Litho-fluid Facies Research Articles

Bayesian joint inversion of seismic and electromagnetic data for reservoir lithofluid facies, including geophysical and petrophysical rock properties

A Bayesian approach is developed to estimate lithofluid facies and other rock properties conditioned on seismic and electromagnetic data for reservoir characterization. Prior distributions are assumed to be facies-related Gaussian modes of geophysical rock properties directly acquired or converted from petrophysical properties by calibrated rock-physics modeling. An original generalization includes two distributions in the same marginalization integral, analytically solved under a linearized Gaussian assumption to provide a facies model likelihood conditioned on geophysical data. Because computing this probability for all possible facies configurations may be impractical, a Markov chain Monte Carlo algorithm efficiently samples models to provide a full posterior distribution. The linearized Gaussian approach allows the computation of the conditional distributions of geophysical and petrophysical rock properties by applying local deterministic inversions over the many sampled facies models. The inversion uses simulated geophysical data from a 1D synthetic model based on the geologic scenario and a well from a selected marine oil field. Two other wells from the same reservoir are used to gather prior distributions. Data from the well, calibration of the rock-physics modeling, and facies matching between the priors and the synthetic model are presented and discussed. Numerical tests validate nonlinear forward-modeling adaptations based on the assumed linearized Gaussian approach. The simulated stand-alone and joint geophysical data sets are then inverted for lithofluid facies models under different prior inputs. Two challenging geoelectric scenarios also are tested, one with lower resistivity contrasts and another with a misguided background model. All results demonstrate a gain in precision and accuracy when associating these geophysical signals to estimate the oil column. Facies-conditioned inversions for the rock properties also indicate potential for quantitative reservoir interpretations.

Quantifying thin heterogeneous gas sand facies of Rehmat gas field by developing petro elastic relationship in fine stratigraphic layers through bayesian stochastic seismic inversion

The Lower Goru Formation (LGF) sand intervals are prolific producers in the Central Indus Basin (CIB), Pakistan. The B-Interval of LGF is a proven reservoir of the Rehmat gas field, having heterogeneous and thin sand layers (14–17 m) below seismic tuning thickness. To quantify these potential layers, a petro-elastic relationship is developed using seismic and well data. Well data although having high resolution, lacks in many aspects, such as poor density logs due to bad boreholes along with the missing shear sonic (Vs), adding more challenges to characterize a complex reservoir. Therefore, an extended cemented sandstone model is established by incorporating the petro-elastic models (PEMs) of the reservoir's litho-fluid facies in the rock physics modeling (RPM) framework. RPM compensated for the lithological variability, quartz cementation impact, missing and bad density logs. The missing low and high frequencies of band-limited seismic data are retrieved through constraining wells by geostatistical parameters for enhanced visualization of thin-bedded heterogeneous sand. Bayesian stochastic seismic inversion (BSSI) addressed reservoir challenges by incorporating the PEMs responses along with consigning the frequencies limitation, through a low-frequency model and variograms from the sparse well data for high frequencies. For optimized reservoir characterization, a 3D seismic dataset is inverted in a fine-scale stratigraphic grid of 0.1 ms interval yielding high-resolution elastic properties. These enhanced elastic attributes are integrated with petrophysical relationships simulated through probability density functions that permit elastic attributes to be translated into reservoir characteristics (geological facies). So, an integrated technique of rock physics modeled logs with BSSI successfully identified thin heterogeneous gas sands through several high-resolution realizations with associated uncertainty.

Estimation of Litho-Fluid Facies Distribution from Zero-Offset Acoustic and Shear Impedances

Seismic data are considered crucial sources of data that help identify the litho-fluid facies distributions in reservoir rocks. However, different facies mostly have similar responses to seismic attributes. In addition, seismic anisotropy negatively affects the facies predictors extracted from seismic data. Accordingly, this study aims at estimating zero-offset acoustic and shear impedances based on partial-stack inversion by two methods: statistical modeling and a multilayer feed-forward neural network (MLFN). The resulting impedance volumes are compared to those obtained from isotropic simultaneous inversion by using impedance logs. The best impedance volumes are applied to Thomsen’s anisotropy equations to solve for the anisotropy parameters Epsilon and Delta. Finally, the shear and acoustic impedances are transformed into elastic properties from which the facies and fluid distributions are predicted by using the logistic regression and decision tree algorithms. The results obtained from the MLFN show better matching with the impedance and facies logs compared to those obtained from isotropic inversion and statistical modeling.

