Signal amplitude, colour and apparent shape must not be converted directly into a confirmed void, defect severity, affected volume, load capacity, repair quantity or repair design.
1. Use an evidence hierarchy
- Data integrity: are files, line directions, coordinates and coverage reliable?
- Waveform repeatability: does the response persist locally and across reasonable processing?
- Spatial corroboration: is it repeated on adjacent lines and genuinely intersected by cross-lines?
- Context: how does it compare with reference ground on both sides?
- Competing explanations: which causes remain physically plausible?
- Specific-cause confidence: what independent evidence supports identification?
Starting with a presumed void and processing toward it reverses this hierarchy and invites confirmation bias.
2. Separate observation from interpretation
An observation describes what the data show: for example, a repeatable change in reflector continuity between stated coordinates. Interpretation explains what may contribute: disturbed fill, moisture variation, an interface change, utility, drainage feature, reinforcement effect or unresolved loss of support.
The conclusion should show the link between them. Avoid naming a cause when the evidence only supports a local contrast.
3. Test spatial support
Record whether the response repeats on neighbouring longitudinal lines, at what coordinates and with what similarity. Confirm whether cross-lines physically pass through the same location. Cross-lines outside the response zone do not close its boundary.
Distinguish the surveyed footprint, the radar area of interest, an advisory investigation buffer and any physically confirmed affected area. These are not interchangeable.
4. Keep competing hypotheses visible
Depending on the site, plausible influences include variable backfill, moisture, drainage, slab or base thickness changes, joints, reinforcement, utilities, chambers, edges, surface coupling, ringing, multiples and an unresolved void. Compare each explanation against geometry, timing, known features, adjacent lines and other evidence.
Drain leakage or washout should remain a hypothesis unless inspection, testing or exposure confirms it. Equally, failure to identify a distinct target does not prove intact ground.
5. Grade different uncertainties separately
Detection
Does a variation exist?
Consider repeatability, signal quality and processing dependence.
Location and extent
Where does it occur?
Consider positioning, line spacing, intersections and unsurveyed gaps.
Cause and depth
What is it?
Consider calibration, physical alternatives and independent verification.
Use plain confidence levels with reasons. Example: confidence is moderate that a local response variation exists, moderate in its along-line coordinate, low in a closed lateral boundary because cross-lines do not intersect it, low in depth without velocity calibration, and low that it specifically represents an open cavity.
6. Treat depth as model-dependent
Two-way travel time becomes depth only through an estimated electromagnetic velocity. A known-depth target or site-specific velocity analysis can improve the estimate. If the site is variable or wet, one velocity may not represent the entire profile.
Report the method and uncertainty. Do not infer a cavity roof, base, thickness or volume from a single reflector without calibrated velocity and corroborated geometry.
7. Recognise common interpretation traps
- Calling warm colours “severe” when they are display amplitudes.
- Drawing a closed boundary from separated lines.
- Treating a hyperbola or reflection as unique material identification.
- Using a known surface symptom as proof of the subsurface cause.
- Interpreting a quiet lower profile as intact ground.
- Ignoring metallic congestion, edges or coupling changes.
- Reporting more coordinate or depth precision than acquisition supports.
8. Write a decision-useful conclusion
State what was observed, where, how repeatable it is, which alternatives remain, the confidence in each part of the interpretation, what is unresolved and what evidence would change the decision. The purpose is to narrow uncertainty—not to make the image sound definitive.