Seismic Interferometry

A geophysical method that correlates or convolves seismic recordings at different receivers to recover an approximate impulse response as though one receiver were a virtual source.

Version

v1 · 2026-09-28 · History

Domain-specific #

7756

Domain group

Natural Sciences

Origin domain

Geology & Earth Sciences

Subdomain

Seismic Imaging → Geology & Earth Sciences

Aliases

SI, Seismic virtual-source interferometry

Core Idea

Seismic Interferometry reconstructs the wave response between observation points by combining recordings made at those points, most commonly through cross-correlation and, in appropriate formulations, convolution.^[1] One receiver is thereby treated as a virtual source and another as its receiver.^[2] Under the required wavefield and sampling conditions, the result approximates the Green's function or impulse response that an actual impulsive source at the virtual-source position would have produced.^[3]

The crucial abstraction is not “using noise as data.” It is the transformation of a pair or ensemble of receiver traces into a source–receiver response through their relative phase and travel-time structure.^[4] Ambient seismic noise is especially important because it can supply energy without a controlled active shot, but active-source and modeled wavefields can also be used.^[5] Stacking correlations over sources or time suppresses incoherent contributions and reinforces repeatable propagation paths.

The recovered response is conditional rather than magical.^[6] Ideal derivations invoke reciprocity, adequate illumination around the receivers, source independence, and a sufficiently diffuse or equipartitioned wavefield. Scattering can help approximate some of these conditions, while uneven source distributions, attenuation, finite record length, and correlated noise introduce bias or artifacts. A correlation peak is therefore evidence to interpret, not automatic proof of a physical reflector.^[7]

Structural Signature

Sig role-phrases:

• the recorded wavefield ensemble — passive noise, active-shot, drill-bit, or modeled seismic traces retain coherent propagation information across observations
• the receiver-pair geometry — two or more known recording locations define the path whose inter-receiver response is sought
• the interferometric operator — cross-correlation or an appropriate convolution combines trace pairs under a declared lag or frequency convention
• the stacking ensemble — integration over sources or time reinforces repeatable path information and suppresses contributions that do not add coherently
• the virtual-source assignment — one receiver location is re-described mathematically as the source of the response observed at another
• the representation conditions — reciprocity, source independence, illumination, wavefield content, attenuation assumptions, and preprocessing delimit Green’s-function retrieval
• the recovered propagation response — the processed quantity approximates an inter-receiver impulse, surface-wave, or reflection response for later imaging, inversion, or monitoring
• the fidelity qualification — causal/acausal symmetry, phase, amplitude, record length, and source-direction tests state which features of the response are trustworthy
• the association-only limit — trace similarity or a correlation peak without a justified propagation-response interpretation is not Seismic Interferometry

What It Is Not

• Not ordinary waveform similarity. Cross-correlation becomes constitutive here only when the receiver-pair geometry, lag convention, stacking ensemble, and virtual-source assignment support reconstruction of an inter-receiver propagation response; an association peak alone reaches the structural signature's explicit limit.

• Not a physical shot fired at the virtual-source location. The recovered response is a mathematical re-description of recordings under an interferometric representation, and it retains source, illumination, preprocessing, and amplitude qualifications that an actual controlled source experiment need not share.

• Not exact Green's-function recovery under arbitrary noise. The approximation depends on reciprocity or a related representation, adequate wavefield content and illumination, source independence, record length, and treatment of attenuation; directional or correlated sources can bias branches or create artifacts.

• Not a reflector or medium change certified by one correlation peak. The fidelity qualification requires tests of causal/acausal symmetry, phase, source-direction sensitivity, preprocessing, and persistence before the recovered feature receives a subsurface interpretation.

