http://www.nytimes.com/ref/business/20070527_COKE_GRAPHIC.html

Two related thoughts:

(1) If mobile phones can be equipped with weather sensors (temperature/humidity/pressure) and if the data can be collected and mapped at no coast to the phone owner, this might provide an accurate and cheap weather forecasting system that is complementary to satellite data.

(2) Does Pepsi occupy a different ecological niche in Africa? We need to know the distribution densities of competing players in the African market. Would it be worthwhile to consider bottled water distribution as well, to reach urban farmers and gardeners?

15 years of maize trial data from around the world….

In the future, I think much breeding should be crowdsourced to farmers.

Forget about Fisherian statistics, “stability” (without reference to the environmental conditions themselves!) and that kind of nonsense .

This is my recipe:

Step 1. Stick two little bags of crop seeds with unique codes on every bottle of Coke in Africa. Add a little explanation on the bottle and also spread the message through local radiostations.

Step 2. Farmers grow the seeds and then send their opinions with their mobile telephones, sending just the code of the winning seeds. Organize a lottery to provide an incentive. Also get the location of every SMS message from the GSM network (or a GPS coordinate — with newer phones).

Step 3. Link the data with environmental data (remote sensing, interpolated wheather station data, soil data). Delete double entries (the codes are unique, and both varieties cannot be best at the same time.) Then separate the signal from the noise using machine learning. Get info on variety performance in space.

Step 4. Based on the outcomes of Step 3, send variety suggestions for seeds to the mobile phones of all farmers who participated (and suggesting where they may buy them).

Step 5. Do the lottery and send out your free boxes of Coke. Take pictures for the newspaper and agro.biodiver.se.

Step 6. Define a new group of varieties and where they should be evaluated based on the outcomes of Step 3.

Step 7. Go to step 1.

If the breeding program has a robust multi-location trial program, isn’t phenotyping the most expensive part of a breeding program? Especially for public research that depends on NARS (and in places with less commercial interest)? The seed, DHL, the agronomist to conduct the trial, all the infrastructure of the experiment station – the total costs for all involved it is not insubstantial.

I know of a breeding program that is doing great work, but funding constraints meant they had to decide what to cut. They scaled back phenotyping. Because their NARS partners would need too much support. This is especially true when you are evaluating traits other than yield — traits that need special trial management (e.g. drought tolerance) or a high level of phenotyping skill. But, man, with their fancy biotech labs, they sure can genotype!