AVO: Definition, Classes, and Direct Hydrocarbon Indicators

AVO, short for Amplitude Variation with Offset, is one of the most powerful seismic analysis techniques available to exploration geoscientists. It describes the systematic change in the amplitude of a seismic reflection as the angle of incidence increases from near-offset to far-offset traces. Where a conventional seismic stack compresses all offsets into a single average, AVO analysis preserves and interrogates that offset-dependent amplitude behavior to extract rock-physics information about the subsurface that a stacked section simply cannot reveal. When hydrocarbons fill the pore space of a reservoir rock, they change the elastic properties of that rock in measurable ways: compressional-wave velocity (Vp) drops, shear-wave velocity (Vs) may stay approximately constant or rise slightly relative to Vp, and bulk density decreases. These contrasts in elastic properties across a reflector translate directly into an offset-dependent reflection response. Properly interpreted, that response becomes a direct hydrocarbon indicator (DHI), guiding drill decisions worth hundreds of millions of dollars.

Key Takeaways

• AVO measures how seismic reflection amplitude changes with source-receiver offset (or angle of incidence), exploiting the sensitivity of elastic contrasts to pore-fluid type.
• The Zoeppritz equations govern the exact partitioning of energy at a reflecting interface; the Shuey two-term approximation simplifies this to an intercept (R0) and a gradient (G), which are the workhorses of practical AVO analysis.
• Four standard AVO classes (I through IV, plus IIb) each define a different intercept-gradient relationship and correspond to different geologic settings and reservoir impedance contrasts.
• Attributes derived from AVO, including fluid factor, Lambda-rho (LR) and Mu-rho (MR), and scaled Poisson's ratio change, help discriminate gas-saturated rock from brine-saturated rock and from lithology-related anomalies.
• AVO analysis has its limits: the fizz-water problem, seismic noise, overburden anisotropy, and poorly consolidated thin-bed tuning can all generate false positives or suppress real anomalies. Integration with wireline logs and reservoir characterization models is essential.

How AVO Works: The Zoeppritz Foundation

When a compressional (P-wave) seismic wavelet strikes a boundary between two elastic half-spaces, energy is partitioned into four wave types: a reflected P-wave, a transmitted P-wave, a reflected converted S-wave, and a transmitted converted S-wave. The exact amplitudes of each depend on the angle of incidence and on the elastic properties of both layers: Vp, Vs, and bulk density (rho). The Zoeppritz equations (1919) describe this partitioning exactly. In practice, the full matrix solution is computationally unwieldy for interpretation workflows, so geophysicists rely on linearized approximations. The most widely used is the Shuey (1985) two-term approximation, which expresses the P-to-P reflection coefficient R as a function of incidence angle theta:

R(theta) = R0 + G sin^2(theta)

Here R0 is the zero-offset (normal-incidence) reflection coefficient, often called the intercept, and G is the gradient, which controls how rapidly the amplitude changes with angle. R0 is primarily sensitive to acoustic impedance contrast (the product of Vp and density), while G is sensitive to the change in Poisson's ratio across the boundary. Because Poisson's ratio depends on the Vp/Vs ratio, and that ratio is highly sensitive to pore-fluid content (gas lowers Vp strongly, barely affecting Vs), the gradient G carries the fluid-discrimination signal. A large negative gradient on a sand-shale interface signals a drop in Poisson's ratio from shale to sand, which is diagnostic of gas saturation. For higher-angle analysis or stronger impedance contrasts, the three-term Shuey or Aki-Richards approximations add a curvature term (C) to account for density and far-angle behavior, though beyond about 45 degrees the approximations themselves begin to break down and the full Zoeppritz solution must be used.

In practice, seismic gathers are sorted into angle ranges (typically near: 0-15 degrees, mid: 15-30 degrees, far: 30-45 degrees, and sometimes ultra-far beyond 45 degrees). The intercept and gradient volumes are extracted by fitting a least-squares line to the amplitude-versus-angle curve at every sample in every gather. These volumes are then cross-plotted and spatially interpreted to identify anomalous combinations of R0 and G that depart from the background wet-sand and shale trend, ideally clustering in the regions of the crossplot expected for gas sands.

AVO Classes: A Practical Taxonomy

Rutherford and Williams (1989) introduced the classification of AVO anomalies into three classes based on the sign of the intercept R0 and the behavior of amplitude with offset. A fourth class and a variant (IIb) were added later. Understanding which class applies to a given play is critical because the seismic signature looks completely different across classes, and confusing them is a common source of false negatives and false positives.

