You are on page 1of 8

INTERPRETERS CORNER

Coordinated by REBECCA LATIMER

Typical AVO response as a function of depth and uid pressure gradient: Gulf of Mexico shelf
RICK BOX and ERIC DOSS, Hunt Petroleum, Houston, USA

he eect of uid pressure gradient on AVO response is not easy to discern in the literature and is awkward to communicate. Most explorationists cannot predict how log responses and AVO signatures will change if the uid pressure gradient is dierent at a prospective area than at known control (even if their depths are the same): An increase in uid pressure gradient means a decrease in the hardness values of the acoustic properties (VP, VS, and Rho) for both the shales and the sands, but since AVO varies in a nonlinear way with the contrast in these six variables, it is dicult for us to intuit what the overall eect on AVO will be. The depth may also dier, making the problem even more complex. In Tertiary basins, the most important driver of acoustic properties is compaction (which is primarily a function of depth, but is also aected by uid pressure gradient). Most modeling packages provide somewhat automated ways to calculate the eect of changes in uid content, thickness, and porosity but not uid pressure gradient. Equations and methods exist (Gassmann uid substitution, Greenberg-Castagna equations, Gardner equation, Widess diagrams, etc.) to investigate the eects of changes in density, compressional velocity, shear velocity, thickness, uid content, bandwidth, and many other variables, but uid pressure gradient is conspicuously absent. To help ll in this gap, we present a new method of modeling the eects of uid pressure gradient on AVO, along with some new types of diagrams to illustrate the results. The results of this study suggest that uid pressure gradient is as important as depth in controlling AVO.

study of the area Denition of one standard lithological scenario, including sand thicknesses, bed boundaries, and uid contents.
Log trends Our study area contained 13 wells from the Louisiana Gulf of Mexico shelf, from West Cameron to Main Pass, including East Cameron, Vermilion, South Marsh Island, Eugene Island, and South Timbalier. The wells were selected on the basis of their extensive logged range, nearly complete log suites, relative verticality, and interpreted quality. This large area (approximately 270 100 miles) illustrates the lateral consistency of rock properties. A smaller area would have less variation, and a correspondingly tighter denition of typical. The wells were puried (that is, the logs were petrophysically edited using a vendors proprietary technique); a lithology fraction curve has been derived from lithology logs (GR, SP, etc.), a uid pressure gradient curve has been derived from the shape of the resistivity, sonic, and density curves; and the sonic and density logs were edited to contain no values that contradict other curves. The 13 wells contain more than 200 000 samples. For each acoustic variable, it is necessary to perform a dual-histogram analysis to separate the eect of pressure from the eect of depth on compaction. For example, the analysis of the VP acoustic variable for shale is shown in Table 1. The upper left bin shows that the average sonic transit time for all points whose uid pressure gradient falls between 0.45 and 0.55 psi/ ft, and whose depth falls between 1500 and 4500 ft is 142 us/ft. The mechanics of this process are illustrated in Figure 1a, where all points whose lithology fraction falls between 0% and 20%, in the rst pressure column (0.45 < uid pressure gradient < 0.55) have their puried sonic transit time plotted against depth. The results form a trend curve, showing how sonic velocity increases smoothly with depth (because of compaction) when other variables are constant. The central values were extracted every 3000 ft to form inputs to Table 1, column 1. The nal two entries in this column are none because no shales of such low pressure exist below about 12 500 ft (arguably, the values could be extrapolated from the trend curve, if there were reason to believe that such rocks existed in some undrilled place). Four more graphs like Figure 1a were required to ll in the four remaining columns of Table 1. Not surprisingly, the top two entries in the nal column are none because no shales of such high pressure exist above 7500 ft. Figure 1b shows a potential pitfall. Plotting puried sonic log values against depth for the 13 wells for shales of all pressures results in a trend that is less clear. It would be very easy to conclude that there is a poor relationship due to the combination of wells from a wide area which may encompass very diverse geology; but Figure 1a demonstrates that holding

Concept To understand the eects of depth and uid pressure gradient on AVO, the range of uid pressure gradient values (0.45 to 0.95 psi/ft) may be split into ve equal increments, analyzed, and modeled separately. Within each pressure range, ve equal depth intervals may be analyzed and modeled. Within each depth interval, the acoustic parameters (V P, VS, and density) may be linearly interpolated from typical values at the top and bottom of the interval. The typical V P and density values for wet sand and shale may be estimated from trend curves (dened below); the VS values may be computed using Greenberg-Castagna; the values for oil and gas cases may be computed using uid substitution. Values other than typical ones are possible, but may be ignored here, because the purpose is to hold all variables except depth and uid pressure constant. The result is a 5 5 grid of AVO models which demonstrates the relative changes in AVO caused by changes in depth and uid pressure gradient. Determining the input parameters for each grid cell requires:

