Search

Two end-member approaches to 3-D reservoir characterization modeling exist. The “Frankenstein” approach integrates hard and soft conditioning data while incorporating geologic concepts and interpretations, to construct a highly detailed deterministic model. The Frankenstein approach focuses on what is known about the reservoir, and is so named because construction of the model can be time-consuming, which can negatively impact project timelines, sometimes “killing” its creator. The “Gilligan” approach uses available data to build suites of models that are simple, and strives to address what is not known about the reservoir. Gilligan models are simplified, at least in the view of the Frankenstein modeler, and focus on quantitative characterizations of uncertainty. Reservoirs usually become more complex with new data from wells and production, a concept referred to here as “the Law of Increasing Reservoir Complexification”. The Law of Increasing Reservoir Complexification should influence earth modeling workflow choices. Currently, the Gilligan probabilistic modeling approach is more commonly employed in early phases of projects. The appearance of overly complex Frankenstein models when probabilistic models are needed remains an ongoing problem. Mature reservoirs typically possess uncertainties that could influence business decisions, and may warrant a Gilligan probabilistic approach as well. No single recipe exists for choosing the best 3-D modeling approach: a fit for purpose or Goldilocks modeling approach should depend on business needs and requirements. Recognition that uncertainty is a fundamental and prevalent driver of business value for all reservoirs in all phases of development demands widespread adoption of probabilistic or Gilligan approaches.

Unconventional opal-A and opal-CT diatomaceous reservoirs in the Monterey Formation of the San Joaquin Valley in California have produced for more than 30 yr. These opal-A reservoirs are characterized by high porosity (>50%) and low permeability ( 45% [Cymric and McKittrick fields]) or are very thick (>1000 ft thick [Lost Hills field]). These insights provide general guidance about characterization of unconventional diatomaceous reservoirs and the identification of factors that most dramatically influence performance.

The upper Miocene Kern River Formation at Kern River field is a thick succession of braided stream deposits. Because the oil field has been penetrated by more than 21,000 wells, with well spacing typically of 2.5 ac (104 m2), high-quality imaging of the reservoir can be accomplished using well logs. The field contains more than 650 observation wells recording temperatures and saturation, such that heating and drainage of the reservoir can be characterized over time. Oil and gas saturations in a 360-ac (1.5 × 106-m2) study area within the Kern River field were modeled in three dimensions over time, creating a four-dimensional (4-D) earth model. Static lithology and porosity properties were first created. Then, temperature, oil, and gas saturations were kriged using data from observation wells at different time increments. From these kriged models, difference models were created to show changes in oil and gas saturation with time. Changes in oil volumes in the 4-D model matched produced volumes within a 10% error. Next, categorical descriptions of net oil saturation were made (best oil, background oil, partially drained, and residual saturations), and temporal evolution of the volumes of the different categories was modeled. By characterizing the reservoir in detail using the well-log data to define a reservoir framework and the observation wells to characterize changes over time, significant challenges in steam-flood reservoir management can be addressed, including determining which parts of the reservoir are producing the most and recognizing parts of the reservoir that are not actively draining.

