Semblance

Semblance in geophysical signal processing is a normalized measure of the coherence or similarity between multiple traces (waveforms) in a gather or array, calculated as the ratio of the squared sum of the traces to the sum of the squared individual traces, producing a value between zero (completely incoherent signals, no similarity) and one (perfectly coherent signals, identical waveforms); semblance was introduced by Taner and Koehler in 1969 as a measure for quantifying the degree to which a set of seismic traces align (are coherent) along a specified moveout trajectory in a seismic common-midpoint (CMP) gather, enabling the automated picking of normal moveout (NMO) velocity by searching over a range of stacking velocities and identifying the velocity that produces the highest semblance value at each two-way travel time; in the context of velocity analysis, the semblance function is displayed as a color-coded panel (the velocity spectrum or semblance panel) with stacking velocity on the horizontal axis and two-way time on the vertical axis, and the peaks in the semblance panel indicate the velocity-time pairs that produce the most coherent alignment of reflection events across the offset traces in the CMP gather, from which the velocity analyst picks the interval velocity function used for NMO correction and seismic migration; the same semblance concept is applied in array acoustic (sonic) logging through the slowness-time coherence (STC) processing algorithm, where the semblance of the multi-receiver waveform dataset is computed over a grid of slowness and arrival time values to identify the wave modes (compressional, shear, Stoneley) and their propagation velocities from the downhole sonic tool data; semblance is also used in borehole image log processing, VSP processing, coherence-based seismic attribute computation, and diffraction imaging as a general measure of waveform consistency across a set of related traces.

Key Takeaways

  • Semblance velocity analysis for seismic reflection processing computes the semblance function S(v,t) = [sum over offsets of (sum over tau of A(x,tau))]^2 / [N * sum over offsets of (sum over tau of A^2(x,tau))], where A(x,tau) is the amplitude of trace x at time tau along the hyperbolic moveout trajectory defined by velocity v and two-way zero-offset time t, N is the number of traces, and the summation over tau is performed over a short window of typically 20 to 80 milliseconds centered on the reflection event; the resulting semblance spectrum peaks where the hyperbolic NMO trajectory through the CMP gather passes through the center of a coherent reflection event at the correct moveout velocity for that reflection, allowing the analyst to identify the stacking velocity function that maximizes the constructive interference of the reflected energy during stack; the semblance velocity spectrum is the most widely used velocity analysis display in seismic processing and is computed at spatial intervals of typically one to five CMP gathers (every 25 to 125 meters along a 2D seismic line or in a grid pattern across a 3D survey) to track lateral velocity variations caused by changes in lithology, fluid content, or structure; the semblance resolution (the minimum velocity difference that produces distinct semblance peaks at a given time) is inversely proportional to the maximum offset-to-depth ratio in the CMP gather, making deep reflection semblance velocity analysis inherently less resolved than shallow velocity analysis for the same source-receiver offset range.
  • Semblance in array sonic logging (slowness-time coherence, STC) applies the same coherence concept to the multi-receiver waveform dataset from a downhole acoustic tool, computing the semblance of the recorded waveforms over a grid of slowness p (microseconds per foot) and arrival time t0 (the arrival time at the first receiver): for each (p, t0) combination in the search grid, the algorithm time-shifts each receiver trace by the amount p * dx (where dx is the receiver distance from the first receiver in the array) and computes the semblance of the shifted traces within a short time window, with the result mapped onto the slowness-time coherence plot as a two-dimensional function of p and t0; peaks in the STC map identify the dominant wave modes and their propagation slownesses, with the compressional wave typically appearing as a peak at low slowness (40-100 microseconds per foot for fast formations) and early arrival time, the shear wave appearing at higher slowness (60-200 microseconds per foot depending on formation type) and intermediate arrival time, and the Stoneley wave appearing at the highest slowness (typically 200-500 microseconds per foot, close to the mud slowness) and latest arrival time; the semblance-based STC processing requires a minimum of three to five receiver stations in the array to produce reliable coherence estimates, which is why modern array sonic tools use eight to thirteen receivers rather than the two receivers of the older borehole compensated sonic.
  • Semblance as a seismic attribute for stratigraphic interpretation is computed from the similarity between adjacent traces in a 3D seismic volume to create a coherence volume that highlights lateral discontinuities in the seismic reflection pattern associated with faults, fractures, channel edges, and other stratigraphic boundaries: the coherence or semblance attribute is computed for each sample in the 3D volume by taking a small spatial analysis window (typically a 3x3 or 5x5 trace aperture in the inline and crossline directions) and a short time window (two to four samples, 4 to 8 milliseconds), computing the semblance of the traces within the spatial window along the local seismic dip direction, and assigning the semblance value to the center trace at the center time of the window; high semblance values (near 1.0) indicate regions where the seismic reflections are laterally continuous and coherent across the analysis aperture, corresponding to undeformed layered stratigraphy; low semblance values (near 0) indicate lateral discontinuities where adjacent traces do not match, highlighting fault planes, fracture zones, channel margins, and erosional unconformities that appear as dark features on the semblance attribute volume when it is displayed alongside or co-rendered with the seismic amplitude volume; the semblance attribute has largely been replaced by more sophisticated coherence measures (eigenstructure coherence, gradient structure tensor similarity) in modern seismic interpretation workflows, but remains in common use because of its computational simplicity and because it is available in essentially all commercial seismic interpretation software packages.
  • Semblance limitations in velocity analysis arise from ambiguities between multiples and primaries, cycle-skipping in the velocity spectrum, and interference between closely spaced reflection events that reduce the reliability of automated semblance velocity picking in complex geological settings: seismic multiples (reflections that have bounced more than once between interfaces) have different moveout velocities than the primary reflections from the same two-way time, and if the multiple velocity coincides with a primary reflection velocity at a different time, the multiple can produce a high-semblance peak in the velocity spectrum that is mistakenly picked as a primary velocity; in areas with strong multiples (shallow water, hard seafloor, salt flanks), the semblance velocity spectrum contains both primary and multiple velocity peaks that must be distinguished by the analyst using moveout curvature, frequency content, or model-based demultiple comparisons; cycle-skipping occurs when the semblance window straddles two half-cycles of a reflection rather than one complete cycle, causing the semblance to peak at the correct velocity for the half-cycle alignment rather than the correct velocity for the full waveform alignment, introducing systematic velocity errors equivalent to half the period of the dominant frequency; the application of autopicking algorithms to semblance velocity spectra in areas with complex geology and multiples requires careful quality control to prevent these artifacts from propagating into the velocity model used for depth imaging.
  • Semblance in VSP processing and refraction seismic tomography applies the coherence concept to measure the alignment of first-break arrival times across a set of receivers at known positions, enabling automated determination of near-surface and interwell velocity models from the arrival time pattern: in VSP semblance velocity scanning, the first arrivals recorded at each depth level in the borehole are correlated against a set of candidate travel time curves computed for a simple layered velocity model, and the semblance of the actual arrival times against each candidate model identifies the best-fit velocity model by finding the model that produces the most coherent alignment of the first arrivals; in refraction seismic tomography for near-surface model building, the semblance of first-break times across a receiver array is used to determine the dominant refractor velocity (the velocity of the formation just below the refractor surface) from the observed time-offset gradient, providing a quality-controlled estimate of the refractor velocity that is more robust than the slope of a hand-drawn T-X line through noisy arrival time picks; the generalization of semblance as a coherence measure across different geophysical measurement contexts (surface seismic, borehole acoustic, VSP, electromagnetic) reflects the universality of the underlying mathematical concept: wherever signals that share a common origin are recorded at multiple positions, semblance provides a model-independent measure of how well a proposed propagation model explains the observed time or phase relationships between the recordings.

