Marcellus Shale Lithofacies Prediction by Multiclass Neural Network Classification in the Appalachian Basin

Marcellus Shale is a rapidly emerging shale-gas play in the Appalachian basin. An important component for successful shale-gas reservoir characterization is to determine lithofacies that are amenable to hydraulic fracture stimulation and contain significant organic-matter and gas concentration. Instead of using petrographic information and sedimentary structures, Marcellus Shale lithofacies are defined based on mineral composition and organic-matter richness using core and advanced pulsed neutron spectroscopy (PNS) logs, and developed artificial neural network (ANN) models to predict shale lithofacies with conventional logs across the Appalachian basin. As a multiclass classification problem, we employed decomposition technology of one-versus-the-rest in a single ANN and pairwise comparison method in a modular approach. The single ANN classifier is more suitable when the available sample number in the training dataset is small, while the modular ANN classifier performs better for larger datasets. The effectiveness of six widely used learning algorithms in training ANN (four gradient-based methods and two intelligent algorithms) is compared with results indicating that scaled conjugate gradient algorithms performs best for both single ANN and modular ANN classifiers. In place of using principal component analysis and stepwise discriminant analysis to determine inputs, eight variables based on typical approaches to petrophysical analysis of the conventional logs in unconventional reservoirs are derived. In order to reduce misclassification between widely different lithofacies (for example organic siliceous shale and gray mudstone), the error efficiency matrix (ERRE) is introduced to ANN during training and classification stage. The predicted shale lithofacies provides an opportunity to build a three-dimensional shale lithofacies model in sedimentary basins using an abundance of conventional wireline logs. Combined with reservoir pressure, maturity and natural fracture system, the three-dimensional shale lithofacies model is helpful for designing strategies for horizontal drilling and hydraulic fracture stimulation.     

Seismicity Characterization of Iran: A Multivariate Statistical Approach

This paper presents the results of two multivariate analysis techniques—principal component and cluster analysis—as they are applied to the seismicity characterization of Iran. The seismic data used in this study covers a period of 50 years, from the beginning of 1957 to the end of 2006. The values of eight seismic variables were calculated on a grid of equally spaced points at one geographic degree spacing in both latitude and longitude. The data matrix was analyzed using principal component and cluster analysis. Principal component analysis identified two significant components, introduced in this study as the Seismic Frequency Index (SFI) and the Seismic Severity Index (SSI), responsible for the data structure. The SFI and SSI explain 34.34 % and 32.33 % of the total variance of the data set, respectively, and allowed grouping of the selected variables according to their common features. The standardized data matrix was analyzed using Ward’s clustering method. The resulting seismicity pattern recognition maps of the region at three levels of similarity are presented. From these maps, differentiated seismic zones are outlined in detail and compared quantitatively. Comparison between the seismic zoning maps obtained in this analysis and the general tectonic map of the region indicates that the seismic zones are consistent with the tectonic zones of the region. This study presents the necessity and usefulness of multivariate analysis in evaluating and interpreting seismic data catalogues with the goal of obtaining more objective information about the seismicity pattern of regions.     

沉积微相测井资料神经网络判别方法研究

不同的沉积微相可以由不同的相标志组合识别,相标志与沉积微相之间的关系可以采用神经网络通过许多基本处理单元间并行的相互作用建立。沉积微相相标志既可以由地质资料的观察、岩芯分析直接获得,也可以由测井资料间接地求得     

Comparison of automatic classification methods applied to lake geochemical samples

Geochemical samples from part of Lake Geneva were analyzed for 29oxides and trace elements. The variables and samples were subjected to R- and Q-mode analyses. The following techniques were applied in sequence: data transformation (normalization and standardization), data reduction (principal component and factor analysis), and automatic classification (dendrograph). The data were treated using various combinations of these techniques, and the resulting classifications evaluated by means of several criteria. The best classification of the samples is given by a cluster analysis performed on four principal components computed from standardized variables. The discriminatory power of the variables also was measured and determined to depend on their degree of intercorrelation. As a final result, the 29original variables were reduced to four components and the sediment samples classified into four facies, leading to easily interpretable geochemical maps.     

