This study focuses on the contrasted architectural elements of the two late Pleistocene alluvial fill sequences (T1 and T2 attributed to the Late Saalian and Weichselian) of the upstream Seine River alluvial plain, la Bassee, and constitutes the first estimation of the sediment volumes associated to these glacial cycles along this river. Estimations of the incised valley and valley fill spatial distribution are obtained by cokriging of data from 546 wells collected from the French national repository and a digital elevation model (DEM) of the present topography. The well stratigraphic attribution is given by the homogenised geomorphological map of la Bassee. We calculated the sediment volume of eroded material for the T1 sequence to be 1.70 10(9) m(3) and of valley fill deposits to be 1.18 10(9) m(3). This corresponds to a mean thickness of 6.1 m. By contrast, volumes of the T2 sequence are one-third smaller, 1.23 10(9) m(3). The T2 mean thickness is very similar with 5.8 m. The alluvial plain is divided into four reaches representing a typical assemblage of the T1 and T2 incision style and valley fill deposits. Along the downstream reaches (1 - Montereau-Bazoches, 2 - Bazoches-Grisy), the T2 proportion per kilometer are close to the alluvial plain mean value (reach 1: 76%; reach 2: 69%, la Bassee: 71%), the T2 mean incised valley width corresponds to three quarters of the T1 (4 km). In the middle upstream reach (3 - Grisy-Courceroy), the T2 proportion is the largest (92%), the T2 mean incised valley width is the largest and comparable to the T1 (5.5 km). Along the upstream reach (4 - Courceroy-Rornilly), the T2 proportion is the lowest (53%) and the T2 incised valley (2 km wide) is enclosed within the T1 incised valley (4 km wide). The rate of exported material, though, is the same for each sequence in any reach (31%). The spatial pattern of the T1-T2 deposits coincides largely with the presence of a knickpoint along the T2 paleothalweg (downstream part of reach 4, slope face 1.5 parts per thousand), suggesting that the knickpoint had a large influence on the fluvial processes. The toe of the knickpoint (reaches 3 to 1, slope 0.24 parts per thousand) was the site of the most efficient erosion of the T1 deposits. We propose that the change of slope favored diversion of the channel path. Upstream of the knickpoint face (upstream part of reach 4, thalweg slope 0.34 parts per thousand), the incision processes were dominated by incision, avulsions or multithreading with limited lateral migration as indicated by the numerous T1 relicts preserved across valley. The sediment export corresponds to a mean incision rate of 40 m/Ma during the last 300 kyr; a fairly low value for the Paris basin, suggesting a low uplift area north of the Massif Central. Such contrast between the fluvial processes during the T1 and T2 sequences is most likely to be looked for in the climatic conditions that existed at the onset of the incision as no local controls for the knickpoint location could be identified such as active faults, substrate heterogeneity, or flow modifications. The lower sediment flux and the knickpoint associated to the T2 sequence suggest a transient river profile and valley incision at the end of the climatic transition associated to the Weichselian (T2). (C) 2018 Published by Elsevier B.V.

One of the main challenges of Geostatistics in reservoir characterization is to populate a portion of 3D earth model with rock properties. Their spatial characteristics are analyzed along the stratigraphic reference system where the variogram is modeled. A base case estimation (kriging) may be sufficient, or complemented by stochastic simulations for a sensitivity analysis. Sometimes properties present multimodal distribution reflecting the heterogeneity within the studied stratigraphic domain. Therefore it makes sense to look for explanatory co-variables, such as the lithotype. Then we simulate the spatial organization of different lithotypes first (using categorical simulation) and afterwards populate each lithotype with the rock property following the rock- dependent spatial characteristics. The previous workflow, which relies on the independence between lithotype and rock property, may not be relevant if the samples show a border effect when crossing the lithotype border. Some statistical tools are introduced to check the relevance of this assumption, such as the transition probabilities and the contact analysis. The border behavior is also checked after the initial simulation outcome has been upscaled. Some stochastic joint models of lithotype and rock property are conceived to reproduce the presence or absence of borders, conditioned to sample data.

