What the marine seismic reflection display actually represents
Did you know that a marine seismic reflection image is not a photograph of the seabed interior? A source produces acoustic energy, receivers register arriving waveforms, and processing arranges those recordings into an interpretable section. The bright and dark bands primarily reflect changes in acoustic impedance and the wavelet’s shape. A dark line cannot independently tell you that a stratum is weak, permeable, old or suitable for a foundation. To translate a section accurately, identify the acquisition geometry, travel-time axis, velocity model, processing sequence and evidence behind each geological label. That sequence prevents the most persuasive colours on a display from outrunning what was actually measured.
A reader may encounter a shot gather, common-midpoint gather, stacked section, time-migrated section, depth-migrated section or interpreted horizon map. These are not six interchangeable views. A gather retains many traces whose source and receiver positions carry offset information; stacking compresses selected, corrected traces into one response per spatial bin; migration relocates energy under a velocity assumption. A report can be precise only when it says which product is being interpreted. The word translate here means turning specialist notation and images into statements that keep the underlying measurements, assumptions and uncertainties intact.
This is a learning guide, not a survey certificate or a substitute for a licensed geophysicist’s project-specific evaluation. The arithmetic examples are explicitly hypothetical and use idealized geometries. Real interpretation requires navigation logs, instrument calibrations, field records and often multidisciplinary constraints. The purpose is to teach an independent reader to distinguish observed arrivals from a processing artefact and a plausible geological explanation from a verified subsurface fact.
Source, receiver and offset: begin with geometry
A marine acoustic source might be an airgun array or a different source appropriate to the survey objective, while a multichannel streamer carries many receiver groups behind a vessel. A trace is normally associated with a source location, receiver location and acquisition time. The source–receiver separation is its offset, which is not the same as distance from vessel GPS to receiver. For a straight geometry one may take offset as the horizontal separation, but real navigation can involve streamer feathering, crossline displacement and changing tow depths. Saying that an event was seen at a particular offset without defining this geometry can hide a substantial positioning error.
For a hypothetical source at along-line coordinate 100 m and a receiver at 700 m, the horizontal offset is 600 m and the geometric midpoint is at 400 m. If the receiver instead drifts 80 m crossline, total horizontal separation is the square root of 600 squared plus 80 squared, roughly 605.3 m. That distinction becomes important when nonstraight streamers are forced into a two-dimensional interpretation. Check coordinate reference systems and navigation files before converting a distance between channels into a seismic offset. An impressive velocity curve cannot repair wrong geometry.
What two-way travel time does and does not say
Most seismic reflection vertical axes show two-way travel time, often in seconds or milliseconds. A wave propagates from source toward a reflector and returns to a receiver. For a zero-offset, flat-interface, constant-velocity illustration, reflector depth below the reference surface is approximately velocity multiplied by two-way time divided by two. The division by two is essential. Converting 0.4 s at 1,500 m/s into 600 m would be a one-way-time error; the corresponding ideal depth is 300 m. Real seawater and sediment have different velocities, so a single constant velocity is not a reliable depth model for both.
For layered media, work with interval travel times and interval velocities rather than applying one convenient number to the whole record. If the seabed arrives at 0.40 s and a deeper horizon at 0.70 s, the extra sub-bottom two-way time is 0.30 s. Using an illustrative sediment interval velocity of 1,800 m/s gives approximately 270 m of sediment thickness along an assumed vertical ray path. The result is not a measured thickness independent of velocity, dip or ray geometry. Preserve this conditional wording when translating a depth conversion.
Why the reflector is an impedance contrast, not a material label
The basic normal-incidence reflection coefficient in an ideal elastic-acoustic setting can be written as the difference in acoustic impedance divided by its sum: R = (Z2 − Z1)/(Z2 + Z1), where impedance Z is density times compressional-wave speed. The sign and magnitude depend on the contrast, not on a single rock name. A change in porosity, compaction, fluid or lithology may contribute. A strong event might be the result of multiple causes that are not uniquely distinguishable from one trace. Phase convention, source signature and polarity also affect what the positive or negative excursion looks like on screen.
For a synthetic boundary with Z1 = 2.4 million kg/(m²·s) and Z2 = 3.0 million in the same units, R = 0.6/5.4, about 0.111. This predicts a modest positive ideal coefficient under the stated convention; it does not imply an eleven-percent increase in rock density. Increasing velocity alone could produce the contrast. To describe a bright reflection as a specific formation, look for cores, wells, regional mapping or other appropriate ground truth. Distinguish a seismic facies observation from a lithological assertion.
The source wavelet limits the apparent sharpness of layers
A reflection trace is not a direct series of perfect spikes at interfaces. The source pulse, receiver response, propagation, attenuation and processing combine into a wavelet convolved with the reflectivity structure. Two closely spaced interfaces can interfere constructively or destructively. A thick-looking bright band may represent a thin bed whose top and bottom responses overlap, while a geologically real boundary can be visually weak because its impedance contrast is small. Temporal resolution therefore depends on frequency content and velocity, not simply on zooming into a digital picture.