A continuous oil accumulation is one that is pervasive throughout a large area and is not affected by natural hydrodynamic influences. Three source rocks in the San Joaquin Valley of California are actively producing hydrocarbons and represent potential continuous oil accumulations: the Monterey, Kreyenhagen, and Moreno Formations. The Energy Information Administration announced in 2014 that there are potentially 15 billion bbl of recoverable oil in the Monterey Formation in California, spiking huge interest. Such a resource would make the Monterey by far the largest continuous oil accumulation in North America. This number has since been reduced dramatically to 600 million bbl of oil for the state and 21 million bbl of oil within the San Joaquin Valley. In this study, the subsurface character of the Bakken Formation of North Dakota is compared with the three source rocks in the San Joaquin Valley at oil window depths. To characterize these reservoirs, hundreds of well logs, core descriptions, and mud logs were studied. A technique to rank character of oil show data was developed in which interesting, possibly interesting, or not interesting wells were located on thermal maturity maps. Interesting wells have significant oil shows, whereas not interesting wells have minor or no shows. In the Bakken Formation, the character of the oil show correlates with well productivity. Applying this same classification to the San Joaquin Valley source rocks leads to a more disappointing conclusion. Although there are oil shows in the source rocks of the San Joaquin Valley at oil window depths suggesting the presence of a continuous oil accumulation, the distribution of shows is both laterally and vertically heterogeneous and not predictable. Moreover, recent attempts to produce from source rocks at these depths have not been economically successful. We conclude that the three source rocks in the San Joaquin Valley represent heterogeneous and discontinuous oil accumulations at oil window depths in the subsurface. Likely there are billions of barrels of oil in these discontinuous oil accumulations. Source rocks in the Bakken are rated as world class; source rocks in the San Joaquin Valley are good to excellent quality. However, the quality of the oil shows in the San Joaquin Valley appears poorer than the Bakken Formation. It is possible that effective drainage between the source rocks and the up-dip reservoirs has left large volumes of the source rocks at oil window depths with only residual oil saturation. Complex structural and stratigraphic architecture, heterogeneity, and continuity create issues of predictability for optimal areas to target. Rapid rates of subsidence over the past few million years and accompanying thrusting and folding resulted in a complex subsurface pressure regime. The lack of clear hydraulic fracture targets, analogous to the middle Bakken, further complicates drilling decisions and likely deliverability. In addition, the oil windows in the San Joaquin Valley are significantly deeper than the Bakken Formation, which would result in substantially higher well cost. Recent drilling results support this study and suggest that heterogeneous and discontinuous oil accumulations in the San Joaquin are unlikely to become economic without dramatic changes in technology.

The Lombardi reservoir at San Ardo oil field in California is extremely sand rich (>90%) and appears to be a relatively simple reservoir several hundred feet thick. The purpose of this study is to characterize stratigraphic complexities in the Lombardi reservoir that may affect later production and interpret how changes in stratigraphic architecture are tied to local syndepositional tectonism. Two parasequence sets are present in the Lombardi: the reservoir exhibits clinoform geometries along the western edge of the field and stacking patterns within the reservoir change across the field. Based on stacking pattern, core, and correlation studies, two possible sequence boundaries are recognized in the Lombardi sand. A lower sequence boundary is proposed approximately in the middle of the Lombardi sand, and an upper sequence boundary is proposed at the top of the Lombardi sand. In the southern part of the field, the Lombardi sand consists of two progradational parasequence sets, separated by the mid-Lombardi sequence boundary. The upper sequence boundary clearly truncates underlying folded Lombardi sand. Just a few miles (∼5 km) away, in the northwest part of the field, the Lombardi sand consists of a lower progradational parasequence set and an upper retrogradational parasequence set, separated by the mid-Lombardi sequence boundary. The upper sequence boundary at the top of the Lombardi sand shows little evidence of truncation of underlying units and is recognized by onlapping of younger strata. The difference in reservoir architecture between the south and northwest part of the field, a distance of only a few miles (∼5 km), is interpreted to be related to syndepositional uplift accompanying tectonism in the south. In core, the Lombardi sand consists mostly of homogeneous sand and bioturbated sand, arranged in upward-coarsening successions. The Lombardi sand is interpreted to be the product of deltaic deposition based on (1) the presence of homogeneous and laminated sandstones interpreted to represent mass flow deposits and (2) its restricted geographic distribution within the Salinas basin. No evidence of deposition in a wave-dominated or tidal-dominated shoreline is indicated. The interpretation presented here is that the sediments were deposited on the front of a delta emptying into a wave-protected estuary or bay, with minimal tidal effects. Complexities in stratigraphic architecture along the western edge of the field do not currently affect production but will likely become important during later stages of field development. That is, in the developed part of the field, reservoir continuity is not an issue, because of high sand content and sheet-like geometries. However, at the margins of the field, which will be produced in the future, clinoform geometries and sandstone tongues clearly affect reservoir continuity and will create vertical compartments during draining.