Attribute: Seismic Attribute Definition, DHI Analysis, and Reservoir Characterization
A seismic attribute is any measurable property derived from seismic data that provides information about subsurface rock properties, fluid content, or geological structure beyond what is visible in the conventional reflectivity amplitude display of a processed seismic section or volume. The concept of seismic attributes emerged in the 1970s with the commercial availability of digital seismic data and computing power sufficient to calculate instantaneous phase, frequency, and amplitude envelope (the complex trace attributes introduced by Taner and Sheriff in 1977), and has since expanded to encompass hundreds of distinct attribute types organized into two broad classes. Instantaneous or single-trace attributes are computed sample-by-sample from the complex analytic signal at each time sample on each individual trace, without reference to neighboring traces; they include the amplitude envelope (also called the reflection strength, the magnitude of the complex trace), the instantaneous phase (the angular position of the wave cycle at each sample, sensitive to reflector continuity regardless of amplitude), and the instantaneous frequency (the time derivative of instantaneous phase, sensitive to bed thinning and fluid effects). Multi-trace attributes are computed from spatial relationships among groups of neighboring traces, capturing lateral variations in reflector geometry and amplitude character that are invisible on single-trace displays; they include coherence and similarity (measuring trace-to-trace waveform correlation, highlighting faults, channels, and discontinuities by their low-coherence signatures), curvature (the second derivative of the reflector surface, a proxy for natural fracture intensity and structural style), and amplitude-variation-with-offset (AVO) attributes such as intercept and gradient that exploit the angle-dependent reflectivity to distinguish gas-saturated from brine-saturated and lithology-contrast reflections. A third broad class, model-based or inversion-derived attributes, transforms the seismic data into rock physics parameters such as acoustic impedance, Vp/Vs ratio, and density through model-based seismic inversion workflows that constrain the solution with well control. Seismic attributes are the primary analytical currency of reservoir characterization: they bridge the gap between the geometrical structural interpretation of a seismic section and the quantitative petrophysical characterization required to build a reservoir model that drives volumetric estimates, well placement decisions, and field development plans.
Key Takeaways
- Instantaneous attributes: amplitude envelope, phase, and frequency from the complex trace: Instantaneous attributes are computed from the analytic signal A(t) = x(t) + i times H(t), where x(t) is the real seismic trace and H(t) is its Hilbert transform. The amplitude envelope (reflection strength) is the magnitude of A(t), representing the total energy of the seismic signal at each sample regardless of polarity and is sensitive to acoustic impedance contrasts, gas saturation, and lithological changes. The instantaneous phase is the arctangent of H(t) divided by x(t) and records the wave cycle position at each sample independent of amplitude; it is particularly useful for tracking lateral reflector continuity across amplitude anomalies and for identifying stratigraphic terminations where amplitude fades. The instantaneous frequency is the time derivative of instantaneous phase and is sensitive to bed thinning (frequency decreases toward tuning thickness), gas saturation (low-frequency shadow beneath gas columns from high intrinsic attenuation), and processing artifacts. Instantaneous attributes are computed rapidly and cheaply on any seismic dataset, but they are sensitive to noise and cycle-skipping artifacts that require careful data conditioning (signal-to-noise enhancement, zero-phase wavelet extraction) before reliable geological interpretation.
- Horizon-based amplitude attributes: RMS, max, average, and the DHI workflow: Horizon-based attributes are extracted from a seismic volume within a time window defined by an interpreted reflection horizon plus or minus a specified number of milliseconds (the extraction window). Root-mean-square (RMS) amplitude, maximum amplitude, average absolute amplitude, and sum of positive (or negative) amplitudes are the standard choices for a flat-spot or bright-spot direct hydrocarbon indicator (DHI) screening workflow: gas-saturated sands in Class III AVO settings (sub-critical incidence, low-impedance reservoir) generate bright-spot amplitude anomalies on the near-angle stack and crossover or polarity reversal at the gas-oil or gas-water contact on the full-offset stack. A well-designed DHI analysis workflow extracts RMS amplitude from the near-offset stack (approximately 0 to 15 degrees incidence) and far-offset stack (approximately 25 to 45 degrees) separately, computes the intercept (near-stack amplitude at zero offset) and gradient (change of amplitude with offset) from AVO crossplot analysis, and co-renders the amplitude map with the AVO gradient map to separate gas sand anomalies (bright near, negative gradient) from lithology contrasts (bright near, near-zero or positive gradient). In the Montney and Cardium plays of the WCSB, horizon-based RMS amplitude maps are routinely used to identify the highest-amplitude sweetspots for targeting horizontal well pads within the broader pool license area.
- Coherence and curvature: multi-trace geometric attributes for fault and fracture mapping: Coherence (also called similarity, semblance, or continuity depending on the specific computation algorithm) measures the degree of similarity between neighboring traces in a moving spatial window, producing a volume where reflectors show high coherence (near 1.0) and faults, channels, and fracture corridors show low coherence (near 0) from the lateral discontinuity of the reflection. Time slices and horizon slices through coherence volumes reveal the plan-view geometry of structural and stratigraphic features at a resolution determined by the seismic dominant wavelength, typically 20 to 50 metres in the WCSB at Montney target depths. Curvature is the second spatial derivative of the reflector surface, measuring how much the reflector deviates from planarity at each grid point; positive curvature identifies anticlinal (convex upward) bends and negative curvature identifies synclinal (concave upward) bends. Most-positive curvature and most-negative curvature are the components used in fracture characterization: in tectonic settings where natural fractures formed in response to bending stresses during folding, the highest curvature values (both positive and negative) correlate spatially with the highest natural fracture intensity, providing an indirect map of fracture density that guides horizontal well azimuth selection (wells drilled oblique to fracture strike maximize fracture intersection per unit length) and hydraulic fracture stage spacing design in naturally fractured Duvernay and Nisku carbonate reservoirs.
