On-Farm Trials

Excerpt from Agronomy Guide for Field Crops (Chapter 1)

Order OMAFRA Publication 811: Agronomy Guide for Field Crops

Table of Contents

1. Introduction
2. Site Selection to Reduce Variability
   • Trial Positioning
   • Orientation of Replications
   • Trial Design
   • Treatment Selection
3. Data Collection
4. Data Assessment
   • Strip Trial Variability Example
   • Example of Differences in Yield Between Two Treatments (Hybrid A and B)
5. Updates on On-Farm Trials

Introduction

On-farm testing of new production techniques provides an opportunity for growers to evaluate new methods under controlled conditions without investing a large amount of acreage or costs. In this manner, new techniques can be investigated for adaptability to each individual's farming situation. The following information will provide a framework for making these evaluations in comparison to the present methods and techniques of production being used.

Standardization in on-farm plot design and set-up will allow for simple comparison of new technologies with those presently used on the farm. Standardization of plots will aid implementation, data collection and analysis, and will hopefully reduce the variation (error) that can occur with field scale plots.

| Top of Page |

Site Selection to Reduce Variability

Trial Positioning

If a field site has variations in elevation, other topography issues or obvious soil-type variation on visual inspection, position the plot to avoid these areas (see Figure 1-2, Orientation of Treatments to Field Topography).

Figure 1-2.
Orientation of Treatments to Field Topography

All plots should be located in fields where the intended plot area was treated as a single field unit in previous years (see Figure 1-3, Locating On-Farm Trials in Relation to Previous Year(s) Cropping Practices). This would include the same crop, herbicide, fertility, manure application programs or other agronomic practices. This will ensure that no underlying variation from previous years can cause inconsistencies in the plot results this year.

Figure 1-3.
Locating On-Farm Trials in Relation to Previous Year(s) Cropping Practices

| Top of Page |

Orientation of Replications

Where possible, plots should be located perpendicular to known sources of potential trial variation such as tile lines, primary tillage dead furrows, gullies or low spots that cut across the plot area, as shown in Figure 1-4, Orientation of Trial Treatments in Relation to Tile Lines or Other Potential Sources of Variation. The intention is to reduce variation over the whole plot and where variation can not be removed, to at least have it occur equally across all treatments. Otherwise variation may affect only one or a few treatments in a trial leading to incorrect conclusions from results. It is understood that this alignment of plots may not be possible.

Figure 1-4.
Orientation of Trial Treatments in Relation to Tile Lines or Other Potential Sources of Variation

Where possible, apply herbicide, fertilizer or other inputs to the plot area perpendicular to the direction of treatments, as shown in Figure 1-5, Direction in Which Inputs Should Be Applied on Plots to Reduce Inherent or Applied Variation. Where not practical, the application of inputs to the whole plot area should be done as accurately as possible. This will help reduce the potential of introducing variation into the plot because one or a few treatments received more or less fertilizer or a high or low rate of herbicide, etc., then adjacent treatments. By applying inputs perpendicular to treatments, all treatments have the same opportunity to experience an application error. The goal is to reduce inherent or applied variation in the whole plot so that the treatment differences observed are real and not due to some underlying variation.

Figure 1-5.
Direction in Which Inputs Should Be Applied on Plots to Reduce Inherent or Applied Variation

| Top of Page |

Trial Design

Trials that are established with fewer treatments and more replication provide better results by reducing variation. Trials having three to four replicates are preferred, while trials having only one replicate are least desirable for reducing variation. Where replication is not possible, duplication of a single treatment within the plot will help provide an estimate of the variation that may be present at a site.

Placing a trial in the middle of a field has benefit. This reduces the potential for variation such as compaction, weed pressures, change in soil pH, etc., that may be present near field edges. Also, if the plot is placed in the middle of a field, the edges of the plot can be used as check strips for the plot (See the third plot layout in Figure 1-6, Three Potential Plot Layouts). A single treatment equal to the overall field program (check) can then be included in the middle of the trial, providing three check strips for estimating the plot variation.

Depending on the type of plot being conducted, it may be necessary to account for potential edge effects. The plot must be wide enough that the region of an individual treatment where assessments are going to be taken will not be influenced by adjacent treatments.

