The Global Seed Vault and Why It Is Not Enough

The Svalbard Global Seed Vault was formally opened on February 26, 2008, the product of a collaboration between the Norwegian government, the Global Crop Diversity Trust, and the Nordic Genetic Resource Center. The political timing was deliberate. The year before, a landmark report from the FAO had documented the accelerating pace of crop genetic erosion globally. The vault was, among other things, a statement — a visible symbol that the international community recognized the problem and had taken action.

Symbols are necessary but not sufficient. To understand why the vault is insufficient, it helps to understand what gene banks in general can and cannot do.

The global gene bank system consists of approximately 1,750 collections worldwide, including national gene banks, international centers associated with the CGIAR (Consultative Group on International Agricultural Research), and regional collections. Together they hold an estimated 7.4 million seed samples. Of these, many are duplicates — the same variety held in multiple locations. The actual number of distinct genetic accessions is significantly lower, and the proportion of global crop diversity that has been collected at all is unknown but certainly incomplete. Swaths of crop diversity in West Africa, Central Asia, the Amazon basin, and highland Southeast Asia have been inadequately surveyed and collected.

Svalbard serves as a backup for this system — a duplicate copy of samples held in national and international gene banks. If a gene bank is damaged by fire, flood, conflict, or institutional collapse, its samples can in principle be recovered from Svalbard. This is valuable. Since 2008, there have been multiple gene bank failures — in the Philippines, in the United States, in several African countries — where Svalbard backups would have been, or were, important.

But backup of a system that is itself incomplete is not the same as comprehensive conservation. The gap between what has been collected and what exists — or existed, before it was lost — is not recoverable through Svalbard or any other vault. The vault can only hold what was collected. It cannot reverse the extinction of varieties that were never sampled.

The condition of seeds in long-term storage is another underappreciated issue. Seeds are not perfectly stable. Even under optimal cold storage conditions — minus eighteen degrees Celsius is the Svalbard standard — viability decreases over time. The rate of decrease depends on the species, the initial moisture content, the packaging, and various biochemical processes. Some species store well for decades or centuries; others degrade significantly within ten to twenty years. Gene bank management requires periodic regeneration — growing stored seeds in field conditions to produce fresh, viable samples. This regeneration is expensive, labor-intensive, and scientifically demanding. It requires agronomic knowledge specific to each variety. Many gene banks lack the resources to maintain adequate regeneration schedules, meaning that portions of their collections are in declining viability and are at risk of loss even in storage.

The CGIAR gene banks have estimated that a significant fraction of their collections require urgent regeneration. The cost of adequately funding global gene bank maintenance is estimated at several hundred million dollars per year. Actual funding is substantially lower. The gap between what is needed and what is provided is itself a form of genetic erosion — a slow loss of viability that does not make headlines.

The representation problem — the fact that gene bank samples do not capture the full within-variety diversity of traditional landraces — has been studied most carefully in crops with high outcrossing rates, such as maize and rye, where genetic variation within a population is very high. A sample of a few hundred seeds from a landrace maize variety carries only a fraction of the allelic diversity present in the full population. When that variety is stored from a single collection event, the genetic diversity of the stored accession is permanently narrower than the diversity that existed in the field. This matters for breeding programs that rely on gene bank material to introduce novel traits. The narrower the representation, the less likely that the specific allelic combinations needed are present in the stored sample.

The knowledge problem is, in some ways, the most significant. Plant genetic resources are not fully separable from the human systems that maintain them. The performance of a traditional variety in a specific environment depends on management practices that have co-evolved with the variety. The timing of planting relative to rainfall patterns, the density of planting, the soil amendments used, the harvesting and threshing practices that affect seed quality — all of these are elements of a technology that accompanied the seed. This is ethnobotanical knowledge, and it disappears with the communities that hold it. Linguists estimate that a language becomes extinct approximately every two weeks. The agricultural knowledge embedded in those languages is not separately catalogued.

The institutional infrastructure surrounding seed use is equally vulnerable. ICARDA's withdrawal from Svalbard in 2015 was ultimately successful because ICARDA as an institution survived — in reduced form, relocated to Morocco and Lebanon. Researchers with the skills to work with those seeds were still employed. The breeding programs that would use the retrieved material still existed. In a scenario where the institutional disruption was more comprehensive — as could happen in a large-scale conflict, a pandemic affecting research institutions, or a collapse of international funding — the retrieved seeds would be available but the capacity to deploy them would not. Seeds without plant breeders, agronomists, and extension networks are potential, not food.

There is also the question of access. The International Treaty on Plant Genetic Resources for Food and Agriculture establishes a multilateral system of access and benefit sharing for forty-four major food crops and twenty-nine forage crops. Accessing material from the gene bank system for these crops involves a Standard Material Transfer Agreement — a contractual arrangement that governs the terms of use. For crops not covered by the Treaty, access can be more complex, requiring bilateral negotiations that can take years. In a genuine food security emergency, this legal architecture creates delays. The material exists; the path to using it is not always clear.

The alternative architecture — which Svalbard was partly intended to support but cannot substitute for — is a distributed system of active seed saving and community gene banks. In this model, genetic diversity is maintained in the same communities that use it, updated continuously through natural selection and farmer selection, accompanied by the knowledge needed to cultivate it effectively, and accessible without institutional intermediaries. This model is described further in the companion article on decentralized seed saving. What needs to be understood here is the relationship: the vault is not a replacement for living diversity. It is, at best, a last resort when living diversity has already been lost.

The planning implication is clear. Investing in Svalbard and the gene bank system is necessary and worth doing. But it is not an alternative to maintaining living seed diversity in fields. Every dollar spent on cold storage that substitutes for — rather than supplements — investment in community seed systems is a misallocation. The vault buys time. Living seeds in farmers' hands are the real insurance. Both are needed, but the second is primary.

The Svalbard vault has a permafrost foundation that has already experienced one flooding event as the mountain's permafrost thaws — a direct consequence of Arctic warming. The infrastructure requires continuous monitoring and intervention. Even the vault's physical premise — the stability of Arctic cold — is no longer a given. The backup has vulnerabilities. The living seed is the original.