A useful first approximation for vertical bed-resolving scale is a quarter-wavelength criterion, with wavelength equal to speed divided by dominant frequency. At 1,800 m/s and 60 Hz, the wavelength is 30 m and the quarter-wavelength scale is 7.5 m. This is a rule of thumb for tuning, not a universal ability to separate every seven-and-a-half-metre bed. Bandwidth, noise, interference and processing matter. Do not use a single frequency printed in a survey brochure as a guarantee that the processed section resolves all beds at that thickness.
Shot gathers are the recordings before spatial regrouping
A shot gather collects recordings from many receivers following one source activation. The horizontal trace order can represent channel number, offset or another declared coordinate; the vertical axis is recording time. The direct water wave, seabed reflection, refracted arrivals, multiples, bubbles and other noise can overlap. Distinct arrival slopes across receiver positions are part of the evidence used to identify propagation paths. A steep event on one gather is not necessarily a steep subsurface stratum. It may reflect acquisition geometry or wave propagation through water.
Before describing a shot gather, write down source location, streamer configuration, near offset, far offset, sampling interval and whether any traces were muted or edited. Ask whether amplitudes were normalized trace by trace. Automatic gain can make a tiny late arrival appear equally bold as a strong early arrival. When comparing source tests or a noisy channel, retain a version with an interpretable amplitude scale. A reader who skips these questions may confuse a display enhancement with a new physical signal.
Common-midpoint sorting: why many paths share a nominal location
Common-midpoint, or CMP, sorting gathers traces whose source–receiver midpoints fall within the same spatial bin. In a simple horizontal, approximately one-dimensional earth, reflections associated with a common subsurface point can be compared at different offsets. This is the practical foundation for moveout analysis. The word midpoint is a geometric acquisition concept; common reflection point is a more restrictive geological approximation. Dipping interfaces, lateral velocity changes and three-dimensional structure break the convenient equivalence. Some reports use common-depth-point terminology loosely, so the definition should be stated.
A single CMP gather might contain offsets of 200, 400, 600 and 800 m, all assigned to a particular midpoint bin. Each trace can show the same reflecting horizon at a different arrival time. Those travel-time differences help estimate effective propagation velocity. They do not mean four independent borings measured one material at that location. The stacked product is a processed combination of those paths, with correlations and model dependence. Preserve the distinction between fold, spatial coverage and independent ground truth.
Fold and bin size are different quantities
Fold commonly refers to how many traces contribute to a CMP bin before editing, with effective fold often falling after noisy or unsuitable traces are removed. Bin size describes the spatial grouping scale. A survey advertised as high-fold may still have poorly distributed offsets, irregular coverage or narrow azimuths; a small bin may contain too few usable traces for stable velocity estimation. Do not translate fold as image resolution or certainty. It is one acquisition and processing characteristic among several and must be read together with source interval, receiver interval, streamer geometry and survey boundary effects.
In a deliberately simplified straight-tow, split-spread-type counting example, 12 usable traces in a bin mean an effective fold of 12 at that bin. If four are muted for unusable near offsets or contaminated events, the remaining fold is eight. That example says nothing about actual statistical independence or geological accuracy. At the edge of a survey, fold may taper rapidly. A map showing nominal fold without effective fold can make the least constrained margin look as trustworthy as the center. Ask for actual fold and offset-distribution plots where project decisions depend on imaging quality.
Hyperbolic normal moveout: the central calculation
For a flat reflector in a homogeneous isotropic approximation, reflection time at horizontal offset x follows t(x)² = t0² + x²/v², where t0 is the zero-offset two-way time and v is the relevant effective velocity. An offset trace arrives later because its path is longer. Normal moveout correction, NMO, uses a velocity model to shift traces toward an equivalent zero-offset arrival before stacking. This relationship is exact only for the simple stated model. Nonhyperbolic moveout, anisotropy, dip, strong lateral variation and elastic complexity require more careful analysis.
Suppose t0 = 1.20 s, offset x = 900 m and v = 1,800 m/s. Then x/v = 0.50 s and t(x) = square root of 1.44 + 0.25, or 1.30 s. The illustrative NMO adjustment is therefore 0.10 s. A report that calls the measured 1.30 s the zero-offset event has skipped the correction. A report that substitutes 900 ms for 900 m inside the equation has mixed physical dimensions. The units themselves provide an immediate diagnostic: x/v must be time before it can be squared and added to t0².
