Zero Point Three Two Until the Aluminum Said Otherwise

Borrowed a 400 MHz ground-penetrating radar unit yesterday. Handheld sled, built-in display, single-button operation. Plug-and-play subsurface imaging. Except the depth scale is wrong out of the box, and unless you calibrate it against your specific soil, every depth reading is a guess dressed up as precision.

The GPR measures time, not distance. It sends a radio pulse into the ground at 400 MHz—UHF band, shorter wavelength than the 2-meter amateur allocation but same fundamental RF physics—and timestamps the echo when it bounces back from a buried object or layer boundary. The display shows depth in metres, calculated from round-trip travel time divided by two. But that calculation requires knowing the propagation velocity, and propagation velocity depends entirely on what the electromagnetic wave is traveling through.

Free space: 0.3 metres per nanosecond, the speed of light. Dry sand: 0.15 m/ns. Wet clay: 0.06 m/ns. Ice: 0.17 m/ns. Concrete: 0.12 m/ns. The dielectric constant of the material slows the wave. Higher dielectric constant means slower propagation. Water has a dielectric constant of 80. Dry soil might be 4. Wet soil might be 25. The composition isn’t uniform, the moisture content changes with weather and depth, and the factory default setting—usually calibrated for 0.1 m/ns, a middle-ground compromise—is wrong for almost everywhere.

The seismic refraction work last week used the exact same discipline. Acoustic wave velocity depends on material density. Loose topsoil: 380 m/s. Clay hardpan: 2100 m/s. You can’t measure depth accurately until you characterize velocity for that specific site. Ground truth calibration is mandatory.

The Calibration Target Method

Standard practice: bury a metal plate at a known depth, scan over it, measure the two-way travel time to the reflection, solve for velocity. If the plate is 0.5 metres down and the GPR shows the reflection at 8.0 nanoseconds round-trip, the wave traveled 1.0 metre total (down and back). Velocity is distance divided by time: 1.0 m / 8.0 ns = 0.125 m/ns. That’s the propagation velocity for electromagnetic waves in that soil, at that moisture level, today. You enter 0.125 into the GPR settings, and now the depth scale is calibrated.

I dug a hole this morning. Not deep—40 centimetres down into sandy loam behind the garage, where the clay layer sits at 1.8 metres but the topsoil is relatively uniform above it. Placed a 20 cm × 20 cm aluminum sheet flat at the bottom. Aluminum is a strong radar reflector—high conductivity creates a large impedance mismatch at the boundary, so most of the electromagnetic energy bounces back instead of penetrating. Backfilled carefully, tamped the soil to eliminate air pockets, smoothed the surface. Air voids scatter the signal. Compaction matters.

GPR sled positioned directly over the buried plate. Factory setting still at 0.1 m/ns. Single-button trigger. The radargram—vertical cross-section display showing depth vs. horizontal position—shows a bright hyperbolic reflection. Peak of the hyperbola marks the plate’s position directly beneath the antenna. The display reads 0.32 metres depth.

Wrong. The plate is at 0.40 metres. The GPR is interpreting the 5.3 nanosecond round-trip time (measured, displayed in the raw data view) as 0.32 m using the 0.1 m/ns default. Real velocity must be slower. I calculate: 0.80 m total travel distance / 5.3 ns = 0.151 m/ns. That’s the number I need.

Entered 0.151 into the velocity setting. Rescanned. The depth reading now shows 0.40 metres. Correct. The hyperbola is still there—same shape, same reflection strength—but the vertical scale has stretched. What was labeled 0.32 m is now 0.40 m. Every other depth reading on the display scales proportionally.

Hyperbolic Arrivals

The hyperbola itself is diagnostic. Point-source reflectors—pipes, buried objects, voids—create hyperbolic signatures as the antenna moves past them. When the antenna is directly over the target, the two-way travel time is minimized. As the antenna moves away, the slant distance increases, travel time increases, and the reflection plots deeper on the radargram even though the object hasn’t moved. The curve is a hyperbola with the apex marking the target’s true depth and lateral position.