Knowledge of the typical acoustic parameter values for sand and shale in each cell determined from a well log
1252 The Leading Edge October 2008

INTERPRETERS CORNER

Figure 1a. Puried sonic logs versus depth for 13 wells, allowing normally pressured (0.450.55 psi/ ft) shales (020% lithology fraction), color coded by well. There is a single trend, as expected. Variations from the trend are mainly due to di erences in sand quality (grain size, sorting, etc.). Di erences between wells are small: about the same as di erences within wells. Five of these 2D trends (for the ve uid pressure gradient intervals studied) combine to form a coherent 3D whole (compare to Figure 2).

Table 1. Sonic transit times in shales (0.00 to 0.20 lf).

pressure constant removes this impression. This is often the casewhat appears to be a lateral variation in rock properties disappears when the log data are puried, and pressure is accounted for. Figure 1c extends this idea. Plotting puried sonic log values against depth for the 13 wells for all rocks (shale and wet sand) for all pressures results in a trend that is marginally less clear than Figure 1b. Analyzing sands and shales together (something most people would instinctively avoid) causes less trend obfuscation than failing to account for pressure. Togeth-

er, Figures 1a, 1b, and 1c illustrate that pressure and depth are rst-order controls on sonic velocity, while lithology is secondary, and lateral variations are tertiary. This accords with laboratory experiments, physical intuition, and results we have observed in clastic basins around the world. The numbers in Table 1 were contoured, as shown in Figure 2, in order to conrm that the sonic trends in shales vary smoothly with respect to pressure, as they do with respect to depth (as shown in Figure 1a). Similar trend curve analysis was done for wet sand transit
October 2008 The Leading Edge 1253

INTERPRETERS CORNER

Figure 1b. Puried sonic logs versus depth for 13 wells, allowing shales (0.2 lithology fraction) of all pressures, color coded by well. The trend of traveltime versus depth is unclear. We will show that the reason is not because wells from a wide area (270 100 miles) were combined; it is because shales of various pressures were combined. Compare to Figure 1a.

(a at oil) were also held constant to allow comparison between models. Standard scenario for modeling An arbitrary geologic scenario was chosen, consisting of a 3000-ft interval with seven distinct sands separated by shale intervals (Table 2). Within each interval, the lithology varies somewhat randomly to simulate the jitter in typical lithology logs due to the ne-scale layering of the Earth. All other logs in the suite were calculated from this standard lithology log, parameterized by the rock property results (such as Figure 2). The wet sands (A, C, E, and G) are identical to each other in thickness, lithology, and uid content, but since they occur at dierent depths, they have dierent acoustic properties. This illustrates the eect of compaction on the AVO background trend lines. Sands B and D contain gas and have identical sharp bases; but B has a transitional (shaling upward)

Table 2. Standard scenario for modeling.

times, wet sand densities, and shale densities. The results were also stable, conrming that the values are representative of the region, and appropriate for use in AVO modeling. There was not enough shear log data, or data in gas and oil sands, to allow analysis of these cases by the method shown above. Shearvelocity values were computed using the standard GreenbergCastagna method. Fluid substitution was used to nd the gas-sand traveltime and density due to the insucient log data over productive zones. Parameters used for gas (a standard dry gas) and oil
1254 The Leading Edge October 2008

INTERPRETERS CORNER

Figure 1c. Puried sonic logs versus depth for 13 wells, allowing all lithologies and all pressures, color coded by well. The trend of traveltime versus depth is unclear. Compare to Figure 1b.: Plotting sands and shales together obscures the trend, but not as severely as disregarding pressure did.

top, while D is blocky, lithologically identical to A, C, E, F, and G. Comparisons of B to D allow investigation of whether the crispness of the top has appreciable eect on the AVO signature. This model, unlike most published AVO models, accounts for the eects of compaction. Each parameter was interpolated at each intermediate depth. For example, given a shale traveltime of 124 us/ft at a depth of 3000 ft and 114 us/ft at 6000 ft, the traveltime at 4500 ft depth will be 119 us/ft. All sands are thinner than a quarter of the wavelength of the wavelet, so tuning is a factor in the amplitudes (as it is in most gas elds in the Gulf
1256 The Leading Edge October 2008