Stacking velocity is not necessarily sediment interval velocity
Velocity analysis identifies trial velocities that best align selected reflection events across offsets. Stacking velocity is an effective correction parameter for a particular event and geometry. In simple horizontal isotropic layers, it often resembles root-mean-square velocity rather than an interval velocity of the deepest layer. Dispersion of velocities from semblance picks does not automatically measure changes in rock type. Dip, anisotropy, multiples, insufficient offset range and noisy data can all bias picked peaks. Translating a stacking-velocity panel into a geological stratigraphy without these qualifications is a common overreach.
Imagine a panel where trial velocities from 1,500 to 2,100 m/s are tested and the chosen event aligns best around 1,850 m/s. That number is an interpretation supported by an observed coherence maximum under the assumed processing. It is not proof that every traversed material has speed 1,850 m/s. Inspect whether the peak is broad, whether conflicting events cross the window and whether neighboring CMPs show a smooth, geologically plausible trend. Report the pick with its time, location, confidence and limitations rather than one context-free number.
Semblance panels measure coherence, not geological truth
Semblance is a coherence statistic used during velocity analysis. In a typical display, colour intensity indicates how consistently corrected traces line up at candidate time–velocity combinations. A sharp peak can indicate a well-supported coherent event; a broad ridge can indicate weak velocity discrimination; several peaks can arise from crossing events, multiples or noise. Display colour ranges differ, so bright colours are not universal probabilities. Nor is a semblance value interchangeable with percent confidence. The calculation describes trace agreement under a search window and model, not the probability that a named geological formation exists.
Take a hypothetical event that yields similarly strong semblance from 1,800 to 1,950 m/s. The responsible translation is that the velocity pick is poorly constrained across that range, subject to the analysis method. It would be misleading to publish a velocity of exactly 1,875.000 m/s merely because that pixel appears brightest. Investigate aperture, fold, time sampling, offset distribution and frequency. A meaningful confidence statement may require alternative velocity functions and comparison of stacked or migrated outcomes.
Stretch mute protects against distorted far-offset wavelets
NMO does not only move arrival times; it can stretch wavelets, particularly at shallow times and large offsets. A stretch mute excludes portions of traces where the correction would produce unacceptable waveform distortion. The threshold should be documented because different settings change the fold and amplitude-versus-offset information. A clean stack may partly reflect aggressive removal of troublesome data, not improved physical conditions. The absence of a far-offset event after muting cannot be taken as evidence that the reflector is absent at that distance.
When translating a processed line, inspect the trace count remaining at shallow times and compare it with the nominal fold. An example record that begins with 20 traces but retains only seven after stretch mute is effectively supported by seven for that event. A narrow low-noise stacked band can be useful, yet the missing thirteen traces should not vanish from the uncertainty statement. Report both the chosen mute criterion and its consequences for offset coverage. Revisit it if a scientific conclusion depends on angle or amplitude behaviour.
Multiples are not additional sediment horizons
A multiple is energy that has reflected more than once along its travel path. In marine settings, water-layer reverberations and interactions with the free surface can generate arrivals later than the primary seabed reflection. Their apparent continuity can resemble real geological layering. Predictive demultiple methods attempt to suppress them, but suppression can be incomplete or remove genuine primary energy. Translating every continuous event into a stratigraphic horizon invites phantom geology. A disciplined interpretation compares the event with anticipated multiple travel times and acquisition geometry.
If the ideal seabed primary two-way time is 0.40 s, a simple water-bottom-related multiple may appear near 0.80 s, depending on the propagation path. That numerical coincidence is a screening clue, not a proof of identity. Check moveout, polarity, continuity and modelled multiple responses; make a comparative panel before and after processing. If an event disappears under a plausible multiple-removal procedure, document that fact and test whether it was a primary damaged by overaggressive suppression. Interpretation must retain processing alternatives where uncertainty survives.
Frequency filtering improves legibility while changing evidence
Band-pass filtering reduces selected low- and high-frequency contributions, including some noise, but also changes the wavelet and can remove real geological information. The four corner frequencies in a tapered band-pass filter are not four geological boundaries. Spectral plots should show what energy was retained. In USGS processing reports, filters, sample intervals and other operations are documented as part of the data provenance rather than hidden presentation choices. A translated report should make that processing sequence readable to someone who did not operate the software.
Suppose a target reflection is visible only after a narrow filter is applied. That is not automatically a false reflection, but its robustness is weaker than an event that persists through reasonable alternative bands and neighboring lines. Compare raw or minimally processed traces, filtered gathers and stacked results. Note whether apparent bed continuity improves at the expense of vertical resolution. A high-frequency component may sharpen a boundary but become noise-dominated at depth; a low-frequency component may penetrate farther without resolving closely spaced beds. One filter cannot optimize every scientific question.
