2006 Annual Report for GNC03-020
Development of Appropriate Participatory On-Farm Trial Designs for Sustainable Precision Agriculture Systems
Summary
Research is nearing completion. Seven farms from Illinois, Indiana, Kentucky, and Canada collaborated in on-farm research. Spatial analysis was conducted on select farms and all farmers were the subject of a case study of the 1) use of yield monitor technology, 2) use of on-farm trial data in farm management decisions, 3) confidence level of on-farm trial data analyzed with appropriate statistical methods and 4) perceptions of alternative field-scale experimental designs. The graduate student completed his Ph.D. and continues to work with farmer collaborators and other farmers and makes local, state, and national presentations concerning on-farm research and spatial analysis.
Objectives/Performance Targets
Short-term outcomes include:
Farmer-collaborators receive assistance in farm management decision making.
Identification of alternative experimental designs conducive to on-farm planned comparisons.
Farmer-collaborators provided with the opportunity to experience the advantages of appropriate on-farm trial designs.
Identification of appropriate statistical analysis methods appropriate for research using precision farming technologies.
A renewed relationship among farmers, researchers, and Extension developed.
Intermediate outcomes:
Farmer-collaborators are able to make better decisions from participatory research.
Farmer-collaborators empowered to use local information instead of depending on external information sources for large geographic regions.
Farmer-collaborators are able to decide whether industry claims are true, improving confidence in production information.
Reduction of over-application of inputs, reducing pollution.
Strengthened community relationships will emerge from cooperation between farmer-collaborators and universities.
Long-term systemic changes:
More farmer-collaborators conduct on-farm trials.
More farmers conduct on-farm research in partnership with universities.
Reduction of reliance upon generalized recommendations based upon large geographical scales.
Increased confidence in farm management decisions based on localized information.
Accomplishments/Milestones
The farmer-collaborators and researchers gained experience in what works for this type of research over the three year project on on-farm planned comparisons. Challenges from both field implementation and computer statistical analysis were overcome and are being documented. From experiences in 2004 and 2005, more practical experimental designs were used in the 2006 season that were not only logistically easier to implement at the farm level but also more conducive to spatial analysis.
The graduate student has completed and successfully defended his Ph.D. dissertation, completing his Ph.D. in December 2006, however continues to work with the farmer-collaborators and Farmer Advisory Panel as well as other farmers and research in his new role as an Assistant Professor and Extension Economist with University of Arkansas Cooperative Extension Service.
Short-term outcomes:
* Identification of alternative experimental designs conducive to on-farm planned comparisons.
The graduate student’s Ph.D. dissertation (Griffin, 2006) expanded the earlier simulation work on alternative experimental designs as presented in Griffin et al. (2005). Traditional and alternative farm-level experimental designs have been implemented on farmer fields. Datasets are have been analyzed and results communicated with farmer-collaborators (for examples see Griffin (2006)). Some farm management decisions are made prior to harvest including hybrid seed purchases, thus final decisions on hybrids must be made even before the end of harvest to secure discounts and reserve limited supplies.
On the basis of 2004, 2005, and 2006 experiences, more practical experimental designs were described in Griffin (2006) that was easier to implement at the farm level and easier to analyze. Benefits of each experimental design were documented from both the farmers’ and analysts’ perspectives. Some of the problems cited for designs derived from small plot experimentation such as split-planter trials are that the analyst may be required to count passes if the design is not electronically recorded with software and GPS and that mathematical problems arise from defining which observations are considered neighbors for purposes of spatial analysis. In addition, the spatial variability is not overcome even with the many replications from split-planter trials as determine from simulations reported in Griffin et al. (2005). Split-field designs seem to work the best, but some opposition by traditional field scientists has emerged due to the lack of replication, which are deemed by some to be universally required.
* Farmer-collaborators provided with the opportunity to experience the advantages of appropriate on-farm trial designs.
The Farmer Advisory Panel is scheduled for a final meeting on February 28, 2007 in conduction with the Yield Monitor Data Analysis Workshop on March 1, 2007 on Purdue University campus similar to the workshop held November 13-14, 2005 (Erickson, 2005).
* Identification of appropriate statistical analysis methods for on-farm research using precision farming technologies.
Several spatial and traditional analysis methods were conducted on farmers’ datasets to compare the differences regarding benefits and ease. From theoretical simulations, several methods have been proven to be superior to others under field-scale conditions (Griffin et al., 2005; Griffin, 2006). These statistical models were evaluated in this analysis to demonstrate the potential erroneous decisions that the farm managers would have made if traditional analysis were used rather than the appropriate spatial analysis. Hence, the question of what is the cost of a wrong decision was addressed. Statistical diagnostics are used to ascertain which models fit the data and to correctly specify the experimental model. Spatial statistical methods used in spatial analysis include: general spatial model, spatial regimes, spatial error model, spatial lag model, cross regression, and geostatistical methods.
Intermediate outcomes:
Each farmer-collaborator was interviewed face to face as the final case study data collection phase for this research this year.
* Farmer-collaborators are able to make better decisions from participatory research.
Theory and simulation has shown that spatial analysis of on-farm trials lead to better decisions than non-spatial analysis of the same data. Therefore, farmers using spatial analysis have information appropriate for addressing farm management decisions. In addition to knowing analysis results are better, farmers using spatial analysis stated that they had more confidence in their data and their decisions than before spatial analysis.
