The Gold Standard of Archaeology | Electrical Resistivity Tomography, Inversion Artefacts and Depth Ambiguity

1. Define the archaeological target before laying out electrodes

Back to contents

State the expected size, depth, orientation and physical contrast of the target, together with geology, soil moisture, topography and modern services. A stone foundation may be resistive relative to moist fill, while a water-filled ditch may be conductive, but local conditions can reverse or erase that contrast. Choose whether the question requires a profile, a grid of parallel lines or true three-dimensional coverage. A beautifully inverted section cannot answer a target the array scarcely sensed.

2. Understand apparent resistivity and the forward problem

Back to contents

Each four-electrode measurement samples a broad volume of ground. Current electrodes create an electric field, potential electrodes observe a voltage difference and geometry converts the result to apparent resistivity. That value is not the resistivity of one cell at one depth. The forward model predicts measurements from a proposed earth model; inversion works backward to find models that reproduce the observations. Many different subsurface structures can fit nearly the same data.

3. Choose array geometry for sensitivity, not convention

Back to contents

Wenner, dipole–dipole, pole–dipole and gradient arrays distribute sensitivity differently. Some provide stronger signal; others emphasise lateral change or support faster acquisition. Electrode spacing governs a trade-off between shallow resolution, survey extent and depth sensitivity. Combine arrays only with a clear rationale and comparable error treatment. Record every electrode coordinate rather than assuming equal spacing across obstacles or uneven ground.

4. Measure data quality before inversion

Back to contents

Inspect contact resistance, reciprocal measurements, repeat readings, current levels and obvious cultural interference while still in the field. Dry or stony ground can produce unstable contacts; fences, pipes and buried cables can redirect current. Reciprocal error helps reveal inconsistent data but does not detect every systematic bias. Removing outliers after seeing an attractive image risks circularity, so set rejection and weighting rules in advance and archive both raw and filtered data.

5. Treat inversion as a family of conditional models

Back to contents

Smoothness constraints favour gradual cell-to-cell change, while robust or blocky constraints can preserve sharper boundaries. Regularisation strength balances data fit against model complexity. The starting model, mesh, topographic correction, error model and bounds can all change the section. Report more than one plausible inversion when interpretation depends on these choices. A lower misfit is not automatically a truer archaeological model if it follows noise or creates implausible structure.

6. Worked example: a resistive band beneath a courtyard

Back to contents

In an invented survey, 48 electrodes at one-metre spacing cross a courtyard. A smooth inversion shows a broad resistive body centred near 1.8 metres depth. A blocky inversion places a thinner body at 1.2 metres, while removing three high-error readings breaks the feature into two patches. A perpendicular line intersects only the eastern patch. The team therefore interprets a probable zone of masonry or dry rubble, not a continuous wall at an exact depth. A small core finds rubble beginning at 1.1 metres and wet clay beneath, supporting the contrast while correcting its geometry.

7. Read depth through sensitivity and resolution

Back to contents

Display sensitivity or depth-of-investigation information alongside the model. Resolution generally weakens with depth and away from electrodes; inversion can smear shallow anomalies downward or pull deep contrasts upward. Cell centres should not be read as precise interfaces. Synthetic recovery tests—placing a target in a plausible background and reinverting simulated noisy data—show what the acquisition could recover, although they test only the assumed scenario.

8. Changed-context transfer check: sloping ground and a buried service

Back to contents

Move from a level courtyard to a hillside crossed by a metal water pipe. The transferable workflow still records electrode positions, reciprocal error and alternative inversions. The context adds topographic current distortion and a strong modern conductor that can cast artefacts around archaeological targets. Include measured elevation in the mesh, map the service independently and test models with and without its likely geometry. No cosmetic filter can restore sensitivity lost to a poor layout.

9. Know the limits and common misconceptions

Back to contents

ERT does not reveal material identity directly, and a colour boundary is not automatically a wall face. Depth labels are model coordinates with uncertainty, not excavation levels. Smooth inversions blur boundaries; blocky inversions can manufacture sharpness. Two-dimensional lines assume limited out-of-plane variation, so nearby three-dimensional objects can be misplaced in the section. Agreement with a preconceived plan is not validation unless acquisition and processing alternatives were assessed independently.

10. Audit the model from electrodes to archaeological claim

Back to contents

Archive electrode coordinates and elevations, array sequence, instrument settings, contact and reciprocal errors, exclusions, topography, mesh, forward solver, regularisation, error model, starting model, convergence history and software version. Save raw data and every reported inversion. Link cores, trenches or other geophysics to predictions made before validation, and record disconfirming checks. Another analyst should be able to reproduce both the section and the uncertainty around the interpretation.

11. Frequently asked questions

Back to contents

Is the inverted section a photograph? No; it is a constrained model that predicts measured data. Does greater depth always require wider spacing? Wider spacing can increase depth sensitivity while losing shallow resolution. Which array is best? It depends on target geometry, noise and logistics. Does low misfit prove the structure is real? No. Can ERT identify stone? It detects resistivity contrast; material identity needs contextual or direct evidence.

12. The gold standard makes inversion dependence visible

Back to contents

A credible archaeological ERT result begins with a target-led layout, tested contacts and transparent error handling. It compares reasonable inversion choices, shows where sensitivity declines and treats depth as an interval supported by model resolution. Validation then checks predicted locations without turning one core into universal proof. The gold standard is not the smoothest subsurface picture; it is an evidence chain that reveals which parts of the picture the data can actually support.