Joint impedance and facies inversion of time-lapse seismic data for improving monitoring of CO2 incidentally stored from CO2 EOR

Time-lapse seismic monitoring is an effective and proven technology for mapping the distribution of CO2 in a subsurface reservoir. When injected CO2 displaces other reservoir fluids, porous-medium properties are changed and thus the seismic impedance changes, causing time-lapse seismic amplitude differences in the injection zones. The analysis and interpretation of images created from these amplitude differences can provide information about reservoir architecture and the CO2 migration within the reservoir. Incorporating seismic inversion and rock physics into the interpretation of time-lapse seismic data can considerably improve the modeling and monitoring to detect and assess the location of CO2 over time. The joint inversion method presented in this paper has an integral representation of the geology in the inversion algorithm using elastic facies, which provides information about the spatial distribution of the geologic heterogeneities controlling the movement of fluids in the reservoir. The method was successfully applied to time-lapse seismic data from a mature oil field undergoing CO2 enhanced oil recovery. The estimated seismic acoustic impedances and facies reflect the characteristics of individual geologic facies and fluid conditions of the reservoir subject to CO2 injection. The probabilities estimated by the joint impedance and facies inversion for the reservoir's litho-fluid facies can be used for forecasting CO2 saturation and pressure changes within the target reservoir.

Rock-physics machine learning toolkit for joint litho-fluid facies classification and compaction modeling

Correct lithofacies interpretation sourced from wireline log data is an essential source of prior information for joint seismic inversion for facies and impedances, among other applications. However, this information is difficult to interpret or extract manually due to the multivariate and high dimensionality of wireline logs. Facies inference is also challenging for traditional clustering-based approaches because pervasive compaction trends affect a number of petrophysical measurements simultaneously. Another common pitfall in automated clustering approaches is the inability to account for underlying diagenetic processes that correlate with depth. Here, we address these challenges by introducing a rock-physics machine learning toolkit for joint litho-fluid facies classification. The litho-fluid types are inferred from the borehole data within the objective framework of a maximum-likelihood approach for latent facies variables and rock-physics model parameters, explicitly accounting for compaction and depth effects. The inference boils down to an expectation-maximization (EM) algorithm with strong spatial coupling. Each litho-fluid type is associated with an instance of a particular rock-physics model with a unique set of fitting parameters, constrained to a physically reasonable range. These fitting parameters in turn are inferred using bound-constrained optimization as part of the EM algorithm. Outputs produced by the toolkit can be used directly to specify the necessary prior information for seismic inversion, including per-facies rock-physics models and facies proportions. We present an example application of the tool to real borehole data from the North West Shelf of Australia to illustrate the method and discuss its characteristic features in depth.

Reservoir facies design and modeling using probabilistic rock-physics templates

We have developed a technique to design and optimize reservoir lithofluid facies based on probabilistic rock-physics templates. Subjectivity is promoted to design possible facies scenarios with different pore-fluid conditions, and quantitative simulations and evaluations are conducted in facies model selection. This method aims to provide guidelines for reservoir-facies modeling in an exploration setting in which limited data exist. The work includes two parts: facies-model simulations and uncertainty evaluations. We have first derived scenarios with all possible fluid types using Gassmann fluid substitution. We designed models with different numbers of facies and pore-fluid conditions using site-specific rock-physics templates. Detailed facies simulations were conducted in the petroelastic, elastic, and seismic domains in a step-by-step framework to preserve the geologic interpretability. The use of probabilistic rock-physics templates allowed for multiple realizations of each facies model to account for different types and magnitudes of errors and to infer facies probability and uncertainty. For each realization, we used Bayesian classification to assign facies labels. Comparisons between the predicted and true labels provided the success rates and entropy indices to quantify the prediction errors and confidence degrees, respectively. This workflow was tested with well-log data from a clastic reservoir in the Gulf of Mexico. We simulated models with five to seven facies with different pore-fluid parameters. From the petroelastic, elastic, and seismic domains, the uncertainty of facies models significantly increased due to well-log measurement errors, data-model mismatch, and resolution differences. The facies model consisting of oil sand, gas sand, and shale was the optimal set based on the high success rates and low entropy indices. Facies profiles estimated from this optimal model presented significant consistency with well-log interpretations. The techniques and results demonstrated here could be applied to different types of clastic reservoirs, and they provide useful constraints for reservoir facies modeling during early oilfield exploration stages.