Deconvolution and phase can alter waveform appearance
Deconvolution attempts to compress, stabilize or otherwise account for the effective seismic wavelet and reverberation, subject to assumptions about noise and reflectivity statistics. Results depend on operator design, stability choices and available bandwidth. A narrower-looking pulse is useful only if ringing and spurious oscillation are controlled. Phase rotations can change the visible position of peaks and zero crossings without moving a real interface. Thus a horizon picked on a peak in one phase convention may not match a zero crossing used in another interpretation.
A useful translation identifies the wavelet convention, whether displayed polarity is standard for that dataset and how phase was estimated. If a horizon shifts by several milliseconds after wavelet shaping, investigate whether the interpreted event marker changed rather than declaring actual stratigraphic motion. Compare ties at wells or independent horizons if they exist. Without a seismic-to-well tie, avoid phrases suggesting a particular coloured lobe is the exact top of a formation. Vertical precision in a digital pick is not the same as certainty about geological interface position.
Gain control can conceal genuine relative amplitudes
Time-varying gain compensates for geometric spreading and attenuation, while automatic gain control equalizes amplitude over a selected window. Both make weak deep events visible, yet both alter relative amplitude relationships. A late event shown as dark as the seabed does not necessarily reflect equivalent physical contrast. Likewise, trace-by-trace normalization can make unusually noisy channels appear normal in a colour display. If amplitude interpretation matters, ask for an appropriately preserved amplitude-processing branch, not only the aesthetically strongest image.
Consider two illustrative reflectors whose recorded amplitudes differ by a factor of ten before gain. After independent normalization, both could appear equally bright. A descriptive writer should then speak about continuity, time and geometry rather than a direct ten-to-one impedance interpretation from the display. Report the gain method and window length, and separate qualitative structural reading from quantitative amplitude analysis. This distinction is especially important if someone is tempted to identify gas, fluid change or material strength solely from a vivid patch.
Stacking raises coherence but discards some distinctions
Once NMO-corrected traces are aligned, stacking combines them into a single trace per bin. Coherent signal may reinforce while uncorrelated noise tends to cancel. The benefit depends on usable fold, phase consistency and residual moveout. Stacking can also blur offset-dependent amplitude behaviour, suppress genuine events if alignment is poor, or retain coherent noise such as multiples. Calling a stack an average of independent measurements is too simple because traces share source and propagation conditions and are often correlated. It is a computational imaging product.
If eight traces contain a shared reflection but independent noise, a simple ideal argument suggests noise reduction proportional to the square root of eight, about 2.83, under restrictive assumptions. That factor is not a guaranteed signal-to-noise improvement on a real survey, particularly where noise is coherent. Inspect pre-stack gathers for residual alignment and image the variation along the line. A stack that looks cleaner after muting more traces may also lose diagnostic information. A fair report says which version was used and whether key features survive plausible alternatives.
Migration relocates energy; it does not verify a fault
In an unmigrated section, dipping reflectors and diffraction energy may appear displaced relative to their true subsurface origins. Migration uses a propagation-velocity model to relocate and focus energy. Time migration operates principally within a travel-time imaging framework; depth migration relies on a depth-velocity model to position reflectors more directly. Both can improve structural geometry when models and assumptions are appropriate. Incorrect velocities can overmigrate, undermigrate or produce false-looking structures. Migration is an inference step, not a certificate that every sharpened line represents a fault.
A point diffractor may create a hyperbolic event in an unmigrated section. Correct migration seeks to collapse that energy toward the scattering location. If the event remains a broad smile, revisit velocity, acquisition aperture and processing. If an apparently sharp fault termination appears only in one migration run, test velocity sensitivity before promoting it to a tectonic conclusion. In language suitable for a manager: the image supports a possible discontinuity under a stated model, and its geometry should be tested against alternative velocities and adjacent profiles.
Time migration and depth migration answer different questions
Time migration can sharpen a section while leaving the vertical axis in two-way time. A time-migrated horizon is therefore not automatically at an observed physical depth. Depth migration integrates a velocity field but is only as credible as that model. Converting every time sample with a single average velocity is dangerous where water depth, sediment consolidation or lateral velocity structure vary. Translators should preserve the word time or depth in figure captions and specify the vertical datum, units and whether vertical exaggeration is applied.
A 1.0 s two-way-time event drawn at the same apparent height in two time sections may correspond to different depths if the velocity structures differ. Conversely, a depth slice at 900 m is not necessarily a reflection-time slice at one universal value. When comparing sections, use consistent reference surfaces and verify the velocity functions. A cross-section marked metres that was derived using a preliminary model should be labeled interpreted depth, not measured penetration. Ask whether the geological hypothesis persists when reasonable velocity perturbations are applied.
The velocity model carries geological assumptions
Water velocity responds to temperature, salinity and pressure; sediment velocity generally changes with composition, compaction and other state variables. A project may combine water-column observations, semblance analysis, well information, tomography and external constraints into a velocity model. Interpolation between sparse control points can create smooth-looking variations unsupported by dense data. A contour in the final velocity field is not independent evidence for a sedimentary unit. Keep inputs, interpolation methods and sensitivity tests visible in the interpretation.
