Category: GIS

It turns out that if you keep a blog going long enough, you start to notice that you’re not so much writing new posts as filing updates to old ones. Looking back over two decades of the blog, a few obsessions definitely surface again and again, some healthy, others not so much. One of my main ones has been: Why do we still not have a proper early warning system for genetic erosion? You tell me if it’s healthy or not.

I first raised the question in 2009, in a post cheekily titled “Reporting threats to agrobiodiversity: A modest proposal”. I was inspired by ProMED, the disease-outbreak reporting system, and by a participatory mapping tool called pestMapper. Oh, and by Jonathan Swift.

The analogy seemed obvious. If a new pathogen appears somewhere, people can report it, someone can assess the information, and the reports can be aggregated to give us a picture of where the threat is spreading, or indeed where there are multiple threats. So why not do something similar for genetic erosion? What if losing a crop or variety in a particular place triggered the same kind of alarm as the outbreak of a new pathogen?

I even sketched out what a putative…wait, what shall we call it? The Global Agrobiodiversity Surveillance Programme — GASP for short? Yes, let’s go with that. What might GASP look like? Well, for a start: a way for people to report diversity that was disappearing, being replaced or becoming difficult to find; a system capable of dealing, somehow, with the inconvenient fact that the same variety might have several names, and similar issues; and maps and other visuals that could reveal trouble before it became irreversible.

And perhaps, I suggested, we could bundle in something contributors to the system might actually directly benefit from — like a way to get seeds, for example — so that it didn’t become another worthy database that nobody visited.

Worthy indeed. Seventeen years on, nobody has built the thing. Yet.

It has, however, been mildly gratifying, in a slightly irritating told-you-so sort of way, to see the idea keep resurfacing in other people’s work, dressed up in new technology each time.

For example, at one point I got excited about a Slow Food appeal for information on Italian vineyards under threat of destruction. Could crowdsourced alarm-raising like that be the closest thing yet to the system I had been going on about?

I don’t think I blogged about it, but there was also a project in Puglia that used Open Data Kit and Google Fusion Tables to collect information, including GPS readings and photographs, on vegetable crops at risk, then map and share it. Well, that sounds a bit familiar.

But there was kind of a fundamental problem. Before you can warn about genetic erosion, you need to know what genetic erosion looks like. And that’s not always easy.

This is where the notorious “75% of crop diversity has been lost” statistic comes in. I spent a lot of time on the blog — maybe too much time? — taking that claim apart, because we didn’t seem to know what the 75% was 75% of, or how either the starting point or the loss had been measured. We can’t build an early warning system out of made-up numbers. As I put it at the time, we should admit that we don’t know how much crop diversity has been lost overall, and go out and get better numbers. Fifteen years later, we do have some better numbers, indeed we have a review of all the numbers, I’m still “100% sure that 75% is incorrect”. But I still don’t know what headline number for the crop diversity we have lost to give to a journalist.

Yes, genetic erosion is annoyingly slippery. A farmer can stop growing a variety without losing the seed. A variety can survive in one village after disappearing from ten others. Two varieties can acquire the same name. One variety can acquire several names. A new variety can replace an old one without reducing the total number of varieties. The number of varieties can stay constant while their genetic complement and composition changes.

An example of this comes from two studies of sorghum in Ethiopia I blogged about in 2008. One found that farmers had stopped growing some varieties, but that the number of distinct variety names in the area had actually increased. Another found important varieties disappearing or becoming marginalised, while overall phenotypic diversity did not show a decline. The picture was one of “tremendous dynamism”: local losses, but also gains and turnover.

A snapshot tells you what exists. You need a time series to tell you what is changing.

And because farmers’ diversity doesn’t exist in isolation, monitoring those changes would need to connect with the processes that actually make diversity move. In other words, we shouldn’t just monitor the plants. We should monitor the systems that determine which plants get grown.

More recently, the technology for doing that has become rather tantalising. Satellite data on global human modification of ecosystems had me daydreaming yet again about a real-time version focused specifically on farming landscapes — “a perennial favourite here,” as I put it at the time, by which point I wasn’t even pretending to have an original idea anymore.

Then a Brainfood roundup mentioned remote monitoring of drought stress. If satellites can give us early warnings of drought, why shouldn’t we eventually be able to detect environmental and other changes that put crop diversity at risk? At this point, any spatial dataset that smells vaguely of early warning gets roped into the running list.

And it isn’t only satellites. Just a few weeks ago I pointed to FAO’s work on tools for capturing wild foods in dietary surveys: local inventories, market surveys, harvest calendars and identification aids. Why not adapt those methods to crop diversity? If they get widely adopted, you just need to mine the resulting information and aggregate it.

In “Listening for genetic erosion”, I wrote about a longitudinal study in which researchers returned to the same Australian livestock producers and landscapes after about fifteen years, using repeat photographs and interviews to understand what had changed. I was particularly taken by their use of “I-Poems”: first-person statements assembled from interviews to capture how people themselves experienced change. Why not do something similar with crop diversity? And feed the poems into the system.

A variety doesn’t usually go from “important” to “extinct” in a single bound. There is a story in between, and those stories may provide some of our best early warnings.

Put all this together and, after twenty years, I think I have a better idea of what GASP might actually look like. Not one giant database, and not a map with little red flashing lights, but a collection of imperfect sensors: threat reports, repeated field surveys, seed-system information, farmer interviews, spatial data, genebank records, remote sensing, all monitored in real time — presumably by Skynet — helping to identify places where the probability of important diversity being lost is increasing.

We keep half-finding the tool. Someone builds a citizen science app. Someone else maps a global index of something or other. Someone develops a new way of surveying crops from space. Someone puts a clever machine-learning model on top of a mountain of data. Someone creates a fancy dashboard.

And I write a post going: “Ooh. Is this it?”

It never quite is, and I go back to waiting for the next piece of the puzzle.

It’s like Twain said. Making predictions is always difficult, particularly about the future. Spotting genetic erosion is genuinely hard, especially before it happens. But I’m pretty sure we’ll eventually crack it. The technology will get better, the datasets will start talking to each other and someone will finally build something that truly deserves to be called GASP.

And then, of course, we’ll have a new problem.

Because knowing that crop diversity is in danger of disappearing is not the same as doing something about it. An early warning system can tell us that a landrace is becoming rare, that a landscape is changing, or that some other signal deserves attention. It can tell us where to look and perhaps even how urgent the situation is. But then someone has to decide what happens next.

Is the diversity already safely conserved somewhere? Should it be collected? Should farmers be supported to keep growing it? Does the seed system need strengthening? Is there someone who needs to be told about it? Or is this actually a perfectly normal bit of turnover that doesn’t require a conservation response at all?

In other words, once we finally solve the problem of seeing genetic erosion coming, we’ll have to get good at responding to it.

Which is probably the chief lesson of twenty years of discussing this on the blog. The problem keeps moving one step further downstream. First we needed to know what had been lost. Then we needed to know what was in danger of being lost. Now we need to work out how to turn a warning into a solution.

I’m quite looking forward to that problem, though. It seems like a healthier one to have.