Calcrete conglomerate, case-hardened conglomerate and cornstone - a comparative account of pedogenic and non-pedogenic carbonates from the continental Siwalik Group, Punjab, India

The occurrence of authigenic carbonates formed in three different environmental situations, within the continental Siwalik Group, has been used to compare the lithological and petrographic characters of the contrasted lithofacies. The three lithofacies are: (1) calcrete conglomerate, (2) case-hardened conglomerate, (3) cornstone (pedogenic, nodular calcrete). The calcrete conglomerate facies laterally intertongues with the channel conglomerates. It consists of pisolites which are interpreted to have formed from carbonate-rich spring waters emerging on to the gravelly substrate of dry, abandoned channels. The laminae characteristics of these pisolites are distinctly different from those of marine origin and also from comparable biogenic materials. Case-hardened conglomerate occurs in the youngest part of the Siwalik stratigraphic column, in boulder conglomerates having limestone as the principal clast component. This lithofacies has resulted from cementation of the conglomerate through continued dissolution and re-precipitation of calcite, by meteoric water, downwards from the surface. It displays a coarsely crystalline, sparry calcite cement with no evidence for displacive growth or replacement by calcite. Cornstones (nodular calcrete) occur in several sedimentary cycles of the Middle Siwalik Sub-Group. These are immature and commonly associated with thinly-bedded sandstones (levée) and red shales (overbank). This lithofacies is a result of concentration of carbonate through capillary action associated with pedogenic activity. Ooids developed in cornstone are essentially micritic in nature and usually composed of less than five indistinct laminae.     

Joint Consistent Mapping of High-Dimensional Geochemical Surveys

Geochemical surveys often contain several tens of components, obtained from different horizons and with different analytical techniques. These are used either to obtain elemental concentration maps or to explore links between the variables. The first task involves interpolation, the second task principal component analysis (PCA) or a related technique. Interpolation of all geochemical variables (in wt% or ppm) should guarantee consistent results: At any location, all variables must be positive and sum up to 100 %. This is not ensured by any conventional geostatistical technique. Moreover, the maps should ideally preserve any link present in the data. PCA also presents some problems, derived from the spatial dependence between the observations, and the compositional nature of the data. Log-ratio geostatistical techniques offer a consistent solution to all these problems. Variation-variograms are introduced to capture the spatial dependence structure: These are direct variograms of all possible log ratios of two components. They can be modeled with a function analogous to the linear model of coregionalization (LMC), where for each spatial structure there is an associated variation matrix describing the links between the components. Eigenvalue decompositions of these matrices provide a PCA of that particular spatial scale. The whole data set can then be interpolated by cokriging. Factorial cokriging can also be used to map a certain spatial structure, eventually projected onto those principal components (PCs) of that structure with relevant contribution to the spatial variability. If only one PC is used for a certain structure, the maps obtained represent the spatial variability of a geochemical link between the variables. These procedures and their advantages are illustrated with the horizon C Kola data set, with 25 components and 605 samples covering most of the Kola peninsula (Finland, Norway, Russia).     

Estimating airborne heavy metal concentrations in Dunkerque (northern France)

This work aims to estimate the levels of lead (Pb), nickel (Ni), manganese (Mn), vanadium (V) and chromium (Cr) corresponding to a 3-month PM₁₀ sampling campaign conducted in 2008 in the city of Dunkerque (northern France) by means of statistical models based on partial least squares regression (PLSR), artificial neural networks (ANNs) and principal component analysis (PCA) coupled with ANN. According to the European Air Quality Directives, because the levels of these pollutants are sufficiently below the European Union (EU) limit/target values and other air quality guidelines, they may be used for air quality assessment purposes as an alternative to experimental measurements. An external validation of the models has been conducted, and the results indicate that PLSR and ANNs, with comparable performance, provide adequate mean concentration estimations for Pb, Ni, Mn and V, fulfilling the EU uncertainty requirements for objective estimation techniques, although ANNs seem to present better generalization ability. However, in accordance with the European regulation, both techniques can be considered acceptable air quality assessment tools for heavy metals in the studied area. Furthermore, the application of factor analysis prior to ANNs did not yield any improvements in the performance of the ANNs.     