An interval velocity rising from 1,650 to 2,000 m/s with depth may be consistent with increasing compaction, but the same numerical trend does not prove a single lithology. For a reported depth, distinguish the part of uncertainty attributable to picked time from that due to velocity. If t = 0.40 s and v = 1,800 ± 100 m/s in an ideal vertical conversion, depth is 360 m and a velocity-only half-width is about 20 m, before other errors. Avoid listing 360.000 m as if the display’s decimal precision overcame geological uncertainty.
Residual moveout is a useful diagnostic after correction
After NMO, a flat reflection under the assumed model should approximately line up across usable offsets. A systematic curved residual may indicate an incorrect velocity, anisotropic effects, dip, lateral variation or misidentified events. Random scatter can arise from noise, acquisition errors or interference. A single flat stack does not reveal these differences because stacking has already combined the traces. Retain pre-stack QC displays for critical interpretation. They help identify what the chosen velocity failed to explain.
A downward-curving event after correction does not translate automatically into ‘velocity too low’ or ‘velocity too high’ without checking axis orientation, which time is plotted downward and the correction convention. Instead describe the observed residual pattern, state the trial model and verify the sign by testing a small velocity perturbation. If the event flattens consistently, record the preferred range. If it changes unpredictably, investigate geometry, multiples or low fold. Wording that preserves the diagnostic sequence is stronger than a confident but untested direction of error.
Dipping beds violate the simplest midpoint intuition
On dipping reflectors, the actual reflection point is not necessarily the midpoint directly beneath the source and receiver. Normal moveout can be affected by dip and azimuth, especially in two-dimensional data over complex geology. A feature observed at one CMP bin may originate away from its nominal vertical projection. Migration partially addresses imaging displacement when the model is appropriate. The first explanation for a misplaced event should not automatically be unusual tectonics; survey geometry might be sufficient.
When a seismic line crosses a slope obliquely, an apparently low-angle reflector may actually represent a three-dimensional surface intersected at an awkward angle. Two crossing profiles can constrain the geometry better than one. Compare map orientation, line direction and acquisition azimuth before assigning structural dip. The narrative should say apparent dip in the profile until true dip has been constrained. Overstating true dip from a single section can mislead downstream engineering decisions about slope stability or route alignment.
Vertical exaggeration changes the story told by the picture
Seismic sections often exaggerate the vertical axis to make shallow features visible. In a time section, even a one-to-one printed aspect ratio does not make time directly comparable to horizontal distance. In a depth section, one can calculate vertical exaggeration from the plotted physical units. A gently sloping surface can look steep and a thin unit can seem massive. Captions should specify the domain and plotting ratio rather than relying on appearance. Readers need to know whether a ‘cliff’ is geological or graphical.
Consider a depth section with 100 m vertically occupying the same page distance as 1,000 m horizontally. The vertical exaggeration is ten. An apparent visual slope of 45 degrees can correspond to a true slope of about arctangent(0.1), approximately 5.7 degrees. This is not a universal conversion for all published sections; it applies to the described scale pairing. A sound explanation uses actual distance and depth, not adjectives from the picture alone. If numerical slopes are important, calculate them from coordinates.
Seabed picks and sub-bottom horizons need separate checks
The water-bottom reflection often provides an important timing reference. It may also contain strong interference, rough-topography effects or nearby multiples. An error in the seabed pick propagates into sub-bottom thickness estimates. A deeper horizon should not be assigned directly to a borehole layer without checking the tie and velocity model. The water column is not an optional blank region on the section; its two-way travel time, source–receiver geometry and sound-speed variations influence every deeper interpretation.
In a hypothetical vertical model, a seabed event at 0.50 s with water speed 1,500 m/s corresponds to 375 m of water. A deeper event at 0.74 s has an extra 0.24 s two-way through sediment. At 1,800 m/s that interval implies 216 m sediment thickness. The deeper interface is then around 591 m below the reference surface, ignoring model complexity. The arithmetic should show each segment, not multiply the whole 0.74 s by sediment speed. Misusing one velocity for both layers produces a plausible-looking but biased number.
Seismic facies is a disciplined description, not a rock name
Reflection continuity, amplitude character, frequency, configuration and terminations can be described without claiming direct lithology. Chaotic reflectivity may be consistent with deformation, mass movement, gas interference or heterogeneous deposits. Parallel reflections might record layered sediments, but similar appearances can result under different physical conditions. Translate a facies observation in terms of visible attributes first, then state competing geological explanations and what evidence would distinguish them. This makes the analysis useful to readers unfamiliar with the survey’s regional history.
