The Gold Standard of Archaeology | Magnetic Gradiometry, Background Removal and Feature Validation

1. Define the survey question and expected contrasts

Back to contents

Decide whether the survey must locate enclosures, map settlement density, avoid services or test an evaluation area. Review soils, geology, land use, known archaeology and modern infrastructure. Expected targets differ: a fired kiln may form a strong compact anomaly, while a shallow ditch filled with magnetically enhanced soil may appear as a weaker linear response. A negative survey cannot answer a question for which the method has little contrast or insufficient resolution.

2. Record sensor geometry and acquisition conditions

Back to contents

Document instrument, sensor separation, carrying height, traverse spacing, sample interval, grid or GNSS control, walking direction and calibration. A gradiometer measures a field difference over a baseline; changing height or orientation can change the response. Note weather, ground obstacles, fencing, power lines, vehicles, footwear and operator equipment. The acquisition log is part of the dataset because many later artefacts follow the way readings were collected.

3. Inspect raw data before processing

Back to contents

Plot raw values, histograms, traverse profiles and grid boundaries. Look for spikes, striping, alternating traverse offsets, drift, clipping and areas of missing or poor coverage. Compare anomalies with field notes and modern features. A dramatic processed image can hide whether a line was present in the measurements or introduced by interpolation. Preserve the raw, minimally converted data as an immutable reference.

4. Remove artefacts with reversible, declared steps

Back to contents

Despiking can limit isolated extreme readings, de-staggering can correct positional offsets between traverses and de-striping can reduce line-to-line bias. Each operation encodes a diagnosis. Apply it only when the artefact and parameter choice are documented. Keep intermediate outputs and quantify what changed. Broad high-pass filtering may suppress geology or large archaeological features; aggressive clipping may hide ferrous evidence that matters for interpretation.

5. Separate background variation from target geometry

Back to contents

Magnetic background can vary with geology, soil formation, cultivation and topography. A local trend removal may clarify small anomalies but can also subtract a genuine broad feature. Compare several display ranges and unfiltered profiles. Archaeological interpretations should use morphology, orientation, association and landscape context rather than brightness alone. Parallel stripes aligned with survey traverses deserve an acquisition explanation before an archaeological one.

6. Worked example: a rectangular anomaly after de-striping

Back to contents

An invented 40-by-60-metre grid shows north–south banding and a faint rectangular response. Raw traverse medians differ by up to 1.8 nanotesla because one sensor drifted. Median line correction reduces the banding, but a rectangle becomes much clearer. The team checks the raw profiles: the rectangle crosses multiple traverses, has consistent edge positions and was already present before correction. A second processing route using conservative polynomial background removal retains it. Trial trenches then find a shallow ditched enclosure on two sides, while the third side corresponds to a modern drain. Validation confirms part of the interpretation and corrects the tempting complete-rectangle story.

7. Use anomaly categories before feature names

Back to contents

Describe positive linear, dipolar, weak diffuse or strong discrete anomalies before calling them ditches, hearths or graves. Interpretive confidence should reflect shape, amplitude, context, repeatability and alternative sources. Ferrous dipoles may mark modern litter, but they can also be archaeologically relevant. A clean map with every anomaly assigned to one feature class hides uncertainty and discourages targeted follow-up.

8. Changed-context transfer check: an urban brownfield

Back to contents

Move from an open agricultural field to a brownfield crossed by pipes, reinforced concrete and demolition rubble. The transferable workflow still logs acquisition, inspects raw data and validates processing. The context changes the signal-to-background ratio: intense modern ferrous responses can mask subtle archaeology and cause instrument saturation. Smaller grids or alternative methods may help, but no processing filter can reliably recover information that was never measured beneath saturation.

9. Know the limits and common misconceptions

Back to contents

Gradiometry does not measure depth directly, and anomaly strength does not map simply to feature size or importance. A blank area can reflect weak magnetic contrast, depth, disturbance or unsuitable sampling. Interpolation adds pixels, not observations. Background removal is not neutral cleaning; it changes spatial frequencies and therefore possible feature shapes. Ground truth can also be incomplete when trenches miss a narrow or mislocated target. Avoid treating excavation as a perfect binary answer.

10. Audit raw data, processing and validation

Back to contents

Archive survey design, control, raw readings, acquisition logs, processing software and version, every operation and parameter, intermediate grids, display scales, interpretation polygons and validation outcomes. Reprocess a subset independently and check that key anomalies survive reasonable alternatives. Link trenches or other tests back to predicted coordinates with positional uncertainty. Report false positives, false negatives and interpretations revised after fieldwork.

11. Frequently asked questions

Back to contents

Why use a gradiometer rather than one magnetometer? The sensor difference suppresses some regional variation and emphasises near-surface contrasts. Does de-striping always improve data? No; it can remove real linear structure. Can the map show exact feature depth? Not by itself. Are strong anomalies always modern metal? No. Does excavation validate the whole survey? Only the sampled predictions. Should processed images replace raw data? Never; raw and intermediate data are needed to audit the result.

12. The gold standard keeps processing answerable to measurement

Back to contents

A credible gradiometry survey connects its interpretation to the instrument path. It records how measurements were acquired, diagnoses artefacts in raw data, applies reversible processing for stated reasons and tests features against independent evidence. The best image is not the smoothest or most dramatic. It is the one whose archaeological patterns remain visible across reasonable processing choices and whose failures teach the next survey where confidence should narrow.