• Not optical interferometry merely moved to seismic instruments. The family resemblance lies in phase-sensitive combination, while Seismic Interferometry specifically reconstructs a seismic propagation response between receiver positions through correlation or convolution and virtual-source interpretation.^[8]

Scope of Application

Seismic Interferometry is a precondition-bounded geophysical method: it applies where seismic recordings at known receiver locations contain enough shared propagation information for correlation or convolution and stacking to recover a qualified virtual-source response between them.^[9] Every habitat must state acquisition regime, receiver geometry, operator and lag convention, stacking ensemble, frequency band, reciprocity or representation assumptions, source distribution, illumination, attenuation and preprocessing, response target, and fidelity limits; waveform correlation without Green's-function or propagation-response interpretation is outside scope.

• Ambient-noise surface-wave retrieval. Long passive records at station pairs are correlated and stacked to estimate interstation surface-wave responses and travel information.
• Ambient-noise tomography inputs. Virtual responses from many receiver pairs supply dispersion or travel-time observations for inversion only after directional illumination and retrieval uncertainty are assessed.
• Regional and continental arrays. Dense or long-running networks reconstruct responses over large station separations when source distribution, seasonal variability, and causal–acausal asymmetry are controlled.
• Earthquake-coda interferometry. Repeated scattered wavefields are compared to retrieve or monitor stable propagation features under explicit source and medium assumptions.
• Near-surface velocity analysis. Passive or active receiver records recover direct or surface-wave responses used to estimate shallow shear-wave structure.
• Ground-roll estimation. Interferometric responses characterize coherent surface waves so exploration processing can model or suppress them without treating the correlogram itself as the subsurface image.
• Engineering seismology in structures. Sensor records in buildings or engineered sites support estimates of shear-wave velocity, attenuation, or response change when structural geometry and excitation are declared.
• Volcano imaging and monitoring. Ambient or event-generated records retrieve paths through volcanic regions to study scattering, velocity structure, or temporal change under nonuniform source conditions.
• Active-source redatuming. Controlled-shot data are reorganized so a receiver position acts as a virtual source, with the physical source wavelet, coverage, and representation theorem retained.
• Reflection-response retrieval. Direct and reflected arrivals are correlated or convolved and then migrated or inverted to infer subsurface reflectors rather than read from a raw peak alone.
• Borehole virtual-source imaging. Downhole or surface arrays reconstruct responses referenced to a borehole receiver when direct arrivals and geometry provide the needed common paths.
• Drill-bit seismic interferometry. Persistent drilling energy supplies recordings that can be reorganized into a virtual downhole source without requiring the drill location to emit a known impulsive wavelet.
• Subsalt and steep-dip exploration. Virtual sources near complex salt geometry improve illumination of adjacent dipping sediments when conventional surface acquisition leaves coverage gaps.
• Reservoir time-lapse monitoring. Correlations from repeated windows are compared for small travel-time or waveform changes only after source-distribution and preprocessing drift are separated from medium change.
• Induced-fracture and unknown-source studies. Interferometric formulations help locate sources or map fracture-related responses when event timing, station geometry, and source assumptions support that inverse problem.
• Synthetic and modeled wavefields. Computer-generated seismic responses test retrieval, multiples, source imbalance, attenuation, geometry, migration, and failure modes before claims are transferred to field data.
• Environmental and hazard monitoring. Repeated passive recordings track shallow or regional changes when the claimed observable is a validated propagation-response change rather than generic noise correlation.

Clarity

A clear claim states whether correlation or convolution is used, identifies the virtual-source and receiver coordinates, declares preprocessing and normalization, and names the response intended for retrieval. “We correlated noise” is incomplete without the physical mapping from correlation lag to causal and acausal propagation.

The report should separate mathematical output from physical approximation. It should state the illumination assumption, record duration, frequency band, station geometry, stacking procedure, and known departures from ideal source distribution. Changes in those choices can alter amplitude, symmetry, and apparent arrivals without changing subsurface structure.

Manages Complexity

Seismic Interferometry compresses a distributed wavefield into virtual source–receiver experiments indexed by receiver pair, lag or frequency convention, correlation or convolution rule, stacking ensemble, and frequency band. A fixed array can therefore be reorganized into many virtual-shot gathers, and a long record of ambient or active energy can be reduced to approximate inter-receiver impulse, surface-wave, or reflection responses. Arrival timing, causal and acausal symmetry, and changes across repeated windows then become readable without specifying every contributing source waveform individually.

