This study presents results of outcrop characterization and modeling of lithologic heterogeneity within a well-exposed point bar of the Williams Fork Formation in Coal Canyon, Piceance Basin, Colorado. This deposit represents an intermediate-scale depositional element that developed from a single meandering channel within a low net-to-gross ratio fluvial system. Williams Fork outcrops are analogs to petroleum reservoirs in the Piceance Basin and elsewhere. Analysis and modeling of the point bar involved outcrop measurements and ground-based high-resolution light detection and ranging data; thus, the stratigraphic frameworks accurately represent the channel-fill architecture. Two- and three-dimensional (2-D and 3-D) outcrop models and streamline simulations compare scenarios based on different lithologies, shale drapes, observed grain-size trends, petrophysical properties, and modeling methods. For 2-D models, continuous and discontinuous shale drapes on lateral-accretion surfaces result in a 79% increase and 24% decrease in breakthrough time (BTT), respectively, compared to models without shale drapes. The discontinuous shale drapes in the 2-D and 3-D models cause a 30% and 107% decrease, respectively, in sweep efficiency because they focus fluid flow downward to the base of the point bar. For similar reasons, 2-D models based on grain size exhibit 67–267% shorter BTT and 44–57% lower sweep efficiency compared to other model scenarios. Unlike the 2-D models, the continuous shale drapes in the 3-D models cause the fluid front to spread out and contact more of the reservoir, resulting in 42–53% longer BTT and 41–52% higher sweep efficiency compared to the other models. These results provide additional insight into the significance of intermediate-scale heterogeneity of fluvial reservoirs.
Artificial intelligence methods have a very wide range of applications. From speech recognition to self-driving cars, the development of modern deep-learning architectures is helping researchers to achieve new levels of accuracy in different fields. Although deep convolutional neural networks (CNNs) (a kind of deep-learning technique) have reached or surpassed human-level performance in image recognition tasks, little has been done to transport this new image classification technology to geoscientific problems. We have developed what we believe to be the first use of CNNs to identify lithofacies in cores. We use highly accurate models (trained with millions of images) and transfer learning to classify images of cored carbonate rocks. We found that different modern CNN architectures can achieve high levels of lithologic image classification accuracy (approximately 90%) and can aid in the core description task. This core image classification technique has the potential to greatly standardize and accelerate the description process. We also provide the community with a new set of labeled data that can be used for further geologic/data science studies.
This study addresses the field-scale architecture and static connectivity of fluvial sandstones of the lower Williams Fork Formation through analysis and reservoir modeling of analogous outcrop data from Coal Canyon, Piceance Basin, Colorado. The Upper Cretaceous lower Williams Fork Formation is a relatively low net-to-gross ratio (commonly <30%) succession of fluvial channel sandstones, crevasse splays, flood-plain mudstones, and coals that were deposited by meandering river systems within a coastal-plain setting. The lower Williams Fork outcrops serve as proximal reservoir analogs because the strata dip gently eastward into the Piceance Basin where they form natural gas reservoirs. Three-dimensional architectural-element models (3-D reservoir models) of the lower Williams Fork Formation that are constrained to outcrop-derived data (e.g., sandstone body types, dimensions, stratigraphic position) from Coal Canyon show how static sandstone body connectivity is sensitive to sandstone body width and varies with net-to-gross ratio and well spacing. With a low well density (e.g., 160-ac well spacing), connectivity is low for net-to-gross ratios less than 20%, connectivity increases between net-to-gross ratios of 20 to 30%, and levels off above a net-to-gross ratio of 30%. As well density increases, static connectivity increases more linearly with an increasing net-to-gross ratio. For a 20-ac well spacing, static connectivity can range from approximately 35 to 75% and 45 to 80% for net-to-gross ratios of 10 and 15%, respectively, depending on sandstone body width. Given the lower net-to-gross ratio and continuity of lower Williams Fork deposits, this underscores the importance of representative sandstone body statistics (e.g., sandstone body type, dimensions) to aid in subsurface correlation and mapping and to constrain reservoir models.