The layout of an on-farm trial is important in ensuring that meaningful data can be collected, interpreted and reported.

| Top of Page |

Treatment Selection

On-farm trials should be kept simple. Choose a small number of treatments to compare that endeavour to answer a limited number of questions. The more questions that a trial is asked to answer, the more difficult it will be to determine direct treatment effects. Treatments should accurately reflect a comparison of new and existing technologies or methods. Design the trials carefully to ensure that they are answering the questions being asked and that the treatments chosen can generate those answers.

Identify up front the types of information wanted from the trial. From the beginning, know what data needs to be collected, how to make the assessments and when to record the data.

Plots should be kept as consistent as possible. It is important to have an accurate measure of plot size, especially the sampled or harvested area. Sampling should be consistent from plot to plot and should occur across all plots in the shortest time possible. Where an individual on-farm trial is a portion of a larger project with several locations, try to use the same input types and equipment. This removes the possibility of input source, equipment and/or field operation from contributing to trial variation.

| Top of Page |

Data Collection

Data to be recorded should be identified at the beginning of the trial, and every effort should be made to record the data in an orderly and timely manner. The type and timing of data collection will vary with the nature of trial.

Since a great deal of on-farm research and many demo trials are ultimately concerned with the yield impacts of various treatments, the method of harvest and the appropriate equipment and people involved in yield data collection should be identified.

Communication between those involved in trials is important to ensure that an entire season's work is not lost because a weigh wagon or the right people are not present when the combine is in the field.

In general, the data that should always be recorded for an on-farm trial include:

• planting date
• harvest date
• cultivar or hybrid
• sampling/assessment dates (tissue, soil samples, height, silking/flowering dates, etc.)
• soil type, soil condition at planting
• tillage system used
• fertility amount, type, time and method of application
• weed control program or other inputs used
• daily temperature (minimum and maximum)
• rainfall
• notes on site with reference to any factors that could contribute to plot variation

Generic forms and other information on conducting on-farm research and demonstration trials is available on the OMAFRA crops Web pages at www.omafra.gov.on.ca/english/crops.

Always make a plot map showing the type and placement of treatments. This is valuable to the grower and any person assisting with the trial. Include reference and distances to permanent markers such as fence rows, buildings, trees, etc. Include an indication of direction. Refer to sample forms at Field Record Form, Field Scouting Form and Project Data Form.

| Top of Page |

Data Assessment

Check plots are necessary in trials so the amount of variability can be evaluated. It is important to review trial results in light of the variation present in the trial area. The error variation present in a trial comes from soil type, fertility, topography, drainage and/or other factors that are not uniform across the whole trial area. Therefore, the difference in treatment results (e.g., yield) may be partially due to random variation in the trial area and not solely to treatment (e.g., hybrid) differences.

Strip Trial Variability Example

The following charts give examples of hybrid strip trial harvests where the amount of variability is different despite average yields across the checks being equal.

¹Calculation of range around average yield:

t/ha: 7.21 x 0.9 = 6.49 and 7.21 x 1.1 = 7.93
bu/ac: 115 x 0.9 = 103.5 and 115 x 1.1 = 126.5

In this case, Examples 1 and 2 would be considered to have too much variability across the trial and the data would be rejected. Example 3 has all the Check plot yields within the 10% range of the average yield suggesting that variation is minimal or evenly distributed over the plot area. This data would therefore be accepted. Differences seen between other hybrids in the strip trial could be considered real.

By comparing the amount of variability between the check plots in a trial, you can determine if the treatment differences observed are actually due to treatment (e.g., hybrid) and not some other unrelated factor(s) (such as fertility, drainage, etc.).

If the difference in yield (or other parameter being evaluated) among the check plots is greater than +/- 10% of the average of the check plots, the variability is probably too great, and the trial results should be discarded.

If in reviewing the data from a trial, the difference in yield between two treatments (e.g., hybrids) is less than the largest difference between the checks, then it is unlikely that the treatments (e.g., hybrids) are different.

| Top of Page |

Therefore, in this case, Hybrid A would not be considered different from Hybrid B because the difference between the two (0.5 t/ha) is less than the difference between the check hybrids (0.76 t/ha).

| Top of Page |

Updates on On-Farm Trials

No updates available at this time.

| Top of Page |