Policy

September 25, 2026September 27, 2026

Joining the dots on food plant diversity

We have access to more kinds of food plants, yet the world’s diets and production systems have never been more alike. That is the paradox at the heart of Food plant diversity in a rapidly changing world, an unusually broad review by Colin Khoury (very much friend of the blog) and colleagues. It has all the latest data, but it also does several other things that make it more than a conventional state-of-the-art paper.

For one, it brings together diversity at several levels that are too often discussed separately: species, varieties and genes, of course, but also the diversity of diets, farming systems, foodways and the knowledge associated with them. It traces the movement of cultivated food plants over the past 12,000 years alongside the much more recent trend towards homogenisation. The result is the weird world we live in today: more kinds of crops available globally, but less difference between what people in different places eat.

Colin and friends also take a hard look at what we don’t know. We have surprisingly little evidence for how much food-plant diversity has actually been lost, how vulnerable today’s production systems are because of increasing uniformity, or what is happening to less visible forms of diversity, such as nutritional and chemical variation. The authors call for better monitoring, including an early-warning system for crop vulnerability. Now where have I heard that before?

And the paper connects the dots between conservation approaches that all too often live in isolation: genebanks, botanic gardens, community seed banks, on-farm initiatives, protected areas, breeding programmes and markets. Genebanks are all well and good, but diversity also needs to keep going in fields, landscapes, cuisines and communities. Again, not particularly new, but it’s good to have it repeated, brought up to date, and placed in the broadest possible context.

A couple of those dots are discussed in more detail in a companion opinion paper, Enhance food plant diversity conservation and use through greater coordination between genebanks and botanic gardens, which asks what might become possible if two of the world’s major plant conservation communities worked a bit more closely together. Which must be possible because they are nicely complementary, so neither should fear the other as a competitor.

The review paper maps the problem. The opinion piece explores a possible route towards a more joined-up conservation system. Which leaves us with its own small paradox: conserving food plant diversity requires flattening some of the differences within a remarkably diverse conservation community. Happy Crop Diversity Day 2026.

September 23, 2026September 21, 2026

Nibbles: China genebank, China blueberries, PhilRice, Italy olive genebank, GIAHS, Mangelwurzel, Papaw, Bean AI

China is building its germplasm collection at the National Frigid Zone Crop and Soybean Germplasm Resources Mid-term Bank in Harbin. Cool.
Driscoll’s helped build China’s blueberry industry, then discovered that knowledge, infrastructure and genetics have a habit of escaping. Not so cool.
Meanwhile, the Philippines wants its rice collection to hit 21,000 varieties by the end of the year. IRRI unavailable for comment.
Italy is establishing a major global olive genebank. Unclear where it fits in with other major global olive genebanks, but some redundancy is good.
FAO showcases Globally Important Agricultural Heritage Systems, where crops, landscapes and traditional knowledge have survived together. Like olives in Italy, I guess.
The mangelwurzel is making a comeback, because apparently forgotten root vegetables just need better publicity, and not a fancy GIAHS.
Ohio growers are trying to save the pawpaw just as Americans are discovering it, partly because of better publicity.
BeanGPT is bringing AI to breeding and farming beans. Maybe it can help the mangelwurzel and pawpaw too?

September 21, 2026September 20, 2026

Brainfood: From data to action

A three-decade-long analysis shows rising threatened species and responsibility gaps in global supply chains. Data says that we must make responsibility for biodiversity traceable from production to final demand.
Development of an evidence-based integrated germplasm prioritization framework for conservation and breeding. Data says: here’s how to extract some defensible conservation and breeding priorities from this large collection.
A methodological framework for the standardised evaluation of olive genetic resources: GEN4OLIVE harmonized protocols. Data says that we should standardise phenotyping across countries to make olive germplasm more useful in breeding.
High-throughput phenotyping of Bambara groundnut [Vigna subterranea (L.) Verdc.] seed morphometrics using videometer. Data says that we should use multispectral imaging to really speed up seed characterization and breeding in Bambara groundnut.
Predicting the cross-continental spread of the cassava brown streak disease epidemic in sub-Saharan Africa. Data says that we must strengthen surveillance and delivery systems for clean planting material before cassava brown streak disease reaches Nigeria.
The Sustainability of Robusta Coffee Under Global Change—A Review. Data says that we must match robusta production to genuinely suitable environments, plus invest in lower-input cultivation and alternative coffee species.
Borassus Akeassii Facing Societal and Environmental Challenges: Case Study of the Gouin Society of Southwestern Burkina Faso. Data says that we need to conserve this multipurpose palm once central to Gouin livelihoods together with the local knowledge and practices that make it useful.

