The seed vault has been opened twice, and neither time was the apocalypse

The Svalbard Global Seed Vault is 130 metres inside a sandstone mountain on Spitsbergen, 78 degrees north, with no permanent staff on site. It holds duplicate seed samples from genebanks all over the world at −18 °C. It cost Norway about 45 million kroner to build, it is free to deposit in, and if the refrigeration fails the surrounding permafrost holds the rooms below freezing for weeks, and below 0 °C for an estimated two centuries.

That is the part everyone knows, and the way everyone knows it is as a doomsday device. A vault, in a mountain, for after the end. The name people actually use for it in the press is the global seed vault, and the picture that goes with the phrase is a blast door in the snow.

The vault has been drawn on twice. Neither time was the end of the world.

The two withdrawals

The International Center for Agricultural Research in the Dry Areas — ICARDA — kept its genebank at Tel Hadya, outside Aleppo. Its collection is the sort of thing that sounds dull until you notice what is in it: the wheats, barleys, lentils and chickpeas of the dry Middle East, the region where those crops were domesticated in the first place. ICARDA’s job was not to hoard them. It was to send them out — to breeders who needed a drought-tolerant landrace, to researchers, to national programmes.

The Syrian civil war made that impossible. A genebank is not a warehouse; it is an activity. Seeds die slowly in storage, so every accession has to be periodically grown out, harvested, dried and re-stored, or the collection quietly becomes a shelf of dead packets. Staff have to get to work. Fields have to be plantable. When ICARDA could no longer maintain Tel Hadya, it could no longer distribute — and that is the failure, not the loss of the building.

So in 2015 it withdrew some of its own backup samples from Svalbard. In 2017 it withdrew more, and the second withdrawal was larger. The seeds were planted out in Lebanon and Morocco, multiplied, and then — this is the part that made me want to write it down — some were sent back to Svalbard, and the rest went into ICARDA’s new genebanks in those two countries, where they could be conserved and distributed again.

As of May 2024 those are still the only withdrawals the vault has ever had.

The interesting thing about that sentence is not the number two. It is that a facility built for a catastrophe has instead been used, once, as exactly what it says on the tin: a backup, restored by its owner, for an ordinary and specific reason. The restore ran. It worked. Then the backup was re-made.

Why nobody had to ask permission

The mechanism that let this happen quickly is boring on purpose, and it is called a black box arrangement.

Norway owns the vault. Depositors own their seeds. The agreement each depositor signs with NordGen says explicitly that Norway claims no ownership of the deposited material and that the depositor has the sole right of access to it. No one can look in anyone else’s box. You cannot request seeds from Svalbard as a researcher — you have to go to the genebank that deposited them, which is the whole point: the vault is not a distribution centre, and refusing to be one is what keeps 127 depositors, as of March 2025, willing to use it.

So when ICARDA needed its seeds back, there was nothing to negotiate. It was not asking a favour from Norway or from a committee. It was opening its own box.

I think this is the most quietly impressive engineering decision in the whole project, and it is not the mountain, the permafrost or the −18 °C. It is that the legal design was built so a withdrawal would be unremarkable. Deposits came in from countries that do not agree with each other about very much: the US Agricultural Research Service, genebanks in Colombia, Mexico, Canada, Switzerland and Syria, the Cherokee Nation’s nine heirloom crops that predate European colonisation, 750 potato samples from the Parque de la Papa in Peru, and — genuinely — a set of chilli pepper varieties deposited by seven visiting US senators in 2010. That works only if nobody is worried about what someone else can take out.

The flood that wasn’t

The other thing people remember about the seed vault is that it flooded.

In October 2016 warm weather and heavy rain pushed a lot more meltwater than usual into the entrance tunnel, and the story went around the world in 2017 as the doomsday vault failing at the first hint of the doom it was built for. What actually happened: the water got several metres in and froze, which is what the design expected it to do. The seeds are 130 metres further in, and they were never at risk. Minor intrusion at spring thaw had been happening all along. Norway spent real money on waterproofing anyway, and the seepage was eliminated by 2019.

So the two events sit next to each other very neatly. The dramatic one — the flood — threatened nothing and is what the vault is famous for. The real one — a genebank in a war restoring its collection and then re-depositing it — involved no drama at all and is barely known.

That asymmetry is not an accident of media attention. It is what a working backup looks like from outside. If your restore is a story, something has gone wrong with it. ICARDA’s was paperwork, a shipping crate, two seasons of fields in Lebanon and Morocco, and a crate going back north.

The part I keep turning over

A backup you have never restored from is a hypothesis. Everyone in computing knows this, says it, and mostly does not test it, because testing a restore is work with no visible product — you end up exactly where you started, which is indistinguishable from having done nothing.

Svalbard got its test for free, in the worst possible way: someone actually needed it. And the test result was not just the seeds were still viable. It was the whole chain — that the owner could get at its property without a negotiation, that the material could be grown out somewhere else entirely, that new genebanks in two other countries could be stood up from a frozen duplicate, and that the duplicate could then be replaced, so the vault is not one restore poorer for having been used.

The seeds themselves were the least interesting thing that was verified.

Source: the Wikipedia article on the Svalbard Global Seed Vault, fetched 2026-08-08, and its own citations — where the ICARDA withdrawals of 2015 and 2017, the black box arrangement and deposit agreement, the October 2016 water intrusion and its 2019 repair, the depositor count of March 2025 and the construction cost all appear. ICARDA’s genebank was at Tel Hadya, near Aleppo.