This study addresses the field-scale architecture and dimensions of fluvial deposits of the lower Williams Fork Formation through analysis of outcrops in Coal Canyon, Piceance Basin, Colorado. The lower Williams Fork Formation primarily consists of mud rock with numerous isolated, lenticular to channel-form sandstone bodies that were deposited by meandering river systems within a coastal-plain setting. Field descriptions, global positioning system traverses, and a combination of high-resolution aerial light detection and ranging data, digital orthophotography, and ground-based photomosaics were used to map and document the abundance, stratigraphic position, and dimensions of single-story and multistory channel bodies and crevasse splays. The mean thickness and apparent width of the 688 measured sandstone bodies are 12.1 ft (3.7 m) and 364.9 ft (111.2 m), respectively. Single-story sandstone bodies ( N = 116) range in thickness from 3.9 to 29.9 ft (1.2 to 9.1 m) and from 44.1 to 1699.8 ft (13.4 to 518.1 m) in apparent width. Multistory sandstone bodies ( N = 273) range in thickness from 5.0 to 47.1 ft (1.5 to 14.4 m) and from 53.2 to 2791.1 ft (16.2 to 850.7 m) in apparent width. Crevasse splays ( N = 279) range in thickness from 0.5 to 15.0 ft (0.2 to 4.6 m) and from 40.1 to 843.3 ft (12.2 to 257.0 m) in apparent width. These data show that most sandstone bodies are smaller than the distance between wells at 10-ac spacing (660 ft [201 m]). Analyses of interwell sandstone-body connectivity suggest that even at 10-ac spacing, only half of the sandstone bodies are intersected and few are intersected by more than one well.
Petrophysical data from dolomite outcrops of the Mississippian Madison Formation at Sheep Canyon, Wyoming, exhibit three scales of lateral variability in single rock fabric units. These include a near-random component (nugget effect), a short-range structure, and a long-range cyclic trend (hole effect). The nugget effect is high and accounts for 31–39 and 48–50% of the variance in porosity and permeability, respectively. Short-range lateral variability is reflected by correlation lengths of 6.5–16 ft (2–5.5 m). Laterally, long-range periodicities are equivalent to approximately 10% of the petrophysical variance and have wavelengths of 31 and 140 ft (9.5 and 42.6 m) for porosity and permeability (55 ft [16.8 m] for log10 permeability), respectively. Cross sectional and plan-view petrophysical models and streamline simulations explore the effects of these scales of heterogeneity on fluid flow. Although short-range variability accounts for most of the petrophysical heterogeneity, the longer range trends can significantly affect fluid-flow behavior. Results indicate that breakthrough time and sweep efficiency vary depending on the magnitude of the lateral, long-range, petrophysical variability that exists in a dolomite reservoir. As the component of the long-range periodicity (hole effect) increases from approximately 0 to 25% of the total petrophysical variability, a corresponding increase in breakthrough time and sweep efficiency occurs. However, as the magnitude of the lateral, long-range, petrophysical variability increases beyond 25% of the total petrophysical variability (e.g., from 25 to 50%), a corresponding reduction in breakthrough time occurs because the spatial continuity of permeability is greater. Results indicate that heterogeneity caused by lateral petrophysical cyclicity should be incorporated into dolomite reservoir models for hole effect magnitudes that are greater than 10% of the petrophysical variance. To properly characterize and model these scales of variability in a petroleum reservoir, outcrop analogs are essential to provide accurate quantitative descriptions of lateral variability in dolomite rock fabrics.
The distribution and orientation of faults, fracture intensity and seismic-reflection characteristics of the Mesaverde Group (Williams Fork and Iles formations) at Mamm Creek Field vary stratigraphically, and with lithology and depositional setting. For the Mesaverde Group, the occurrence of faults and natural fractures is important as they provide conduits for gas migration, and enhance the permeability and productivity of the tight-gas sandstones. The Upper Cretaceous Mesaverde Group represents fluvial, alluvial-plain, coastal-plain and shallow-marine depositional environments. Structural interpretations based on three-dimensional (3D) seismic-amplitude data, ant-track (algorithm that enhances seismic discontinuities) seismic attributes and curvature attributes are utilized jointly to understand the complex fault characteristics of the Williams Fork Formation. This study reveals that the lowermost lower Williams Fork Formation is characterized by NNW- and east–west-trending small-scale thrust and normal faults. Study suggests that the uppermost lower Williams Fork Formation, and the middle and upper Williams Fork formations, exhibit NNE- and east–west-trending arrays of fault splays that terminate upwards and do not appear to displace the upper Williams Fork Formation. In the uppermost Williams Fork Formation and Ohio Creek Member, NNE-trending discontinuities are displaced by east–west-trending events and the east–west-trending events dominate. Fracture analysis, based on borehole-image logs, together with ant-track and attenuation-related seismic attributes, illustrates the spatial variability of fracture intensity and lithological controls on fracture distribution. In general, higher fracture intensity occurs within the southern, southwestern and western portions of the field, and fracture intensity is greater within the fluvial sandstone deposits of the middle and upper Williams Fork formations. More than 90% of natural fractures occur in sandstones and siltstones. In situ stress analysis, based on induced-tensile fractures and borehole breakouts, indicates a NNW orientation of present-day maximum horizontal stress ( ), an approximate 20° rotation (in a clockwise direction) in the orientation of with depth and an abrupt stress shift below the Williams Fork Formation within the Rollins Sandstone Member.