Mapping out the problem of crop mapping

Time to celebrate the fact that I’ve spent a surprisingly large part of the life of this blog trying to answer what sounds like a simple question: Where are crops actually grown?

In Where do bananas grow anyway?, for example, I compared several supposedly authoritative maps of banana distribution and found that they disagreed in all sorts of interesting ways. That was followed by Yes, we have more banana distribution data, which summarized the discussion that followed that post and tried to tie up some loose ends. I dread to think how many of those datasets are still available.

Why bother, you may ask? Well, because knowing where a crop can be found is necessary if genebanks want to plan gap-filling collecting: map where the crop is grown, subtract what the genebanks already have, and explore the gaps. Alas, the reality was rather less tidy, hence Gap-filling may be harder than we thought. Tell me about it.

I kept returning to the problem, though. In The geography of black rice, I wrestled with the practical business of joining passport and characterization data just to get diversity onto a map. In Ground-truthing SPAM, a global crop-distribution model confidently put crops on my mother-in-law’s farm that she and her neighbours didn’t actually grow. Bambara groundnut produced another set of contradictory maps.

Years later, the tools were a bit better: I explored Afghan wheat in QGIS, and overlaid Kenyan bean accessions on a new crop atlas.

And by earlier this year Colin Khoury was able to summarize the whole historical trajectory of gap analysis in a guest post. A trick he then repeated elsewhere.

Looking back over all these endeavours, what strikes me is how often the answer seemed to be just one step away. A better crop map, better coordinates, a more complete genebank database, a cleverer GIS analysis. And each time the new thing helped, but also exposed another problem: the crop map was wrong here, the accession wasn’t georeferenced there, or the apparent gap didn’t mean what we thought it meant. I guess that’s science.

A lot has changed since I started writing about this problem, but the basic challenge remains. We want to know where diversity is, where it might be missing, and where it is worth looking for more. Getting all the relevant data into the same map is only the beginning, but it needs to be done. There’s more data now, and making maps is easier to do. But, unless you’re a GIS expert, it’s still way too hard.