September 17, 2026September 20, 2026

What seventy years of adoption research can teach genebanks

There’s a paper out that’s worth getting to grips with if you are interested in genebanks having an impact, even though it never actually mentions genebanks. David Pannell’s Reflections on Adoption of New Agricultural Practices distils more than 70 years and 5,000 studies of why farmers take up, adapt, abandon or reject innovations, be they new practices or technologies. It’s not specifically about plant genetic resources, but think of bringing back lost landraces, releasing a new variety or giving a neglected crop a second chance as an innovation and some of Pannell’s conclusions will seem very relevant to genebanks and their users. I’ll just pull out what I think are the top lessons.

Perhaps the main point Pannell makes is that testing something out isn’t adoption.

That sounds pretty obvious, but agricultural projects routinely give farmers something to try and then apparently count the resulting trial as adoption. Pannell points out that what is in fact being measured may be nothing more than farmers trying to figure out whether the innovation is good enough to keep using once the project support disappears.

The farmer planting that accession from the genebank or that new variety from the breeder is evidence of interest, not adoption. The more meaningful result is whether she is still using it, without being paid or encouraged to do so, several seasons later.

And speaking of trying things out, how easy that is to do matters for adoption, and it varies enormously among innovations.

Pannell notes that a new crop variety has an important advantage over other agricultural innovations: farmers probably already grow the crop. They can try the new variety right alongside the older ones without redesigning their whole farming system. In contrast, innovations that require systemic change can take decades to reach widespread adoption.

So, a package of sorghum landraces, or a new disease-resistant sorghum variety, being presented to sorghum farmers is one type of adoption problem. Promoting an opportunity crop that most farmers have forgotten how to grow is quite another. That may require new knowledge, equipment, markets and consumers, as well as a convincing reason to grow it. It’s bound to take longer.

Next, and perhaps more surprisingly, extension cannot compensate for a weak advantage.

Pannell is pretty clear: extension services can speed up awareness and evaluation early on, but they cannot manufacture adoption if and when farmers discover that an innovation isn’t sufficiently good.

That puts the emphasis back on the thing being offered. Does the new variety really outperform what farmers already have? Does the old landrace being revived have drawbacks? Does the neglected but really nutritious opportunity crop provide a return that compensates for the risks and additional work involved? If the answer is no, the best communications operation in the world won’t fix it.

Finally, every innovation is unique.

Pannell shows that different innovations being considered by the same farmers, and the same innovation being considered by different farmers, can have radically different drivers of adoption. Farmers aren’t a uniform population, and neither are innovations.

That is a useful reality check. The story of one neglected crop becoming an agricultural opportunity will never be repeated exactly. Each innovation has its own combination of advantages, constraints and players. There is no generic path to impact, though there may be recurring questions.

So what does this mean for genebanks trying to have — and measure — impact?

I’ll go with four fairly simple things. First, don’t treat germplasm distribution or a farmer field day as the end point: follow up and find out what breeders actually used, and farmers actually kept using. Yes, I know that takes time and resources, but it needs to be a routine part of the genebank’s work. Second, design projects tightly around the particular crop diversity involved, rather than assuming that the same adoption machinery will work across crops, countries, ecologies, whatever. Third, spend at least as much effort understanding why farmers might want the diversity in the first place as on getting it in front of them. If an advantage isn’t there, no amount of factsheets and posters and field days will rescue it. Finally, and maybe most importantly, remember that non-adoption isn’t always failure. Genebanks exist in large part for option value: material nobody wants now may be exactly what everyone wants in 2060. Plan the project so that even a failure to adopt teaches you something.

I’ve written plenty of posts here celebrating some interesting new diversity getting into farmers’ hands. I’ve written far fewer checking back to see what happened next. But that’s the more important story to tell.

September 9, 2026October 4, 2026

Missing use, or missing middle?

Genebank data seems to be in the air. Just in the past few days, I pontificated about descriptors, Mike Jackson has reminisced about the early days of rice molecular markers and CGIAR announced a big new initiative.

More data is always good to have, I’m sure. But, but, but… Genebanks are underused and more data will solve the problem? We’ve been saying something much like this for a long time now. Leave aside for the moment the question of whether CGIAR genebanks really are “underused.” Underused by whom, according to what metric, compared to what? But let’s just assume for the sake of argument that they are, indeed, underused. What if… Hear me out: what if that problem cannot be solved just by making genebank databases bigger, deeper and easier to search?

Ok, that will need some unpacking. But remember as you read this that I’m not saying that genebanks working with others to generate and share more data on nutritional quality or disease resistance is a bad thing. Just that it’s not enough. Never has been. Probably never will be.

Continue reading “Missing use, or missing middle?”