zero-offset data, Oil & Gas Glossary

Zero-offset data is seismic data in which the energy source and the receiver occupy the same surface location, so the offset, meaning the horizontal separation between shot point and geophone, is zero. In this idealized geometry a seismic wave travels straight down to a reflector and straight back up along the same vertical raypath, which makes the recorded two-way traveltime depend only on depth and velocity rather than on any lateral travel component. True zero-offset acquisition is almost never practical in the field because a physical source and a physical receiver cannot occupy the exact same point at the same instant, and a buried charge or vibroseis sweep placed directly under a geophone would saturate or damage the sensor. For that reason the petroleum industry produces the equivalent of zero-offset data synthetically: traces shot with separated sources and receivers are sorted into common-midpoint (CMP) gathers, corrected for normal moveout (NMO) so that the extra traveltime caused by offset is removed, and then summed, or stacked, to yield a single composite trace that behaves as though the source and receiver had been coincident above the midpoint. The result, a stacked section, simulates a zero-offset section while delivering a far better signal-to-noise ratio than any single live recording, because random noise tends to cancel when many corrected traces are added while coherent reflection energy reinforces. Zero-offset is also the geometry that post-stack time and depth migration algorithms assume as their input: the exploding-reflector model treats every reflector as if it detonated at half the medium velocity and the wavefield was recorded at the surface with zero offset, which is why a properly NMO-corrected and stacked volume is the standard starting point for structural interpretation. In Western Canadian Sedimentary Basin (WCSB) exploration over the Montney, Duvernay, and Cardium plays, multi-channel 3D surveys recording thousands of offsets are routinely processed to a zero-offset-equivalent stack before any horizon picking, fault mapping, or amplitude-versus-offset (AVO) decomposition begins. The accuracy of the zero-offset simulation hinges entirely on the velocity field used for NMO: too slow a stacking velocity over-corrects far offsets, too fast under-corrects them, and either error smears the stacked reflection and degrades the zero-offset approximation. Depths derived from a zero-offset stack are expressed in two-way traveltime measured in milliseconds, then converted to metres or feet using an interval-velocity model, so a Duvernay reflector at roughly 3,400 m (about 11,150 ft) might appear near 2,100 ms two-way time depending on the overburden velocity.

Key Takeaways

Coincident Source and Receiver: Zero-offset means the shot point and the geophone share one surface position, so the seismic raypath is vertical and the recorded two-way traveltime reflects only depth and overburden velocity. Because a real source would damage a co-located receiver, this geometry is essentially never acquired directly; it is reconstructed in processing from offset data.
CMP Stacking Builds the Equivalent: Traces sharing a common midpoint but recorded at many source-receiver offsets are gathered, NMO-corrected, and summed. The stacked trace simulates a coincident-source-receiver recording and typically raises signal-to-noise by the square root of the fold; a 60-fold WCSB 3D survey can improve coherent-to-random ratio by roughly eight times.
Normal Moveout Is the Critical Step: NMO removes the extra hyperbolic traveltime that offset adds, flattening reflection events across the gather. The correction depends on a stacking-velocity field, and a velocity error of even a few percent over-corrects or under-corrects the far offsets, smearing the zero-offset approximation and distorting later amplitude analysis.
Foundation for Migration: Post-stack time and depth migration assume zero-offset input under the exploding-reflector model, where reflectors are treated as sources firing at half the true velocity. A clean zero-offset-equivalent stack is therefore the standard input for structural imaging of Montney and Duvernay horizons before fault interpretation.
Time, Not Depth, Is Recorded: A zero-offset stack is measured in two-way traveltime in milliseconds. Converting to a metre or foot depth requires an interval-velocity model; a reflector near 2,100 ms two-way time can correspond to roughly 3,400 m (11,150 ft) in a typical WCSB overburden, with the exact figure set by the velocity profile.

Why True Zero-Offset Acquisition Is Avoided

Placing a high-energy source directly beneath a geophone would clip the sensor and bury the early reflection under source-generated noise, ground roll, and the direct arrival. Single-fold zero-offset traces would also carry no redundancy, so any near-surface static or random noise would pass straight into the section. Instead, surveys deliberately spread receivers across long offset ranges, often 3,000 m or more in WCSB 3D programs, then collapse that redundancy back to a zero-offset equivalent during stacking. This design captures the moveout information needed to estimate velocities and AVO response while still producing the clean coincident-geometry image that interpreters want for mapping structure.

Stacking Velocity and the Quality of the Approximation

The fidelity of a zero-offset stack is governed by the stacking-velocity model applied during NMO. Velocities are picked from semblance analyses on selected CMP gathers, then interpolated across the survey. Where overburden is laterally variable, such as channelized Mannville sands above a Duvernay target, a single velocity function flattens near offsets but leaves far offsets mis-corrected, smearing the stacked reflection. Processors mute the most distorted far offsets and iterate velocity picks to sharpen the event. A well-tuned velocity field can hold reflection timing errors to a few milliseconds, preserving the zero-offset character that downstream migration and inversion depend on.

Fast Facts

The common-midpoint method that makes zero-offset stacks possible was patented by Harry Mayne of Petty Geophysical in 1956, and the resulting signal-to-noise gains were so dramatic that CMP stacking became the dominant land seismic technique within a decade. Modern WCSB 3D surveys routinely record fold counts above 100, meaning more than 100 individual offset traces are summed into each zero-offset-equivalent output trace, which is why faint Duvernay and Montney reflections buried in field noise emerge as continuous, mappable events after processing.

Zero-offset data is one node in a connected chain of seismic concepts. Normal Moveout is the correction that removes offset-dependent traveltime and is the prerequisite for building a zero-offset stack. Common Midpoint describes the gather geometry whose traces are summed to simulate coincident source and receiver. Offset Vertical Seismic Profile sits at the opposite end of the geometry spectrum, deliberately separating source and downhole receiver to image away from the wellbore, and Seismic Migration consumes the zero-offset stack to reposition dipping reflectors to their true subsurface locations.

Real-World WCSB Scenario: Duvernay 3D Over Kaybob

An operator shooting a 3D survey over a Duvernay light-oil block near Kaybob, Alberta, acquires roughly 110-fold data with offsets reaching 3,200 m to capture moveout for both velocity estimation and AVO. The raw field records show the target reflector near 2,050 ms two-way time, corresponding to about 3,350 m (10,990 ft) of depth, but it is invisible on any single offset trace because of ground roll and source noise. The processing crew picks stacking velocities on semblance panels every 500 m, applies NMO, mutes far offsets distorted by the shallow Mannville channel system, and stacks to a zero-offset-equivalent volume at a cost on the order of 1,200 CAD per square kilometre for processing alone.

The resulting zero-offset stack resolves the Duvernay as a continuous, mappable trough, letting the interpreter pick the horizon and adjacent faults with confidence before passing the volume to pre-stack depth migration. Because the velocity field was iterated rather than picked once, residual timing error stayed under three milliseconds, and the well subsequently drilled landed within 4 m of the predicted reservoir top, validating the zero-offset image used to plan it.