A good report could say that a package contains discontinuous high-amplitude reflections with localized blanking and uncertain basal geometry. A weak report would declare that the package consists of unstable gas-charged mud solely because its colour looks disrupted. The better version invites checks against well logs, core recovery, geochemical evidence, alternative seismic frequencies or independent geophysical data. Interpretive restraint is not indecision; it is a clear boundary between what a waveform shows and what a material classification requires.
Gas blanking, attenuation and fluid inference
Gas in marine sediment can change acoustic properties and make some shallow sections difficult to image. Acoustic blanking may be consistent with gas, but absorption, roughness, processing, insufficient illumination and equipment limitations can produce similar-looking low-reflectivity zones. A bright spot is also not an automatic hydrocarbon discovery. Relative amplitudes depend on processing, tuning and impedance contrasts. Especially in educational writing, separating a possible fluid-related anomaly from confirmed fluid content prevents readers from overreading specialized graphics.
If one line shows a low-amplitude chimney-like region and a crossing line does not, consider acquisition orientation, source bandwidth, velocity, stacking and whether the feature is genuinely three-dimensional. Do not select the most dramatic section as the entire truth. Phrase the hypothesis as a possible gas-related response pending corroboration. Where environmental or safety actions hinge on the interpretation, require a documented professional assessment rather than converting a visually interesting patch into an operational decision.
Faults and reflector terminations require geometric corroboration
A fault interpretation may rely on displaced horizons, truncations, drag, associated deformation and continuity across several profiles. Diffraction, imperfect migration, velocity artifacts and erosional channels can imitate breaks. A single discontinuity is therefore an observation that invites hypothesis testing, not automatically a measured slip. When drawing a fault line across weak-data intervals, differentiate the imaged portion from the extrapolated part. This distinction may be visually subtle yet consequential for a reader interpreting a subsurface-risk map.
Suppose one horizon appears offset by 12 ms of two-way time across a candidate discontinuity. At an illustrative local interval velocity of 1,800 m/s, a naive vertical conversion is around 10.8 m, but that is not necessarily fault throw: dip, imaging, travel path and relative reference must be handled. It is more accurate to report an observed time separation and a model-dependent apparent displacement. Additional lines, structural context and a depth model are needed before treating the difference as physical movement.
Navigation and timing errors can imitate geological change
Navigation from the vessel, streamer positioning, source firing time and receiver clock synchronization determine where each trace belongs. Modern acquisition systems manage these with specialized logs and corrections, but a derived section can still inherit mistakes in coordinate reference, sign convention, time origin or shot indexing. One shifted trace may produce a discontinuity that appears geological. A systematic navigation error can misplace an entire reflector map. The correction record should be part of the interpretive evidence.
If a source location is stored in projected metres but read as geographic degrees, the resulting geometry is not a small rounding mistake. Likewise, swapping eastings and northings may create an apparently plausible line in an incorrect location. Before interpreting an isolated structural feature, inspect adjacent shot positions, missing-shot handling and navigation continuity. Translate ‘positioned’ carefully: it means positioned under a stated geodetic reference with a known accuracy, not simply that a file contains latitude and longitude columns.
Sampling interval, aliasing and signal bandwidth
Digital seismic traces contain samples separated by a time interval, often expressed in milliseconds. The Nyquist frequency is one-half the sampling rate, but a Nyquist value is not proof the useful seismic bandwidth reaches that frequency. Anti-alias filtering, source bandwidth, attenuation and noise set additional limits. A trace sampled every 2 ms has a sample rate of 500 Hz and Nyquist frequency of 250 Hz. Describing it as ‘250 Hz resolution’ would be inaccurate because temporal sampling and geological resolution are different properties.
Spatial aliasing can also occur if receiver, shot or CMP sampling is too coarse to represent steeply dipping energy at relevant frequencies. Migration may then produce coherent-looking artifacts. Use acquisition sampling and event dip jointly when interpreting such features. Report time sample rate, trace spacing and processing resampling separately. Adding interpolated traces can smooth a display but does not create the same independent evidence as collecting more field measurements.
Quality control is an argument, not a decorative appendix
A credible marine seismic interpretation should make a traceable chain from navigation through editing, geometry, velocity picking, moveout, stacking and migration to the final horizons. Each operation should have a reason and a way to reveal failure. Compare selected shot gathers, CMP gathers, fold maps, spectra, velocity panels, migration tests and crossing-line ties. A processing flowchart without QC results is insufficient. Equally, a smooth final section cannot substitute for the missing logs because smoothness can be manufactured by filtering.
Record what was excluded and why: dead channels, noisy shots, contaminated intervals, insufficient offsets and low-fold zones. Separate a feature absent because no usable trace illuminated it from a feature absent because geology lacks an interface. Where a report extrapolates across poor data, mark that boundary visibly and verbally. A concise sentence naming the evidence, the processing assumptions and the unresolved uncertainty often communicates more than a paragraph of adjectives praising image clarity.
