On-Farm Field Trials - Plot Protocol
years, corn producers have been participating in (or conducting) field trials
to provide information that would make production practices more sustainable
and profitable. In some cases, farmers have worked with university or government
personnel to establish complex, replicated experiments. In a large number
of situations, strip trials have been established. Generally these strip trials
are simple, field length strips. As well, they are unreplicated that is, each
variety, fertilizer, etc., appears only once in the trial.
Controversy has existed for years over these two approaches to gathering information. Those with a leaning towards scientific research reject the notion of plots not replicated. They claim you have no way of telling whether the yields are a result of position within the field (e.g., down the strip that received extra manure two years ago) or the actual treatment imposed. Replication within the fields is the only way to have confidence in the results, they say.
On the other hand, small, replicated plot research has often been criticized for relying on plots that are not representative of normal field conditions. In addition, small plot research has generally been conducted in a very small number of locations, and so the adaptability of the results to other environments has been questioned. Field length strips unreplicated, but conducted across many environments is the only way to have confidence in the results, according to replicated plot research critics.
Fortunately, the marrying of these two positions is possible when we consider a few basic ideas.
In-field variability should be a concern to all those involved in gathering crop production information. Replication can often be unmanageable and time consuming. A reasonable alternative to replication is planting a check strip down the left, centre and right side of each trial. In a hybrid trial these check strips would, of course, be planted with the same hybrid. As the results are obtained from each strip, producers can simply evaluate the variability within the check strips. If these three yields are relatively close - within 10 to 15 per cent of each other - then we can have some confidence the data will contribute to our understanding. If the check strips yields vary greatly, the results for the entire test have little value and should be discarded.
A friend tells you he is sure the cashier has just handed him a two-headed coin, but he won't allow you to look at it. He decides to flip the coin repeatedly and challenges you stop him when you are 100 per cent confident that it is, in fact, a two-headed coin. After two or three flips (all heads), you realize it is still quite possible that chance alone has produced the result. However, after some number of replications (5, 10, 15, etc.), you become confident it is two-headed, even though you realize there is still an outside possibility that it's a normal coin. Replication is necessary to have confidence in test results. In some research, this replication occurs at specific sites with many small plots. In strip trials, it is often accomplished by having a large number of sites.
Interpretation of results
Variety selection is most suited to unreplicated strip trials. It is generally easiest to generate a large number of variety strip trials (replication) and in most cases, little other information is gathered beyond yield data. Our goal is to pick a winner; we are somewhat less concerned about why it won. However, if we are interested in comparing two types of fertilizer, things change considerably. First, it generally becomes more difficult to find co-operators for these types of projects, so our total data base gets smaller and we need to consider replicating at each site to build confidence in the results. As well, if accurate soil tests are not taken from each strip, it becomes difficult to interpret the results and predict when a grower might expect a yield advantage for Fertilizer A over Fertilizer B.
In Ontario, the Ontario Corn Committee conducts small-plot research to evaluate hybrid performance. The value of these performance trials lies in the fact that they are replicated, field variability is carefully assessed and the vast majority of available hybrids for a particular heat unit rating are present in the same test. The main weakness of these trials is that a range of hybrids is only tested at three or four locations. So while we may be confident that a certain number of hybrids are superior at these particular sites, we have considerably less confidence in deciding whether Hybrid A will outyield Hybrid B across a wide range of environments.
In variety strip trials, these head-to-head comparisons can be made with greater confidence, provided that a large number of trials actually have the two hybrids present that you want to compare. In the past decade, seed companies have been big promoters of this type of analysis. Recently, the Ontario Soil and Crop Improvement Association, with the assistance of Agriculture and Agri-Food Canada, launched its Ontario Field Trials Online, where strip trials conducted by farm co-operators from across the province can be entered into a common data base. The potential is excellent for this data base to complement the performance trials and to assist in hybrid selection.
The process of making better decisions begins now as you plan your on-farm trials for 2000.