Thread by @Jacobvanetten

Continuing the chat on our article in @PNASNews "Crop variety management for climate adaptation supported by citizen science" doi.org/10.1073/pnas.1… Today, I tweet about #agrobiodiversity and #climateadaptation and the limitations of agricultural science 1/16

#Agrobiodiversity is often mentioned as an important way to adapt to changing climates. But there are important challenges to realise this promise. The most important one is that this whole idea relies on massive information that climate change is actually destroying. 2/16

Farmers can only effectively use of #agrobiodiversity if they know which species/varieties are well-adapted. Climate change is making such local knowledge quickly obsolete, as climates shift. How can we generate new information quicker than climate change destroys it? 3/16

Farmers also use #agrobiodiversity to reduce climate risk growing combinations of crops/varieties. For example, two varieties that flower at different times are never vulnerable to extreme heat at the same time. Again, information is key to #agrobiodiversity use 4/16

The inconvenient truth here is that agricultural science is pretty bad at dealing with environmental variation. Crop breeding is mainly focused on "broad adaptation" or "stability" which means that varieties should do well under any and all conditions. 5/16

The analysis of environmental influences has the goal to minimize the influence of the environment as "noise" that hides the "signal" of the genetic data that the breeder is interested in. 6/16

(As a little aside, statistical analysis in agriculture builds on the work of Sir Ronald Fisher, who introduced complex randomizations but showed little love for collaborative work with farmers.)

This is not to say that breeders do not have other ways to look at environmental influences. They breed for abiotic stress tolerance. There is work on target environment characterization. But in experimental work, environment is in the year and location are dummy variables! 7/16

Not breeders/agronomists but economists have done most work on showing how variety combinations can be used to manage risk. In Mexico for example,wheat variety portfolios work well. jstor.org/stable/41960521 Variety blends are used in practice, about 10% of wheat in Kansas. 8/16

But ironically, this work relies on data from research stations, data generated to show that these varieties are broadly adapted. The variation these varieties are exposed to does not always reflect farm reality, especially in low-input environments. 8/16

The economists use multi-year data and estimate distributions from that. But ten data points is normally not enough for climate risk analysis or any risk analysis. Only a climatic analysis with a longer time period of data can a good job. 9/16

So this is where citizen science can play a role. Instead of controlling environmental variation experimentally and hoping that it hits the target, the idea is to bring that variation into the experiment. Less experimental control! 10/16

The citizen science format means that farmers do their own thing in the tests. So we get a spread of planting dates, soil fertility and so on. Instead of controlling environmental variation, recording variation. 11/16

By pulling the variation into the analysis, we can link the farmer-generated data to climate data. For example, even within the same season and location, the spread in planting dates causes a lot of variation. 12/16

Then we can link this variation to climate databases with decades of data. This allows a proper risk analysis. We have focused on a metric called "regret". We found that combinations of two varieties could be interesting in many of the locations. 13/16

Also, we can extrapolate variety recommendations across space, using the same climate data. In this way, we can really map the different niches for the different varieties. This is a recommendation map from the article for Nicaragua. 14/16

So in this type of analysis of environmental variation is no longer noise, but becomes the engine of climate adaptation. Without environmental variation within seasons and areas, we would not know how to adapt to changes over longer time periods! 15/16

Also, we cut through a lot of crop modelling and complicated analyses that would usually fill this gap between experimentation and real-world farming. Citizen science starts at the other end in building understanding, at the farm. More about farmers later! 16/16