Datapages, Inc.Print this page

Click to view article in PDF format. 

 

Rock Property Data Volumes from Well Logs*

By

L.R. Denham1 and H. Roice Nelson, Jr. 2

 

Search and Discovery Article #40268 (2007)

Posted November 3, 2007

 

*Adapted from extended abstract prepared for poster presentation at AAPG Annual Convention, Long Beach, California, April 1-4, 2007

 

1Interactive Interpretation & Training, Inc., Houston, TX ([email protected])

2Geokinetics Processing & Interpretation, Houston, TX    

 

Abstract 

Geology is sampled densely by wells vertically, but sparsely and irregularly horizontally. A modern seismic survey has sparse samples vertically, but close and uniform sampling horizontally. Seismic data is more useful for regional trends, but does not convey petrophysics, and is difficult to relate to wells. Inverting seismic data to resemble well data is almost impossible. Can we make well data resemble seismic data, not just a single-trace, but a complete volume?  

We reduced well vertical sample interval to 60 m, computing numerous petrophysical properties. Then we computed vertical arrays of values for each property on a regular grid, using the samples computed at each well location. Wells within a specified radius were used, weighted inversely with distance, and well samples over a limited depth range, weighted inversely with depth difference from the sample depth. The computed three-dimensional array was written to disk in SEG Y format, and loaded into a seismic interpretation system.  

In the initial project we generated data volumes for shale and sand P-wave velocities and densities, sand percentage, and pressure (mud weight), for most of the Gulf of Mexico. These volumes showed regional trends within the basin. The method has some problems in areas of sparse well information, and does not account for major structural features.  

Petrophysical data volumes generated from well information allow the geologist to integrate information from thousands of wells using standard interpretation systems. So far, the technique seems to be more suited to regional analysis rather than to prospect development.

uAbstract

uFigure captions

uIntroduction

uGeneration of volumes

uResults

uUse

uConclusions

uReferences

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

uAbstract

uFigure captions

uIntroduction

uGeneration of volumes

uResults

uUse

uConclusions

uReferences

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

uAbstract

uFigure captions

uIntroduction

uGeneration of volumes

uResults

uUse

uConclusions

uReferences

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

uAbstract

uFigure captions

uIntroduction

uGeneration of volumes

uResults

uUse

uConclusions

uReferences

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

uAbstract

uFigure captions

uIntroduction

uGeneration of volumes

uResults

uUse

uConclusions

uReferences

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

uAbstract

uFigure captions

uIntroduction

uGeneration of volumes

uResults

uUse

uConclusions

uReferences

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

uAbstract

uFigure captions

uIntroduction

uGeneration of volumes

uResults

uUse

uConclusions

uReferences

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

uAbstract

uFigure captions

uIntroduction

uGeneration of volumes

uResults

uUse

uConclusions

uReferences

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

 

uAbstract

uFigure captions

uIntroduction

uGeneration of volumes

uResults

uUse

uConclusions

uReferences

 

Figure Captions

Figure 1: Sand density in the Gulf of Mexico at 10,000 ft (3050 m) below the sea floor.

Figure 2: Sand-shale ratio at 5000 ft (1525 m) (a) and 15,000 ft (4570 m) (b); number of sand samples per trace (c). In (a) and (b), 50% sand is green and 0% sand is magenta and red. The gaps in the volume increase with depth. This is partly explained by the total number of sand samples per trace, shown in (c), in which largest values are red; smallest are blue.

Figure 3: Quality control plots. (a) Mud weight at 5000 ft. (b) Mud weight at 10,000 ft. Black areas are excluded by a limiting polygon. (c) Sand P-wave velocity at 10,000 ft (d) Crossline through mud weight volume

Figure 4: Sand P-wave velocity. (a) 11,900 ft (3630m) below sea floor. (b) Green Canyon to Sabine Pass (A-A’).

Figure 5: Sections showing shale P-wave velocity. (a) Texas to Florida (B-B’). (b) Texas to Louisiana (C-C’). Vertical scale in feet below sea floor.

Figure 6: 10,000 ft (3050 m) below sea floor (a) Temperature in degrees F. (b) Shale density. (c) Standard deviation of sand velocity.  

 

Introduction 

One of exploration’s perennial problems is relating well-derived geological information to seismic data. An additional tool for this is now available: data volumes of rock properties in SEG Y format, generated from well logs. These volumes are compatible with all standard seismic-interpretation systems and can be used by the interpreter to constrain interpretation of seismic data by giving the probable properties of rocks in an undrilled prospect.

 

How The Volumes Are Generated 

The starting point for the rock-property volumes is the standard suite of well logs. The standard set of rock properties in a dominantly clastic sequence is derived from velocity, density, and resistivity logs. Properties computed using fluid replacement require measurements or assumptions about properties of the fluids, such as oil and gas density, water salinity, and gas-oil ratio. Temperatures are based on measurements made while logging, and formation pressures are estimated from drilling mud weights.  