Independent worked example: from gather to a defensible statement
Consider a fully hypothetical CMP event at t0 = 1.00 s, with offset 800 m and effective moveout velocity 1,600 m/s. The offset term x/v is 0.50 s. Therefore t(x) = square root of 1.00 plus 0.25, approximately 1.1180 s. The moveout difference is about 0.1180 s. If the observed event is near 1.12 s, it is compatible with the simplified model within a few milliseconds; if it lies near 1.35 s, investigate picking, geometry or model assumptions instead of silently forcing the trace into alignment.
After velocity analysis, suppose a corrected gather uses 11 traces out of 16 original candidates because five are muted. Describe the output as an eleven-trace effective stack at that event, not sixteen independent confirmations. If a migrated image contains a strong continuous reflector at approximately 1.00 s time, say it is an interpreted seismic horizon rather than a measured rock boundary. A complete explanation includes survey geometry, vertical time domain, effective velocity, mute policy and whether the event survives alternative processing.
Independent worked example: comparing thickness estimates
Imagine a deeper marine sediment interval with measured two-way time difference 0.28 s and a plausible interval speed range from 1,700 to 1,900 m/s. Its ideal vertical thickness estimate spans 238 m to 266 m, centered around 252 m at 1,800 m/s. That range addresses velocity uncertainty only. Pick uncertainty, lateral variations, interface dip and waveform phase can increase the total uncertainty. A report saying exactly 252.0 m from this information alone is false precision. The responsible translation gives a range and names what it does not include.
Now suppose the picked time interval itself could vary by ±0.01 s while the speed stayed at 1,800 m/s. That source alone changes the thickness by approximately ±9 m. For a rigorous combined interval, use an appropriate uncertainty model instead of automatically adding independent error terms without considering correlation. This example teaches a general lesson: seismic depth and thickness estimates should preserve both time-pick and velocity uncertainty, with the relation between their errors made explicit when material.
Independent worked example: compare two velocity interpretations
A coherence panel supports either a velocity of 1,700 m/s or 1,900 m/s for a deeper event, depending on the analyst’s picking window. For t0 = 0.80 s and x = 600 m, the ideal moveout times are approximately sqrt(0.64 + (600/1700)^2) = 0.8744 s and sqrt(0.64 + (600/1900)^2) = 0.8601 s. The predicted difference is roughly 14 milliseconds. If trace timing or event ambiguity is of similar scale, the available offsets may not distinguish the models convincingly.
Rather than treating the two interpretations as a vote to average speeds, inspect the data and calculate what measurement quality would resolve their predicted divergence. Consider larger usable offsets, more coherent events or well-control constraints. The model-selection decision should be driven by data fitness and geological plausibility. This kind of translation retains an intelligible quantitative question: is the evidence precise enough to choose? A colourful velocity panel alone cannot answer that question.
Independent worked example: interpreting apparent dip responsibly
Suppose a depth-migrated reflector rises from 600 m depth to 540 m over 1,000 m horizontal distance. Its vertical change is 60 m, giving absolute gradient 0.06 and an apparent dip angle of arctangent 0.06, about 3.43 degrees. This is an apparent profile dip, subject to depth model and section orientation. If the printed figure exaggerates vertical scale fivefold, it may look more dramatic, but the calculated gradient remains 0.06 using the actual coordinates. Be explicit about the reference datum and coordinate direction.
If the reflector is actually part of a three-dimensional surface, its true maximum dip may be steeper than the section’s apparent dip. A crossing survey line can help resolve orientation, but differences in imaging quality and navigation must still be reconciled. The useful written statement is not ‘a steeply dipping layer’ but ‘the interpreted horizon rises about 60 m along this kilometre of section under the stated depth model.’ It is a claim someone can check rather than an adjective someone must trust.
Transfer exercise 1: unit discipline
A reflector appears at 600 ms two-way time in an ideal 1,500 m/s uniform medium. Convert to seconds first, then calculate ideal depth. Answer: 600 ms is 0.600 s and 1,500 × 0.600 / 2 = 450 m. The calculation is not a claim that real marine sediments have a uniform 1,500 m/s speed.
Transfer exercise 2: identify the variable
A table lists t0, x and v for moveout. Which field is offset and what units should it have? Answer: x is source–receiver separation, ordinarily metres in an SI calculation; v must then be metres per second and t0 must be seconds. The dimensional test rejects a common mistake before a plot is drawn.
Transfer exercise 3: do not confuse fold and resolution
One CMP bin has 24 input traces and only 10 survive editing. Does its 24-fold brochure description tell you the effective fold of a target event? Answer: no. The event’s effective fold is ten under this example, and resolution requires separate bandwidth, velocity and sampling evidence.
