The Gold Standard of Archaeology | Array Aperture, Spatial Aliasing and Direction-of-Arrival Resolution

1. Begin with the archaeological scale and target depth

Back to contents

Define the feature or stratigraphic question before choosing an array. A buried wall, palaeochannel, deep basin edge and general shear-wave profile demand different frequency and wavelength ranges. Passive arrays estimate properties of the recorded wavefield; they do not directly draw an archaeological plan. Link the expected depth and velocity contrasts to the wavelengths that would carry useful sensitivity.

Separate reconnaissance from parameter estimation. A short deployment may reveal dominant noise directions, while a defensible dispersion curve needs stable recording, geometry and quality checks. State whether the aim is a one-dimensional profile, directional comparison or evidence for lateral structure.

2. Relate aperture and spacing to wavelength

Back to contents

The largest inter-station dimensions influence resolution of small wavenumbers and therefore longer wavelengths. Increasing aperture can extend the low-frequency range, but it does not guarantee energy exists there or that the site behaves as assumed. Minimum spacing helps define the high-wavenumber limit before repeated phase differences become ambiguous.

No single diameter describes every two-dimensional array. Effective aperture changes with azimuth, and irregular layouts have a direction-dependent response. Report sensor coordinates and examine the full array response rather than quoting only maximum and minimum distance.

3. Distinguish resolution from spatial aliasing

Back to contents

Resolution asks whether nearby wavenumbers or arrival directions can be separated. Spatial aliasing asks whether different wavefields produce indistinguishable sampled phases at the sensors. A large aperture can narrow the main lobe while wide or repetitive spacing creates grating lobes that support false peaks. The useful band lies between limits imposed by both effects.

A visually narrow beam is not enough. Check the sidelobe structure and frequency-dependent response of the actual layout. When a picked dispersion branch approaches the array limits, label it uncertain or unsupported instead of extending a smooth curve for appearance.

4. Read the array response in two dimensions

Back to contents

Construct the theoretical array response or wavenumber-resolution function from the station geometry. The central lobe shows the neighbourhood over which two wavenumbers are hard to distinguish; sidelobes show plausible false solutions. A circular or nested layout may provide more even azimuthal performance than a line, but site access often produces gaps.

Inspect slices by azimuth and frequency. A geometry can be strong for waves crossing one axis and weak for waves travelling parallel to another. If the response is elongated, the direction-of-arrival uncertainty should be elongated too.

5. Measure directionality and wavefield composition

Back to contents

Ambient noise can be dominated by roads, machinery, coastlines, wind or intermittent human activity. Frequency–wavenumber methods may estimate arrival direction and apparent phase velocity, while spatial autocorrelation methods rely on different assumptions about the wavefield. A dominant source direction can bias a method that assumes an isotropic or sufficiently diffuse field.

Use time windows, polarization, beam power and azimuthal plots to see whether the wavefield changes. Compare day and night or operational and quiet periods. Directional stability may identify a useful source, but it may also mean one source overwhelms the modes needed for a structural interpretation.

6. Worked example: one array, two convincing peaks

Back to contents

Nine sensors are placed in a nearly linear corridor beside a buried precinct. At 8 Hz, the frequency–wavenumber spectrum shows two peaks corresponding to apparent velocities of 180 and 360 metres per second. The tempting story is two surface-wave modes. The array-response map shows a strong grating lobe at the second wavenumber, aligned with the corridor, and synthetic plane waves reproduce both peaks from one source.

The team redeploys a nested two-dimensional layout with shorter baselines and records during two noise regimes. The false branch disappears, while the remaining arrival direction rotates with a road source. Only the frequency band inside the resolution and aliasing limits is retained for inversion.

7. Use nested apertures and complementary methods

Back to contents

Several arrays of different size can cover a wider wavelength range than one compromise layout. Overlapping bands allow consistency checks and help reveal branches that occur only where one array is weak. Keep station timing, orientation and coordinate accuracy consistent so differences are not created by instrumentation.

Compare frequency–wavenumber, high-resolution beamforming and autocorrelation-based results when their assumptions are plausible. Agreement does not prove truth because methods can share data and wavefield limitations, but disagreement is valuable evidence about aliasing, directionality or mode mixture.

8. Changed-context transfer check: a constrained urban trench

Back to contents

Move from an open field to a narrow urban corridor where sensors can only form an L-shape. The transferable work is wavelength planning, response analysis, direction-of-arrival diagnostics and synthetic testing. The changed geometry has strong azimuthal inequality and traffic noise concentrated along streets.

Do not pretend the L-shape has circular resolution. Report directional uncertainty, repeat the layout after rotation if possible, and combine passive results with boreholes, ground-penetrating radar or active profiles. A constrained array can still be useful when its blind directions are explicit.

9. Validate picks before inversion

Back to contents

Estimate array limits, inspect spectra across windows, reject transients when inconsistent with the target, and test synthetic wavefields passed through the actual geometry. Verify clock synchronisation and station coupling. Plot dispersion picks with beam power, uncertainty and any alternative maxima rather than only the chosen curve.

Before inversion, test whether mode labels, lateral heterogeneity or directional velocity differences can explain the observations. A one-dimensional shear-wave model fitted to a contaminated branch may look stable because the inversion is solving the wrong input very precisely.

10. Know the limits and common misconceptions

Back to contents

Larger aperture does not improve every frequency, and denser spacing does not replace a sufficient footprint. More sensors help only when their placement adds independent spatial information and timing is reliable. Spatial aliasing is not ordinary random noise; averaging aliased peaks can make the error look more stable.

Direction-of-arrival is not automatically the direction of a buried feature. It describes the incoming wavefield, which is shaped by sources, propagation and structure. Passive-array results remain model-dependent where multiple modes, body waves, topography or strong lateral changes violate assumptions.

11. Audit geometry, processing and uncertainty

Back to contents

Publish station coordinates, elevations, sensor types, clocks, sampling rate, recording times, rejected windows and environmental context. Retain array-response maps, wavenumber limits, processing parameters, spectra, picks, alternative peaks and synthetic tests. Name the software and version.

Map which frequencies are supported by which aperture and method. Re-run with different windows, grids and source subsets. If the archaeological interpretation depends on a branch close to an aliasing lobe or a single arrival direction, treat that dependence as a result to be tested, not a detail to hide.

12. The gold standard connects geometry to every claim

Back to contents

A credible passive array does not begin with a familiar layout and end with the smoothest dispersion curve. It begins with a wavelength and direction problem, designs geometry around that problem, and shows where resolution ends and aliases begin.

The gold standard uses multiple apertures or methods where needed, diagnoses source direction, validates picks synthetically and carries uncertainty into inversion. Archaeology gains a defensible subsurface constraint, not a coloured image whose precision came from ignoring the array.