Use On-Farm Research to Evaluate Profitability
February 16, 2009
Everyday you see, read or hear about farming practices that hold promise for increasing the profitability of your farming operation. The question is, “Will they work on my farm?”
Farmers participating in the Nebraska Soybean and Feed Grains Profitability Project will share the results of their 2008 on-farm research March 4. Learn what they've learned.
To answer this question you may need to experiment. How you conduct the experiment can determine the credibility of the resulting information and how much you trust it and can rely on it.
Planning Your On-Farm Research
Typically, there is much variation in crop performance within a field. A side-by-side comparison of two practices does not give reliable information because we cannot know if yield differences or lack of differences are due to the different management practices or to different growing conditions in the two strips or parts of the field. There are some experimental procedures that should be applied to obtain reliable information on alternative practices on your farm. Following these procedures will insure that you, your neighbors, business associates, and others can rely on your results and will see the value in conducting scientifically based on-farm research. You can easily conduct valid, field-scale on-farm research on your land using your equipment and management practices.
On-farm research has three basic components:
- formulating a question,
- testing the hypothesis with experimentation, and
- drawing a conclusion based on the data.
Forming a Question
Simply develop a well-defined research question that can be answered with data that you can collect from the field. Try to keep the question focused on one practice. Will changing this practice, e.g. decreasing planter speed, affect crop performance and profitability? Formulate this into a question to be tested.
Following are several examples of questions used for on-farm research:
- Strip tilling soil at or before planting field corn increases grain yield.
- Treating bean leaf beetles on early-planted soybeans reduces the incidence of bean pod mottle virus and ultimately increases soybean yield.
- Incorporating agricultural lime will increase soil pH and enhance corn and soybean yields faster than non-incorporated lime applications.
- Soybeans planted in early April will have higher yields than those planted in May.
- No-till corn and soybeans will yield better than conventional tilled crops on Lutton Clay soil in the Missouri River Bottom.
Testing the Question
Here you decide which treatments (farming practices) to compare, design the field layout of the experiment, and determine what you will measure. Two experimental layouts are typically used in on-farm research. The proper design is determined by the number of treatments to be compared. Two-treatment experiments use the paired-comparison layout. Experiments with additional treatments are designed using the randomized complete block layout.
Using these two layouts, you can measure differences between treatments with confidence. Randomization and replication are important components of a well-designed, on-farm research experiment. Randomization ensures that favoritism is not given toward a treatment. Replication reduces the possibility that results are due to chance rather than the treatment. These two factors separate demonstration plots from on-farm research experiments, which are designed to draw conclusions with confidence and ultimately lead to wise business decisions.
Paired comparison designs should have six to seven replications (Figure 1). If you lose a pair or two due to pest infestations, weather damage, or other circumstances, you will still have sufficient pairs (minimum of five) to analyze the experiment with confidence. Two harvest weights are measured in each treatment area and then compared to the alternative treatments on either side.
The Randomized Complete Block Design. The randomized complete block design (Figure 2) is used for on-farm comparisons that require three or more treatments. At least five blocks are recommended (four minimum) to conduct reliable statistical analysis. Treatments are randomized within each block to remove favoritism in the comparison.
Measuring and Marking Comparison Area
Information such as row, planter, combine, sprayer and fertilizer rig, and other relevant implement widths must be known prior to the comparison design. Incorporating this data into the experimental layout will improve the efficiency of managing the experiment as treatments are applied and harvest is conducted.
Be sure to mark the treatment locations well with field flags, wooden stakes, global positioning system (GPS) equipment,or a combination of these tools. Draw a sketch of the experiment layout for reference at harvest. Buffer areas between treatments are typically important to ensure that treatments do not influence each other. Including a field border/buffer is important to eliminate influence from compacted end rows, fence line grasses, field roads, etc.
Harvest weights may be collected using a properly calibrated weigh-wagon or yield monitor. Moisture and test weights should be collected as indicated in the research design. Use the attached worksheet to record harvest data and calculate crop yields.
In addition to crop yield, grain moisture and test weight you may want to collect additional data relevant to your research question. Examples of this data include soil fertility, plant height, insect counts, weed densities, planting and harvest populations, grain protein analysis, etc. This data can be statistically analyzed if collected according to the research design.
Keeping a diary of crop and weather conditions throughout the growing season is an invaluable resource when the time comes to draw conclusions from the experiment. A photographic record of observed differences during the growing season also may be useful.
Data collected from a well-designed experiment can be statistically analyzed and interpreted to determine whether real differences are present among treatments. It’s difficult to draw conclusions by simply looking at the raw data. Statistical analysis determines probabilities that the differences were caused by treatments rather than random variation.
Various confidence levels are used to analyze agricultural data. University researchers generally select a minimum confidence level of 95%, which means that there is a 95% probability that experimental differences measured are due to the treatments rather than to chance. There is a 5% possibility that the treatment differences are due to chance. Many farmers readily accept a confidence of 90% depending on the nature of the experiment. By following the experimental guidelines suggested here, the odds of making a scientifically based decision are in your favor.
Once yield and/or other treatment differences are determined, focus on economics. What is the cost:benefit ratio? Is the advantage of a given treatment worth the cost difference among the treatments? Non-tangible benefits such as improved soil quality, environmental improvement, and efficiency also should be considered. Conclusions should generally be drawn from comparisons repeated over time that facilitate evaluation under varying growing conditions.
On-farm research is a powerful decision-making tool for farmers. The time and effort required to design, implement and analyze a sound on-farm research comparison is worth the confidence that you and other farmers will have in the results. Results should be taken a step further by analyzing the economics of the comparison. A yield difference is important, but may not be relevant if the cost of the yield increase is more than the return on investment.
Contact your local Extension Educator today to plan your on-farm research agenda.
UNL Extension: On-Farm Research Web site, which includes information on two multi-year and multi-county Nebraska on-farm research projects, impacts from farmer research, and Frequently Asked Questions (FAQ).
On-Farm Research Guide, written by Jane Sooby, technical program coordiantor of the Ohio Farming Research Foundation. (2001)
Guidelines For On-farm Research (ANR-001-97), written by Alan Sundermeier and published by Ohio State University Ag and Natural Resources. (1997)