Comparison of Three Un-mixing Based Hyperspectral Image Fusion Techniques

With the availability of multi sensor data in many fields, such as remote sensing, medical imaging or machine vision, sensor fusion has emerged as a new and promising research area. It is possible to have several images of the same scene providing different information although the scene is the same. This is because each image has been captured with a different sensor. A non-negative matrix factorization (NNMF) un mixing based fusion technique with vertex component analysis (VCA) based end member initialization and simple multiplicative update to improve the spatial resolution and to preserve the spectral resolution of the hyper spectral image is proposed. Its performance is analyzed with different number of iterations and end member initializations. A Constrained Non Negative Matrix Factorization unmixing based fusion technique is developed by adding a regularization term to the objective function to preserve the spectral resolution of the hyper spectral image, and its performance is analyzed with different number of iterations and end members. A rank two NNMF and hierarchical clustering based end member initialization and block principal pivoting algorithm based abundance estimation technique, for fusing hyper spectral image and simulated multispectral image is proposed and its performance is analyzed for different overlapping and non overlapping group of multispectral and hyper spectral bands. The performance of the above three methods are compared and analyzed. The obtained results show that the performance of rank two NNMF hierarchical clustering based fusion technique is better than the other two constrained and unconstrained NNMF un mixing based techniques. Also, the performance of these three proposed multi sensor image fusion techniques are compared with an existing image fusion technique.     

Neural network based prediction of ground surface settlements due to tunnelling

Ground surface settlement due to tunnel excavation varies in magnitude and trend depending on several factors such as tunnel geometry, ground conditions, etc. Although there are several empirical and semi-empirical formulae available for predicting ground surface settlement, most of these do not simultaneously take into consideration all the relevant factors, resulting in inaccurate predictions. In this study, an artificial neural network (ANN) is incorporated with '113' of monitored field results to predict surface settlement for a tunnel site with prescribed conditions. To achieve this, a standard format (a protocol) for a database of monitored field data is first proposed and then used for sorting out a variety of monitored data sets available in KICT. Using the capabilities of pattern recognition and memorization of the ANN, an attempt is made to capture the rich physical characteristics smeared in the database and at the same time filter inherent noise in the monitored data. Here, an optimal neural network model is suggested through preliminary parametric studies. It is shown that preliminary studies for generating an optimal ANN under given training data sets are necessary because no analytical method for this purpose is available to date. In addition, this study introduces a concept of relative strength of effects (RSE) [Yang Y, Zhang Q. A heirarchical analysis for rock engineering using artificial neural networks. Rock Mechanics and Rock Engineering 1997; 30(4): 207–22] in sensitivity analysis for various major factors affecting the surface settlement in tunnelling. It is seen in some examples that the RSE rationally enables us to recognize the most significant factors of all the contributing factors. Two verification examples are undertaken with the trained ANN using the database created in this study. It is shown from the examples that the ANN has adequately recognized the characteristics of the monitored data sets retaining a generality for further prediction. It is believed that an ANN based hierarchical prediction procedure shown in this paper can be further employed in many kinds of geotechnical engineering problems with inherent uncertainties and imperfections.     

