LE NIR, ISABELLE and GONZALO RUIZ*. Schlumberger-RPC

Abstract: From Images and Open Hole Logs to Sequence Recognition and Sedimentary Environments

Introduction

An accurate evaluation and description of sequences is required for reservoir characterization and modeling.

Often enough, equivalent lithological features are not contemporaneous, resulting in diachronous correlations. The conceptual framework provided by sequence stratigraphy solves this problem. Yet the issue remains complex because sequence stratigraphy, as usually done by geologists or geoscientist, is a time consuming procedure.

Proposed procedure

A sequence stratigraphy interpretation and a logical procedure for facies definition applied to the study of well logs is very useful in order to reduce analysis time. The proposed approach is a combination of traditional "hand made" (Figures 1, 2) and computerized methods (Figure 3).

Using electrical images, dipmeter data and openhole logs it is possible to characterize facies in a key well. Surfaces can be identified by characteristic signatures on open hole logs as peaks. Sequence stratigraphy analysis are based on surfaces and stacking pattern definition. Stacking patterns are interpreted and defined by the geologists and when surfaces are already found this can be done relatively quickly.The facies and sequence stratigraphy framework analyzed in a key well can further be propagated to other wells using computer methods (neural networks). Inputs used for training and interpolation by neural network techniques are borehole intervals where log signatures are characteristic for each facies. These methods are fast, and only a short amount of time is required for testing the validity and applicability of the results, allowing for quick corrections and refinements of the model.

Application for correlation of a set of wells

After a general overview (Figure 1) and observation of available data and the elaboration of a working hypothesis in terms of general depositional setting, the procedure followed in this example is:

Selection of the key well.

This well should gather the best representation of series and sequences as well as subsurface data with good calibration, electrical images of quality, dipmeter information...

Facies and sequence definition in the key well.

Facies are defined on electrical images, logs (mainly gamma ray (GR) and Spectroscopy Gamma Ray (SGR) as well as the associated curves of Uranium (U) Thorium (Th) and Potassium (K)) and on paleocurrent information provided by dipmeter (Figure 2). Six facies are distinguished: three types of sandstones, a high-thorium silt, delta front shales and transgressive-high uranium shales.

Sandstone interpretation is mainly based on electrical images analysis.The first type of.sandstone is considered as tide dominated. These sandstones are formed by large sets of cross bedding alternating regularly with thin shale levels. Some weak internal erosions are observed. The regular alternance of sandstones and shales is similar to those described in some tidal bars, with sandstone foresets draped by shales. The small scale of internal erosions and the consistent foresets direction can be explained by strong asymmetrical tides.

The second type of sandstone is interpreted as channel and/ or bar subenvironment in a delta setting.These sandstones are always located at the top of the elemental sequence. These sequences are shale decreasing upwards.The top of each sequence is formed by a relatively thick (if compared with the sequence thickness) cross bedded package of clean and coarse sandstone.At the base of these levels some small wave structures are observed.Wave structures disappear as coarse sandstones become more important. This indicates a transition from a wave dominated to a nearshore (fluvial to upper shoreface) environment in a prograding setting.

The third type of sandstone is overlying the second one and is located below shales of clearly marine origin.These packages are formed by sands and shales (shaly matrix) not organized in a regular pattern but as a mixed lithology. Cross bedding is observed in this facies. The position of this third type of sandstones in the elemental sequence as well as its higher shale content indicates an increased marine influence.These sandstones are interpreted as transgressive reworking of the second type of sandstones.

Silts are observed along the well, both alternating with or at the base of the second type of sandstones.These silt levels are formed by very fine grained sands displaying gentle cross laminations. Some of these laminations present a wavy pattern. The high thorium content is due to the presence of mica.

Shales are well developed along the borehole. Based on the SGR log, two types of shales are identified: a high thorium-low uranium shale and a high uranium - low thorium one.

Sequence recognition lies on the identification and definition of bounding surfaces such as maximum flooding surface and sequence boundary. In the present example, the maximum flooding surface is represented by a peak in the uranium curve derived from the SGR. The high proportion of uranium is due to the increase of organic matter during transgression. This peak can be identified and traced in the rest of wells, sometimes as a surface, sometimes represented as a thin interval of relatively high uranium content, and separates two well differentiated sequences.

Identification of the elemental sequences.

In this study, several coarsening upwards (GR decreasing upwards) sequences have been defined.Two types of elemental sequences can be recognized. Sequences located above the maximum flooding surface are slightly shalier and uranium richer than those located below the maximum flooding surface (Figure 1). Sandstone bodies are less developed above the maximum flooding surface than below.

Identification of facies forming the"elemental sequence" in the key well (Figure 2).

The elemental stratigraphic unit is composed of transgressive shales, delta front shales alternating with high thorium silts and overlaid by sandstones.After defining the set of facies, their distribution along the key well is tested by neural networks techniques.The elemental stratigraphic unit and facies succession can vary slightly from the set of facies used for final propagation.

Selection of facies (and number of facies) and distribution in the key well (Figure 3).

