Average Velocity: Definition, Depth Conversion, and Seismic Tie

Average velocity (symbol: Vavg) is the bulk seismic propagation velocity calculated by dividing twice the depth to a reflector by the two-way travel time (TWT) recorded at the surface. In mathematical form:

Vavg = 2D / TWT

where D is the depth in metres or feet and TWT is measured in seconds. Average velocity is the foundational parameter for converting a seismic section from the time domain, in which reflectors appear as horizontal bands at measured two-way travel times, into the depth domain, in which those reflectors are placed at their actual subsurface positions in metres or feet. Because nearly all drilling decisions, casing point selections, and reserves calculations are made in depth, the accuracy of average velocity estimates has direct operational and financial consequences. It differs from interval velocity, RMS (root-mean-square) velocity, and instantaneous velocity, each of which describes wave propagation differently and is measured or derived by different methods.

Key Takeaways

• Average velocity is defined as Vavg = 2D / TWT, the simplest form of time-to-depth conversion used in seismic interpretation.
• It represents the harmonic average of all interval velocities along the travel path and is always less than or equal to the RMS velocity for the same depth.
• A 1% error in average velocity propagates directly to a 1% error in converted depth, which at 4,000 metres (13,120 feet) equals 40 metres (131 feet) of structural uncertainty, enough to misplace a casing shoe or a gas-water contact.
• The well-seismic tie, which uses a sonic log to build a synthetic seismogram and then compares it to the nearby seismic trace, is the primary quality-control step for verifying average velocity accuracy in drilled areas.
• Depth conversion from acquisition-quality seismic requires a velocity model that integrates seismic velocity analysis, well ties, and often vertical seismic profile (VSP) data.

Velocity Types and Their Relationships

Average velocity, interval velocity, RMS velocity, and instantaneous velocity are related but distinct concepts, and confusion between them is a common source of errors in depth conversion. Interval velocity (Vint) is the velocity within a specific rock layer, defined as the layer thickness divided by the one-way travel time through that layer. Interval velocity reflects the mechanical properties of a specific formation, specifically its bulk modulus, shear modulus, and density, and is the physically meaningful quantity for rock-physics analysis. It varies continuously with depth because different lithologies have different elastic properties, and it can change sharply across unconformities or fluid contacts.

RMS velocity (Vrms) is the root-mean-square of all interval velocities from the surface to a given depth, weighted by the one-way travel time through each layer. Because squaring the velocities before averaging gives greater weight to high-velocity layers, RMS velocity is always greater than or equal to average velocity for any realistic Earth model in which velocity increases with depth. RMS velocity is what seismic velocity analysis actually estimates from the normal moveout (NMO) correction applied during processing; it is therefore called stacking velocity or NMO velocity in processing terminology, though stacking velocity incorporates additional effects from dipping reflectors and lateral velocity heterogeneity that make it an approximation of true RMS velocity.

Instantaneous velocity V(z) is the velocity at a specific point in the subsurface as a function of depth z. It is measured directly by the acoustic log (sonic log), which records the interval transit time (DT, in microseconds per foot or microseconds per metre) over a 0.6-metre (2-foot) sampling interval. The relationship is V(z) = 1,000,000 / DT (for DT in microseconds per metre and V in metres per second). Integrating the instantaneous velocity log from the surface to a given depth yields the one-way travel time, which is the basis for building the synthetic seismogram used in the wireline log-to-seismic well tie. The array sonic logging tool provides both P-wave and S-wave interval velocities and is the standard source of instantaneous velocity data in modern well-log suites.

How Average Velocity Is Derived From Seismic Data

In an exploration or appraisal setting without nearby wells, average velocity is derived from seismic data through a sequence of processing and analysis steps. During data acquisition, each common midpoint (CMP) gather contains traces recorded at multiple source-receiver offsets. A reflection from a horizontal interface at depth D arrives at the zero-offset position at time T0 = 2D / Vrms and at a far-offset position at a later time that follows the NMO hyperbola. Velocity analysis is the process of scanning over a range of trial velocities and selecting the velocity that best flattens the NMO hyperbola across the offset range, maximizing coherence (semblance) in the corrected gather. The velocity that maximizes semblance is the stacking velocity, which approximates the RMS velocity at that two-way time.

Converting stacking velocities to interval velocities uses the Dix equation:

Vint² = (Vrms2² x t2 - Vrms1² x t1) / (t2 - t1)

where Vrms1 and Vrms2 are the RMS velocities at times t1 and t2 bounding the interval of interest, and Vint is the interval velocity of that layer. The Dix equation assumes horizontal, isotropic, laterally homogeneous layers, conditions that are often not met in complex structures or in areas with significant velocity anisotropy. Errors in the stacking velocity picks propagate strongly through the Dix inversion because the equation involves differencing terms that can be of similar magnitude, amplifying pick uncertainty into large interval velocity errors especially for thin layers. Once interval velocities are obtained, average velocity to any depth is calculated as the harmonic mean travel-time-weighted combination of the interval velocities above that depth:

1 / Vavg = (1/D) x SUM[ (hi / Vi) ]

where hi is the thickness of layer i and Vi is its interval velocity. This is equivalent to the original definition Vavg = 2D / TWT when TWT is computed by summing the two-way travel times through all layers above depth D.