- AVO attributes: intercept, gradient, and fluid factor for DHI classification: Amplitude variation with offset (AVO) attributes exploit the angle-dependent reflection of seismic waves at lithological interfaces (governed by the Zoeppritz equations, simplified by the Shuey approximation for practical use) to extract information about the Vp/Vs ratio and density contrast across the reflector, which in turn discriminate fluid type (gas versus brine) and lithology (sand versus shale). The intercept (A in the Shuey two-term linearization R(theta) = A + B times sin2(theta)) is approximately equal to the zero-offset P-wave reflectivity and represents the acoustic impedance contrast. The gradient (B) is proportional to the contrast in Vp/Vs ratio and is the critical AVO discriminator: gas-saturated low-impedance sands generate large negative gradient (Class III AVO), while brine-saturated or tight sands generate near-zero or positive gradient. The fluid factor (F = delta Vp/Vp minus (Vp/Vs)squared times delta Vs/Vs) derived from AVO intercept and gradient is sensitive to departures from the brine-saturated mudrock trend in the A-B crossplot space, making it a direct hydrocarbon indicator. AVO analysis requires true-amplitude-preserved, offset-binned prestack seismic gathers, free from AGC gain, which destroys amplitude variation information, and free from residual multiple contamination, which produces spurious AVO gradients.
- Spectral decomposition and acoustic impedance inversion as reservoir characterization tools: Spectral decomposition (frequency decomposition) computes the seismic response of the subsurface at a series of discrete frequencies within the seismic bandwidth, producing a frequency-decomposed volume or set of horizon slices at each frequency. Because tuning thickness (the minimum resolvable bed thickness, approximately one-quarter wavelength) varies inversely with frequency, different frequencies resolve different bed thicknesses: a 20 Hz frequency component reveals the response of beds 30 to 50 metres thick, while a 60 Hz component reveals beds 10 to 15 metres thick at Montney velocities. Co-rendering three frequency slices as red-green-blue (RGB) color blends produces vibrant maps that highlight stratigraphic architecture (channel bodies, lobe geometries, pinchout trends) at resolution approaching 5 to 10 metres. Acoustic impedance (AI) inversion transforms the seismic reflection coefficient series into a rock property volume calibrated to well-log acoustic impedance data, enabling the seismic interpreter to read relative porosity and lithology changes directly from the inverted volume rather than from reflectivity. AI inversion workflows range from bandlimited inversion (preserving only the seismic bandwidth from approximately 8 to 100 Hz) to model-based inversion (which adds low-frequency AI from well logs and trends to extend the bandwidth to near-DC) and simultaneous inversion for P-impedance, S-impedance, and density using prestack gather data.
Attribute Extraction, Quality Control, and Integration Workflows
Effective attribute analysis requires careful attention to seismic data conditioning before any attribute is computed. The quality of every attribute is bounded by the quality of the underlying seismic data: random noise amplifies in coherence algorithms as spurious low-coherence features; residual multiples create false AVO gradients; AGC processing destroys amplitude attributes; and phase errors from wavelet non-stationarity distort instantaneous phase and frequency attributes. Data conditioning steps applied before attribute extraction typically include: noise attenuation (structure-oriented filtering or dip-steered median filtering to enhance signal-to-noise ratio without degrading reflector continuity); multiple attenuation (SRME and Radon demultiple as described under the Attenuate entry); bandwidth enhancement (Q-compensation and spectral whitening to broaden usable frequency content); and phase correction (deterministic or statistical zero-phasing to ensure a consistent symmetric wavelet throughout the volume). Only after these conditioning steps can the geoscientist interpret attribute anomalies with confidence that they reflect geological signal rather than data artifacts.
The attribute extraction workflow for DHI screening follows a logical sequence from structural to stratigraphic to fluid indicators. The interpreter begins by generating a coherence volume to identify all major faults and stratigraphic discontinuities (channels, reef edges, slope failures) that segment the reservoir and might control hydrocarbon trapping or production compartmentalization. The structural map is then used to define extraction windows for amplitude attributes at the reservoir level, ensuring the window tracks the formation of interest rather than time-slicing obliquely through multiple formations. AVO intercept and gradient are extracted from pre-stack gather data and crossplotted to identify anomalies deviating from the shale compaction trend, which are candidates for DHI classification. Fluid factor and Vp/Vs maps derived from prestack inversion are then overlain to confirm or refine the DHI anomaly outline. This multi-attribute integration approach reduces the probability of false-positive DHI classification (mistaking a lithology or tuning effect for a gas anomaly) and is the workflow mandated in most major oil company exploration technical standards, including those at Equinor, Shell, and Chevron, before a well is approved for drilling on an amplitude-based prospect.