This step is very important for further propagation. Several possibilities must be tested. In this example, definition of only one sandstone lithology results in a quick correlation and sequence identification. Definition of several sandstone facies (sandstone d, sandstone e, etc. Figure 2) rather than a single "sandstone lithology" is more realistic, but gives more complex results.A balance is needed between these two approaches.

A first evaluation was done grouping all the sandstone facies into one sandstone lithology. This kind of definition is very simplistic but very useful. Propagation of lithologies, instead of facies, highlights the sequence stratigraphy framework and results in an accurate "a priori" indication of the position of facies.The position of shales and sandstones helped us to test the validity of the sequence stratigraphy model.

Regarding the propagation of facies, the first run was done after selecting input facies intervals representing the sandstone facies, silt and shales observed in the key well, and using GR, SGR,RHOB (density) and DT (sonic) curves. Furthermore several runs were done, varying the number of logs used in facies definition. Classifications using several log combinations were done (GR and Thorium, GR and Potassium, GR and Uranium, Uranium and RHOB, etc.). The spectroscopy gamma ray (SGR) in this environment gives a good set of logs for facies discrimination. Some of the sand intervals have a high gamma ray response, equivalent to a shale, due to the presence of micas.This ambiguity can be lifted when comparing uranium (U), thorium (Th) and potassium (K) values over the same interval which clearly define sands. The introduction of sonic (DT) in our model is not advisable as in this case DT evidences the gas content of the formation more than its facies.

Three sandstone facies were in fact propagated. The first sandstone facies, interpreted as tidal in the key well is located exclusively at the base of the stratigraphic interval. The second and third sandstone facies appear at several levels in the well. The second sandstone can be interpreted as amalgamated delta mouth bar and channel sandstones, always in a progradational context. The third sandstone appears in retrogradational intervals, overlying the second type of sandstones. The last type is interpreted as a transgressive reworking of these second type sandstones.

Propagation and testing in all the wells (Figure 3).

Facies of the key well are then propagated to other wells using neural network techniques. Distribution of propagated facies must agree with the sequence stratigraphy framework. Characteristics of facies in the new wells are compared with those at the key well.

On cross plots it is possible to compare input facies with the resulting estimation. If the cloud of points representing the estimated facies differs greatly from the cloud of points used in the definition of the same type of facies in the key well, an inaccurate definition or a lateral facies substitution must be suspected. In our example a divergence was observed in the thorium - rich silts and in the delta front shales between definition and propagation. A tuning of these facies was necessary (based on a sharper definition).

Summary of observed results

The definition of a sequence stratigraphy framework highlighted the existence of two intervals with a different content in uranium.These two sequences are separated by a high uranium correlated interval interpreted as the maximum flooding surface.

The definition of the elemental sequence was based on the SGR. These elemental sequences are bounded by relative flooding surfaces.

The automatic recognition of two different types of shales defined a transgressive uranium-rich shale and a delta front shale characterized by low uranium content and high thorium.

High thorium silts were automatically distinguished.

Differentiation of three different sandstones became possible by addition of the density (RHOB) to the logs used in propagation of facies by neural network techniques.

Distribution of facies along the field highlights the sedimentary subenvironments distribution.

Conclusions

Inherent to the success of the sequence stratigraphy is the ability to regionally correlate individual surfaces. Sequence stratigraphy and facies analysis are essential for correlation, interpretation and modeling.The correct selection of a key well is also the "key" for next steps. Once a first and quick study of the wells is carried out, efforts can be focused on the key well for facies and sequences definition and distribution along the well.

If uncertainties remain at this point they are related to the data set used for definition (logs and input facies intervals) more than to the computerized propagation.

In detritics, for quick sequence and lithology definitions, GR and SGR are very useful. Thus, a first propagation should be done using these logs.The results highlight the existence of sequences and some of their characteristic features.

The model can then be refined by adding new logs to the data set used for computation. The model will be more and more complicated as new logs are added. The final (refined) result should fit with the boundaries defined in the previous low-refinement propagation.

From a general point of view (different sets of wells in different sedimentary environments and geological settings) the extent to which the diagnostic features (such as the maximum flooding surface) can be traced from one well to another will depend on the quality of data, but we should use whatever and however many methods are necessary to confidently make lateral correlations; i.e. cross plots at different depth intervals of GR vs.Thorium, GR vs. Uranium, GR vs. dip azimuth, RHOB, DT and so forth. These cross plots are very easy to construct and display, allowing for a quick visualization of intervals with special features. Working in this way, and using the conceptual sequence stratigraphy framework, traceable surfaces, intervals and facies can be identified.

Combining the information of surfaces and intervals with sequence stratigraphy concepts, a valid model can be built.

Efforts can be focused on the key well, defining facies and refining boundaries definitions for further propagation by neural networks techniques.

The success of this procedure lies in the repeatability of the process. Results can be propagated very quick to a newly drilled well and defined facies are automatically highlighted. Comparing the characteristics of facies as primarily defined in the key well and in the new well by cross plots (same or different distributions in the crossplot) allows to identify intervals were facies are properly defined and intervals were new facies must be interpreted.

AAPG Search and Discovery Article #90923@1999 International Conference and Exhibition, Birmingham, England