A petrophysicist classifies the rocks penetrated by each well, separating intervals into water-filled sand, shale, and all other lithologies (salt, coal, limestone, hydrocarbon-filled sand, etc.). These last intervals are excluded from the analysis.  

Wells are then divided into uniform depth intervals, using an interval large enough to contain significant quantities of both shale and water-filled sand, but small enough to adequately describe systematic variations. If the chosen interval is too small, many of the intervals will contain only sand, or only shale. If the interval is too large, there may be significant differences in rock properties from top to bottom, due to the difference in compaction, and more depth samples will include rocks with widely varying depositional environments. For the Gulf of Mexico examples described here, the interval chosen is 200 ft (61 m).  

For each interval, the averages of the fundamental properties of sand and shale are computed, along with the amounts of sand and shale within the interval and the variation of each property within the interval (recorded as standard deviation). Additional rock properties can be computed from the fundamental properties using standard procedures such as the Greenberg-Castagna technique (Greenberg and Castagna, 1992) for computing shear-wave velocities, inverse Gassmann’s equation (Gassmann, 1951) for computing dry-rock properties, and Gassmann’s equation along with the dry-rock properties to compute the properties of hydrocarbon-filled sands (Hilterman et al., 1999, Hilterman, 1990; Hilterman et al., 1998).

 

Once the well database is constructed, the SEG Y data volumes can be generated. There are several points to consider carefully:

  • What trace interval should be used for the volume? A close trace interval is likely to be more useful in comparing rock properties with seismic data, but may give a misleading impression of reliable detail. A trace grid exactly matching that of an existing 3D survey may be particularly useful. Most of the work done so far involves regional data volumes with a trace spacing much larger than normally used for seismic data. Where logs are available from a large number of wells in a developed field the horizontal sampling by the wells may be comparable to the seismic sampling. In such cases, the detail in the well data volume may be as good as that in the seismic volume.

  • What map projection should be used? The well locations are defined in latitude and longitude, but a SEG Y 3D data volume must be defined in a projection. For a regional volume, the differences between volumes can be quite noticeable: we have generated volumes over most of the Gulf of Mexico using both the Louisiana South projection and Universal Transverse Mercator Zone 15. In both cases the area covered goes well beyond the area normally used for the projection.

  • How far should we interpolate between wells or extrapolate from a single well? In areas with many wells, this is not a critical decision, but in the deep-water areas of the Gulf, for example, where wells are widely spaced, it is an important parameter. Even when interpolation is adequate at shallow depths (Figure 2a), it may not be deeper (Figure 2b). A plot of the valid samples for each trace (Figure 2c) may help the user choose the best compromise: using too large a distance rapidly increases the computation effort and may give the impression of reliable information where there is none; and using too short a distance leaves large gaps in the data volume.

  • How far should we interpolate or extrapolate vertically? Wells are often missing log data from part of their depth range, and sand properties may be missing over a depth range simply because there is no sand for several hundred feet. The well database is carefully constructed to leave gaps where data is missing, but by producing traces on a regular grid we always generate values where there are no data. How far do we want to carry this process?

  • What vertical sample interval should be used? The wells are sampled at a fixed interval, but there is no reason why the volume generated should use the same interval. A closer interval will give a smoother transition in areas where there are abrupt changes in properties with depth. The volume could also be generated in reflection time, to match seismic data, if desired. In most cases there will be adequate velocity control from the well information alone to do this.

  • What depth range should we use for the data volume? So far, we have generated volumes with a sea-floor datum, typically starting close to the sea floor, and going to the depth of the deeper wells in the area. There is no point in going shallower than the shallowest data, or deeper than the deepest data, and there is little point in generating samples to a depth reached by only a very small number of wells or logs.

  • Should the area of the volume be limited (by a lease line, for example, or to restrict extrapolation into areas of no interest or little data, as in Figure 3)?

 

When these questions are answered, the volume is generated. The process follows these steps for each trace:

  • Compute the location of the trace in map projection X and Y, using the specified grid: origin, orientation, and spacing along inlines and crosslines.

  • Convert the location to latitude and longitude (the only uniform location information in the well database for all wells is the geographic location: the map projection used for X and Y coordinates varies with state and zone).

  • Check that the trace is within the area of interest (if defined by a limiting polygon).

  • Identify all wells within the specified extrapolation distance.

  • For each sample:

    • Search the identified wells for data within the vertical interpolation distance specified.

    • Compute a weighted average value of the desired rock property, weighting the well data inversely with distance, and inversely with difference in depth from the depth of the sample.

 

As each trace is completed, it is written in 32-bit floating point format to a standard SEG Y format file (Barry et al., 1975) which can be loaded into any seismic-interpretation system.  