Modeling and prediction of ventilation methane emissions of U.S. longwall mines using supervised artificial neural networks

Methane emissions from a longwall ventilation system are an important indicator of how much methane a particular mine is producing and how much air should be provided to keep the methane levels under statutory limits. Knowing the amount of ventilation methane emission is also important for environmental considerations and for identifying opportunities to capture and utilize the methane for energy production.Prediction of methane emissions before mining is difficult since it depends on a number of geological, geographical, and operational factors. This study proposes a principle component analysis (PCA) and artificial neural network (ANN)-based approach to predict the ventilation methane emission rates of U.S. longwall mines.Ventilation emission data obtained from 63 longwall mines in 10 states for the years between 1985 and 2005 were combined with corresponding coalbed properties, geographical information, and longwall operation parameters. The compiled database resulted in 17 parameters that potentially impacted emissions. PCA was used to determine those variables that most influenced ventilation emissions and were considered for further predictive modeling using ANN. Different combinations of variables in the data set and network structures were used for network training and testing to achieve minimum mean square errors and high correlations between measurements and predictions. The resultant ANN model using nine main input variables was superior to multilinear and second-order non-linear models for predicting the new data. The ANN model predicted methane emissions with high accuracy. It is concluded that the model can be used as a predictive tool since it includes those factors that influence longwall ventilation emission rates.     

主成分分析与Bayes判别法在突水水源判别中的应用

准确有效地判别突水水源是解决矿井水害的前提条件。基于淮北袁店二矿各含水层共59个水样水质化验资料,利用主成分分析法,计算各水样的因子得分,并进行系统聚类,剔除错误样本。利用剩余水样作为学习样本,检验Bayes判别函数的判定准确性,得出准确率为92.5%,并进行交叉验证。利用该判别函数对某工作面底板下一富水区水样进行判别,结果与实际情况吻合。结果指示基于主成分分析与Bayes判别法较单一Bayes判别法更加准确,能够消除样本变量之间的相互影响,实现对突水水源的快速有效判别。      

Prediction of the Rock Mass Diggability Index by Using Fuzzy Clustering-Based,ANN and Multiple Regression Methods

Rock mass classification systems are one of the most common ways of determining rock mass excavatability and related equipment assessment. However, the strength and weak points of such rating-based classifications have always been questionable. Such classification systems assign quantifiable values to predefined classified geotechnical parameters of rock mass. This causes particular ambiguities, leading to the misuse of such classifications in practical applications. Recently, intelligence system approaches such as artificial neural networks (ANNs) and neuro-fuzzy methods, along with multiple regression models, have been used successfully to overcome such uncertainties. The purpose of the present study is the construction of several models by using an adaptive neuro-fuzzy inference system (ANFIS) method with two data clustering approaches, including fuzzy c-means (FCM) clustering and subtractive clustering, an ANN and non-linear multiple regression to estimate the basic rock mass diggability index. A set of data from several case studies was used to obtain the real rock mass diggability index and compared to the predicted values by the constructed models. In conclusion, it was observed that ANFIS based on the FCM model shows higher accuracy and correlation with actual data compared to that of the ANN and multiple regression. As a result, one can use the assimilation of ANNs with fuzzy clustering-based models to construct such rigorous predictor tools.     

Using principal components analysis (PCA) with cluster analysis to study the organic geochemistry of sinking particles in the ocean

Principal components analysis (PCA) is a multivariate data analysis tool that can be used to recombine the variables of a large multivariate dataset in such a way that the first few variables of the reconstructed dataset account for the majority of the variance in the data. Application of PCA in marine geochemistry has become quite common in recent years. In this study, we illustrate the use of PCA through examples that arose while investigating the geochemistry of sinking particles during the MedFlux project. The examples presented do not simply repeat the analyses of the original study, but instead extend them in the context of simultaneous application of PCA and cluster analysis. Our results show that constructing a one dimensional (1D) “degradation index” using only the first principal component (PC) is in most cases oversimplified, and that constructing 2D or 3D “degradation trajectories” with the first 2 or 3 PCs is more informative. Use of the first three PCs is indicated when the variance explained by the third PC is comparable in magnitude to that explained by the second PC in the reconstructed dataset. We also discuss the use of scree plots and cluster analysis in helping decide whether the third PC is needed to capture the essential information in the dataset.     

