Every acre can be a research acre, said Nicole Philp to farmers at this years CropSphere in Saskatoon.
Farmers interested in testing new products and practices can create powerful data sets with a little co-ordination, said Philp, a Canola Council of Canada agronomist.
But how can you make sure you get good data out of your on-farm research trial? Here are eight tips to running a trial.
1. Keep it simple
It might be tempting to run several trials at a time, given the flood of new technology and products washing over the market. But when it comes to on-farm trials, it’s better to keep things simple.
Murray Hartman, oilseed specialist with Alberta Agriculture and Forestry, said that farm trials with several treatments have many disadvantages:
- Logistics of running several treatments.
- Time commitment.
- Risk — treatments could cost money and lower yield.
- Complicated statistics.
Results are only specific to that field in that year. Unless farmers repeat each trial at several locations, the results may not apply to all their fields.
Hartman suggested running 10 paired strip comparisons of one treatment for two years.
“That’s going to give me about 90 to 95 per cent confidence, whatever that answer is, it’s going to apply to my fields in the future. Unless you have a year like last year,” Hartman said.
Small plot trials, typically run by universities and Agriculture and Agri-Food Canada, are better suited to testing several treatments, said Kristen Podolsky, agronomist with Manitoba Pulse and Soybean Growers. For example, researchers can try 10 different seeding rates in small plots. If two or three show promise, researchers can roll them into on-farm trials, she explained.
2. Follow protocol
Podolsky said the Manitoba Pulse & Soybean Growers, along with their research partners, set up protocols for their on-farm trials.
For example, they wanted to see if soybeans needed a double inoculant if fields had a soybean history. When they looked for cooperating farmers, one condition was that the farmers’ fields had grown at least two soybean crops already.
Once they’ve developed the protocol and signed up cooperating farmers, their research partners review that protocol with farmers, said Podolsky.
3. Pick the right fields
Farmers will want to pick fields that are uniform so they know any differences in results are due to the treatment, Philp said. She also recommended leaving a check strip in an area representative of the field, rather than the field’s edge or headlands.
Hartman suggested placing paired strips beside each other, in an area where they’ll be covering similar variations. Topography and yield maps can help farmers pick areas to place the strips, he added.
4. Don't confound data
Farmers should make sure the only difference between the checks and treatments is the actual treatment, Hartman said. Otherwise they risk “confounding” the data, making it impossible to tell if differences are due to the treatment.
For example, a farmer comparing two canola varieties might seed them both at five lbs. But say the thousand kernel weight for Variety A is four grams, and six grams for Variety B. The farmer would be planting 50 per cent fewer seeds of Variety B, Hartman explained.
Hartman also warned against applying fungicide to treatments in boron trials, and skipping the fungicide treatment on the check.
“Now you get a difference in results. What is it due to? Is it due to the boron or is it due to the fungicide? You can’t separate them,” he said.
5. Take notes
Philp, Hartman, and Podolsky all recommended checking the trial through the growing season. “Don’t put it in the ground and come back at harvest,” said Hartman.
Philp suggested physically marking the check strips and treatments. “GPS is great, but it’s nice to see where your treatments are in the field when you’re going out and taking your notes,” she said.
Hartman suggested checking the trials at different crop stages, such as bolting and flowering. Hartman also said farmers should note whether conditions are fair in all the strips — for example, are there drowned-out spots?
All three recommended collecting in-season data, such as hail, frost, excessive heat, and humidity. Such information can explain yield increases or decreases, Podolsky said.
The Canola Council of Canada has data collection sheets online, to record everything from seeding to scouting to harvest. The sheets also calculate the treatment’s profitability, “because at the end of the day, that’s what it comes down to, is using these products to add profitability to your farm,” said Philp.
Farmers can download those data collection sheets from the Canola Council of Canada website. (see "Related" link below)
Hartman said farmers may also want to sort data based on criteria such as soil type. “If you have all that information, you can sort it out later and say, ‘Where is this response happening?’”
6. Harvest the trial
“In my experience, the most patience is needed at harvest. Especially when it’s a poorer year at harvest and you’re way behind,” said Hartman.
Already hit your stress limit during harvest? Then you’d best skip the on-farm trials next year, Hartman said. Otherwise, you increase your chances of unraveling on family or hired hands during harvest, he said.
Philp recommended harvesting the entire trial the same day. Try to harvest at the same speed throughout the trial, she added.
All three suggested using a calibrated weigh wagon to measure yield. “We’re not quite confident enough in yield monitor technology yet,” said Philp.
7. Analyze the results
Before doing any data analysis, farmers should make sure they followed protocol at all trial sites. The Manitoba Pulse & Soybean Growers had to discard data from one site, Podolsky told growers. The co-operating farmer realized his field had only seen one soybean crop instead of the required two crops, she explained.
That made a big difference, as the yield dropped, she said. The yield drop underlined the importance of double inoculation in new soybean fields, she added.
Farmers looking at results will likely notice yields bouncing around within the same treatments, Hartman said. So how can growers figure out whether their results are due to the treatments or just flukes?
That requires figuring out the statistical significance of results. Results that aren’t statistically significant could be due to chance rather than the treatment.
Hartman recommended analyzing results from paired strips using a paired t-test. Farmers can find t-test calculators online (see the "Related" link below for one online example). Once they plug in the numbers, the software will calculate the probability of the differences between treatment and check being due to the treatment.
Farmers can select a paired t test using a calculator like this one at graphpad.com. The calculator will state whether or not the results are statistically significant.
Once farmers have chucked results from any sites that didn’t follow protocol, and determined the results are statistically significant, they can apply the economics, Podolsky said.
8. Share and collaborate
It’s quite possible for many farmers to conduct their own trials these days.
“But just remember, you don’t have time to make all the discoveries and the mistakes yourself,” said Hartman. “That’s why it’s so helpful to collaborate.”
Instead of doing 10 paired strips on one farm, two farms could split the workload, by doing five each, Hartman said. Farmers could also work together to trial different products and share results, he added.
Hartman added extension people could “provide some useful discussion” during the planning and data analysis. He also likes to see on-farm trial results, to have in the back of his mind when looking at scientific stuff.
Farmers can also check with their grower groups about getting involved with research projects. The Canola Council of Canada has been running field-scale trials through the Ultimate Canola Challenge. The Manitoba Pulse & Soybean Growers also runs field-scale trials with co-operating growers and other research partners.