The objectives of this paper are to address the problem of integration of data corresponding to different physics and different supports to construct a geological static model. In fact, diagenesis and petrophysical properties are characterized and quantified from laboratory analyses, observations of thin sections, tests on cores while values are needed to populate cells of the geological model (metric to decametric size). At present, mean characteristics of these properties are generally attributed to the dominant facies in these cells, despite their more complex distribution existing from experimental measurements at a smaller scale. We illustrate on different examples derived from real data some ways to integrate this variability to obtain more realistic distribution of the reservoir properties. Moreover, the hypotheses on the physical process of diagenesis and on its continuity through different facies can be tested through different spatial model in order to quantify the final uncertainty on the resulting model. These points will be addressed using the Pluri-Gaussian Simulation method with parameters adapted to each type of uncertainty.

Sandstone-type uranium deposits contain approximately 28% of the world uraniumresources. Many of these deposits are located below the water table in weakly lithified or non-consolidated sands, and therefore they can be exploited using in situ leach (ISL) technology. Such technology is based on dissolving uranium minerals directly in their host rocks (in situ) by reactive solutions that are injected through drill holes and then pumping the dissolved solution to the surface through some discharge drill holes. Uranium grade is determined by down-hole geophysics, in particular the prompt fission neutrons, or PFN, technique, coupled with sampling and assaying of the drill core. The drill grid which is used in Kazakhstan for definition of ISL uranium resources are as follows: (i) Measured: average 506100 m (range from 25650 to 506100 m); (ii) Indicated: average 506200 m (range from 506100 to 506200 m); and (iii) Inferred: average 506400 m (range from 506400 to 1006800 m). Estimation and reporting of uranium resources for ISL projects differ from hard rock mining projects in the need for quantitative estimation of the geotechnical and hydrogeological parameters which are specific for ISL technologies. The main parameters which need to be considered are as follows: (i) grade and geometry of mineralisation are estimated with accuracy sufficient for supporting the remote mining; (ii) if grade is estimated using the gamma logging technique secular disequilibrium should be studied and reported; (iii) hydrogeological confinement of the mineralised horizon; (iv) permeability of the mineralised horizon; (v) composition of the host rocks, in particular the carbonate content, in order to estimate if uranium mineralisation is amenable to dissolution by acid or alkaline solutions; (vi) groundwater flow; (vii) aquifer salinity; and (viii) rate of the in-situ dissolution of the uranium minerals. Hydrogeological and geotechnical information is obtained by testing the drill core samples and in the field, using the pump tests and the down-hole piezometers. Modifying factors for conversion resources to reserves are verified and corrected using field leach tests of uranium. This test is a strict requirement for feasibility studies of the ISL uranium projects in Kazakhstan.

A possible risk of geomechanical nature related to deep injection of CO2 is the shear reactivation of faults, hence potentially leading to the creation of new leakage pathways and eventually inducing earthquakes felt at the surface. Current practices to evaluate fault stability in the domain of CO2 storage still remain limited regarding two issues: 1. Faults are complex and heterogeneous geological systems, which do not correspond to discrete surfaces as already postulated by many authors. Reservoir-scale faults in a priori low-deformed reservoirs targeted for CO2 storage can present high complex architecture, which might influence the hydro-mechanical behaviour of the fault system; 2. Chemical interactions (dissolution and precipitation processes, chemically-induced weakening, etc.) between CO2-enriched brine and the minerals constituting the fault zone can affect the mechanical stability and the transport properties of the faulted/fractured system. The research project FISIC (www.anr-fisic.fr, funded by the French National research Agency) intends to overcome those limitations by accurately modelling the hydro-chemo-mechanical complexity of a fault zone. The main goal is to improve the stability analysis of a fault both undertaking pressure increase and alteration due to the presence of an acidic fluid. The progress of this research project is presented here.

The impact of microscopic rock heterogeneities on elastic properties in an unconsolidated sandstone reservoir in the Orinoco Heavy Oil Belt in Eastern Venezuela is studied at 3 wells using rock physics models. Differences in grain and pore sizes and the presence of shale/clay and cement, lead to differences in elastic properties values at same porosity. This allows to detect two different consolidation levels in the reservoir: (1) unconsolidated and (2) weakly consolidated due to cement and smaller grains/pores, where the first one has better rock quality. Being able to detect the two consolidation levels proves the interest of rock physics modelling in our case as tool to relate geology and seismic and to perform a quantitative seismic interpretation.

