The Gold Standard of Archaeology | Crosshole Seismic Tomography, Ray Coverage and Inversion Smearing

1. Begin with a question that velocity can answer

Back to contents

Define the target in physical terms before designing the survey. A void, saturated fill, masonry body or fractured zone matters only if it produces a velocity contrast large and extensive enough to affect the measured wavefield. State the expected depth, thickness, orientation and competing geological explanations. Crosshole tomography is strongest between accessible boreholes; it says less outside their connecting volume. If the archaeological question concerns a very shallow isolated artefact, another method may be more appropriate. A precise target model guides source frequency, borehole spacing, receiver interval and the independent evidence needed for interpretation.

2. Measure geometry as carefully as arrival time

Back to contents

The inversion assumes known source and receiver coordinates. Boreholes can deviate from vertical, collars can be surveyed incorrectly and trigger delays can shift every pick. Measure borehole trajectories where possible, synchronise clocks, test repeatability and document source depth and receiver coupling. A small coordinate error can create coherent travel-time residuals that look geological. Reciprocal shots, with source and receiver sides exchanged, help reveal timing and geometry problems. Preserve raw waveforms rather than only picked arrivals, because later review may change the phase or first break judged to be the correct pick.

3. Understand how ray coverage is created

Back to contents

Each source–receiver pair contributes a path through the model. Dense sampling does not guarantee uniform information: many paths can overlap near boreholes while the centre, corners or shallow edge remain weakly sampled. Plot ray density, angular coverage and sensitivity, not merely a colourful velocity section. Features are easier to constrain when crossed by rays from several directions. With two near-parallel boreholes, most paths have similar orientations, so elongated anomalies can be poorly resolved. Coverage is a property of geometry and physics, not a decorative quality score added after inversion.

4. Pick arrivals with uncertainty attached

Back to contents

First arrivals may be weak, dispersive, clipped or preceded by noise. A single precise-looking time can hide ambiguity of several samples. Use consistent picking rules, examine frequency content and assign uncertainty or weights based on signal quality. Automated picks should be checked against representative waveforms and failure cases. Removing difficult traces can improve residuals while selectively erasing paths through the target. Report exclusions and test whether the anomaly persists under plausible repicks. Later arrivals contain information too, but a travel-time inversion based on first arrivals should not silently claim to use the full waveform.

5. See the inverse problem behind the image

Back to contents

Tomography seeks a velocity model whose predicted travel times resemble observations. Many models can fit nearly equally well, especially where coverage is sparse. Regularisation chooses among them by favouring smoothness, small departures from a starting model or another prior structure. Strong smoothing broadens anomalies and joins separate bodies; weak smoothing can amplify noise into spots and streaks. The colour scale therefore displays a regularised estimate, not uniquely measured material boundaries. Show the starting model, cell size, objective function, regularisation strength and data residuals together.

6. Worked example: two walls or one broad anomaly

Back to contents

Two masonry walls, each 0.8 metres thick and separated by 1.2 metres of soil, lie between boreholes 12 metres apart. Synthetic travel times are generated, then realistic picking noise is added. A strongly smoothed inversion returns one broad high-velocity zone because the separation is smaller than the effective resolving power of the dominant rays. A weaker penalty produces two peaks but also streaks near the boreholes. Checkerboard and spike tests show that two targets can be recovered only in the central depth range with crossing rays. The defensible conclusion is a high-velocity zone consistent with masonry, with the number of walls unresolved until excavation or denser geometry supplies evidence.

7. Diagnose smearing and streaking

Back to contents

Smearing spreads an anomaly along directions that the data constrain poorly. In straight-ray inversions it often follows dominant ray paths; in refracting media, incorrect path assumptions can shift or bend the spread. Compare resolution matrices, point-spread tests, alternative parameterisations and synthetic recovery experiments shaped like the expected target. A narrow reconstructed stripe is not proof of a narrow wall if a point anomaly produces the same stripe under the survey geometry. Discuss lateral and vertical resolving power separately and identify edge regions where the model is controlled mainly by regularisation.

8. Changed-context transfer check: a suspected burial chamber

Back to contents

Suppose low velocity between boreholes beside a tomb is interpreted as an air-filled chamber. Rays skirt a real low-velocity zone, and straight-ray inversion smears the anomaly upward. A curved-ray model moves the centre deeper, while attenuation and ground-penetrating radar suggest wet disturbed fill rather than air. The transferable method checks target physics, path assumptions, coverage and independent evidence. The archaeological context adds ethical limits on drilling and excavation. The correct report may narrow possible locations without naming a chamber, preserving uncertainty while still improving the next non-invasive survey decision.

9. Know the limits and common misconceptions

Back to contents

A low velocity does not uniquely identify a void, and high velocity does not uniquely identify masonry. Moisture, porosity, compaction, weathering and lithology can generate similar contrasts. Smaller cells do not create finer resolution when wavelengths and ray geometry do not support it. Low travel-time residuals can coexist with a wrong model because the inverse problem is non-unique. Boreholes disturb ground and limit where the method can be used. Two-dimensional inversion can misplace out-of-plane structures. Treat sharp colour boundaries, isolated edge anomalies and untested labels with particular caution.

10. Audit acquisition, inversion and archaeological interpretation

Back to contents

Archive raw seismograms, trigger tests, source and receiver coordinates, borehole deviation surveys, pick files with uncertainty, excluded traces and field notes. Retain the velocity starting model, mesh, ray tracer, damping and smoothing parameters, convergence history and software version. Publish coverage and recovery tests using the actual geometry. Re-run plausible pick sets, starting models and regularisation strengths, then distinguish stable anomaly location from unstable shape or amplitude. Link the final interpretation to borehole logs, excavation, reflection, refraction or electrical evidence without forcing all methods to agree.

11. Frequently asked questions

Back to contents

Does every source–receiver pair follow a straight line? No; rays bend when velocity varies, and the modelling choice matters. What does a good residual mean? The model predicts travel times well, not that its geometry is unique. Can checkerboard tests prove resolution? They are useful but pattern-dependent; target-shaped tests are also needed. Why do anomalies blur near the edges? Paths and crossing angles are often poor there. Can tomography identify material directly? Usually it constrains velocity contrasts that must be interpreted with geology and archaeological context.

12. The gold standard reports information and ignorance together

Back to contents

A credible tomogram is accompanied by its geometry, picks, coverage, regularisation and recovery tests. The analyst asks not only where velocity differs, but which aspects of the anomaly the data actually resolve. Stable position may coexist with uncertain thickness; a broad zone may contain multiple features; edge structure may be prior-driven. Crosshole tomography earns archaeological meaning when those limits remain visible and independent evidence tests the physical explanation.