NMO: Normal Moveout Correction, Stacking Velocity, and CMP Gather Flattening in Seismic Processing
NMO, short for normal moveout, is the difference in arrival time of a reflected seismic event recorded at different source-to-receiver offsets, and the NMO correction is the processing step that removes that offset-dependent delay so traces from a common midpoint can be summed into a single high-quality stacked trace. When a seismic source fires, the reflected energy from a flat subsurface interface reaches a near receiver sooner than a far receiver simply because the far receiver's raypath is longer, even though both reflections come from the same point on the reflector. Plotted as a function of offset, the arrival times trace out a hyperbola whose curvature is set by the depth of the reflector and the velocity of the rock above it. The NMO correction stretches the time axis of each trace to collapse that hyperbola flat, making every trace in the common midpoint gather look as if it had been recorded at zero offset, directly above the reflection point. Once the events are flattened, the traces are added together in a process called stacking, which reinforces the coherent reflected signal while averaging out random noise, dramatically improving the signal-to-noise ratio of the final section. The correction is governed by the NMO equation, in which the travel time at a given offset depends on the zero-offset two-way time, the offset distance, and a velocity term known as the stacking or NMO velocity. That velocity is not measured directly; it is estimated from the data itself through velocity analysis, in which processors test a range of trial velocities and pick the value that best flattens each event, building a velocity-versus-time function for every part of the survey. Choosing the velocity too low overcorrects the event so it bows upward at far offsets, while choosing it too high undercorrects so it still curves downward, and the correct velocity produces a flat alignment. A practical side effect is NMO stretch, a distortion that lengthens the wavelet at far offsets and shallow times, which is why processors apply a stretch mute to discard the badly distorted portions of the far traces before stacking. NMO assumes small-offset, near-hyperbolic moveout over a horizontally layered earth; dipping reflectors require an additional dip moveout correction, and strongly anisotropic or laterally varying media demand higher-order or non-hyperbolic moveout terms to flatten the gather accurately. Despite these limitations, NMO remains one of the foundational steps in the conventional seismic processing sequence, sitting between deconvolution and stack, and the stacking velocities it produces feed directly into time migration and into the interval velocity estimates used for depth conversion. For any explorationist interpreting a seismic line over a target such as the Montney or Duvernay, the quality of the NMO correction and the velocities behind it underpins how trustworthy the imaged reflectors, and the depths assigned to them, ultimately are.
Key Takeaways
- Offset-Dependent Time Delay: Normal moveout is the extra travel time a reflection accumulates as source-to-receiver offset increases, because the far-offset raypath is longer than the near-offset path to the same reflection point. Plotted against offset, the arrival times form a hyperbola whose curvature depends on reflector depth and overburden velocity, and the NMO correction removes that curvature.
- Flattening for Stacking: The correction stretches each trace's time axis so the hyperbolic event becomes flat, mimicking a zero-offset recording. Once flattened, all traces in a common midpoint gather are summed during stacking, which reinforces coherent reflected signal and suppresses random noise, raising the signal-to-noise ratio of the final seismic section substantially.
- Stacking Velocity Drives the Correction: The NMO equation relates offset travel time to zero-offset time, offset, and a velocity term. That stacking or NMO velocity is estimated from the data through velocity analysis, testing trial velocities to find the one that best flattens each event and producing a velocity-time function used downstream for migration and depth conversion.
- Over- and Undercorrection: A velocity picked too low overcorrects the event, bowing it upward at far offsets, while a velocity too high undercorrects, leaving residual downward curvature. Only the correct velocity yields a flat alignment, so velocity picking accuracy directly controls stack quality and the reliability of the velocities used for later imaging.
- NMO Stretch and Muting: The time-axis stretch distorts and lengthens the wavelet most at far offsets and shallow times, an artifact called NMO stretch. Processors apply a stretch mute to remove the badly distorted far-trace portions before stacking, trading some far-offset fold for cleaner, less smeared reflection events in the stacked output.
The NMO Equation and Velocity Analysis
The hyperbolic moveout relation states that the squared travel time at offset x equals the squared zero-offset time plus x-squared divided by the square of the NMO velocity. Because the only unknown is that velocity, processors run velocity analysis on selected gathers, computing semblance across a fan of trial velocities and picking the value that maximizes coherence for each reflector. The result is a velocity function that varies with time and location across the survey. These picks are interpretive: a noisy gather, a weak reflector, or interfering multiples can all bias the chosen velocity, which is why experienced processors cross-check semblance panels against the flattened gathers before committing the function to the stack.
Limitations: Dip, Anisotropy, and Higher-Order Moveout
Standard NMO assumes a horizontally layered earth and small offsets. Dipping reflectors break the simple common-midpoint geometry and require a dip moveout, or DMO, correction before stack. In shales such as the Duvernay, intrinsic anisotropy makes velocity depend on propagation direction, so far-offset moveout departs from a simple hyperbola and a fourth-order or anisotropic NMO term is needed to flatten the gather. Ignoring these effects leaves residual moveout that smears the stack, mis-times events, and feeds erroneous interval velocities into depth conversion, pushing structural picks tens of metres off true depth.
Fast Facts
The hyperbolic moveout idea predates digital processing: it grew out of the refraction and reflection seismograph work of the 1920s and 1930s, when crews on the US Gulf Coast first exploited travel-time differences to map salt domes. What changed with the common-midpoint method, formalized by Harry Mayne in the 1950s and patented in 1956, was the realization that many traces sharing one midpoint could be NMO-corrected and summed, turning a single noisy reflection into a robust stacked trace and multiplying the effective signal-to-noise ratio of every survey shot since.
Related Terms
NMO sits inside a chain of processing concepts. The common midpoint gather is the trace collection that NMO flattens, and stacking is the summation step it enables. The stacking velocity produced by velocity analysis is the parameter that controls the correction and later feeds depth conversion, while seismic migration takes the stacked, NMO-corrected section and repositions dipping events to their true subsurface locations. Together these steps turn raw field records into an interpretable image of formations like the Montney or Cardium.
Real-World WCSB Scenario: Velocity Picking Over a Montney Target
A processing contractor reprocessing a 3D survey over a Montney play in northeast British Columbia, on acreage held by an operator such as ARC Resources, finds that the original stacking velocities were picked too coarsely, leaving the deep Montney reflector at roughly 2,500 m slightly undercorrected. The far offsets still curve downward, smearing the event in the stack and pushing the time-to-depth tie about 15 m too shallow against well control. Reprocessing with denser velocity analysis and an anisotropic moveout term costs the operator on the order of CAD 120,000 for the survey.
The payoff is a tighter, better-focused Montney image and a depth conversion that ties the offset wells within a few metres, sharpening the horizontal well landing window. On a multi-well pad where a mis-landed lateral can cost CAD 1 million or more in lost productivity, the reprocessing pays for itself many times over by getting the NMO velocities, and the depths that depend on them, right.