Evaluation of graphical and multivariate statistical methods for classification of water chemistry data

A robust classification scheme for partitioning water chemistry samples into homogeneous groups is an important tool for the characterization of hydrologic systems. In this paper we test the performance of the many available graphical and statistical methodologies used to classify water samples including: Collins bar diagram, pie diagram, Stiff pattern diagram, Schoeller plot, Piper diagram, Q-mode hierarchical cluster analysis, K-means clustering, principal components analysis, and fuzzy k-means clustering. All the methods are discussed and compared as to their ability to cluster, ease of use, and ease of interpretation. In addition, several issues related to data preparation, database editing, data-gap filling, data screening, and data quality assurance are discussed and a database construction methodology is presented. The use of graphical techniques proved to have limitations compared with the multivariate methods for large data sets. Principal components analysis is useful for data reduction and to assess the continuity/overlap of clusters or clustering/similarities in the data. The most efficient grouping was achieved by statistical clustering techniques. However, these techniques do not provide information on the chemistry of the statistical groups. The combination of graphical and statistical techniques provides a consistent and objective means to classify large numbers of samples while retaining the ease of classic graphical presentations. Electronic Publication     

Teaching Aid: Minimum/Maximum Autocorrelation Factors for Joint Simulation of Attributes

Multivariate conditional simulation is used to assess the multivariate grade risk in mineral deposits. With the presence of several spatially correlated attributes, it is important to ensure that their joint simulation is carried out properly and that the observed spatial correlation is reproduced in the realizations. The method of minimum/maximum autocorrelation factors (MAF) is a well established and practical technique that can be used for this purpose. MAF offers tremendous advantages over standard full cosimulation, principal component analysis, and stepwise techniques. In what follows, a detailed review of the MAF technique, its applications, and examples are provided to guide the practitioner on its use.     

Slope failure susceptibility zonation using integrated remote sensing and GIS techniques: A case study over Jhingurdah open pit coal mine,Singrauli coalfield,India

The susceptibility of slopes in open pit coal mines to various modes of failure (i.e., plane, wedge, circular and toppling failure) could be envisaged by virtue of processing and analysis of pertinent satellite data. The aim of the present study was to integrate thematic maps generated using remote sensing image processing techniques, in order to finally produce slope failure hazard zonation maps in and around Singrauli coalfield, India. The various failure-inducing factors, variables and parameters can be extracted from different satellite data and imageries. The data acquired by different sensors such as TM, ETM+, etc., of LANDSAT series and CARTOSAT of ISRO Bhuvan was used in this study. All these data were subsequently used to create different thematic maps such as slope map, lithological map, land use/land cover map, principal component analysis map, digital elevation model (DEM), etc. An advanced analysis for extraction of lineament attributes was also undertaken.     

Stepwise Conditional Transformation for Simulation of Multiple Variables

Most geostatistical studies consider multiple-related variables. These relationships often show complex features such as nonlinearity, heteroscedasticity, and mineralogical or other constraints. These features are not handled by the well-established Gaussian simulation techniques. Earth science variables are rarely Gaussian. Transformation or anamorphosis techniques make each variable univariate Gaussian, but do not enforce bivariate or higher order Gaussianity. The stepwise conditional transformation technique is proposed to transform multiple variables to be univariate Gaussian and multivariate Gaussian with no cross correlation. This makes it remarkably easy to simulate multiple variables with arbitrarily complex relationships: (1) transform the multiple variables, (2) perform independent Gaussian simulation on the transformed variables, and (3) back transform to the original variables. The back transformation enforces reproduction of the original complex features. The methodology and underlying assumptions are explained. Several petroleum and mining examples are used to show features of the transformation and implementation details.     