The generation of these volumes takes time, so we generate graphical progress reports (updated every 1000 traces), allowing the user to check that the values used for interpolation and limits on the area covered are realistic without waiting for the job to finish. These plots are of two forms: maps (Figures 1, 2, 3a-c) and sections (Figure 3d).

 

Results 

The data volumes are loaded into standard seismic-interpretation systems (in this example, the Halliburton Landmark SeisWorks application), where they can be manipulated in the same way as ordinary seismic data. Figure 4a shows the P-wave velocity of water-filled sand in the western Gulf of Mexico (Texas to Alabama) at a depth 11,900 ft (3630 m) below the sea floor. The patches of background color left of the middle of the figure indicate areas where there is no well data available to this depth. The arcuate edges of data along the southern limits come from the distance limit on extrapolation from widely separated wells.  

Figure 4b shows a section through the same data volume, running from the middle of the Green Canyon area on the left to the Sabine Pass area on the right. The gaps in the bottom of the section mark areas with no deep wells (or no deep logs). The missing data at the top of the section at the left is where velocity logs were not available at depths less than 7500 ft (2290 m) below water bottom (the last 5% of the section depends on a single well). The white line marks the depth of geopressure as interpreted by examination of each well.

 

Uses for the Volumes 

These volumes have great potential for increasing an explorationist’s productivity and for defining more closely the risk of a prospect.  

Suppose, for example, you have identified a potential prospect on an OCS block in the Gulf, miles from the nearest existing well, and want to know whether the AVO anomaly associated with the prospect is what would be expected in that location at that depth, for either oil or gas. The usual solution is to model the AVO response. But the modeling program requires values for shale P-wave velocity (Figure 5a) and density (Figure 6b), sand P-wave velocity (Figure 3c, 4), sand density (Figure 1) and thickness, as well as depth (which can be determined from the seismic interpretation), mud weight (Figure 3), temperature (Figure 6a), gas density, oil density, gas-oil ratio, salinity and water saturation: a total of twelve unknowns. The new tool can provide a data volume derived from well data for six of those unknowns, so values can be extracted almost instantly. Only gas and oil density, gas-oil ratio, salinity, water saturation, and sand thickness remain, and hydrocarbon densities, gas-oil ratio, and salinity tend to vary relatively slowly from region to region. The interpreter can now concentrate on varying sand thickness and water saturation in the model, looking for a match to the observed AVO response.  

The variability of rock properties is important in estimating the probability of success for a prospect. Figure 6c shows the standard deviation of water-filled-sand velocities at 10,000 ft (3050 m) below the sea floor. This is one indicator of the variability of sand properties at this depth. Similar volumes can give actual measurements of the variability of other properties used for estimating probable reserves for a prospect.  

At a simpler level, the interpreter may need to know whether an observed change in amplitude at an apparent fluid contact is compatible with a change from water to oil. This question could be answered by comparing the difference in values from an oil sand reflectivity volume and a wet sand reflectivity model with the change in amplitude observed in the real seismic data in an intercept stack volume. This would be a deterministic solution analogous to the probabilistic solution described by Denham and Johnson (2006).  

On an even more basic level, a gross overview can be quickly accessed, with mud weight (Figures 3a and 3b), for example, showing regional variations in geopressure at any depth.

 

Conclusions 

By combining two universally-used exploration tools – well logs as actual measurements of rock properties, and workstations for viewing three-dimensional data volumes – the explorationist can improve productivity and reduce risk by making better use of existing data. The missing link between the two tools is the uniformly-sampled data volume in a standard format, generated from irregularly-scattered well data.

 

References 

Barry, K.M., Cavers, D.A., and Kneale, C.W., 1975, Report on recommended standards for digital tape formats: Geophysics, v. 40, no. 2, p. 344–352.

Denham, L.R., and Johnson, D., 2006, Estimating probability of hydrocarbon content from seismic amplitude anomalies: Soc. Explor. Geoph. 76th Annual Meeting, INT3.3.

Gassmann, F., 1951, Elastic waves through a packing of spheres: Geophysics, v. 16, no. 4, p. 673–685.

Greenberg, M.L., and Castagna, J.P., 1992, Shear-wave velocity estimation in porous rocks: Theoretical formulation, preliminary verification and applications: Geophys. Prosp., v. 40, no. 2, p. 195–210.

Hilterman, F., Sherwood, J.W.C., Schellhorn, R., Bankhead, B., and DeVault, B., 1998, Identification of lithology in the Gulf of Mexico: The Leading Edge, v. 17, no. 2, p. 215–222.

Hilterman, F., Verm, R., Wilson, M., and Liang, L., 1999, Calibration of rock properties for deepwater seismic: 69th Ann. Internat. Mtg, p. 65–68.

Hilterman, F., 1990, Is AVO the seismic signature of lithology? A case history of Ship Shoal-south addition: The Leading Edge, v. 9, no. 6, p. 15–22.

 

Return to top.