Fast Facts

The semblance measure was introduced by M. T. Taner and F. Koehler in their 1969 paper "Velocity Spectra: Digital Computer Derivation and Applications of Velocity Functions" in Geophysics, which described the first systematic computational approach to seismic velocity analysis using multi-trace coherence rather than individual trace picking. The method transformed velocity analysis from a labor-intensive manual process on paper records to an automated computational workflow, enabling the routine production of velocity functions from the large CMP datasets that became standard with the adoption of digital seismic recording in the mid-1960s. The same issue of Geophysics that published the Taner-Koehler semblance paper also published the introduction of the common-depth-point stacking method by Barry, Cavers, and Kneale, reflecting the rapid transformation of seismic processing methodology in the late 1960s.

What Is Semblance in Geophysics?

Semblance is a number between zero and one that measures how similar a set of waveforms are to each other, calculated by comparing the coherent energy (what the traces share) to the total energy (what each trace contains individually). A semblance of one means all the traces are identical: perfectly coherent. A semblance of zero means they share nothing in common: pure incoherence. In seismic velocity analysis, semblance is computed for every possible combination of stacking velocity and two-way time to find the velocity that produces the most coherent alignment of reflected energy across a CMP gather, creating the velocity spectrum from which the analyst picks the velocity function for NMO correction. In array sonic logging, semblance is computed over a grid of acoustic slowness and arrival time to identify the wave modes arriving at the receiver array and measure their propagation velocities. In 3D seismic interpretation, semblance computed between adjacent traces reveals lateral discontinuities associated with faults and stratigraphic boundaries. The mathematical concept is the same in each application: find the propagation parameter (velocity, slowness, or arrival direction) that maximizes the coherence of the recordings across the measurement array.