Figure 2. Shale transit time versus depth and pressure. Binned pressures vary from 0.45 to 0.85 psi/ft on the horizontal axis and depths vary from 3000 to 18 000 ft on the vertical axis; together they form a 6 5 grid, with 22 of the 30 cells populated. Acoustic transit times for shale (s/ft) are posted and contoured. Lowest velocities are towards the top right, where high uid pressure and shallow burial combine to produce the least compaction. Each column of values was extracted from a trend curve, such as the one in Figure 1a. The composite of ve 2D trends forms a 3D trend with good consistency, as conrmed by the smooth contours. This map contours shale transit time; similar maps of sand transit time, shale density, and sand density were done and also observed to be smooth.

INTERPRETERS CORNER

Figure 3. An AVO model (depths from 12 000 to 15 000 ft; uid pressure gradient xed at 0.85 psi/ft). Log suites (inferred from trend curve results) were put into a Zoeppritz AVO model, to generate angle gathers for traces from 0 to 50 incrementing by 5 degrees, using a 20-Hz zero-phase Ricker wavelet. A standard set of AVO displays were generated from the gathers. The 010 traces were summed to produce the near trace (rst black trace). The 3045 traces were summed to produce the far trace (second black trace). A standard two-term Aki-Richards curve-t was performed on the gather to compute the intercept (rst blue trace) and gradient (second blue trace). These were crossplotted to dene a chi angle, and coordinates were rotated by the Chi angle to produce the uid (red) trace and the lith (gray) trace. The rotation was chosen to make the uid trace zero for wet sands. But, due to compaction, the four wet sands di er somewhat. In this case, none of the sands were completely zeroed, although sand A was closest. Di erent depth and pressure scenarios were run through identical AVO modeling. The results are compared in Figure 4 .

of Mexico). It would be instructive to repeat the modeling for a thicker-than-tuning case. AVO modeling For each scenario, an AVO model was computed using an identical workow. Figure 3 shows the model for the highest uid pressure gradient category (0.850.95 psi/ft) and fairly deep depths (12 00015 000 ft). Shale fraction, uid pressure, sonic, density, and resistivity logs were generated by the interpolation method described in the previous section, using parameters derived from trend curve analyses (such as the example shown in Figure 2). Shear logs were generated via Greenberg-Castagna methods (there was insucient measured shear data to grid and contour them similarly to the density and compressional velocity). Seismic angle gathers (red and blue color-density displays in Figure 3) were modeled using a full Zoeppritz solution and a constant 20-Hz zerophase Ricker wavelet. A two-term Aki-Richards t through
1258 The Leading Edge October 2008

the gathers produced the intercept and gradient traces (blue). A crossplot of the intercept and gradient was made for each model. The four wet sands clustered more or less together, and a regression was t through these points. The angle, chi, of this regression line was used to rotate the axes. After coordinate transformation, the max-uid and max-lith traces emerge. The objective of the transformation is to make the response on the max-uid trace in wet zones be near zero. The hydrocarbon zones remain bright and therefore stand out. Assembling the pieces The completed models were assembled into a poster (not shown) that is tremendously valuable for understanding the eects of compaction on AVO. The details of such a large display are not visible at page size, therefore, summary displays were made. Figure 4 reduces the clouds of points on the crossplots to their centroids: a shale-over-wet-sand point (blue), a shale-over-oil-sand point (green), and a shale-over-gas-sand

INTERPRETERS CORNER

Figure 4. Grid of AVO clouds. This is a clock hands display. Within each cell, intercept is plotted on the horizontal axis, and gradient on the vertical. The cases are shale-over-gas-sand (red), shale-over-wet-sand (blue), and shale-over-oil-sand (green). Th e e ect of uid pressure gradient (independent of depth) on AVO is visible by comparing columns within a row.