This article presents a method to construct a three-dimensional (3D) block within units whose spatial characteristics are similar to those of a uranium roll-front deposit. In order to mimic the particular uranium roll-front deposits, a specific stochastic model has been developed, using some geostatistical simulation concepts. They refer to the well-established PluriGaussian Simulation model which has been rephrased in order to produce the specific ore body shapes.

Abstract The relationships between sedimentological facies and diagenetic properties differ depending on burial and geological history. In this paper we propose several examples where different quantitative expressions of diagenetic imprints lead to different workflows to be applied, to obtain final simulations reproducing both the sedimentary facies and the diagenetic trends in reservoir models. A first case is based on an outcrop study of mixed siliciclastic-carbonate facies. The relationships between sedimentary facies and diagenesis have been used to define truncation rules for the plurigaussian algorithm, which aims at simulating both the distribution of sedimentary facies and the diagenetic overprint. Another workflow has been developed on this example, using the bi-plurigaussian simulation algorithm to address the problem of heterotopic bivariate conditional simulation (the two variables are known at different locations). A second example is related to early diagenesis characterization. In this example (Madison Formation, Wyoming, USA), diagenetic imprints have been quantified, corresponding to a succession of diagenetic phases (micritization, calcite cementation, dolomitization...), that can coexist in the same sedimentary facies. The workflow using plurigaussian and nested simulation algorithms shows the impact of such joint simulations, as the diagenesis may completely modify the distribution of reservoir properties only based on a facies simulation. In a third example, the modeled Aptian/Albian formation includes hydrothermal dolomites (HDT) associated to fractures and faults hosted in reefal and slope limestones. Simulations succeed in redistributing the HTD correctly at the surface and in the subsurface. Hence, in the deeper parts of the Aptian/Albian reefal rock unit, the dolomite is inherently associated to fractures, while at shallower levels, dolomitization seems to be more lateral, affecting probably more permeable facies. These results provide insights and numerical means for estimating volumes of dolomites (vertically and laterally) in an excellent case study of fracture-related HTD. Diagenesis alters the original relationships between depositional facies and petrophysical properties in a reservoir. The quantitative integration of sedimentological facies and diagenetic properties is a key point to obtain a realistic geological model in reservoir studies.

Classically, the non stationary lithofacies distribution inside reservoirs is described by the 3D distribution of their proportions. This approach is very attractive because proportions have a physical meaning. Moreover their 3D distribution reflects the qualitative information coming from geology or the quantitative constraints derived from seismic attributes. In models such as the truncated gaussian, the proportions are directly used in the simulation process. In object-based models, such as the Boolean model, the problem is more complex because the proportions are the results of two sets of parameters: the object description (shape and dimensions) and their 3D distribution. The non stationarity in an object-based method can be reproduced either by using a non stationary object description or through a regionalized distribution of the objects. In this paper, we focus on the latter approach. The main contribution of the proposed method is the fact that the fit of the intensity point distribution is obtained globally in one computation step for any non stationary facies distribution. The interest is to constrain, a priori, the lithofacies simulation by a given 3D proportion distribution and not by convergence during the simulation process.

