The soil seed bank

Why does a burned hillside green up with plants nobody has seen there for decades?

A chaparral slope burns in September. By the following spring it is not bare: it is carpeted with fire poppies and whispering bells, annuals that nobody walking that slope for the previous thirty years had ever seen there. Where did they come from?

The tempting answer is that they blew in. But test that. If the seed arrived after the fire, the flush should be thickest at the burn's edge and thin toward its middle, and it should mirror whatever grows just outside. Instead the flush is dense right through the interior, and it can include species absent from every neighbouring unburned slope. So the plants did not arrive. They were already in the ground — and they had been for decades, doing nothing.

Start by asking what "nothing" means for a seed. A seed is not a sleeping plant waiting on warmth and water; if it were, the first wet winter would have emptied the ground. What it carries instead is a block that has to be picked before water and warmth count at all.

So ask what picks it. For many fire-following species the cue is the fire itself, and there turn out to be at least two independent keys. One is heat: hard-coated seeds, especially legumes and rockroses, need a brief pulse of high temperature to crack the coat so water can finally get in. The other is chemical. In 2004 the compound in charred plant material that triggers germination was isolated and named a karrikin — a small molecule produced when vegetation burns, active on seeds at astonishingly low concentration. Some species respond to smoke-derived nitrogen oxides instead. Notice what these cues have in common: none of them can be faked by a mild wet winter. They are specific to the canopy being gone, which is precisely the moment when a small annual has light and bare ground and no competition.

Now the part that actually answers the question. Why does the supply last thirty years? Because the bank is not one seed with a long fuse — it is very many seeds, and they are not identical. They differ in coat thickness, in burial depth, in how much dormancy the mother plant loaded into them, in exactly how much heat or how much karrikin it takes to move them. So any given event — a hot fire, a cool fire, a landslip, a gopher — germinates a fraction of the bank and leaves the rest locked. What looks like patience is really variation plus numbers.

That is the redundancy argument, and it is worth stating in its strongest form: the population survives not because any one seed is durable but because there are thousands of independently triggered copies, so no single event can take them all. Is thirty years even the limit? An experiment begun in 1879, in which bottles of seed were buried to be dug up at intervals, was still producing moth mullein seedlings when a bottle was opened in 2021 — more than a century underground. Longevity in real soil is harder to measure than that, since you cannot tell an old seed from a new one, and honest estimates for most species come with wide error bars.

Three refinements hold the picture together. First, dormancy is not one state. A seed can be shed already blocked, lose that block over months of particular temperatures, and then be re-blocked if it is buried into darkness and low oxygen — so a single seed cycles in and out of readiness for years rather than counting down a timer.

Second, "seed bank" is a claim about persistence, not merely presence. Ecologists separate a transient bank, whose seeds germinate or die within a year, from a persistent one that carries the record of the site across decades. The fire-followers are the persistent extreme; most grassland species are not.

Third, this explains a conservation trap. Because the standing vegetation and the buried record can disagree, a hillside can look intact while its bank has quietly been exhausted by too-frequent fire — burning again before the annuals have set seed spends the copies faster than they are replaced. The redundancy is deep, but it is finite, and it is invisible until the year it fails to answer.

• How fire-return intervals that are too short convert a resilient chaparral into a grassland, and how ecologists detect that before it happens.
• Whether seed banks can be used to restore a site — and why sowing usually fails where the buried bank would have worked.
• The same logic in other systems: dormant bacterial spores, diapausing eggs in a desert pond, latent viruses in nerve cells.

Every term the collection defines is gathered in the glossary.