# How Long Would the Doomsday Seeds Last? Frozen seeds can remain alive for decades — but not forever. Explore 40 years of real seed-bank germination data, see which crops lasted longest and discover why seed vaults must keep testing and regenerating their collections. Canonical: https://mariuscomper.uk/how-long-do-seeds-last-seed-vault/ ## The short answer There is no single lifespan for seeds in a seed bank. In a 40-year dataset from the Australian Grains Genebank, some mung-bean and sorghum seedlots still achieved at least 90% germination after about 45 years at −20 °C, while other seedlots deteriorated much sooner. Professional seed banks therefore test stored seeds periodically and regenerate accessions when viability falls. ## There isn't one answer. Below are six real crops. Each number is the oldest seedlot that was still germinating at 90% or better the last time anyone checked — in 2024, after decades in a freezer at −20 °C. These are the oldest highly viable seedlots observed in this dataset — not universal expiry dates. ### Mung bean (Vigna radiata (L.) R.Wilczek) 45.5 years, ≥90%, n=680, 2024 ### Sorghum (Sorghum bicolor (L.) Moench) 44.4 years, ≥90%, n=1,812, 2024 ### Common bean (Phaseolus vulgaris L.) 37.8 years, ≥90%, n=334, 2024 ### Soybean (Glycine max (L.) Merr.) 34.4 years, ≥90%, n=500, 2024 ### Adzuki bean (Vigna angularis (Willd.) Ohwi & H. Ohashi) 29.1 years, ≥90%, n=339, 2024 ### Pigeon pea (Cajanus cajan (L.) Huth) 25.2 years, ≥90%, n=196, 2024 ## How long should a seed bank wait before checking again? Sorghum: 2,901 → ~17.0 (when the latest germination result is 98%) Mung bean: 2,075 → ~21.5 A monitoring interval of N years means “test again in N years” — it does not mean the seeds die after N years. ## A seed vault is not a time capsule. It is a maintenance system. 1. STORE — Dried seeds are sealed in foil packets and frozen at −20 °C. 2. WAIT — Years pass. The seeds are alive, and slowly ageing. 3. TEST — A sample is thawed and germinated. This is a viability test. 4. DECIDE — Still above 85% germination? Back into storage. Below it? Regenerate. 5. GERMINATE — The tired accession is sown, not shelved. 6. GROW — Plants grow, flower, and set seed in the field. 7. HARVEST — Fresh seed is harvested, dried, and tested. 8. STORE AGAIN — The new generation is sealed and frozen. The lineage continues. ## The Doomsday Vault Is Not a Freezer Full of Immortal Seeds Seeds are living biological material. Cold slows their chemistry to a crawl, and drying slows it further — but neither stops it. A frozen seed is not paused. It is waiting, and waiting has a cost. How fast a seedlot ages depends on the species, the quality of the seed on the day it went in, its moisture, the storage temperature, the quirks of that particular accession, and everything that happened to it along the way. Two seedlots of the same crop, stored side by side, can age years apart. That is exactly why one universal “seed lifespan” number is misleading — and why the six numbers above come with their n counts, their test year, and their warning label. ## Two vaults, two different jobs ### Australia: What was actually observed - Multi-decade viability records: 24,141 germination observations. - Storage around −20 °C in sealed aluminium packets. - Real germination testing, including 3,861 tests in 2024. - Six tropical crops, managed and retested for decades. ### Svalbard: The famous vault - Stores duplicated collections as a backup, around −18 °C. - Depositors own their boxes; only they may open them. - Svalbard does not test viability — monitoring stays with the depositors. - It is not a no-maintenance substitute for an active genebank. Every germination number on this page comes from Australia. Svalbard has no germination data of its own. ### Svalbard is running a 100-year experiment. Seed samples went into the vault in 2020 to be withdrawn and tested every ten years. The first test is scheduled for 2030; the last for 2120. A separate, older experiment has run in a Svalbard coal mine since 1986. 2020 → 2030 (first test) → every 10 years → 2120 Some of the most famous seeds on Earth are in an experiment whose final answer none of us alive today will see. ## What actually happens when a seed fails? One packet, flagged GERMINATION FALLING. Walk it through regeneration. 1. FLAG — A routine test comes back below 85%. The accession is marked for regeneration. 2. SOW — The remaining old seeds go into soil — in a greenhouse or an isolation field. 3. GROW — Plants grow and flower under watch, kept true to type. 4. HARVEST — New seed is collected at peak maturity. 5. DRY & TEST — Dried to 5–8% moisture, then germination-tested. 