The resulting branches depend on the acquisition regime: passive noise, controlled shots, and drill-bit energy require different stacking and source assumptions; surface-wave retrieval, reflection imaging, and time-lapse monitoring demand different response fidelity; phase or travel-time recovery can remain useful when amplitude is biased. The compression stops at the representation theorem. Uneven or correlated sources, incomplete illumination, attenuation, scattering, receiver coupling, finite records, and preprocessing remain latent in the correlogram, so a peak is not by itself a reflector or medium change. Those conditions must stay attached when the virtual response is migrated, inverted, or compared.

Abstract Reasoning

The central reasoning move is reciprocity-guided re-description. Instead of asking how an unknown source reached two receivers separately, the analyst asks what the relative phase and lag between the recordings imply about travel between the receiver locations. Integration or stacking over many source contributions cancels terms that do not consistently encode that path.

Counterfactual checks are essential. If the ambient source field becomes strongly one-sided, the branch aligned with the illuminated propagation directions is strengthened while the reciprocal branch is weakened or distorted; if receiver geometry changes while the medium does not, recovered lags shift with the new inter-receiver paths and the virtual-source location moves to the new reference receiver. A response that remains stable as record length increases and across defensible preprocessing choices, under sufficiently distributed and uncorrelated illumination, is retrieval-supported. A feature that follows source-direction imbalance, disappears under longer stacking, or moves with filtering rather than with receiver-path geometry is more plausibly an acquisition artifact; when those alternatives cannot be separated, the subsurface inference remains underidentified.

Knowledge Transfer

Within seismology and exploration geophysics, Seismic Interferometry transfers literally among ambient-noise studies, controlled-source surveys, borehole or drill-bit imaging, engineering seismology, and time-lapse monitoring when recordings are combined to recover an inter-receiver propagation response. The mechanism and vocabulary carry intact: cross-correlation or convolution, stacking over time or sources, virtual-source and receiver roles, causal and acausal lags, and an approximate Green's function or reflection response. So do the diagnostics and interventions: vary record length and preprocessing, inspect directional illumination and reciprocal branches, change receiver geometry, and ask whether a recovered feature follows propagation paths rather than source imbalance or acquisition processing.

Beyond seismic wavefields, the honest transfer is (B) shared abstract mechanism. Acoustic and ultrasonic interferometry can use the same correlation-and-reciprocity logic under their own representation theorems, while the broader parent Correlation also applies to records that have no wave-propagation interpretation. The home-bound cargo is seismic: subsurface propagation, station or borehole geometry, seismic source distributions, migration or inversion, and the claim that the retrieved response describes the Earth between receiver locations. Calling a correlation between economic or biological series “Seismic Interferometry” is only analogy (A) because no virtual seismic source or Green's-function retrieval follows. The stopping boundary is the seismic carrier and its validity conditions: when the medium is nonseismic, or when correlation merely measures association without reconstructing an impulse response, the named method does not transfer.

Examples

Canonical

Two surface stations record a long interval of ambient seismic noise. Analysts cross-correlate the traces over positive and negative lags and stack many time windows. Repeatable phase and travel-time structure between the stations reinforces while incoherent contributions tend to diminish, yielding an approximate interstation surface-wave response. One station is then interpreted as a virtual source and the other as its receiver.^[10] Causal–acausal asymmetry and changing source directions are examined before travel times are used, because a visible correlation peak alone does not guarantee a faithful Green's-function retrieval.

Mapped back: The two long traces form the recorded wavefield ensemble, and their known positions define the receiver-pair geometry. Lagged cross-correlation is the interferometric operator, repeated-window accumulation is the stacking ensemble, and re-describing one station provides the virtual-source assignment. The result is the recovered propagation response, qualified through the representation conditions and the fidelity qualification rather than accepted from correlation alone.

