Geophysical Data Processing and Interpretation

GPR raw data arrives as a series of traces, each one a time-series of signal amplitude from a single antenna position. The collection of traces along a transect is assembled into a 2D section (the radargram). Before this section can be interpreted, several processing steps are applied, typically in a fixed sequence.

1. Zero-time (time-zero) correction
   The direct wave between transmitter and receiver arrives at a consistent short delay, but this delay varies slightly between traces due to electronics and surface-coupling variation. Zero-time correction shifts each trace so that all direct-wave arrivals align at time zero, establishing a consistent surface datum across the section.
2. DC removal and dewow
   A low-frequency drift (wow) appears in the early part of each trace, caused by the tail of the direct wave. Dewow applies a high-pass filter to remove this distortion, which otherwise creates a horizontal banding that obscures shallow features.
3. Gain correction
   The signal attenuates with depth, so deep reflections are much weaker than shallow ones. Gain functions (SEC gain, AGC) amplify deeper parts of the section to make them visible. Over-aggressive gain amplifies noise; too little gain leaves deep features invisible. Gain is the processing setting that most changes the visual appearance of a radargram and that most easily misleads a non-specialist.
4. Background removal
   Horizontal banding caused by consistent reflections (the air wave, the direct wave, or a uniform soil layer) can be removed by subtracting the mean trace from all traces. This dramatically improves contrast for sub-horizontal features like grave bases but can remove genuinely flat reflectors if applied too aggressively.
5. Migration
   Once velocity is estimated, migration collapses hyperbolic arches back to their true subsurface positions. The Kirchhoff migration algorithm is most common in forensic GPR software. Migration works best on sections with many hyperbolas from known targets; on a section dominated by diffuse reflections from grave fills, it may offer limited improvement.

Magnetometry data from a walking survey with a fluxgate gradiometer arrives as a series of readings at fixed spatial intervals along parallel transects. Before these readings can be assembled into a plan image, two processing steps are consistently required: de-striping and gridding.

De-striping equalises the mean or median value across adjacent transects, removing the systematic offset that accumulates when the operator walks at slightly different speeds, changes direction at the ends of lines (producing heading errors in the sensors), or when the instrument drifts slightly between line starts. The standard approach is to subtract each traverse's own mean value before assembling the composite image. More sophisticated methods apply a high-pass filter perpendicular to the traverse direction.

After de-striping, the individual traverse readings are interpolated onto a regular grid to produce a continuous plan image. The grid cell size should not be smaller than the data spacing: at 0.5 m traverse spacing and 0.125 m along-traverse sampling, a 0.25 m grid cell is appropriate. Finer gridding invents spatial frequency that is not in the data; coarser gridding loses real detail.

Every forensic geophysicist maintains a mental library of false-positive sources. The list is long and partly site-specific, but certain categories recur across most search environments.

Distinguishing forensic from non-forensic anomalies is the core interpretive skill. The key strategies are: compare the anomaly plan shape and depth profile with the expected target morphology (a grave is approximately 0.4–0.7 m wide and 1.5–2 m long in plan); check whether the anomaly aligns with or is adjacent to surface features that explain it (a tree, a path, a utility marker); and, where possible, use a second method to confirm. A feature that looks like a grave on GPR but produces no corresponding signal on the resistivity survey is more likely to be a geological feature than a forensic target.

Once processing is complete and false-positive sources have been assessed, all anomalies are classified by confidence tier. The Cheetham (2005) framework, the most widely cited published standard in UK forensic geophysics, uses three tiers: high priority, medium priority, and low priority. The tier governs whether excavation is recommended, optional, or deferred.

High-priority anomalies meet all of the following criteria: size and plan shape consistent with a single burial (typically 0.4–0.7 m × 1.5–2.0 m in plan); depth profile consistent with the expected burial depth; positive identification in at least two independent methods; location and orientation that cannot be readily explained by surface features or geology; and no known non-forensic explanation. Excavation is recommended.

Medium-priority anomalies meet most but not all high-priority criteria, or are confirmed by only one method, or have a partially convincing non-forensic explanation. Excavation is recommended if the operational context is high-stakes (homicide investigation, active warrant) but may be deferred in lower-priority searches.

Low-priority anomalies have features inconsistent with a human burial (wrong size, too deep, irregular plan, clear non-forensic attribution). These are recorded but not excavated unless all higher-priority anomalies have been resolved and resources remain.

Geophysical survey results enter criminal or coroner proceedings as the expert evidence of the qualified practitioner who conducted and interpreted them. The data plots (radargrams, greyscale magnetometry plans, resistance grids) are demonstrative exhibits that support the expert's testimony, not raw evidence that speaks for itself. Understanding this distinction matters for how reports are structured and what the expert must be prepared to explain.

The Cheetham (2005) framework provides a report template that includes: site description and search area specification; instrument specifications and calibration records; survey parameters (traverse spacing, sample interval, direction, date, weather); raw data description; processing steps with parameters; annotated data plots showing all anomalies; anomaly catalogue with coordinates, dimensions, confidence tier, and supporting rationale; and conclusions referenced to the excavation result where available.

Cross-examination typically targets four areas: whether the instrument was properly calibrated, whether the survey design was adequate to detect the target at the depth and size claimed, whether the processing steps could have created or suppressed the anomaly, and whether the false-positive sources have been adequately ruled out. An expert who has documented all four areas in the report, with reference to published standards, handles cross-examination from a position of strength.

Reproducibility is the standard against which forensic science is measured, and it applies to geophysical interpretation as much as to DNA profiling or fingerprint comparison. For a geophysical opinion to be reproducible, the raw data, the processing parameters, and the interpretation criteria must all be preserved and documented in a form that allows a second qualified expert to replicate the analysis independently.

In practice this means: the raw data files should be archived in their native instrument format as well as any export format. The processing project file (in whatever software was used) should be archived with all parameter settings. The annotated report plots should show clearly which anomalies were classified at each tier and why. GPS or total-station coordinates should be recorded for all anomaly boundaries so that excavation can be verified against the prediction.

Post-excavation comparison is the most powerful quality-assurance step: the geophysicist should document whether each excavated target confirmed, partially confirmed, or contradicted the geophysical prediction, and this comparison should be included in the final report. Courts and commissioning investigators benefit from this feedback loop, and it contributes to the published evidence base that the next practitioner relies on.