Creating probabilistic 3D models of lithofluid facies using machine-learning algorithms

Mapping facies variations is a fundamental element in the study of reservoir characteristics. From identifying a pay zone to estimating the reservoir capacity, a hydrocarbon field’s development plan depends to a great extent on a reliable model of lithofacies and fluid content variations throughout the reservoir. The starting point usually is creating 1D facies models based on core samples and drilling reports at each well location. Sparse well locations and the inherent heterogeneity in the reservoir properties make it essential to incorporate the resultant 1D models into a 3D model of facies distribution that includes information about the probability of their occurrence. New techniques using machine learning (ML) to build 3D lithofluid facies (LFF) models can incorporate the prediction of different lithofacies regarding their potential hydrocarbon content, along with the uncertainties of the prediction. We have applied a fuzzy inference system, as an expert-oriented approach, and two separate ML algorithms, to different seismic and elastic attributes to model the LFF classes within the Heidrun oil and gas field. The results, compared with the test wells, show that ML methods could successfully predict the distribution of gas and oil sands within the field, in very good agreement with the known fluid contact intervals. Moreover, the predictions of shale and brine sands vary depending on the method but also are consistent with our knowledge of this field. Comparison between the results confirms the higher reliability of ML methods. Importantly, ML methods provide a better way of investigating and quantifying the uncertainty of the predictions. Implementing ML algorithms in reservoir characterization reduces the risk of drilling unnecessary wells due to false discoveries and can lead to more economical development of hydrocarbon resources.

Integrating petroelastic modeling, stochastic seismic inversion, and Bayesian probability classification to reduce uncertainty of hydrocarbon prediction: Example from Malay Basin

Exploring hydrocarbon in structural-stratigraphical traps is challenging due to the high lateral variation of lithofluid facies. In addition, reservoir characterization is getting more obscure if the reservoir layers are thin and below the seismic vertical resolution. Our objectives are to reduce the uncertainty of reserve estimation and to predict hydrocarbon distribution more accurately in such reservoir layers by proposing a new workflow that works better than the conventional one. The approach was performed by integrating petroelastic modeling, stochastic elastic seismic inversion, and Bayesian probability classification in the upper reservoir layer of Group E in the Northern Malay Basin. A robust petroelastic model was initially built to obtain more obvious separation of different lithofluid classes in elastic properties crossplot, that is acoustic impedance versus [Formula: see text] ratio. To achieve reliable distribution of elastic properties per identified lithofluid class, a Monte Carlo simulation was then run and the posterior probability of all classes was computed using Bayesian classification, followed by confusion matrix assessment. Stochastic elastic seismic inversion was carried out on conditioned seismic data to predict elastic properties away from the wells. Using all elastic properties realizations, ranking was calculated and uncertainty was quantified at the blind well location. The most probable scenario is the realization that has a much closer probability to the measured criterion value at the blind well. The computed posterior probability of hydrocarbon-bearing sand was applied on the selected stochastic realization (acoustic impedance and [Formula: see text] volumes) according to the ranking result. Finally, the hydrocarbon distribution probability map was generated and validated with lithofluid facies information of four distributed wells. Such a comparison authenticated the hydrocarbon prediction particularly at the blind well location.

Assessment of machine-learning techniques in predicting lithofluid facies logs in hydrocarbon wells

Wireline log interpretation is a well-exercised procedure in the oil and gas industry with all its added value from exploration to production stages. It becomes even more important when it is one of only a few available alternatives to compensate for the lack of core samples in a study of lithologic and fluid variations in a well. Yet, as with other purely expert-oriented interpretational techniques, there is always a considerable risk of subjective or technical errors. We have adopted a hybrid approach that links a machine-learning (ML) algorithm to the log interpretation procedure to solve these problems. We have applied this approach to two different hydrocarbon (HC) fields with the aim of predicting the HC-bearing units in the form of lithofluid facies logs at different well locations. The values of these logs are labels of classes that are separated based on their lithologic and fluid content characteristics. After training different MLs on the designed lithofluid facies logs, we chose a bagged-tree algorithm to predict these logs for the target wells due to its superior performance. This algorithm predicted HC units in an accurate interval (above the HC-fluid contact depth), and it showed a very low false discovery rate. The high-accuracy rate, speed of analysis, and its generalization ability, even in data-deficient cases, accentuate why including ML algorithms can improve the understanding of the subsurface at every phase of the exploration and production process. The proposed approach of using ML algorithms, trained and tuned based on the expert’s knowledge of the reservoir, can be modified and applied to future wells in a HC field to significantly minimize the risk of false HC discoveries.

