How to interpret GPR radargrams for utility detection
Hyperbolas, ground velocity, depth, utilities crossed at an angle, noise and attenuation: a step-by-step method for reading GPR radargrams on utility surveys.
What a radargram shows
A GPR sends an electromagnetic pulse into the ground and records the echoes that come back from changes in material. A radargram lines up those recordings along the survey line: distance along the line across, two-way travel time in nanoseconds down.
A pipe crossed by the line does not show up as a line but as a hyperbola. The antenna “sees” the pipe before it is above it and after it has passed it; the travel time is shortest straight above. The apex of the hyperbola marks the top of the utility.
From hyperbola shape to depth
Travel time alone does not give depth: you need the velocity of the wave in the ground, v = c / √εr, where c is 0.3 m/ns and εr is the relative permittivity of the soil. Dry sand (εr ≈ 4) gives about 0.15 m/ns; wet clay (εr ≈ 16 to 25) 0.06 to 0.075 m/ns. For the same travel time, the computed depth can change by a factor of two depending on the ground.
The good news: the velocity is written into the shape of the hyperbola. A wide-open hyperbola means fast ground, a narrow one slow ground. Fitting the theoretical curve to the echoes gives the velocity, and then the depth: z = v · t / 2.
The velocity entered by the operator on site is often a default value. Recomputing it from the site's own hyperbolas is the single most valuable correction in the whole interpretation.
The trap: utilities crossed at an angle
A hyperbola only has its “textbook” shape when the line crosses the utility at right angles. Crossed at an angle, the utility stays in view for longer: the hyperbola widens and the apparent velocity becomes v / sin θ, where θ is the angle between line and utility. At 45° the velocity error is over 40 %, and the depth is off by as much.
This cannot be fixed on a single line. You need the direction of the utility, which means linking detections across several lines — one more reason never to conclude from one radargram.
Clean the signal before reading it
- Remove the low-frequency drift (dewow) and the direct wave, which hide the first few centimetres.
- Remove the average background: horizontal layers fade, hyperbolas stand out.
- Apply a time-varying gain to make up for attenuation, without clipping.
- Drop bad traces (spikes, empty traces) before any automatic detection.
Frequency, attenuation and utilities you cannot see
A high-frequency antenna (700 to 900 MHz) resolves small, shallow utilities; a low-frequency antenna (250 to 400 MHz) reaches deeper but sees less detail. In clay or saline ground the signal dies quickly, and the higher channel is sometimes the only one that gives a usable result. With a dual-frequency antenna, compare both channels rather than picking one by habit.
A small plastic pipe in wet ground can be invisible to GPR. An empty plastic pipe returns an echo of opposite polarity to a metal one. Electromagnetic location remains essential for metallic and traceable utilities: the two techniques complement each other.
Linking the lines: the real proof
A stone, a root or an air void also gives a hyperbola. What sets a utility apart is continuity: it shows up from line to line, aligned, at a consistent depth, with the same ground velocity. That requires lines crossing the expected utilities, 0.5 to 1 m apart, and in two crossing directions when the orientation is unknown.
Radar Plan 3D automates this reasoning: hyperbola detection on every line, velocity estimated from the data, lines linked in 3D, correction for utilities crossed at an angle, then for each utility the probability that its alignment is down to chance. That number ranks it as confirmed, probable or to verify. The plan is still checked by the surveyor, who remains responsible for the deliverable.
Updated 23 Oct 2026