The Gold Standard of Archaeology | Surface-Wave Dispersion, Mode Misidentification and Shear-Velocity Inversion

1. Start with the observable, not the velocity model

Back to contents

The field record contains wavefields. A dispersion image estimates energy as a function of frequency and phase velocity. The coloured ridge is not yet a layer boundary or a unique shear-wave profile. Preserve shot gathers, geometry, preprocessing and the full dispersion image before tracing a curve. A picked line should be treated as a documented interpretation with uncertainty, not as raw data.

2. Understand why several modes can carry energy

Back to contents

Layered ground supports a fundamental Rayleigh mode and one or more higher modes. Their relative amplitudes depend on frequency, velocity contrasts, source and receiver geometry, and coupling. A high-velocity layer over slower material can create sharp changes in modal dominance. Energy may move from one branch to another so smoothly that a composite ridge looks like one continuous fundamental-mode curve.

3. Test mode identity before inversion

Back to contents

Inspect multiple offsets, shots, components and processing windows. Overlay predicted modal branches from plausible starting models and look for crossings or abrupt changes in ridge width. Where available, compare active and passive data rather than splicing them without a mode check. Picking the highest-amplitude ridge is useful evidence, but amplitude dominance is not the same as a permanent mode label.

4. Match array design to target depth and wavelength

Back to contents

Receiver spacing, spread length, source bandwidth and near-field effects limit the trustworthy wavelength range. Short spreads may not resolve slow low-frequency energy that constrains depth; wide spacing can alias high frequencies that constrain shallow layers. State the wavelength or penetration rule used and test it against forward models. A very deep interface inferred outside the sampled wavelengths is extrapolation, even when the inversion software returns a number.

5. Invert a family of models, not one preferred profile

Back to contents

Dispersion inversion is non-unique: different layer thicknesses and velocities can fit the same finite band. Use multiple starting models or global searches, retain acceptable ensembles and report data misfit alongside model spread. Parameterisation matters. Too many thin layers can create unsupported detail; too few can force sharp contrasts into broad gradients. Include density or Poisson-ratio assumptions if the forward model uses them.

6. Worked example: a buried ditch or a false bedrock step

Back to contents

A survey crosses a suspected defensive ditch. The picked ridge falls from 220 to 150 metres per second at low frequency. A fundamental-mode inversion places a slow zone above a sharp high-velocity boundary at 3.5 metres, interpreted as bedrock dropping beneath the ditch. Forward modelling reveals that the low-frequency ridge follows the first higher mode after a mode jump. A joint-modal inversion fits the image with shallow fill over continuous bedrock near 2.2 metres. A test pit finds disturbed fill but no bedrock step.

7. Connect velocity to archaeology cautiously

Back to contents

Shear velocity responds to stiffness, density, saturation, fabric and stress; it is not a direct material label. Rubble, compacted surfaces, wet clay and weathered rock can overlap. Use the velocity model to frame contrasts and test hypotheses with topography, stratigraphy, boreholes, excavation or complementary geophysics. Avoid converting a velocity threshold into “wall”, “ditch” or “bedrock” without site-specific control.

8. Changed-context transfer check: engineering site classification

Back to contents

Move from mapping a shallow archaeological feature to estimating an averaged site parameter such as Vs30. The transferable rule is still to verify modes, usable wavelengths and inversion non-uniqueness. The decision target changes: a stable depth average may tolerate details that would be fatal to a claim about a one-metre wall footing. Quality criteria must follow the question, not the method name.

9. Know the limits and common misconceptions

Back to contents

The strongest ridge is not always the fundamental mode. A low misfit does not prove the correct mode or a unique profile. Greater apparent penetration does not guarantee resolution at depth. Lateral variation can violate the one-dimensional assumption used for each spread, while archaeological targets may be smaller than the effective wavelength. Smoothing can erase a real contrast; flexible layering can invent one.

10. Audit the full path from gather to layer claim

Back to contents

Retain coordinates, elevation, source and receiver details, coupling notes, raw gathers, mute and filter settings, transforms, dispersion images, picks with uncertainty, modal assignments, parameter bounds, starting models, accepted ensembles and sensitivity tests. Forward-calculate predicted dispersion for every displayed model. Show where archaeological boundaries remain stable across plausible picks and where interpretation depends on one branch.

11. Frequently asked questions

Back to contents

What is dispersion? Frequency-dependent surface-wave speed. Why are there several modes? Layered media support multiple Rayleigh-wave solutions. Can software identify them automatically? It can assist, but the assignment remains model- and data-dependent. Does a higher mode mean bad data? No; correctly modelled higher modes can improve resolution. Is one best-fitting profile enough? No; non-uniqueness and pick uncertainty require alternatives.

12. The gold standard makes the picked branch contestable

Back to contents

A credible surface-wave interpretation preserves the dispersion image, explains mode identity and carries pick uncertainty into a family of shear-velocity models. It respects acquisition limits and separates stiffness contrast from archaeological label. The gold standard does not ask whether the final profile looks geological; it asks whether plausible modal and model alternatives would change the site conclusion.