmuir 1997). We have shown that open fractures correlate with an increased abundance of ferroan-dolomite cements at a depth of around 10 940 ft in the Piceance Creek Well 1. A similar correlation between open fractures and increased dolomite cement has been recorded in the MXW well of the southern Piceance Basin (Finley & Lorenz 1988; Pitman et al. 1989). We speculate that this correlation results from either the lower availability of pore space for later calcite precipitation in dolomite-cemented layers and/or from the depletion of the pore fluid in Ca²⁺ by ferroan-dolomite precipitation, with ferroan-dolomite acting as a local Ca²⁺ sink, thus diminishing the propensity for later calcite precipitation. In addition, we note that dolomite fracture cement is consistently absent, even in depth intervals containing dolomite pore cement suggesting that, in contrast to prekinematic calcite, early dolomite appears less prone to dissolution and reprecipitation after fracture formation. We thus recommend including only dolomite that can be shown to be post-kinematic into degradation index calculations.

Lastly, the new evidence we provide of controls on calcite precipitation allows us to propose that for fractures of width >1 mm the propensity for fracture sealing can be predicted to first order in the absence of core samples provided that information about host-rock type, mineral composition and the burial diagenetic evolution is available. Information on burial diagenetic evolution needed for this approach includes the types of diagenetic reactions, the scale of mass transport between the host rock and fracture, and the temperature, pressure and fluid compositional conditions. While all of these conditions may not be closely constrained for an individual location or target interval prior to drilling, we anticipate that meaningful qualitative reservoir-scale predictions can be derived with information available from outcrop, seismic and exploratory wells. These diagenetic predictions have to be paired with structural fracture evolution models for meaningful pre-drill predictions of fractured reservoir flow behaviour.

Natural fractures in tight gas sandstones of the Mesaverde Group are generally lined or, if <1 mm wide, completely filled by quartz. Fractures >1 mm wide are typically lined with quartz that is frequently, but not always, followed by late post-kinematic fracture-occluding calcite, with the distribution of calcite varying spatially between adjacent layers. Whether fractures are partially open or sealed can significantly influence fracture permeability, and thus exploration and development decisions of tight sandstone reservoirs.

Using pore and fracture cement petrography, fluid inclusions, and isotopic and elemental analyses, we demonstrate that host-rock calcite distribution and remobilization govern the porosity degradation of fractures by calcite. Fluid-inclusion analyses indicate that calcite cement formed at 135–165°C. Both the albite and calcite content decreases with depth, consistent with a greater fracture porosity in lower parts of the unit. Thus, the host-rock albite content predicts fracture-damaging carbonate cements. ⁸⁷Sr/⁸⁶Sr ratios of calcite cement are consistent with the albitization of detrital anorthitic plagioclase providing the source of Ca²⁺. As the anorthite component dissolves, albite remains as a porous daughter phase and released Ca²⁺ drives the precipitation of carbonate cements.

The pore calcite cement content of the rock mass, quantified by the degradation index, correlates with, and can be used to accurately predict, depths where fractures >1 mm wide are sealed or open. We explain this correlation with the ease of calcite remobilization by dissolution and reprecipitation as fracture-occluding cement. Dolomite pore cement, on the other hand, correlates with open fractures, indicative of the limited remobilization of earlier dolomite cement. The approach of predicting whether fractures are sealed or remain open using the degradation index requires access to core material but attempts to extrapolate observed spatial and depth trends in carbonate cement to other parts in the same basin based on a fundamental understanding of carbonate reactions and on prior knowledge of the depositional facies distribution in the basin.

J.S. Davis and B. Faulkner contributed to the study with discussions, and by coordinating sample collection and data release. K. Milliken (BEG) assisted with the microprobe analyses. We thank the anonymous reviewers for comments and suggestions for clarification. Publication is authorized by the Director, Bureau of Economic Geology, The University of Texas at Austin.

This work was supported by the ExxonMobil–Bureau of Economic Geology (The University of Texas at Austin) Collaborative Study on Unconventional Reservoirs 2008–2010. Additional funding was provided by grant DE-FG02-03ER15430 from Chemical Sciences, Geosciences and Biosciences Division, Office of Basic Energy Sciences, Office of Science, United States Department of Energy, and by the Fracture Research and Application Consortium.