What meteorite collections over-represent

Why do the rocks in museum drawers misrepresent what actually falls out of the sky?

A museum drawer of meteorites looks like a sample of the solar system. It is not. It is a sample of what survived a series of filters, each of which had its own preferences, and none of which cared about being representative.

Here is the cleanest way to see it. Meteoriticists keep two separate tallies. A fall is a rock somebody saw come down and then picked up. A find is a rock recognised as a meteorite on the ground with no witnessed arrival. Among falls, stony meteorites make up roughly 95 percent of the total and irons only about 4 or 5 percent. Among the finds that filled nineteenth- and twentieth-century collections outside Antarctica, irons made up a vastly larger share — in some collections approaching half.

The sky is not delivering different rocks to different collectors. So what is the difference between a fall and a find actually measuring?

Ask what a find requires that a fall does not. A find must be noticed, by somebody who was not watching it arrive, possibly years or millennia later.

So begin with recognisability. What does an iron meteorite look like? A dense, heavy, dark, distinctly un-rock-like lump that sticks to a magnet and does not resemble anything in a European field. What does an ordinary chondrite look like once it has sat outdoors for a century? A grey-brown rounded stone. In a landscape full of stones, it is invisible. The filter is not "is it a meteorite" but "does it look strange enough for a farmer to keep".

Now the second filter, which is the more powerful one. Meteorites weather. Iron meteorites rust, but slowly, and a large one can survive tens of thousands of years in an arid climate. Ordinary chondrites contain metal grains that oxidise, and they crumble in a damp climate within perhaps a few thousand years. And the most fragile classes — the carbonaceous chondrites, especially the CI group, which are soft, water-bearing and will visibly disintegrate in a rainstorm — barely survive at all. Consider a striking consequence: the CI chondrites, arguably the most scientifically valuable meteorites we have because they most closely match the Sun's own composition, are known from a handful of specimens, and essentially all of them are observed falls. Nobody finds one lying about, because there is nothing left to find.

Notice what that means. The rarer and more fragile a class is, the more the find record suppresses it — and fragility correlates with exactly the primitive, volatile-rich material we most want to study. The bias is not random with respect to scientific interest; it runs against it.

Third, terrain. Where can a stone be found at all? Not the sea floor, which is 71 percent of the planet. Not forest, not grassland, not anywhere with soil, vegetation, and an abundance of ordinary rocks. What is left is surfaces that are barren, pale, and stable: hot deserts, where a dark stone shows up on light pavement, and above all the Antarctic ice sheets, where ice flow and ablation concentrate meteorites into blue-ice fields and where nothing else on the surface is rock at all. That is why the modern collection is dominated by Antarctic and Saharan material — and why our sample is now weighted toward two very particular climates, with their own weathering regimes and their own residence times.

And there is a fourth filter, which acts before any of these. What reaches the ground at all is already selected. Weak, porous bodies fragment and ablate away in the atmosphere; strong, coherent ones get through. So even the fall record — our best sample — is not what strikes the top of the atmosphere. It is the durable minority of it, biased again against the fragile material, and biased once more toward daylight hours and inhabited places, because someone had to be there to see it.

Is any of this correctable? Partly, and that is worth saying. Fireball camera networks now record entry, compute orbits, and direct recovery within days, which shortens the weathering window and works in terrain nobody would search on spec. The bias is not being removed so much as measured — which is the honest achievable goal.

The misconception to correct is that a collection bias is a sampling error to be averaged away with more specimens. It is not. Collecting harder along the same channel does not converge on the truth — it converges on a sharper picture of the filter. Ten thousand desert finds tell you a great deal about what survives desert weathering and almost nothing new about what the sky delivers.

The refinement that makes the whole thing usable is that these filters can be distinguished from each other, because they act at different stages and leave different marks. Atmospheric selection acts on strength. Terrain selection acts on contrast and stability. Recognition selection acts on strangeness. Weathering selection acts on time and moisture, and it leaves a measurable weathering grade on the specimen. Because each filter has a signature, you can correct against one at a time — comparing falls to finds isolates the ground-side filters, and comparing hot-desert finds to Antarctic finds isolates climate. The falls are not unbiased, but they are biased differently, and that difference is the lever.

• How weathering grades are assigned, and how they are used to correct a find population statistically.
• Why Antarctic blue-ice fields concentrate meteorites rather than merely preserving them.
• What fireball networks have revealed about how much fragile material never survives entry.

Every term the collection defines is gathered in the glossary.