Applied / In Practice

During drilling, a persistent drill-bit wavefield is recorded by a surface array. A direct arrival serving as a reference is correlated with later reflected energy recorded at other receivers, reorganizing the observations into responses referenced near the borehole. Those virtual responses can then support imaging beside complex subsurface structure where ordinary surface sources illuminate poorly. The acquisition geometry, source behavior, coverage, attenuation, and preprocessing remain part of the interpretation; merely observing that drilling produces noise would not instantiate the method.

Mapped back: Drill-generated traces instantiate the recorded wavefield ensemble, surface and downhole relationships supply the receiver-pair geometry, and direct/reflected trace combination performs the interferometric operator. Referencing the result near the borehole enacts the virtual-source assignment, yielding the recovered propagation response. Requiring a justified imaging response rather than generic trace similarity enforces the representation conditions and the association-only limit.

Structural Tensions

T1: Passive-source access versus illumination control. Ambient recordings provide long, inexpensive observation without firing a controlled source, but their source directions, spectra, independence, and seasonal balance are only partly controlled. The acquisition advantage can therefore create asymmetric or biased virtual responses.

Diagnostic: Does the observed source distribution illuminate the receiver pair and propagation branch needed for the claimed retrieval?

T2: Simple correlation versus conditional Green's-function interpretation. Cross-correlating two traces is computationally direct, while treating the output as an inter-receiver impulse response requires reciprocity or a related representation, adequate wavefield content, and declared preprocessing. Mathematical association alone does not supply the physical meaning.

Diagnostic: Which representation conditions turn this correlation or convolution into a propagation response rather than a similarity statistic?

T3: Travel-time robustness versus amplitude fidelity. Phase or arrival-time information may remain useful even when source strength, directionality, attenuation, normalization, and receiver coupling distort amplitude. Reading a phase-stable correlogram as an amplitude-correct experiment exceeds what many retrievals support.

Diagnostic: Does the downstream inference require amplitude fidelity, or only phase and travel-time features that the acquisition can justify?

T4: Stacking stability versus temporal resolution. Longer integration suppresses incoherent contributions and can stabilize a virtual response, while averaging across changing source fields or medium states can blur the interval a monitoring claim is meant to resolve. Short windows preserve change but increase retrieval variance.

Diagnostic: What window length separates a medium change of interest from sampling noise and source-distribution drift?

T5: Scattered-wave enrichment versus artifact proliferation. Scattering can distribute energy and improve the effective illumination needed for retrieval, yet complex multipathing and correlated energy also create arrivals whose origin is difficult to assign. More wavefield richness is not automatically more interpretable information.

Diagnostic: Which arrival tests show that scattering supports the intended inter-receiver response rather than an unmodeled artifact?

T6: Virtual-source resemblance versus physical-source equivalence. A virtual shot gather can be processed like active-source data, but its signature, illumination, causal–acausal balance, and coverage arise from the recorded ensemble rather than an impulsive source fired at the receiver. Downstream methods must preserve those qualifications.

Diagnostic: Which assumptions of the imaging or inversion step hold for this virtual source, and which require correction or qualification?

T7: Recovered response versus subsurface interpretation. A stable virtual propagation response is an intermediate product; assigning a peak to a reflector or a lag shift to medium change requires controls for preprocessing, geometry, source field, and instrument behavior. Retrieval success does not settle the inverse problem.

Diagnostic: What evidence excludes acquisition-side change or processing artifacts before the recovered feature is given a geological interpretation?

T8: Seismic Interferometry autonomy versus reduction to Transformation (Transformation). The parent Prime carries the portable restructuring of an input through a rule while preserving specified invariants into an output. Every Seismic Interferometry retrieval is a strict kind of Transformation because recorded wavefield traces are cross-correlated or convolved and stacked into a receiver-pair virtual-source response. Reduction loses receiver geometry, propagation phase or travel-time constraints, stacking conditions, and qualified Green's-function interpretation; total autonomy hides the general input–rule–output structure.

Diagnostic: Does the account preserve the seismic inter-receiver response and retrieval conditions as differentia of this Transformation, rather than treating one branch operator such as correlation as the genus?