Class I describes high-impedance gas sands where the sand velocity is higher than the encasing shale velocity. R0 is positive (a peak on zero-phase data), and the amplitude dims with increasing offset because the gradient G is negative and large enough to reduce the positive intercept toward zero or even reverse it at far offsets. Class I anomalies are common in deeply buried, well-cemented reservoirs such as the North Sea Paleocene Forties sandstones or tight Cretaceous sands of the Western Canada Sedimentary Basin. The amplitude brightening on the stack typically disappears when the sand is wet, making the dim-out itself a DHI. Class II sands have near-zero impedance contrast with the surrounding shale. R0 is close to zero, meaning the stack amplitude is very weak, but G is strongly negative. This creates an AVO crossplot anomaly that is invisible on the stack yet clearly visible in gradient volumes or on gradient-enhanced difference displays. Class IIb sands flip the polarity of the reflection from near to far offset, an unambiguous but easily missed DHI. Class II plays are common in Tertiary deltaic sequences of the Niger Delta, offshore West Africa. Class III is the most familiar "bright spot" play: the gas sand has lower impedance than the shale, so R0 is negative and the amplitude increases (becomes more negative, or "brightens") with offset. This is the classic deepwater Gulf of Mexico Miocene channel sand signature. Class III anomalies are the easiest to see on a stack and the most thoroughly documented in exploration history. Class IV sands also have negative R0 (low impedance), but unlike Class III the amplitude decreases with offset because G is positive. This apparently counterintuitive behavior arises when the shear modulus of the overlying shale is unusually high relative to the gas sand. Class IV is less common but is documented in some overpressured shelf plays.

Rock Physics Crossplots and Fluid Discrimination

The most powerful diagnostic tool in AVO analysis is the Ip-Is (P-impedance versus S-impedance) crossplot, constructed from either well logs or AVO-derived inversion volumes. On an Ip-Is plot, brine-saturated sands and shales follow a predictable "mudrock line" (Castagna et al., 1985), described approximately as Vp = 1.16 Vs + 1360 m/s for water-saturated clastic rocks. Gas sands deviate systematically to the left (lower Vp, lower Ip) while shear properties remain relatively unchanged, creating a distinctive separation from the brine trend. This separation is the theoretical basis for the fluid factor attribute (deltaF = deltaVp - (Vp/Vs_ref) * deltaVs), which is designed to be zero for any rock on the mudrock line and negative for gas-saturated rock. In practice, calibration of the reference Vp/Vs ratio is critical and should be done using well-log data from the project area before applying fluid factor maps to the seismic volume.

Lambda-rho (LR) and Mu-rho (MR), introduced by Goodway et al. (1997), decompose the elastic properties into bulk-modulus proxy (lambda * rho) and shear-modulus proxy (mu * rho). The key insight is that lambda, the first Lame parameter, is strongly sensitive to pore-fluid compressibility, while mu is insensitive to fluids. A gas-saturated sand shows dramatically reduced lambda relative to a brine sand at the same depth and porosity. Crossplotting LR against MR typically produces tight, well-separated clusters for shale, brine sand, and gas sand, making it one of the clearest fluid discriminators available from seismic data. Typical values for Gulf of Mexico Miocene targets: gas sands 5-15 GPa*g/cc (LR), 8-18 GPa*g/cc (MR); brine sands 20-35 GPa*g/cc (LR), 10-20 GPa*g/cc (MR); shales 20-40 GPa*g/cc (LR), 12-22 GPa*g/cc (MR). Conversely, carbonates tend to cluster at very high LR and MR values (above 40 GPa*g/cc), placing them far from sand plays on the crossplot.

AVO simultaneous inversion takes this further by inverting pre-stack gathers for Ip, Is, and density volumes simultaneously, using a model-based or basis-pursuit inversion algorithm. The resulting volumes are interpretable in terms of rock-physics properties directly and can be fed into probabilistic facies classification workflows. This is the current industry standard for deepwater and unconventional play evaluation, producing probability volumes for gas sand, brine sand, and shale at every seismic sample. Integration with vertical seismic profile (VSP) data is essential for tying the inversion to well control and validating the low-frequency model, which cannot be derived from seismic data alone.