Analysis of Different Statistical Models in Probabilistic Joint Estimation of Porosity and Litho-Fluid Facies from Acoustic Impedance Values

We discuss the influence of different statistical models in the prediction of porosity and litho-fluid facies from logged and inverted acoustic impedance (Ip) values. We compare the inversion and classification results that were obtained under three different statistical a-priori assumptions: an analytical Gaussian distribution, an analytical Gaussian-mixture model, and a non-parametric mixtu re distribution. The first model assumes Gaussian distributed porosity and Ip values, thus neglecting their facies-dependent behaviour related to different lithologic and saturation conditions. Differently, the other two statistical models relate each component of the mixture to a specific litho-fluid facies, so that the facies-dependency of porosity and Ip values is taken into account. Blind well tests are used to validate the final predictions, whereas the analysis of the maximum-a-posteriori (MAP) solutions, the coverage ratio, and the contingency analysis tools are used to quantitatively compare the inversion outcomes. This work points out that the correct choice of the statistical petrophysical model could be crucial in reservoir characterization studies. Indeed, for the investigated zone, it turns out that the simple Gaussian model constitutes an oversimplified assumption, while the two mixture models provide more accurate estimates, although the non-parametric one yields slightly superior predictions with respect to the Gaussian-mixture assumption.

Estimation of reservoir properties from seismic data through a Markov Chain Monte Carlo-AVA inversion algorithm

We formulate the amplitude versus angle (AVA) inversion in terms of a Markov Chain Monte Carlo (MCMC) algorithm and apply it for reservoir characterisation and litho-fluid facies prediction in offshore Nile Delta. A linear empirical rock physics model is used to link the petrophysical properties (porosity, water saturation and shaliness) to the elastic attributes (P-wave velocity, S-wave velocity and density), whereas the exact Zoeppritz equations are used to convert the elastic properties into AVA responses. The exact Zoeppritz equations allow us to take advantage of the long offset seismic acquisition and thus to consider a wide range of incidence angles in the inversion. The proposed algorithm reliably estimates the non-uniqueness of the solution that is the uncertainties affecting the estimated subsurface characteristics (both in terms of litho-fluid facies and petrophysical properties), taking into consideration the uncertainties in the prior information, the uncertainties in the estimated rock-physics model and the errors affecting the observed AVA responses. A blind test, based on available well log information, demonstrates the applicability of the proposed method and the reliability of the results. In addition, comparisons between the results provided by the implemented MCMC algorithm with those yielded by a linear AVA inversion and an analytical approach to facies prediction, show the benefits introduced by wide-angle reflections in better constraining the inverted parameters and in attenuating the noise in the predicted subsurface models.

A two-step inversion approach for seismic-reservoir characterization and a comparison with a single-loop Markov-chain Monte Carlo algorithm

We have evaluated a two-step Bayesian algorithm for seismic-reservoir characterization, which, thanks to some simplifying assumptions, is computationally very efficient. The applicability and reliability of this method are assessed by comparison with a more sophisticated and computer-intensive Markov-chain Monte Carlo (MCMC) algorithm, which in a single loop directly estimates petrophysical properties and lithofluid facies from prestack data. The two-step method first combines a linear rock-physics model (RPM) with the analytical solution of a linearized amplitude versus angle (AVA) inversion, to directly estimate the petrophysical properties, and related uncertainties, from prestack data under the assumptions of a Gaussian prior model and weak elastic contrasts at the reflecting interface. In particular, we use an empirical, linear RPM, properly calibrated for the investigated area, to reparameterize the linear time-continuous P-wave reflectivity equation in terms of petrophysical contrasts instead of elastic constants. In the second step, a downward 1D Markov-chain prior model is used to infer the lithofluid classes from the outcomes of the first step. The single-loop (SL) MCMC algorithm uses a convolutional forward modeling based on the exact Zoeppritz equations, and it adopts a nonlinear RPM. Moreover, it assumes a more realistic Gaussian mixture distribution for the petrophysical properties. Both approaches are applied on an onshore 3D seismic data set for the characterization of a gas-bearing, clastic reservoir. Notwithstanding the differences in the forward-model parameterization, in the considered RPM, and in the assumed a priori probability density functions, the two methods yield maximum a posteriori solutions that are consistent with well-log data, although the Gaussian mixture assumption adopted by the SL method slightly improves the description of the multimodal behavior of the petrophysical parameters. However, in the considered reservoir, the main difference between the two approaches remains the very different computational times, the SL method being much more computationally intensive than the two-step approach.