The importance of simulating representative images of heterogeneities no longer needs proving. The remaining problem is the choice of the appropriate model. The main characteristics of the models can be split in two aspects, the conditional one and its ability to reproduce the studied heterogeneities. This paper focuses on this second aspect. The truncated pluri-gaussian method makes it possible to obtain a very wide variety of heterogeneities. With very few parameters conditional simulations with complex anisotropies can be obtained. The parameters are classically fitted on well data and seismic attributes [2]. Moreover the sequence stratigraphy analysis constrains the choice of the relationship between the lithofacies, that is defines their arrangement, order, complementarities, correlationsJ This paper presents some characteristics of this method insisting on the relationships between the lithofacies. Examples of simulations in different environments are also presented Introduction: The fluid movements in the reservoir are guided by the internal heterogeneities of the medium. First at a local scale where the movements depend on the grain sizes and repartition, then at the reservoir scale where the sedimentary arrangement plays an important rule. At both scales, it is important to have representative images to analyze statistically the behavior of the fluids: at the local scale to perform the appropriate permeability up-scaling, and at the reservoir scale to locate the permeabilities barriers or the preferential drains. In order to solve both problems,it is necessary to construct representative images of the medium. The characteristics to be reproduced are very variable from one environment to another, for instance clay plugs in sand bodies deposits, or diagenetic effect around fracturationsJ An efficient tool for reproducing these different shapes is the truncated pluri-gaussian method. The truncated pluri-gaussian tool: The truncated pluri-gaussian method is a generalization of the well-known truncated mono-gaussian method [1]. The strength of this approach is that it makes it possible to reproduce very complex relationships between the lithofacies. The basic idea of this approach is to use several (at least two) gaussian variables (G1 and G2 in figure 1) which condition the spatial structure of the two (or more) different sets of lithotypes. The relationships between the lithotypes are determined by the rock type rule (fig 1-b1) that defines the truncation diagram. The resulting image characteristics then depend on one hand on the spatial structures of the underlying gaussian variables and their correlation and on the other hand on the rock type rule. On this rock type rule diagram, each axis represents the gaussian values, and the rectangular areas correspond to the set of gaussian value couples which define the lithotypes. So, in this example, the red color corresponds to the high values of the second gaussian (G2) and the obtained lithotype (fig1-b1) presents the same anisotropy as G2. This anisotropy can be plotted in a graph giving the values of 90% of the semi-variogram sill for all the directions (fig1-b3). In the same way, the green color corresponds to the lower values of G2 and lower values of G1, the obtained lithotype has the same anisotropy orientation as G1, and the display of the sill presents an ellipsoidal shape orientated west-east. The two other colors (lower values for G2 and higher values for G1) have mainly the same anisotropy orientation as G2 with a smaller influence of G1 anisotropy, the displays of the sills present more rectangular shape. AAPG Annual Convention May 11-14, 2003 Copyright A 2003 by AAPG Salt Lake.City, Utah AAPG Search and Discovery Article #90013©2003 AAPG Annual Meeting, May 11-14, 2003, Salt Lake City, Utah

Object-based approaches are used to simulate oil reservoirs, in particular in a fluvial environment, or to simulate non reservoir lenses in a reservoir facies. One problem with this kind of simulations is the non stationary distribution of the objects in the reservoir. Classically in truncated gaussian methods (Beucher, 1993) non stationarity is given by the 3D distribution of the lithofacies proportions. In the case of object based methods, this matrix of proportions represents the non stationarity of the objects both in terms of distribution and size parameters. In this work we focus on the distribution variability considering that the objects belong to the same family over the whole space. Then the problem is to fit the Poisson intensity of the boolean model knowing the 3D proportion matrix. The result is a matrix of intensity, given the distribution law of the objects. The method described in this paper was first proposed by Schmitt (1996) and tested by Benito (2001) for a one dimensional non stationarity.

Proportions curves which were introduced at the end of the 80’s, have proved to be a powerful tool for sequence stratigraphy analysis and for quantifying lithofacies. In this paper we discuss both the qualitative and quantitative aspects via several examples coming either from analogue studies or from real oil fields.

P215 SEISMIC FACIES MAP INTEGRATION IN GEOSTATISTICAL GEOLOGICAL MODEL - A FIELD CASE Abstract 1 B. DOLIGEZ 1 F.FOURNIER 1 G.JOLIVET 1 S.GANCARSKI 2 and H.BEUCHER 3 1 IFP 1&4 avenue de Bois-Préau 92832 Rueil Malmaison Cedex France 2 TFE Exploration Production In the reservoir characterization field numerous geostatistical methods for external constraint integration are available as well as associated methodologies to build this constraint. The truncated gaussian method in a non stationary framework allows to integrate an external information as a constraint on the computation of a 3D matrix of facies proportions. The supposed prerequisite is that the external

Geostatistics offers fast and flexible methods for reservoir studies and heterogeneity simulations. The starting point in every reservoir simulation process is the construction of the geological model. The type of geostatistical simulation used to reproduce the heterogeneity distribution depends on the sedimentary processes at the origin of the reservoir, and on the type of heterogeneity influencing the reservoir behaviour. This paper presents three outcrop examples that illustrate how to produce realistic geological simulations by optimising the simulation parameters, according to different geological settings. Both siliciclastic and carbonate reservoirs will be described.