High-Frequency Sequence Stratigraphy Using Syntactic Methods and Clustering Applied to the Upper Limestone Coal Group (Pendleian, E1) of the Kincardine Basin, United Kingdom

We define a distance between sedimentary successions to compare their dissimilarity formally. Distance definition is based on attributed syntactic representation. One-dimensional successions can be represented by a string of lithofacies symbols sequentially or vertically. Each symbol can also have a vector of attributes that can provide other information on lithofacies such as thickness. The distance of any two successions is then defined consisting of its syntactic and attribute subdistances. Syntactic distance measures difference of vertical lithofacies change between two successions and attribute distance measures difference of thickness of corresponding lithofacies. Clustering is used to test validity of distance definition and its potential application to analysis of cycle-dominated sedimentary successions. Example is from the Namurian-A succession in Kincardine basin, central Scotland. There are 56 cycles in intervals of about 300 m each in two boreholes. Recognition of intermediate cycles depends on correctly determining of types of these short cycles and their vertical stacking pattern. Intermediate cycles have better potential in high-resolution stratal correlation regionally. Syntactic clustering results show that 56 short cycles can be classified into four groups with distinctive geological interpretation, which further helps reveal hierarchical cyclic architecture of the whole succession.     

Facies analysis and depositional environments of the Bhander Limestone (Precambrian), southeastern Rajasthan and adjoining Madhya Pradesh,India

Special interest is attached to the Bhander Limestone because it is the only calcareous formation in the very thick elastic sequence of Precambrian age, designated informally as the “Upper” Vindhyan. The sedimentology of the Bhander Limestone was studied in the Mandalgarh-Singoli area of southeastern Rajasthan and adjoining Madhya Pradesh with a view to interpreting the depositional environments of the formation. This study has an important bearing on the exploration for oil in India and presents one of the few examples of Precambrian limestones of which thorough modern sedimentological analysis has been made.The Bhander Limestone comprises micritic limestones, crystalline dolostones, siltstones and shales that show desiccation structures (horizontal fenestrae, bird's-eye structures, mud cracks), very shallow small channels filled with flat-pebble breccia, algal lamination, palisade structure, and occasional ripple marks, ripple lamination and micro-cross-lamination. The major petrographic constituents are micrite, intraclasts, sparry-calcite cement, pseudospar and replacement dolomite. Seven environmentally significant microfacies have been recognized: micrite, silty micrite, graded micrite, dolomitized micrite, neomorphosed micrite, intrasparrudite and intramicrudite.The Bhander Limestone Formation has been divided vertically into four lithofacies: red argillaceous micritic limestones (lithofacies A), interlaminated blue micritic limestones and red dolomite (lithofacies B), olive calcareous shales (lithofacies C), and black micritic limestones (lithofacies D). Each lithofacies is characterized by certain megascopic sedimentary features and microfacies. The various lithofacies have been interpreted as representing deposition in the different subenvironments of a generally low-energy, marginal marine environment comprising tidal flats and lagoons. The vertical changes from one lithofacies to another are interpreted as reflecting the change from one subenvironment to another brought about by the landward shifting of the boundaries of these subenvironments in response to a transgression.     

Identification of sedimentary facies with well logs: an indirect approach with multinomial logistic regression and artificial neural network

Taking K-successions of the H-Zone of the Pearl River Mouth Basin as a testing example, we used two kinds of approaches to implement the microfacies identification. One is a direct identification, the other is an indirect approach in which we conducted the lithofacies classification first and then identified the microfacies based on previously estimated lithofacies. Both approaches were trained and checked by interpretations of experienced geologists from real subsurface core data. Multinomial logistic regression (MLR) and artificial neural network (ANN) were used in these two approaches as classification algorithms. Cross-validations were implemented. The source data set was randomly divided into training subset and testing subset. Four models, namely, MLR_direct, ANN_direct, MLR_indirect, and ANN_indirect, were trained with the training subset. The result of the testing set shows that the direct approaches (MLR_direct and ANN_direct) perform relatively poor with a total accuracy around 75%. While the indirect approaches (MLR_indirect and ANN_indirect) perform much better with a total accuracy of around 89 and 82%, respectively. This indirect method is simple and reproducible, and it could lead to a robust way of analyzing sedimentary microfacies of horizontal wells with little core data or even are almost never cored while core data are available for nearby vertical wells.