Horizontal layering creates horizontal reflections. Soil strata, bedrock interfaces, water tables—these show up as continuous lines across the radargram because the boundary extends laterally in all directions. A hyperbola means localized geometry. Pipe, not layer. The shape tells you what kind of structure you’re looking at before you dig.

The width of the hyperbola—measured at half the reflection amplitude—encodes the burial depth. Deeper objects create wider hyperbolae. Shallower objects create tighter curves. If you know the velocity, you can measure the hyperbola width and calculate depth geometrically, even if the object is off to the side of the scan line. It’s trigonometry. The wavefront expands spherically, the antenna records the echo, and the hyperbola is the geometric locus of all points equidistant in travel time.

Moisture Invalidates Everything

I ran a second scan after watering the calibration patch. Two litres of water distributed evenly over the 1-square-metre area around the buried plate. Waited ten minutes for it to soak in. The depth reading changed to 0.35 metres.

Velocity dropped. Water has a dielectric constant of 80 compared to dry soil’s 4-6. Even a small increase in volumetric water content—maybe 10% by volume—significantly slows electromagnetic wave propagation. The two-way travel time to the plate increased from 5.3 ns to 5.9 ns. Same aluminum sheet, same burial depth, different velocity. 0.80 m / 5.9 ns = 0.136 m/ns. The soil is now 10% slower.

This is the calibration problem. Velocity isn’t constant. It changes with weather, with depth, with lateral position across the site. A single calibration target gives you one number at one moisture level in one soil type. If you scan across heterogeneous ground—topsoil transitioning to clay, wet areas near drainage, compacted zones under pavement—the velocity changes mid-scan and the depth scale becomes progressively wrong.

Professional surveys address this by burying multiple targets at different depths, by scanning known utilities before the survey to verify velocity, or by using common-midpoint analysis: two antennas separated by increasing distances, measuring how reflection times change with offset, solving for velocity from the hyperbolic moveout. That’s beyond a single-unit handheld system. For backyard archaeology—mapping where that pre-1900 nail might have come from, whether there are footings under the garage—you calibrate once in representative soil and accept that depth uncertainty grows with distance from the calibration point.

What Clay Does

Alberta clay kills GPR. The 1.8-metre hardpan layer I mapped with seismic refraction last week is conductive enough to attenuate the 400 MHz signal within a metre of penetration. Electrical conductivity creates resistive losses. The electromagnetic wave induces currents in the soil, those currents dissipate energy as heat, and the signal strength decays exponentially with depth. Higher frequencies decay faster. 400 MHz might reach 2 metres in dry sand but only 0.5 metres in wet clay. The ground itself determines what’s possible.

I scanned over the area where the seismic geophones were planted. Clear reflections down to maybe 0.8 metres—layering visible, probably the transition from sandy loam to denser subsoil. Below that, noise. The 1.8-metre clay boundary that showed up clearly in the seismic travel-time curves is invisible to GPR. Too deep, too conductive. Acoustic waves don’t care about electrical conductivity. Electromagnetic waves care immensely.

Lower frequencies penetrate deeper but sacrifice resolution. A 50 MHz antenna might reach the clay, but the wavelength in soil is roughly 3 metres. You can’t resolve objects smaller than about one-quarter wavelength, so 50 MHz gives you ~0.75-metre vertical resolution at best. The 2-cm-diameter water pipe buried at 0.6 metres would be invisible. Frequency is always a trade-off. You choose what to sacrifice: depth or detail.

I calibrated the GPR for sandy loam at 0.151 m/ns. That number is valid for the top 0.8 metres of my backyard today, under current moisture conditions, when scanning within 5 metres of the calibration hole. Everywhere else, it’s an approximation. The depth scale is only as accurate as the velocity model, and the velocity model is only as detailed as the time you’re willing to spend characterizing it.

The aluminum plate is still down there. Permanent reference target. I’ll leave it buried.