Seismic reservoir characterization of a class-1 amplitude variation with offset turbiditic system located offshore Cote d’Ivoire, West Africa

We have evaluated a case study, in which a class-1 amplitude variation with offset (AVO) turbiditic system located offshore Cote d’Ivoire, West Africa, is characterized in terms of rock properties (lithology, porosity, and fluid content) and stratigraphic elements using well-log and prestack seismic data. The methodology applied involves (1) the conditioning and modeling of well-log data to several plausible geologic scenarios at the prospect location, (2) the conditioning and inversion of prestack seismic data for P- and S-wave impedance estimation, and (3) the quantitative estimation of rock property volumes and their geologic interpretation. The approaches used for the quantitative interpretation of these rock properties were the multiattribute rotation scheme for lithology and porosity characterization and a Bayesian lithofluid facies classification (statistical rock physics) for a probabilistic evaluation of fluid content. The result indicates how the application and integration of these different AVO- and rock-physics-based reservoir characterization workflows help us to understand key geologic stratigraphic elements of the architecture of the turbidite system and its static petrophysical characteristics (e.g., lithology, porosity, and net sand thickness). Furthermore, we found out how to quantify and interpret the risk related to the probability of finding hydrocarbon in a class-1 AVO setting using seismically derived elastic attributes, which are characterized by having a small level of sensitivity to changes in fluid saturation.

Application of different classification methods for litho-fluid facies prediction: a case study from the offshore Nile Delta

Application of different classification methods for litho-fluid facies prediction: a case study from the offshore Nile Delta

Adding geologic prior knowledge to Bayesian lithofluid facies estimation from seismic data

Using inverted seismic data from a turbidite depositional environment, we have determined that accounting only for rock types sampled at the wells can lead to biased predictions of the reservoir fluids. The seismic data consisted of two volumes resulting from a (multi-incidence angle) sparse-spike amplitude variation with offset inversion. Information from a single well (well logs and petrological analysis) was used to define an initial set of lithofluid facies that characterized rock type and porefill fluid to emulate a typical exploration setting. Based on our geologic understanding of the study area, we have augmented this initial model with lithofluid facies expected in the given depositional environment, yet not sampled by the well. Specifically, the new lithofluid facies accounted for variations in the mixture type and proportions of shales and sands. The elastic property distributions of the new lithofluid facies were modeled using appropriate rock-physics models. Finally, a geologically consistent, spatially variant, prior probability of lithofluid facies occurrence was combined with the data likelihood to yield a Bayesian estimation of the lithofluid facies probability at every sample of the inverted seismic data. Applying the augmented geologic prior probabilities, we were able to generate a scenario consistent with all available data, which supports further development of the field. In contrast, using the initial, purely data-driven lithofluid facies model based on a single well, the Bayesian classification would lead to prospectivity downgrade or suboptimal development of the field. We found that limited well control in quantitative interpretation needs to be counterweighted by geologic prior information based on detailed stratigraphic interpretation, to derisk geologic scenarios without bias.

Petro-Elastic Log-Facies Classification Using the Expectation–Maximization Algorithm and Hidden Markov Models

Log-facies classification methods aim to estimate a profile of facies at the well location based on the values of rock properties measured or computed in well-log analysis. Statistical methods generally provide the most likely classification of lithological facies along the borehole by maximizing a function that describes the likelihood of a set of rock samples belonging to a certain facies. However, most of the available methods classify each sample in the well log independently and do not account for the vertical distribution of the facies profile. In this work, a classification method based on hidden Markov models is proposed, a stochastic method that accounts for the probability of transitions from one facies to another one. Differently from other available methods where the model parameters are assessed using nearby fields or analogs, the unknown parameters are estimated using a statistical algorithm called the Expectation–Maximization algorithm. The method is applied to two different datasets: a clastic reservoir in the North Sea where four litho-fluid facies are identified and an unconventional reservoir in North America where four lithological facies are defined. The results of the applications show the added value of the introduction of a vertical continuity model in the facies classification and the ability of the proposed method of inferring model parameters such as facies transition probabilities and facies posterior distributions. The application also includes a sensitivity analysis and a comparison to other statistical methods.

Reservoir characterization and paleo-stratigraphic imaging over Okari Field, Niger Delta, using neural networks

Reservoir geometry and internal architecture in the Niger Delta can vary over short distances with rapid lateral and vertical changes in lithology and porosity. Understanding such variations is critical to designing an optimum development strategy for prospects in this basin. It was in order to fully understand the variations in reservoir facies and internal architecture over Okari oil field in the Niger Delta that this study was undertaken. Conventional seismic interpretation, attribute analyses, and subsequent drilling had located a stack of reservoirs in a rollover anticline. To unravel the paleo-stratigraphy of the field and fully populate the field with log properties, we used a multilayered feed-forward neural network (MLFN) to predict shale volume and porosity from seismic and well-log data sets. Earlier, rock physics analyses had been undertaken to understand litho-fluid facies associations in the field and assist in further quantitative interpretation and calibration of neural-network predictions of res-ervoir property distribution.