uAbstract

uFigure captions

uIntroduction

uAim & method

uModel construction

  uCore data & log analysis

  uBorehole images

  uModel synthesis

uModel propagation & results

uConclusions

uReferences

uAbstract

uFigure captions

uIntroduction

uAim & method

uModel construction

  uCore data & log analysis

  uBorehole images

  uModel synthesis

uModel propagation & results

uConclusions

uReferences

uAbstract

uFigure captions

uIntroduction

uAim & method

uModel construction

  uCore data & log analysis

  uBorehole images

  uModel synthesis

uModel propagation & results

uConclusions

uReferences

uAbstract

uFigure captions

uIntroduction

uAim & method

uModel construction

  uCore data & log analysis

  uBorehole images

  uModel synthesis

uModel propagation & results

uConclusions

uReferences

uAbstract

uFigure captions

uIntroduction

uAim & method

uModel construction

  uCore data & log analysis

  uBorehole images

  uModel synthesis

uModel propagation & results

uConclusions

uReferences

Figure Captions

Return to top.

Introduction

Recognition of sedimentary lithofacies is of prime importance when considering the management of uncertainties for the geological model of a field. This paper discusses how a facies description can be obtained from a complete well log dataset--including Electrical Borehole Imaging (EBI)- calibrated to core descriptions. This approach is attractive as it reduces the necessity of an extensive and onerous coring program and covers the complete reservoir interval.

Aim and Method 

Our approach aims to provide rapidly the reservoir geologist with a virtual core, obtainable throughout the entire reservoir section and based on the available core data and wireline logs, including EBI and NMR logs. A sedimentary lithofacies may be first described in terms of mineralogical content. In the absence of core, this aspect, as well as reservoir property evaluation, is routinely evaluated by conventional log interpretation. However, a complete facies description also takes into account sedimentary structures, texture, and the arrangement of the different lithologies. In the uncored areas, this complementary facet was traditionally accessed through manual interpretation of EBI and was superficially taken into account. Automated interpretation of borehole images allows faster interpretation (while removing interpreter bias) at a resolution that can approach that of the eye of the geologist looking at the core (Figure 1). 

In order to create a model, a conventional wireline log suite, borehole imaging log, and representative core data are necessary. This model can then be propagated over any number of wells in an equivalent geological setting that proffer the appropriate data set, giving a complete virtual core over the interval of interest.

Model Construction 

Core Data and Conventional Log Analysis 

The fundamental parameter for accurate modeling is the availability of a relevant and representative core description for use as calibration. However, the interpreter must be aware that logging-tool resolution causes a smoothing (up-scaling) of the basic geological information which must be taken into account during model construction. 

All log data are checked and validated (mainly normalized). Wireline log responses are then analyzed to see which help discriminate core facies. These are used as inputs for the clustering phase, the results of which are calibrated to the core facies. A first case study, illustrated in Figure 2, presents the distribution of the different core facies on a Density vs. Neutron cross-plot for a sandstone and shale succession. Despite fuzzy boundaries, the different core groups are well distinguished: 

- Facies 1 (massive shales) indeed corresponds to the most argillaceous lithologies.

- Facies 2 (mud-dominant heterolithic alternations) enriched in quartz compared to Facies 1.

- Facies 3 (sand-dominant heterolithic alternations) shows an increase of effective porosity coupled to a trend of decreasing shale content, up to clean siltstones or sandstones.

- Facies 4 to 6 are identified as the cleanest sandy lithologies.

- Facies 7 (mudstone breccias) is superimposed on the other facies but does not completely belong to the conventional lithology/porosity trend described above. It also appears as an intermediate stage between massive shales and clean, porous sands.

In the case above, comparing the electrofacies classes and the core facies shows that wireline logs permit a confident identification of the distribution of shale, sand, and porosity. In more complex sedimentary environments (e.g., dissolved carbonates), further refinement can be introduced using the T2 NMR spectra (Figure 3), which describe both fluid nature and porosity geometry. For a constant saturation, it is possible to discriminate different porosity network types.

Borehole Images 

Borehole images are first depth matched with the conventional logs. Checking and repairing the image is a crucial step of the process. The borehole images are then analyzed automatically and a set of relevant data is produced to help with facies identification: 

- Image texture attributes classified using a Kohonen 2D neural network.

- Bed boundaries take into account specific image attributes (‘‘connective components’’). This process identifies lithological and/or textural changes, although they are not quantified nor qualified.

- Bedding takes into account all significant dips, which are automatically picked on the image in a manner comparable to a manual picking.

Additional curves are calculated to characterize the image better, such as event continuity (does it fit all the pads?), event contrast (does it fit a resistivity change?), bedding index (abundance of bedding surfaces in a given depth interval), stratification index (abundance of bed boundaries in a given depth interval), image activity (is the image homogeneous or not?). These are then up-scaled and interpolated to conventional log sampling rate.

Model Synthesis 

The interpreters’ role becomes critical when it comes to combining the lithology from conventional wireline logs with the texture and bedding information from EBI. In effect, this boils down to highlighting and optimizing the relationship between raw mineralogy, image features, and core facies. The use of contingency tables quantifies this relationship and model quality.

Model Propagation and Results 

The model can be propagated over all relevant wells with appropriate data sets. Ideally, further comparison should be performed using a cored well kept aside for the purpose of a blind test. Figure 4 shows a complete data set and modeling results: conventional wireline log curves (left side), borehole image (middle track), dipmeter output (bed boundaries in red, bedding in green), and finally the core facies column and the model output (right side).

Conclusions 

A continuous description of the well-bore is obtained in terms of geological facies using core descriptions as a reference, and both borehole images and conventional well logs as model inputs. Good success rates have been achieved in both clastic and carbonate environments. 

Beyond the calculation of a synthetic facies column, the process leads to the production of a homogeneous database containing conventional logs and EBI, available for further geological and reservoir studies (e.g., permeability modeling, fracture interpretation, helping seismic and structural understanding using dipmeter trends).

References 

Ye, S., Rabiller, P., and Keskes, N., 1997, Automatic high resolution sedimentary dip detection on Borehole Imagery - SPWLA 38th annual symposium.

Ye, S., and Rabiller, P., 2000, A new tool for electro-facies analysis: Multi-resolution graph-based custering - SPWLA 41st annual symposium.

Leduc, J.P., et al., 2002, FMI based sedimentary facies modeling, Surmont lease (Athabasca, Canada).

Return to top.