* Farmer-collaborators are empowered to use local information instead of depending on external information sources for large geographic regions.
With decreased funding for localized field research and an overall shift to regional or multiple state research, more reliance on local on-farm trials have occurred. One farmer suggested that his primary source of quantitative information was the on-farm trails.
* Farmer-collaborators able to decide whether industry claims are true, which should improve confidence in production information.
Case study farmers introduced to spatial analysis had more confidence in their on-farm trial data, information, and decisions. Farmers made decisions quicker and made more decisions than before using spatial analysis.
Work is on-going.
* Reduction of over-application of inputs will reduce pollution potential.
Work is on-going.
* Strengthened community relationships will emerge from cooperation between farmer-collaborators and universities.
Farmer-collaborators and researchers contact each other directly at this point. In the future, it is expected that local Extension professionals will become a larger part of the relationship. Due to the yield monitor data analysis workshop on November 14, 2005, farmer advisory panel (FAP) members contact each other directly to gain feedback into their decisions and identify issues and potential solutions. Farmers have begun contacting one another directly for feedback and to share information.
Long-term systemic changes:
* More farmer-collaborators conduct on-farm trials.
On-going: The USDA-ARMS survey has asked farmers for their uses of yield monitors. In the 2002 ARMS survey, on-farms trials ranked third in the uses of yield monitors by farmers. In the future, the USDA-ARMS survey may assist in tracking information on how yield monitors and spatial analysis influence farmers’ use of on-farm trials over time.
* More farmers conduct on-farm research in partnership with universities.
Case study farmers stated their relationship with university Extension and research was strengthened due to involvement with this participatory research.
* Reduce reliance upon generalized recommendations based on large geographic scales.
Case study evidence has indicated farmers rank their on-farm trial information over regional recommendations. Some farmers used on-farm trial data as the primary source of quantitative information.
* Increased confidence in farm management decisions based on localized information.
These long-term objectives will continue to be evaluated after the end of this project.
Impacts and Contributions/Outcomes
Other farmers in the North Central Region may benefit from this research in a number of ways. This project led to a full day Yield Monitor Data Analysis to be held on March 1, 2007 on Purdue University campus as a follow-up to the workshop held November 14, 2005 on Purdue campus. Farmers and consultants from across the Midwest are registered to attend this workshop, with promotion primarily by word of mouth. After evaluation of the March 1 workshop, the workshop may be offered in additional locations.
The planned comparisons the farmers chose to use were important to the farmers and were intended to answer existing production questions. Several farmers were eager to get their yield data to the researcher for analysis and subsequent decision making as quickly as possible to begin planning for the 2005, 2006, and 2007 growing seasons. Wintertime decisions include which varieties to order. Farm management decisions are more urgent than in years past, with substantial price discounts available for hybrid purchases prior to harvest and the need to reserve certain hybrids, many hybrid and planting decisions are made before harvest is complete.
One farmer-collaborator reduced soybean seeding rates from 130,000 seeds per acre on 15 inch row spacing to 100,000 seeds per acre on the higher productivity soils, which have lead to reduced costs of production, increased planting timeliness and improved profitability. On the relatively poor soils, seeding rates remain at the current 130,000 seeds per acre in order to optimize profit. This farm ordered the appropriate amount of seed for the farm, a substantially smaller amount than if this experiment hand not been conducted. Reduced seeding rates increased planting timeliness during the planting season by reducing downtime for filling planters, thus increasing yields by planting in the optimal time period. A mathematical example based on this situation was modeled with linear programming methods and presented as the Base Case at the 2006 Top Farmer Crop Workshop.
In addition to the increased production efficiency that results from the on-farm experimentation for participating farms and their associated grower groups, a renewed relationship among innovative growers and universities is created. Many farmers are already contacting the researchers directly, which creates renewed opportunities for local Extension staff to work with these farmers. The relationships originated from the researcher-farmer, i.e. graduate student and major professor with farmer-collaborators and farmer advisory panel (FAP). Extension’s interest grew once they learned of the relationship and are becoming involved, and Extension is fostering relationships with new farmer-collaborators from their own experience.
As a result of this project, a procedure for analyzing yield monitor data has been developed into a protocol to allow other researchers, consultants, and farmers to conduct their own data handling and spatial analysis. An early version is available entitled “Yield Monitor Data Analysis: Data Acquisition, Management, and Analysis Protocol”. This protocol is what we have decided worked best for us and is available on the Site-Specific Management Center website at http://www.purdue.edu/ssmc. This protocol is currently being updated to reflect advancements in farm software packages made over the last two years as well as advancements in spatial analysis techniques as a direct result of this research. In addition to the protocol, a recommendation to the farm software industry is being prepared to provide suggestion on what farmer software packages need to add to be able to perform yield monitor data analysis from start to finish including spatial statistical analysis.
From the techniques developed for site-specific data analysis gained from this study, two additional graduate student projects, one PhD and one MS, have utilized these techniques for their research. Both projects used precision agriculture data to analyze split-field or paired-field controlled drainage field-scale experiments in cooperation with farmer-collaborators. The graduate student of this project has provided guidance in the technical aspects of data gathering, data handling, and data analysis with GIS and spatial analysis.