Nuts and bolts of on-farm experiments: Neighbours across a control plot provide answers by Cathy Nicoll
DOING AN on-farm trial is more than just planting a test strip in the back paddock, or picking a few treatments and sowing some plots.
According to Victorian TOPCROP Coordinator Margaret Evans, problems such as paddock variability, seasonal variability, changes across a district, and the ubiquitous 'what would have happened anyway' all serve to confound interpretation of anything but a well-designed trial.
While many scientists prefer replicated small plots for valid results, they can also give deceptive results with factors such as the improved airflow of small plots helping control leaf diseases that may be serious in the paddock. Growers have also found that fully replicated plots can be time-consuming and unsuited to use with farm machinery. Small errors in planning can give data that are difficult to interpret.
Large plots can bring their own problems, with paddock variability in some cases being greater than the effects of any treatment if sites are not chosen carefully. Painstaking research from the 1930s in Iowa and Werribee looked for an optimum plot size for experiments on cereal nutrition and varieties and, according to John Angus of CSIRO Plant Industry, the work has not been superseded.
"This work showed errors due to soil variability increased as plots got larger, but at the same time, sampling errors increased with smaller plots," says Dr Angus.
There are some new solutions, like the ' nearest neighbour' on-farm trial design, that can help find the middle ground between rigorous research and on-farm relevance. By sowing plots that are at least one or two seeder-widths wide, the design fits in with the size and manoeuvrability of farm machinery.
The plots are sown next to each other in a strip and the same control treatment is applied to every third plot. The advantage of this is that it allows statistical analysis of the results in a way that takes into account paddock variability. A poorly selected site with high paddock variability will show up in the variability of the control treatment.
A WA version of the same design reduces the workload for growers by setting the control treatment to be the same as the rest of the paddock, and sowing these plots with the rest of the paddock.
The 'nearest neighbour' design was used as early as 1999 in Victorian on-farm trials, and last year Dr Evans saw the design used in the statewide canolanitrogen trials. "Using a consistent approach with the same basic treatments across the entire State gave us valid information that growers could use to make informed canola management decisions," Dr Evans said.
Many growers had enough confidence and enthusiasm for what they were doing to add a few extra treatments of their own, which greatly increased the amount of information that came out of the trial.
Several WA grower groups use the method to evaluate pre-release wheat varieties. Western Australian Department of Agriculture research agronomist Kevin Young says that using 'nearest neighbour' has given growers involved with the South East Premium Wheat-growers Association (SEPWA) program the ability to select such a uniform site that iheir on-farm trials give results that are consistent with small research plots.
"I no longer see growers showing their trial results at meetings, only to say 'well, don't take any notice of this' because they knew it was unreplicated," he says. Instead they have a set of locally relevant data that can't be ignored (see www.sepwa.org.au).