Semblance Analysis¶
A moveout-velocity scan that scores cross-trace coherence after trial alignment to identify plausible propagation-velocity trends.
Core Idea¶
Semblance analysis scans trial moveout velocities to find which one best aligns an event across traces recorded at different source–receiver offsets. After each trial correction, it measures the local coherence of the corrected traces and displays the score by time and velocity. A plausible coherent trend can guide stacking, but its peak does not by itself prove a primary reflection or reveal the interval velocity of every layer.[ref-b8f1e3204cfc][ref-bd14cf1a6c8a][^ref-3f375bb09900]
Scope of Application¶
The method is used in reflection-seismic CMP gathers and has been transferred to ground-penetrating-radar CMP surveys. The wave physics and useful velocity range differ, but both settings compare offset-dependent arrivals under trial models. One published GPR case found semblance faster yet less precise than alternative picking procedures in that case; it did not establish a universal ranking.[ref-b8f1e3204cfc][ref-96823726faef]
Clarity¶
In a constructed three-trace, two-sample comparison, corrected vectors \((1,0),(1,0),(1,0)\) give conventional semblance \(9/[3(3)]=1\); a competing trial with \((1,0),(0,1),(0,1)\) gives \(5/[3(3)]=5/9\). The local scan prefers the first trial, but these are illustrative amplitudes, not measured velocities or proof of a primary reflection. Opposite polarity or systematic amplitude changes can reduce the simple score even for a real event.[ref-bd14cf1a6c8a][ref-61524a3e74fb]
Manages Complexity¶
A time–velocity spectrum compresses many traces and candidate corrections into a surface that helps an interpreter pick a velocity trend. That compression is useful but can hide coherent unwanted arrivals, noise, survey-geometry limits and model mismatch; a peak remains a provisional guide.[ref-b8f1e3204cfc][ref-61524a3e74fb]
Abstract Reasoning¶
Define a gather, time window and moveout family; correct the traces for each candidate velocity; compare their normalized agreement; then test the resulting trend against nearby gathers and physical assumptions. Deriving layer interval velocity requires additional modeling or inversion, which may be sensitive to picking error.[ref-bd14cf1a6c8a][ref-3f375bb09900]
Knowledge Transfer¶
The reusable pattern is to align redundant observations under competing parameter hypotheses and score their agreement. Seismic and radar CMP surveys both support it, but wave speed, depth conversion, survey geometry and expected precision must be justified anew. The named method remains domain-specific to propagation-velocity analysis.[ref-96823726faef][ref-b8f1e3204cfc]
[^ref-bd14cf1a6c8a]: Gan and colleagues, conventional semblance method, equation (5). [^ref-b8f1e3204cfc]: Stanford Exploration Project, Conventional velocity analysis. [^ref-96823726faef]: Jacob and Urban, original GPR CMP study (2016), abstract; full text not checked. [^ref-3f375bb09900]: Stanford Exploration Project, Dix inversion of interval velocity estimation. [^ref-61524a3e74fb]: Fomel, AB semblance study (2009).
Relationships to Other Abstractions¶
Current abstraction Semblance Analysis Domain-specific
Parents (1) — more general patterns this builds on
-
Semblance Analysis is a kind of Measurement Method Domain-specific
Trial moveout alignment and normalized cross-trace coherence constitute a specified measurement method, with propagation-velocity scanning as its differentia.
Hierarchy path (1) — routes to 1 parentless root
- Semblance Analysis → Measurement Method → Measurement
Neighborhood in Abstraction Space¶
Semblance Analysis sits in a sparse region of the domain-specific corpus (93rd percentile for distinctiveness): few abstractions share its structure, so a faithful description tends to retrieve it precisely.
Family — Unclustered & Miscellaneous (2551 abstractions)
Nearest neighbors
- Stabilized Inverse Q Filtering — 0.79
- Reflection Seismology — 0.79
- Infrasonic passive differential spectroscopy — 0.78
- Program Profiling — 0.78
- Pursuit Curve — 0.78
Computed from structural-signature embeddings · 2026-10-08