GPR · Method
Nine tons of what?
Radar images geometry. Turning that geometry into a tonnage requires an assumption about what the material is — and that assumption, not the survey, decides most of the answer.
In 2010 we were engaged by Dennis and Kem Parada of Finders Keepers to run a ground-penetrating radar survey at Dents Run, Pennsylvania, on ground associated with a long-rumoured consignment of Civil War gold. They had been searching there for years before we arrived, and their fieldwork put us where we needed to be.
What the radar found was a large metallic anomaly at depth, clearly differentiated from the surrounding geology. Asked about it at the time, we said there was something completely different down there, and that it could well be precious.
That is as far as radar goes, and the distance between that sentence and the figure of nine tons that later appeared in every account of the story is worth explaining — because the same reasoning governs a great many ordinary jobs that have nothing to do with treasure. What is inside a buried tank. How many drums are in a trench. How large a void really is.
What radar actually measures
Ground-penetrating radar does not detect metal, and it does not detect gold. It transmits a short electromagnetic pulse and records what comes back. A reflection occurs wherever the electrical properties of the ground change — principally the dielectric permittivity, and to a lesser degree the conductivity.
What you record is a two-way travel time. Converting that time into a depth requires a velocity, and the velocity depends on the material the pulse passed through. Measured on site, from a target of known depth or from the shape of a diffraction hyperbola, that conversion is reasonably well constrained. Assumed from a table, it is not.
So the primary output is geometry: how deep a reflector sits, how far it extends laterally, and something about its shape.
Why metal is obvious and gold is not
Metal is the easiest target radar has. It is effectively a perfect reflector: the pulse does not pass through it, the return is strong, and the object is often ringing in the record. On a clean line you can see a metallic target immediately.
What you cannot see is which metal. Gold, lead, iron and steel all reflect the pulse back. The radar response tells you that something highly conductive is present and roughly where its upper surface lies. It carries no information about composition, and because the pulse does not penetrate the object, it carries very little about how thick the object is.
That last point matters more than it first appears. Radar sees the top of a metallic body. The bottom of it is in shadow.
From a reflection to a volume
To estimate a volume you need bounds in three directions. Two come fairly honestly from the survey: the lateral extent across a grid of lines, and the depth to the top surface. The third — the thickness of the body — has to be inferred, because the radar cannot look through it.
Inference here means reasoning from the setting: the dimensions of the space the object sits in, the geometry of the cavity or excavation around it, the extent of the disturbance in the surrounding ground. All legitimate, all interpretive, and all carrying uncertainty that compounds when three dimensions are multiplied together.
And then the density assumption
This is the step where most of the final number is decided, and it is not a geophysical measurement at all.
Mass is volume multiplied by density. The survey gives you a volume. The density has to be assumed, and the assumption you choose changes the answer enormously:
- Gold — about 19.3 g/cm³
- Lead — about 11.3 g/cm³
- Iron or steel — about 7.9 g/cm³
The same volume that yields nine tons if the material is gold yields roughly three and a half tons if it is iron. The radar record cannot tell you which, because a strong reflection looks the same either way.
So a mass figure from a radar survey is properly stated as a conditional: a volume consistent with approximately so many tons, if the material is gold. Compress that into "gold detected" and you have converted an assumption into a finding. The survey has not changed. The claim has.
Where the nine tons actually came from
Not from radar, and not from us.
Eight years after our survey, a different firm was commissioned to carry out a gravimeter survey at the same site. Gravimetry is a different proposition entirely. It does not image a reflector; it measures minute variations in the local gravitational field, which respond directly to mass and to the density contrast between a body and the ground around it. That survey reported a density consistent with gold and a figure of roughly seven to nine tons.
This is the correct method for that question. Radar could never have produced that number, because radar does not respond to mass. Gravity does. Running the method that answers the question being asked, rather than the method you happen to own, is the whole argument.
The excavation that followed reported finding nothing. That outcome is disputed and has been litigated for years, and it is not a dispute geophysics can settle. But there is a lesson in it regardless, and it is not a comfortable one: a gravimeter result is also a model. Mass and density are inferred from a field measurement through assumptions about the shape and depth of the body. A better method narrows the uncertainty. It does not abolish it.
What a radar survey can honestly conclude
At Dents Run, as on any site of this kind, the defensible statements are narrow ones. That a strongly reflective, highly conductive body is present. Its depth and lateral extent. That its response is inconsistent with the surrounding geology. That a volume of a particular order is implied, with stated assumptions about thickness and density.
What it cannot conclude is what the body is made of. Only a physical sample establishes that.
At Dents Run that step was taken by the Paradas, who micro-drilled through the rock above the target and recovered fragments they identified as gold.
That division of labour is the whole point. The radar put the drill in the right place. The drill answered the question the radar could not. A survey that locates and characterises a target is doing its job; a survey asked to identify what the target is made of is being asked for something it does not have.
Why this matters on ordinary sites
Treasure is an unusual target. The reasoning is not.
The same chain of inference runs through a great deal of routine work. A client asks how much product remains in an abandoned underground storage tank; radar can give the tank's position and dimensions, not its contents. A site investigation asks how many drums are buried in a trench; radar gives the extent of the disturbance and the strongest reflectors within it, not a count. A karst assessment asks how large a void is; radar gives the roof, and the floor is often below the depth of useful penetration.
In each case a number can be produced, and in each case the honest version of that number comes with the assumption attached. Where the assumption carries real cost — a remediation estimate, a grouting volume, a decision to excavate — that is the point at which a second method earns its place. Magnetics responds to ferrous mass directly. Microgravity responds to mass itself, which is precisely what radar does not do.
A surveyor who gives you a single confident figure from a single method is not giving you a better survey. They are giving you a less complete account of the uncertainty.
Ground-penetrating
radar — what it resolves and where it stops →
MASW or ReMi: where each one fails →