Weed seed banks
A Socratic walk-through of weed seed banks — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why does a thoroughly weeded field sprout the same weeds again the next year, and the next?
The obvious explanation is reinvasion: seeds blow in from the verge, the neighbour's headland, the bird that perched on the fence. Clean your ground and the world simply refills it.
That story has a testable consequence. If the weeds arrive from outside, the flush should be worst at the field's edges, should reflect what grows next door, and should thin toward the middle of a large block. Mostly it is not. The flush is even across the field, it repeats last year's species rather than the neighbour's, and it appears fastest exactly where the soil was disturbed. So the seeds were already there. The real question is why removing every visible weed for a whole season does not exhaust the supply.
Reasoning it through
REASONING #Count what is actually in the ground. Arable topsoil routinely holds thousands to tens of thousands of viable seeds per square metre — orders of magnitude more than the plants you ever see. A single lamb's quarters or pigweed plant that escapes to maturity can shed well over a hundred thousand seeds. So one missed plant does not merely add to the bank; it can restock a patch of it.
But that alone would not save the weeds. If the whole bank germinated the following spring, a farmer could weed hard for one season and win. Why does it not?
Because most of the bank cannot germinate, even when conditions look perfect. Here the intuitive model quietly fails. We imagine dormancy as sleep — waiting for warmth and water, waking when they come. If that were true, dormancy would be a timer, and the whole cohort would run out together.
What is actually there is a lock with several different keys, and different seeds carry different locks. Some arrive with a physiological block set inside the mother plant before shedding — primary dormancy — which wears off only after months of particular temperature and moisture, at a rate that varies seed to seed. Others break that block, become germinable, and are then re-locked: burial by cultivation puts a seed into darkness, cool and low oxygen, and induces secondary dormancy in a seed that was ready the week before. A seed can cycle in and out of readiness for years.
Now notice what the keys are detecting. A brief flash of light as soil is turned. A wide daily temperature swing, which happens on bare ground and not under a canopy. Nitrate in solution. None of these measure whether it is spring; they measure whether there is a gap — open, uncontested ground. Which is why tillage produces a flush: clearing the field is itself the signal that the field is worth entering.
Still, why does the fraction that responds stay small? Suppose a genotype existed whose seeds all germinated in the first good gap. In most years it would beat its cautious neighbours outright. Then comes a year when the gap closes early — a late frost, a hoeing, a drought after emergence — and the entire cohort dies with no seed set.
Here is the load-bearing arithmetic. A lineage multiplies: this year's population times this year's factor, times next year's, and so on. Multiplicative sequences are ruined by a single zero, however large the other terms. So what selection optimises is not the average annual return but the compounded one, and a strategy that lowers the good years slightly in exchange for never having a zero year wins over the long run. Holding most of your seeds back is the price of never being wiped out. That is bet-hedging, and a seed bank is what it looks like in soil.
The analogy
THE ANALOGY #Think of a family that forbids its members from sailing on the same ship. Every year a few take passage and most stay ashore, so the family always arrives more slowly than a rival that boards together and crosses in one go. Then comes the year of the storm. The rival family is gone entirely; ours has lost a few and still exists — and existing is the only thing that compounds.
the family deliberates and foresees the storm. A seed foresees nothing. The spread of dormancy depths is set by genotype and by the conditions the mother plant happened to experience while filling the seed, and selection retained it only because lineages without it were removed by storms nobody predicted.
Clarifying the model
THE MODEL #Three refinements hold the picture together.
First, weeding acts on the standing plants and barely touches the bank — it removes this year's withdrawal, not the deposit. The bank declines only through germination, decay and predation, and for many arable weeds that annual loss is a modest fraction, so exhausting a bank takes several consecutive seasons of near-zero seed return, and a single escaped plant undoes them.
Second, "the same weeds" is not coincidence. A seed bank is a record of the field's own management history, and repeating the same crop, cultivation date and herbicide selects for species whose emergence window fits that routine. The field is slowly breeding its own weed flora.
Third, longevity can be startling: seeds William Beal buried at Michigan in 1879 still germinated when a bottle was opened in 2021. That is an extreme — most arable weed seeds persist for years or a few decades, not centuries.
One honest caveat: whether variable dormancy is genuinely adaptive bet-hedging, rather than an incidental consequence of the varying conditions under which seeds mature on the mother plant, is still argued. The evidence is strongest in desert annuals, where germination fractions have been tracked against rainfall variance over long runs; for arable weeds the case is more inferred than demonstrated. The mechanism is not in doubt; the evolutionary bookkeeping behind it is less settled than the tidy version suggests.
A picture of it
THE PICTURE #How to readFollow a single seed. It enters at the left, locked. The two-way arrow between Deep and Ready is the heart of it — a seed does not simply wait and then wake, it cycles in and out of readiness, and cultivation pushes seeds both ways at once. Only from Ready can a gap cue trigger emergence, and only a seedling surviving control reaches Plant. Note that every path except one ends in Gone; the single surviving arrow loops all the way back and refills Deep, which is why one escaped weed cancels a season of clean work.
What became clearer
WHAT CLEARED #The weeds are not arriving; they are already present, in a form weeding cannot reach. Dormancy is not a timer that runs down together but staggered locks whose keys signal open ground rather than the season — so clearing the field is itself the cue to emerge. And the staggering is not inefficiency. Because a lineage's success compounds, never having a zero year is worth more than doing well in the good ones, so the bank keeps most of itself underground on purpose. A field is not weedy because it was invaded, but because it is holding a diversified account it never agreed to open.
Where to go next
ONWARD #- Why a stale seedbed — letting a flush germinate and then killing it before sowing — exploits the same cues that cause the problem.
- How the same compounding argument explains diapause in insects.
Key terms
TERMS #| Term | What it means |
|---|---|
| Seed bank | the reservoir of viable, ungerminated seeds held in the soil. |
| Primary dormancy | a block to germination established in the seed before it is shed from the parent plant. |
| Secondary dormancy | dormancy induced in an already-germinable seed by conditions such as burial in dark, cool, low-oxygen soil. |
| Bet-hedging | a strategy that lowers average return in exchange for reduced variance, favoured because lineage growth is multiplicative across generations. |
Every term the collection defines is gathered in the glossary.