The aim of this study was to simulate the geometry of uranium mineralisation in the Rossing deposit. As plurigaussian simulations have been successful in modelling the lithotypes in oil reservoirs (that is, for sedimentary sequences) we wanted to test whether this approach could also simulate intrusive rocks hosted in metasediments. The work has been carried out in two stages. In the first part (Skvortsova et al, 2000) we successfully simulated the overall structure of the deposit. The lithotypes were reproduced in the correct stratigraphic order and in the right quantities, and mineralisation that is concordant with the bedding planes was reproduced correctly. However we were unable to generate mineralised stringers that cut across the bedding planes. This paper shows how this can be done.

The classical approach to constructing reservoir models is to start with a fine-scale geological model that is densely populated with petrophysical properties. Then scaling-up techniques allow us to integrate this detailed information on a coarser grid and to obtain a reservoir model with fewer grid cells, which can be input in a fluid flow simulator.Geostatistical modeling techniques are widely used to build the geological models before scaling-up. These methods provide possible images of the area under investigation that honor the well data and have the same variability computed from the original data. At an appraisal phase, when few data are available or when data obtained from the wells are insufficient to describe the heterogeneities and the petrophysical behavior of the field, additional constraints are needed to obtain a more realistic geological model. For example, seismic data or stratigraphic models can provide average reservoir information with an excellent areal coverage, but with a poor vertical resolution.New advances in modeling techniques allow integration of this type of additional external information in order to constrain the simulations. In particular, two-dimensional or three-dimensional seismic derived information grids or sand-shale ratio maps coming from stratigraphic models can be used as external drifts to compute the geological image of the reservoir at the fine scale. Examples illustrate the use of these new tools, their impact on the final reservoir model, and their sensitivity to some key parameters.

This paper details a methodology for taking into account seismic information into 3D high resolution stochastic simulations of lithofacies. The simulation is based on a non stationary truncated gaussian approach. Within this frame, a 3D matrix of facies proportion is built. It is estimated from both well and seismic-derived proportions. Thus the seismic attributes are previously turned into soft estimates of proportions with different possible statistical calibration techniques. An application of this methodology to the simulation of lithotypes for a carbonate reservoir is described.

A detailed reservoir model is an essential requirement for the most effective recovery of hydrocarbons in place. For this purpose, the internal reservoir characterization has to be based not only on well data but also constrained by seismic information.The methodology for integrating seismic data using the truncated Gaussian method is based on the non stationary approach. This approach, presented at the previous geostatistical congress, consists in using 3D proportion blocks instead of vertical proportion curves in the horizontal stationary approach. The first step of the integration methodology is the calibration of seismic attributes at the reservoir level into geological information related to lithotypes. The second step is the estimation of the corresponding lithotypes proportion blocks using both well data and seismic derived geological information. This paper discusses this second step.Depending on the available information, different geostatistical methods for the proportion image estimations are proposed. These methods have to take account of the following aspects:The seismic derived information either gives proportions of the studied lithotypes or determines zones of fixed vertical proportion curves.The data are related to different supports :punctual lithofacies for the well, with high vertical resolution and a sparse areal coverage, blocks for the seismic data but dense in the 2D space.Then to illustrate the integration of seismic data in lithofacies simulations, an application to a real case is presented.

Abstract Stochastic models are widely used in the petroleum industry for reservoir characterization. Although there is now an extensive literature on the subject either presenting new models or revisiting older ones, most of these papers have been written with the professional geostatistician in mind. This makes it rather difficult for potential users to understand the methods and to work out their strong and weak points. This paper reviews the broad families of stochastic models that are currently available from a practical point of view. Users are faced with a number of questions like whether to use pixels or object based methods, or how to incorporate various type of information. These points will be addressed in this paper.