Structural–Framed Character

Seismic Interferometry is structural-leaning. Its vocab_travels is moderate because correlation, convolution, stacking, and transformation are general, while Green's functions, wavefields, receiver geometry, and virtual sources are geophysical. Its evaluative_weight is low because the method's output is constrained by wave physics, though adequacy is judged against a declared fidelity target. Its institutional_origin lies in a designed processing method and acquisition practice. Its human_practice_bound is moderate because seismic propagation is observer-independent but the virtual-source representation is constructed. On import_vs_recognize, recordings contain propagation information, while the operator, stacking ensemble, and virtual-source assignment impose a retrieval frame.

The smallest reviewed portable skeleton is Transformation: input traces pass through a rule-governed correlation or convolution and stacking operation into a recovered response while selected travel-time or phase relations are preserved. Portable and cross-domain reach belongs to that Prime. Seismic Interferometry adds receiver-pair geometry, seismic wave propagation, illumination and reciprocity conditions, Green's-function interpretation, causal and acausal fidelity, and an association-only boundary.

Its character: structural-leaning because the input–operator–output map is mathematically explicit, while seismic propagation conditions and virtual-source interpretation remain indispensable.

Structural Core vs. Domain Accent

Seismic Interferometry is domain-specific rather than a prime because its transformation is fixed to recorded seismic fields, receiver-pair operations, and virtual-source retrieval.

What is skeletal (could lift toward a cross-domain prime). The portable skeleton is a rule-governed restructuring from input to output with declared invariants: recorded inputs are combined by an operator, selected relations survive the restructuring, and the output acquires a new usable representation. That complete Transformation signature recurs literally in linear-algebra maps, compiler translation, and chemical conversion. Seismic Interferometry is a strict domain-specific specialization rather than a prime because its input, operator, invariants, and output are fixed to a seismic virtual-source retrieval.

What is domain-bound. The recorded wavefield ensemble at the receiver-pair geometry is combined by the interferometric operator and the stacking ensemble; the virtual-source assignment re-describes one receiver as a source, and the recovered propagation response must remain within the representation conditions and the fidelity qualification. The seismic accent includes causal and acausal lags, propagation phase or travel time, Green's-function or reflection-response interpretation, illumination, attenuation, source distribution, station or borehole geometry, and the association-only limit. Correlation is a common branch operator, not the genus, because convolution variants retain this transformation while ordinary correlations lack the recovered seismic-response identity.

Why this does not clear the prime bar. The complete named signature does not recur literally across at least three unrelated domains: linear algebra, compilation, and chemistry preserve Transformation's input–rule–invariant–output organization but not seismic traces, receiver geometry, virtual sources, propagation-response retrieval, or its representation conditions. Knowledge Transfer likewise permits literal transfer within seismic and closely wave-physical settings, while treating correlation without seismic retrieval as a broader mechanism or analogy. Removing the seismic carrier, virtual-source interpretation, and retrieval conditions while retaining rule-governed restructuring leaves Transformation, not Seismic Interferometry; removing the interferometric restructuring and its preserved propagation relations while retaining seismic recordings leaves seismic observation or association, not a strict specialization of Transformation.

Instantiates / Related Primes

This entry is a kind of Transformation.

Instantiates — Transformation (Transformation). Seismic Interferometry takes recorded wavefield traces as its carrier, applies a rule-governed cross-correlation or convolution and stacking construction, and produces a virtual-source impulse-response estimate. Receiver-pair geometry and propagation phase or travel-time information are the constraints preserved through that restructuring, while source position and acquisition form are deliberately changed in the recovered representation. Removing the seismic medium, receiver geometry, and Green's-function interpretation leaves the Prime's input–rule–output structure; removing the trace-to-response restructuring destroys Seismic Interferometry itself. This supports strict subsumption rather than a weaker presupposition relation.