6. RESEAL — Sealed in a fresh foil packet, labelled, and frozen again. A seed bank does not preserve the exact physical seed forever. It preserves the genetic line — by growing old accessions out and banking a new generation. ## Run the clock Drag through the years. Packets age; at 2024 the recorded result appears. Empty years are exactly that — no test on record, not zero germination. - Mung bean: 45.5 years (Observed in 2024 2024); Beyond 2024: unknown future - Sorghum: 44.4 years (Observed in 2024 2024); Beyond 2024: unknown future - Common bean: 37.8 years (Observed in 2024 2024); Beyond 2024: unknown future - Soybean: 34.4 years (Observed in 2024 2024); Beyond 2024: unknown future - Adzuki bean: 29.1 years (Observed in 2024 2024); Beyond 2024: unknown future - Pigeon pea: 25.2 years (Observed in 2024 2024); Beyond 2024: unknown future ## Seal a packet Pick a crop. It goes into the freezer in 2026 — with its evidence-based next check stamped on it. - Mung bean: 2026 → ≈2048 - Sorghum: 2026 → ≈2043 - Soybean: 2026 → ≈2043 A monitoring schedule from published data — not preservation advice. ## The vault survives because it is opened. The popular image of a seed vault is a sealed room waiting for catastrophe. The real system is more interesting. Seeds are tested. Weak collections are grown. Plants make new seeds. Those seeds are dried, sealed, and frozen again. The point is not to preserve one seed forever. It is to keep the lineage alive. ## Take the evidence Every number on this page, with its source, in one file. https://mariuscomper.uk/how-long-do-seeds-last-seed-vault/assets/evidence.json ## Questions ### How long can seeds survive in a seed bank? There is no single figure. In the best-documented long-term dataset — 40 years at the Australian Grains Genebank — some seedlots were still above 90% germination after 44–45 years, while others of the same species fell below the 85% threshold within 20–40 years. Survival depends on species, seedlot, and storage. ### How long do seeds last in the Svalbard Global Seed Vault? Nobody can quote an observed Svalbard lifespan, because Svalbard does not germinate or test its seeds — depositors do that at home. The vault keeps backup copies at around −18 °C. Its own 100-year experiment, started in 2020, will give the first real answers from 2030 onward. ### Do frozen seeds eventually die? Yes. Freezing at −20 °C slows ageing enormously but does not stop it. Every seedlot has a trajectory; testing exists to catch the decline before it matters. ### Why do seed banks test their seeds? Because viability is invisible from the outside. A packet can look perfect and be half dead. Banks germinate samples on a schedule — roughly a third of the estimated time to 85% viability — so failing accessions get regrown while enough living seed remains. ### What happens when stored seeds stop germinating? If a test falls below about 85%, the accession is regenerated: the remaining seeds are grown out, fresh seed is harvested, dried, tested, sealed, and frozen. The physical seeds are replaced; the genetic line continues. ### Which crop seeds lasted longest in the 40-year study? Mung bean (45.5 years) and sorghum (44.4 years) produced the oldest seedlots still above 90% germination — out of 680 and 1,812 seedlots retested in 2024. Common bean reached 37.8, soybean 34.4, adzuki bean 29.1, and pigeon pea 25.2 years. ### Why did older models suggest thousands of years between tests? Early longevity equations, fitted on short experiments, predicted extremely slow decline for some species — implying sorghum could wait 2,901 years between checks. Forty years of real freezer data show decline arriving orders of magnitude sooner, so the evidence-based interval at 98% viability is about 17 years. The old number was a scheduling implication of the model, never an observed lifespan. ### Is 45 years the maximum lifespan of these seeds? No. It is the observational horizon represented by these highly viable seedlots, not a demonstrated maximum lifespan. The dataset simply ends where it ends — and most seedlots fell below the threshold far earlier. ## Sources ### The study - Baum K.J., Norton S.L., Keeble-Gagnère G., Hayden M. (2026). Safeguarding seed longevity under genebank storage. Plant Genetic Resources 24(4), 276–286. ( doi.org ) - 40 years of records; 24,141 observations; 3,861 tests in 2024; −20 °C; foil packets; 85% threshold; intervals at one-third of time to 85%. ### The vaults - Australian Grains Genebank — Agriculture Victoria; −20 °C long-term store, Horsham. - Svalbard Global Seed Vault — backup storage at −18 °C; depositors test their own seeds. - NordGen 100-year experiment — started 2020, tests 2030–2120; plus the 1986 coal-mine experiment .