Full-waveform inversion (FWI) is an advanced processing technique that iteratively updates a subsurface velocity model to match the full waveform of recorded seismic data rather than just the arrival times used in conventional velocity analysis. FWI produces high-resolution velocity models that are more accurate for depth conversion than semblance-based stacking velocity analysis, particularly in geologically complex areas such as sub-salt or sub-basalt settings. The technique is computationally intensive but has become a standard part of deepwater exploration processing workflows, particularly in the Gulf of Mexico and offshore West Africa.

Time-to-Depth Conversion in Practice

Time-to-depth conversion is one of the highest-stakes computations in petroleum exploration because it determines where the drill bit targets the reservoir. The simplest approach, applying a single average velocity to the entire depth range, is rarely accurate enough for well planning because velocity varies laterally and vertically in ways that a single number cannot capture. Instead, industrial-grade depth conversion uses a layer-cake model in which each seismically mapped horizon is assigned its own average velocity derived from the combination of surface-seismic velocity analysis and well calibration. The resulting depth maps honor the well penetrations exactly and extrapolate between wells using the velocity model derived from seismic data.

In high-pressure high-temperature (HPHT) environments, depth accuracy is a safety as well as an economic concern. The casing shoe must be set at a depth that provides adequate formation integrity to withstand the well kicks anticipated in the next open-hole interval. An error of 30-50 metres (100-165 feet) in the depth of a formation with a narrow fracture gradient window can mean the difference between a safe shoe seat and a shallow-water flow or well-control incident. Regulators including the Bureau of Safety and Environmental Enforcement (BSEE) in the US Gulf of Mexico and the Health and Safety Executive (HSE) in the UK North Sea require that depth prognoses be accompanied by documented uncertainty ranges that reflect the accuracy of the velocity model used.

Depth uncertainty from velocity errors follows a simple proportional relationship: a 1% error in average velocity produces a 1% error in converted depth. At a target depth of 3,000 metres (9,840 feet), a 2% velocity uncertainty yields a depth uncertainty of 60 metres (197 feet). At 6,000 metres (19,685 feet), the same 2% uncertainty yields 120 metres (394 feet). These depth uncertainties directly translate to uncertainty in gross rock volume, reserves estimates, and the confidence interval placed around the gas-water or oil-water contact prognosis used to plan perforation intervals. Geostatistical depth conversion methods, which use stochastic realizations of the velocity field to propagate velocity uncertainty into depth-map uncertainty, are standard practice in major exploration companies for material discoveries requiring resource certification.

Well-Seismic Tie and Velocity Calibration

The well-seismic tie is the process of comparing a synthetic seismogram constructed from well-log data to the actual seismic trace recorded at the well location. It is the primary method for verifying that the velocity model used in depth conversion accurately represents the subsurface. The synthetic seismogram is built by integrating the acoustic log interval transit time DT to obtain one-way travel time as a function of depth, then pairing this time-depth relationship with the density log to compute acoustic impedance (AI = Vp x density) as a function of two-way time. Reflectivity coefficients are computed at each impedance contrast, and the reflectivity series is convolved with the seismic wavelet to produce the synthetic trace.

When the synthetic seismogram matches the polarity, timing, and relative amplitude of reflections on the seismic section, the well tie is considered good and the velocity model is validated. A poor well tie, in which synthetic reflections are systematically shifted in time relative to their seismic counterparts, indicates a bias in the average velocity model. The time shift divided by the two-way time at which it occurs gives a fractional velocity error. In practice, well ties must correct for depth of the first log measurement (the sonic log rarely starts at the surface, and the interval from the surface to the first log point must be estimated from drilling records or check-shot data), cycle skipping and borehole effects in the sonic log, and the variable depth of the seismic datum.

Vertical seismic profile (VSP) surveys provide the most direct measurement of average velocity for well-seismic tie purposes. A downhole receiver is placed at known depths in the wellbore while a surface source fires, and the first-break arrival time at each receiver depth is directly used to compute the average velocity to that depth. The VSP average velocity integrates the interval velocities seen by the seismic wave traveling from the surface to the receiver and avoids many of the borehole and logging artifacts that affect the sonic-derived time-depth function. VSP surveys are particularly valuable in complex lithologies (anhydrites, coals, over-pressured intervals) where the sonic log may be unreliable, and in wells where the well tie reveals a significant discrepancy between the log-derived and seismic-derived velocity profiles.