Decline — Correlation (Correlation). Correlation is a constitutive operation in the common cross-correlation branch, not the genus of the method. Seismic Interferometry uses lagged trace products to reconstruct a propagation response, may instead use convolution, and adds stacking, virtual-source assignment, and validity conditions absent from directionless statistical co-variation. The shared operator therefore does not make every instance a Correlation.

Relationships to Other Abstractions

Current abstraction Seismic Interferometry Domain-specific

Parents (1) — more general patterns this builds on

• Seismic Interferometry is a kind of Transformation Prime

  Seismic Interferometry takes recorded wavefield traces as its carrier, applies a rule-governed cross-correlation or convolution and stacking construction, and produces a virtual-source impulse-response estimate.

Hierarchy path (1) — routes to 1 parentless root

• Seismic Interferometry → Transformation → Function (Mapping)

Neighborhood in Abstraction Space

Seismic Interferometry sits in a sparse region of the domain-specific corpus (64^th percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.

Family — Unclustered & Miscellaneous (2551 abstractions)

Nearest neighbors

• Reflection Seismology — 0.87
• Stabilized Inverse Q Filtering — 0.87
• Seismic Inversion — 0.85
• Blind deconvolution — 0.85
• Infrasonic passive differential spectroscopy — 0.84

Computed from structural-signature embeddings · 2026-10-08

Not to Be Confused With

• Cross-correlation. Cross-correlation is a mathematical operation used by a common branch of Seismic Interferometry, whereas the method also requires receiver geometry, stacking, virtual-source assignment, and a justified recovered propagation response. Tell: if the result is only lagged similarity between traces rather than a qualified inter-receiver impulse response, it is cross-correlation without Seismic Interferometry.
• Convolution. Convolution is an alternative operator in some interferometric representations, not the complete seismic method or its validity conditions. Tell: identify whether convolution is merely combining signals or is embedded in a construction that recovers a virtual-source response between receiver positions.
• Ambient-noise tomography. Ambient-noise tomography is a downstream inversion method that may use travel or dispersion information from many interferometrically recovered station-pair responses. Tell: retrieval of the qualified pairwise virtual responses is Seismic Interferometry; inference of a spatial medium model from those responses is tomography.
• Reflection seismology. Reflection seismology is the broader acquisition and imaging practice of using reflected seismic energy to infer subsurface structure, and it need not reconstruct a virtual source from receiver recordings. Tell: ask whether a physical or modeled source directly supplies the survey response or whether trace combinations re-create the response between receiver positions.
• Optical interferometry. Optical interferometry extracts information by combining light waves under optical carriers and validity conditions, whereas Seismic Interferometry operates on seismic wavefields to recover a propagation response through the Earth or an engineered seismic setting. Tell: the wave carrier and claimed response are optical in the former and seismic in the latter.
• Waveform similarity analysis. Waveform similarity analysis compares recorded shapes or lagged association without necessarily assigning one receiver as a virtual source or invoking a Green's-function representation. Tell: a similarity score or peak alone does not become Seismic Interferometry unless representation conditions justify a recovered propagation response.

References

[1] Tutorial on Seismic Interferometry: Part 1 — Basic Principles and Applications registry ↩ Show verification details

[2] Unverified encyclopedia synthesis; claim-specific authoritative support was not established in this verification pass. ↩

[3] Unverified encyclopedia synthesis; claim-specific authoritative support was not established in this verification pass. ↩

[4] Unverified encyclopedia synthesis; claim-specific authoritative support was not established in this verification pass. ↩

[5] Unverified encyclopedia synthesis; claim-specific authoritative support was not established in this verification pass. ↩

[6] Unverified encyclopedia synthesis; claim-specific authoritative support was not established in this verification pass. ↩

[7] Unverified encyclopedia synthesis; claim-specific authoritative support was not established in this verification pass. ↩

[8] Unverified encyclopedia synthesis; claim-specific authoritative support was not established in this verification pass. ↩

[9] Unverified encyclopedia synthesis; claim-specific authoritative support was not established in this verification pass. ↩

[10] Unverified encyclopedia synthesis; claim-specific authoritative support was not established in this verification pass. ↩