Transfer exercise 4: judge an event near the water-bottom multiple
A seabed primary occurs at 0.35 s and another coherent event near 0.70 s. Is the second event certainly a deep horizon? Answer: no. Its timing is also consistent with one simple multiple hypothesis, which must be tested using moveout, processing and neighboring lines before geological classification.
Transfer exercise 5: separate interval from root-mean-square velocity
A semblance pick is 1,900 m/s at 1.2 s. Can you name a 1,900 m/s layer? Answer: not from that number alone. The picked stacking velocity is effective under a model, and interval velocity demands additional processing and assumptions.
Transfer exercise 6: flag phase uncertainty
A candidate horizon is picked at a positive peak on one processed section and a zero crossing on another. Can you compare those depths directly? Answer: only after checking wavelet phase, polarity, processing and tie convention. Otherwise the picking target itself has changed.
Transfer exercise 7: translate a low-fold margin
A survey edge shows a reflector only in three usable traces while the centre has eighteen. What should a caption mention? Answer: edge interpretation carries reduced effective fold and may be less robust. It should not be shown with the same unqualified confidence as the centre.
Transfer exercise 8: test migration dependence
A fault-shaped break appears under one migration velocity but vanishes under a slightly different justified model. What is the claim? Answer: a model-sensitive possible discontinuity, not independently confirmed fault displacement. Keep alternative images available for review.
Transfer exercise 9: separate time and depth
A figure’s vertical axis says 1,000 ms and its caption calls the event ‘1,000 m deep’. What is missing? Answer: a travel-time-to-depth conversion and a velocity model, plus the relevant datum and assumptions. Milliseconds are not metres.
Transfer exercise 10: write a precise summary
A stacked marine line shows parallel reflectors between 0.50 and 0.85 s, but no cores are available. Model answer: ‘The processed time section contains a package of laterally continuous, approximately parallel reflectors over that interval; lithology, physical thickness and depositional interpretation require additional constraints.’ This wording identifies evidence without inventing rock properties.
FAQ: Does a brighter seismic event mean harder rock?
Not necessarily. Event amplitude depends on impedance contrast, wavelet interference, attenuation, offset, geometry and processing. Hardness is not a direct seismic amplitude measurement. Additional material tests and independent interpretation are required before translating brightness into engineering strength.
FAQ: Does CMP stand for a single borehole location?
No. CMP is a geometric binning concept for source–receiver midpoints. Different offsets may sample somewhat different subsurface reflection points, especially with dipping structures or lateral velocity variation. The bin is not a physical sample.
FAQ: Is a migrated section always more accurate?
Migration can improve imaging but depends on the velocity model, aperture and algorithm. An incorrect model can misplace or distort energy. Compare image behaviour under justified velocity variations instead of assuming the newest processing stage must be the truest.
FAQ: Why can two processed seismic sections disagree?
Differences in filtering, phase, gain, velocities, mute design, geometry, migration and interpretation can alter appearance. The disagreement is useful diagnostic evidence. Reconcile processing inputs and objective before selecting one version for a scientific conclusion.
FAQ: Can seismic reflection alone identify gas or petroleum?
Seismic anomalies can motivate fluid-related hypotheses but rarely establish a unique fluid composition. Additional measurements, geological constraints and independent validation are necessary. A bright spot, blank zone or amplitude change is not a direct hydrocarbon assay.
FAQ: How do I tell whether a horizon depth is measured?
Look for an actual depth-domain processing chain, velocity information, datum, ties and uncertainty. A depth printed on a migrated section is still an interpreted estimate. The word measured should refer to the instrument observations, not automatically to every derived geological quantity.
FAQ: What if the survey has very few offsets?
Velocity discrimination may be poor, some moveout diagnostics become weak, and stacking may offer limited signal enhancement. More confident wording cannot substitute for absent acquisition geometry. The report should state what the available offset range can and cannot constrain.
FAQ: What is the most useful sentence to put beside a seismic figure?
Identify the product, location, vertical units, main observed event, key model dependence and unresolved alternative. For example: ‘This time-migrated profile shows a coherent reflection near 0.9 s two-way time; its interpreted depth depends on the velocity model and no lithology has been independently established.’
A final reader checklist before interpreting the picture
Read the axis, verify the coordinate system, state the seismic product, determine whether the display preserves relative amplitude, identify the reflector-picking convention, inspect usable fold and offsets, inspect the velocity analysis, test for multiples, distinguish migration from validation, and retain a defensible account of uncertainty. The result should let an independent reader reproduce arithmetic, recognize the limits of imagery and decide what evidence is needed next. Good specialist translation leaves no hidden jump between a waveform and a real-world conclusion.
Primary methodological sources and useful related lessons
USGS multichannel seismic reflection processing sequence
USGS 2014 post-stack migrated SEG-Y metadata and processing
USGS San Diego Bay reprocessed seismic gathers and migration
Google Search Central: people-first content guidance
eduKateSG: How to Translate Easily series index