HYBRID selection gets a HELPING HAND

A POWERFUL COMPUTER PROGRAM ALLOWS DATA FROM RESEARCH PLOTS AND FIELD SCALE STRIP TRIALS TO BE COMBINED.THE RESULT IS A BETTER BASIS FOR HYBRID COMPARISONS

Most corn growers are keenly aware that selecting winning hybrids is a key step in trying to squeeze profit out of corn produc-tion. Potential losses are large if you
make the wrong choice. It is not unusual to see a spread of 40 to 50 bu./ac. from top to bottom within a hybrid performance trial. Compare that to tillage or fertilizer decisions where 10 to 15 bu./ac is usually as bad a mistake as you can make. Selecting the best hybrids based on data available over the past several seasons has been challenging to say the least. Yield stability was more elusive than normal. Sharp fluctuations in weather conditions caused some hybrids to yield well one year and poorly the next. Harvest moistures were also difficult to interpret. Long growing seasons in 1998 and 1999 tended to compress hybrid trial moistures into a narrow range at the lower end of the scale, while disease pressure in 2000 caused some hybrids to die prematurely, distorting their harvest moisture values. In general, trying to make decisions without drawing on all available data was a fairly risky adventure.

Performance trials	Strip trials
Small plot	VS.	Field-length strips
Complex design	VS.	Simple design
Replicated	VS.	Not replicated
Few locations	VS.	Many locations

When it comes to hybrid selection, you have essentially 2 distinct sources of data in Ontario. These are the performance trials (run by the Ontario Corn Committee) and strip trials (run by seed companies, growers, farm groups, etc.). Each system has distinguishing characteristics and potential strengths and weaknesses.

The seed industry has made a strategic shift over the years to more field-scale strip trial sites with fewer replications. These strip trials are gaining acceptance as they expose hybrids to more environ-ments and offer more opportunities to assess a hybrid's stability and adaptability.

But the performance trial system continues to be a key tool for growers in their selection decisions. Although performance trials are run at a limited number of locations each year, they are replicated and conducted under controlled conditions to ensure accuracy. In addition, they have the advantage of comparing a very large number of hybrids (more than 100) all at the same site.

Table 1. Occurrence of individual hybrids across various strip trials
HYBRID
Strip Trials	A	B C D	E	F G H	I J	K
1
2
3
4
Actual yield value data per site

Performance trial reports lend themselves to complete reports in which all hybrids in a particular heat unit range can be listed and ranked simultaneously. In contrast, strip trial data sets are incomplete as not all varieties are evaluated together at individual sites. This restricts direct head-to-head comparisons to trials where varieties are present at the same sites. As an example, in the trial illustrated in Table 1 you can make 3 head-to-head comparisons between hybrids D and E, only one between E and F, and none at all between F and G.

Weikai Yan and Tony Hunt at the University of Guelph's Plant Agriculture Department have adapted computerized analysis methods to compare performance and strip trial data. They started with the Ontario winter wheat data, which was complete for the 1998, 1999, and 2000 seasons for both strip trials and performance trials.

Table 2. Occurrence of individual hybrids across various strip trials
HYBRID
Strip Trials	A	B C D	E	F G H	I J	K
1
2
3
4
Predicted yield value data per site
Actual yield data per site

Using powerful statistical software, they compared yield performance for each year and trial system (performance and strip). The statistical procedure cal-culates an overall average for all data present. Then it uses the actual data available to calculate predicted yield values for all missing data. It essentially fills in the blanks in Table 1 with predicted yield values, to produce Table 2.

Since different varieties actually appear in the data set in varying num-bers, and the statistical program predicts yield values using the actual data available, it also computes an estimate of the error (variability) associated with each predicted yield value. Using a combination of the predicted yield for a variety and its error estimate, the system creates an overall ranking from a strip trial data set that was essentially "unrankable". The rankings can then be presented as being equal to, significantly worse than, or significantly better than the overall average. Yan and Hunt found that the 2 systems (performance and strip trials) for generating comparisons in winter wheat could be made complementary. Analysis of the wheat data also showed that a variety needed to appear in at least 20 strip trials within the database before it could be ranked with confidence. The amount of strip trial data for corn hybrid comparisons in Ontario is huge, but until now we couldn't make efficient use of all the numbers, which
are summarized and presented in part by the seed companies operating in Ontario. Growers currently use this data to track head-to-head comparisons between 2 hybrids at a time. In addition, the Ontario Soil and Crop Improvement Association is constructing an Internet site where all data held by all companies or individuals can be compiled. This process is underway and the web site exists at www.trials.ontariosoilcrop.org/default.asp. Head-to-head comparisons are possible between 2 hybrids by working through the menu on this web site. This strip trial database could be further enhanced by using Yan and Hunt's statistical procedure to produce a "performance trial-like" table from amalgamated strip trial data. This would provide an indication of overall yield potential of all hybrids within a given heat unit range.

More work is being done to validate the procedure. Consultation and feedback will be sought from all seed companies and other stakeholders to identify issues and develop a strong and stable tool for growers to use in conjunction with other information to make the best possible hybrid choices. If we go to the trouble of measuring yields and collecting data, it makes sense to work toward a system that considers all available numbers and presents findings in a usable format. Growers can enhance strip trial data by designing and running trials properly, and including a good cross-section of both "proven" and "new" hybrids for evaluation.

*Greg Stewart and Ian McDonald are crop technologists with the Ontario ag ministry

Back to Top