point (red). The clockhands display illustrates how AVO changes as a function of depth and uid pressure gradient. A grid cell is empty if the rock properties are unknown at the beginning or Table 3. AVO e ects of pressure and depth. ending depth. Figure 5a shows the ratio of the shale-over-gas-sand amplitude to the shale-over-wet- three points. Locally, for the angular dierence between wet sand amplitude (i.e., the ratio of the length of the red clock and gas vectors, a 3000-ft change in depth is oset by a 0.10hand in Figure 4 to the length of the blue hand). It demon- psi/ft (1.9 ppg) dierence in uid pressure gradient. In some strates that rules of thumb, such as, I never drill a gas prospect portions of Figure 5b, the eect of uid pressure gradient is unless the amplitude is ve times the background, should not even more severe, and in some portions it is less, but in any be used unless calibrated to depth and uid pressure gradient. case it is enough to warn us not to ignore pressure when inFigure 5b shows how the angular dierence (i.e., the dif- terpreting AVO. ference between the gas angle and the wet angle in Figure 4) varies according to depth and uid pressure gradient. Below Conclusions 5000 ft, this is a reliable gas indicator; above 5000 ft, the am- The dicult and nonintuitive questions surrounding how changes in uid pressure gradient alter the AVO plitude ratio (Figure 5a) is preferred. No single AVO indicator response of reservoir rocks in clastic basins were illumiis best everywherethe entire pattern of AVO is important. nated by separately analyzing ve dierent pressure reTable 3 shows data extracted from three key points in Figgimes. ure 5b. Interestingly, these angular dierences were about the same, causing the 30 contour in Figure 5b to pass near all Within each pressure range, velocity and density of wet
October 2008 The Leading Edge 1259

INTERPRETERS CORNER

Figure 5. (a) Grid of amplitude ratios. In Figure 4, the length of the gas sand vector (red) is typically greater than that of the wet sand vector (blue). The ratio of these lengths is a better gas indicator in softer rocks (top right) than in harder rocks (bottom left). (b) Angular di erences versus depth and pressure. In Figure 4, the angle of the gas sand vector (red) is typically di erent from that of the wet sand vector (blue). The angular di erence is a gas indicator that works except around a depth of 5000 ft, where the sign of the di erence is changing.
1260 The Leading Edge October 2008

INTERPRETERS CORNER
sand and of shale may be estimated from multiwell empirical trend results of puried log data. A uid pressure gradient increase of 0.1 psi/ft (1.9 ppg) had an eect on AVO angular dierence similar to that of a reduction in depth of 3000 ft (more or less at different depths and pressures). Therefore, it is perilous to interpret AVO without considering the eect of uid pressure gradient. Rules of thumb, such as, I never drill a gas prospect unless the amplitude is ve times the background, should not be used unless calibrated to depth and uid pressure gradient. Discriminating gas sands from wet sands is easier at shallower depths and higher pressures. The lower the eective stress, the higher the water content, and the more distinct the AVO signature. No single AVO attribute (such as angular dierence or amplitude ratio) distinguished hydrocarbons from brine at all depths and pressures. Even over a signicant (270 100 miles) portion of the Louisiana shelf, lateral changes in rock properties were small compared to changes caused by lithology and uid pressure gradient. Prospect-level AVO studies, especially comparison to analogous situations may be locally superior to the method shown here because they include locally important variables such as atypical shales (with varying calcite content), atypical sands (varying cementation, etc), variations in uid content (low saturation gas, dry versus wet gas, dierent types of oils, etc), atypical shear velocity situations, anisotropy, and so on. However, the method used here (regional analysis of typical sands and shales) allows interpreters to develop better physical intuition about how AVO generally depends on depth and uid pressure gradient over a wide area of study.
Suggested reading. Excellent synthetic seismograms through the use of edited logs: Lake Borgne Area, Louisiana, US by Box (TLE, 2004). AVO inversion: Isolating rock property contrasts by Kelly et al. (TLE, 2001). Nonbright-spot AVO: Two examples by Ross and Kinman (Geophysics, 1995). Pressure and porosity inuences on VP/VS ratio of unconsolidated sands by Zimmer et al. (TLE, 2002). Acknowledgments: The authors thank everyone who helped us complete this work: Denny Loren for pioneering the concept of log trend curves; Rosa Ethridge and Paul Lowrey at Loren and Associates for doing the well log purication; Rob Mayer at CGGVeritas for assistance with using the Hampson-Russell software; Lee Ethetton for assistance with porting data between dierent hardware platforms and software systems; Laura Kay Ethetton for assembling the log data base; Pete Harth and Jeremy Greene for proofreading and advice on making the gures logical; and Tony Curtis for making the gures clear.

Corresponding author: RBox@huntpetroleum.com

1262

The Leading Edge

October 2008

You might also like