Marine reservoir age

Why does a shell taken from a living sea creature radiocarbon-date as though it were centuries old?

Collect a mollusc alive from the shore, kill it today, and radiocarbon date its shell. It comes back several hundred years old. The animal was demonstrably alive; the method is demonstrably sound. So the error is not in the laboratory and not in the specimen.

That is a strange kind of failure, and worth dwelling on. If the offset were random, radiocarbon dating of marine material would be useless. If it is systematic, it is not really an error at all — it is information about something else, and we should be able to say what.

Start with what the method assumes. Radiocarbon is made continuously in the upper atmosphere, cosmic rays knocking neutrons into nitrogen, and it mixes through the air as carbon dioxide. A living thing exchanges carbon with its surroundings and so carries the same proportion its surroundings do. When it dies the exchange stops and the clock starts, the radiocarbon decaying with a half-life of about 5730 years. Notice the hidden step: the method dates the cessation of exchange, and it assumes the reservoir being exchanged with carried the atmospheric proportion.

So ask the obvious question about a shell. What is the shell exchanging with? Not the air. Seawater. And is the carbon in seawater the same age as the carbon in air?

Follow the path it must have taken. Radiocarbon enters the ocean only through the sea surface, and that crossing is slow — carbon dioxide entering water does not simply dissolve but reacts into bicarbonate, and the isotopic composition of that whole pool takes something of the order of a decade to come into step with the air above it. Already the surface lags. But the surface is not a closed film: water sinks out of it at high latitudes, travels the deep ocean for a long time, and comes back up somewhere else. And while it was down there, sealed from the atmosphere, its radiocarbon was decaying with nothing to replenish it.

Now the picture assembles itself. The water a shell builds from is a mixture: some carbon that recently crossed the surface, and some that last saw the sky centuries ago and has been quietly decaying since. Mix young carbon with old and you get an intermediate proportion — one that a dating laboratory, assuming an atmospheric starting point, must report as an age. The shell is not old. Its carbon is.

How large should the offset be? The mixing time of the deep ocean is of the order of a thousand years, and only a fraction of the surface carbon comes from that source, so a few hundred years is the expected scale. That is what is found: a global average of roughly four hundred years is built into the marine calibration curve, and every marine date is corrected against it.

But here is the part that makes this a measurement rather than a nuisance. If the offset comes from old water reaching the surface, then it must be larger where more old water reaches the surface. Ask where that is: coastal upwelling zones, where wind drives surface water offshore and deep water rises to replace it, and the polar seas, where deep water is exchanged vigorously and sea ice can obstruct gas exchange with the air. And that is precisely the pattern — offsets far above the global average off Peru, off Namibia, in the Southern Ocean, and modest ones in the warm subtropical gyres where surface water has been sitting under the atmosphere for a long time.

So the local deviation from the average — conventionally called delta-R — maps how much old carbon is being brought to the surface. What began as an inconvenience for archaeologists is a constraint on ocean circulation.

Two refinements. First, "reservoir age" is not a property of the animal or of the species. It is a property of the water at that place and moment. A shell from a lagoon fed by a river dissolving old limestone will read old for an entirely different reason — the carbon is geological, not merely unventilated — and that hard-water effect must not be folded into the same correction.

Second, and this is where the honest uncertainty sits: delta-R is not a constant through time. It is set by circulation, and circulation changes. Applying a modern regional offset to a sample from the last deglaciation assumes the upwelling of that coast was doing then what it does now, and there is good evidence that in some regions it was not. This is the dominant error term for old marine dates, and it is larger than the laboratory precision by a wide margin.

How is any of this checked? By finding cases where the true age is known independently. Museum collections hold shells collected and labelled with a date before atmospheric nuclear testing flooded the world with bomb radiocarbon in the 1950s, so their true age is written on the drawer. And archaeological contexts sometimes yield a marine sample and a terrestrial one — charcoal, a seed, a bone — that must be the same age because they were deposited together. The prediction is sharp: the marine sample must read older, by an amount that varies with the local oceanography and not with anything else. If paired samples from a single sealed context showed no offset, or if the offset were the same in an upwelling zone as in the middle of a subtropical gyre, the explanation above would be wrong.

One caution against over-reading a marine record on its own. Atmospheric radiocarbon production itself varies, modulated by the solar wind and the geomagnetic field, so a change in a marine sequence could mean the ocean's ventilation shifted or the sky's supply did. The two are only separable by pairing the marine record against a terrestrial one of the same age — which is exactly what tree rings, growing in the air and dated by counting, provide.

Every term the collection defines is gathered in the glossary.