For all but the most precise replicated field experiments, a difference of a few bushels per acre is seldom REAL.
When weigh wagons came into existence some farmers and agronomists started doing on farm experiments. But at harvest time any extra fuss does not go over big with most farmers. It sounds nice when the experiment is planned and laid down in the spring, but when the combines are rolling, the only real measure of productivity is bushels in the bin in the shortest possible time.
Now that combines have reliable yield monitors, GPS and auto steer, it is a whole new ball game. If the combine computer system allows, each treatment of an experiment can be treated as a separate field and the data could be gathered with little or no loss of precious field time.
In my opinion, a field test with adjacent treatments half a mile long and 30-feet wide is perfectly valid. But be careful how you judge the results. In almost all but the most precise replicated field experiments, a difference of a few bushels per acre is seldom REAL. Instead these differences are most often due to normal field variation, not because of a difference in treatment. I use the term “real” in place of “significant,” which is the term usually used. In a statistical sense, significant means you can reasonably attribute
the difference to the treatment applied and not just due to chance. I have always thought that significant is a poor term to use. Most interpret it as being significant to a farmer’s bottom line. If canola is 10 bucks a bushel or more, a difference of two bushels per acre is quite significant to any farmer.
But in my many years of looking at field research data, I have seen few if any experiments that can measure a difference of a few bushels per acre. I recently received the “Prairie Canola 2008
Variety Trial Test Results” and the least significant difference (LSD) between test varieties and the check was mostly between four and eight bushels per acre. One location was as low as three and one as high as 20.
With modern seeders and combines, the possibilities for perfectly valid on farm experiments are endless. I think the potential is there to gather much better data than is done in “ garden patch” agriculture on research stations.
Start this way, assuming straight combining with 30-foot header, and going twice around the field and then forth and back:
1. Pick your most uniform quarter
2. After the first two rounds are done and you have made at least one trip forth and back — keep track of the yield of each pass for about five rounds (10 passes). Take a rest until near the middle of the field and record yield of 10 more passes. Take a rest and do the same thing as you approach the end of the field.
That will give you a total of 30 yields — all from the same quarter that all received the same seed, fertilizer, herbicide, etc. It will give you a good idea of what kind of natural differences
there are in a uniform field treated in a uniform manner. I expect the variation will be much less than we commonly record on “garden patch” agriculture.
3. Repeat the experiment on a rolling quarter with much more natural and obvious variation.
With modern machines with auto steer, you should be able to reproduce a 25-to 36-foot cut consistently with overlaps and misses essentially eliminated. If your swather has GPS and auto steer, the exact same experiments
could be conducted on canola.
Once you have an idea what natural variation is for your situation, you are ready to start running actual experiments. Start simple with a “yes/no” treatment and repeat the yes/no treatment at least twice somewhere in the middle of the field. A comparison of one measurement of a yes/no experiment will seldom be reliable.
Some farmers make the yes/ no comparison very simple. You just leave out treatment X a few times at random across the field and do not mark it in any way. If you can’t find it at harvest time, then it is not likely making you a lot of money.
The big need for on farm experiments in today’s technology is to compare a single fertilizer practice to variable rate. To conduct that
experiment, I will assume you have GPS, autosteer, a 60-foot drill and a 30-foot header. I will also assume that your seeding setup is such that you can push a button or two to switch from single rate to variable rate.
All of the above technology blows the mind of this old fossil but I believe it is all possible.
For the above experiment proceed as follows:
1. Pick an area of the field that represents the differences in soil, topography, etc., that you are trying to address with the variable rate.
2. At seeding time leave at least three check strips (i. e. single rate) and mark them by GPS or other method. In the fall, combine the checks and adjacent strips of variable rate and compare the results.
You might be tempted to leave one check in the middle of the
field and compare it to the field average, but that is seldom a valid comparison.
So there are a few ideas to add to the grand technology mix that is now available. In addition to the technology, farmers now have access to private agronomic advisors who can assist with setting up and recording the results of the many experiments you may want to do.
These are exciting times in practical field research and I look forward to seeing what comes out of the technology we now have available.
J. L. (Les) Henry is a former professor and extension specialist at the University of Saskatchewan. He farms near Dundurn, Sask.