The 2004 Top Farmer
Crop Workshop Adds Yield Map Analysis
Introduction
Since the late 1960s the Top Farmer Crop Workshop at Purdue University has been famous for taking a hard look at new crop technology. No-till, narrow rows, track equipment, glyphosate resistant varieties and many other innovations were debated in the workshop discussions and tested in the computerized budgeting of grower participants. In 2004, the Top Farmer Workshop will add yield map analysis as another way to test the yield and profit implications of new technology.
Top Farmer participants will have the opportunity to bring raw yield monitor data and supporting information (e.g. soil maps, topography, remote images, planting map) for planned comparisons at the workshop. The technologies used in the comparisons will depend entirely on the questions that participating farmers find most important. Participants at the 2003 workshop mentioned hybrid and variety trials, tillage comparisons and insecticide treated seed as potential on-farm trial questions.
Brief History of the
Top Farmer Workshop
The Top Farmer Crop Workshop is an annual event on the Purdue campus that attracts farmers from around the Corn Belt and beyond. Since 1968 it has devoted two and a half days in mid July to crop technology and management. A key tool has been farm linear programming optimization software that helps participants to develop a budget model of their own farm and test new technologies in the computer before trying them for real.
The linear programming is particularly useful in helping participants to understand their machinery options. It was a key tool in helping some farmers decide on expansion plans in the 1970s and retrenchment alternatives in the 1980s. The genetic and information technology innovations in the 1990s and the early 21st Century have been more difficult to evaluate with linear programming. Yield monitor data provides a way to help farmers analyze choices that do not lend themselves to the linear programming approach.
Top Farmer participants are no strangers to precision agriculture. Most of the participants have yield monitors on their combines. Many have tried variable rate fertilizer applications. Site-specific management has been a perennial workshop topic since the late 1980s. In 2003 about 20% of workshop time was devoted to precision agriculture, including presentations on: use of remote sensing to generate yield maps, profitability of site-specific P and K, use of the GPS time stamp to improve farm logistics, linking crop GIS information and farm accounting records, finding low cost GIS layers on the INTERNET and autosteer technology. The first public demonstration of the new Veris pH soil sensor occurred during an evening barbeque at the Purdue Agronomy Farm during the 2003 workshop.
Dr. Howard Doster, professor emeritus, Purdue Department of Ag Economics, was part of the team that created the first workshop and he led the effort since the early 1970s. Doster is retiring and a new multidisciplinary team is taking over led by Jess Lowenberg-DeBoer, Purdue Agricultural Economics Department. This team will build on the strengths of the past and adapt the workshop to the 21st Century. Based on feedback from Top Farmer participants, survey results and case studies of precision agriculture use by Corn Belt farmers plus recent advances in yield map analysis techniques, the team decided that it was time to move from talking about precision agriculture to using some of the tools in the workshop.
Making Use of Yield
Maps
Since the early 1990s some researchers, agribusinesses and farmers have presented technology trial results based on yield monitor and other sensor data. Usually this was presented in the form of yield maps or descriptive statistics, with subjective interpretation. Even researchers struggled to assess the reliability of their yield monitor data results.
Researchers in other disciplines have had a head start in analysis of spatial data. Geologists and mining engineers have many years of experience in mapping the physical characteristics of the earth’s layers. Their techniques are often lumped together under the term “geostatistics”. Similarly geographers and regional economists have worked on identifying relationships in demographic, social and economic information. Their tools can collectively be called “spatial econometrics”. Agriculture has drawn on both geostatistics and spatial econometrics to develop appropriate methods for spatial crop data.
Initially, yield monitor data analysis drew on the geostatistics background of many soil scientists and agronomists, and focused on yield mapping. Although maps are a good communication device to aid in interpretation of large amounts of data, they may not be the ideal management guide for making economic decisions related to farm planning. It is difficult to show more than 3 or 4 geographic information system (GIS) layers on a map (e.g. by using color, hash marks, dots, outlining polygons) and it is even more difficult to visually indicate the relationship between layers. Whether or not a difference shows up on a map depends on the color scheme and the increments of the legend. For example, if a legend is organized in 10 Bu/Ac increments a yield of 160 and 168 may show up as the same color on a yield map. Difficulty in interpreting yield maps has led to wide spread complaints about the usefulness of yield surfaces generated from arbitrary assigned class definitions.
Management decision-making often requires numbers and it demands some assessment of the reliability of those values. For example, if the yield with one hybrid is 160 Bu/Ac and with the other it is 168 Bu/Ac, how often is that likely to occur? Is it a one time event or likely to be a repeatable real difference?
Purdue has been a leader in adapting on-farm comparison methods to yield monitor technology. In 1993, Steve Hawkins, Bob Nielsen and Jess Lowenberg-DeBoer organized some of the first on-farm trials harvested using yield monitors in Dekalb County, IN. In 1998, Mark Kessler and Jess Lowenberg-DeBoer analyzed potassium response in yield data from the Purdue Davis Ag Center. In the Purdue Precision Farming Profitability book, Sylvie Brouder and Bob Nielsen provided guidelines for on-farm trials using yield monitors. Lowenberg-DeBoer worked with Farm Journal magazine and crop consultant Ken Ferrie to analyze the profitability of variable rate NPK and plant population on the Greg Sauder farm near Tremont, IL. Rodolfo Bongiovani and Lowenberg-DeBoer analyzed the yield and profitability effects of a system designed to control soil compaction on the Mclaughin-Dooley Farms, LeRoy, IL. Charlene Finck described the Sauder and Mclaughin-Dooley farm results in Farm Journal articles in 1998 and 2001. Currently, on-going research involves on-farm trials in Indiana, Illinois, Kentucky and Minnesota.
Spatial analysis is roughly in the same stage of development as linear programming was when the first Top Farmer Workshop was held in 1968. Linear programming was developed in the 1940s and first applied to agricultural research in the 1950s. By the late 1960’s, linear programming had become a commonly used research tool, but it was rarely used with farm audiences. Most researchers thought that linear programming was too technical for most farmers. Methods for communicating linear programming results to non-economists were not well developed.
The Top Farmer Workshop experience showed that farmers could effectively use linear programming results to help make management decisions. Linear programming is essentially an automated budgeting technique and most Corn Belt farmers are very familiar with budgets. The Top Farmer Crop Workshop team felt that Purdue had a comparative advantage in spatial analysis of yield monitor data and that it was time for that technique to move out of the researcher’s office into the extension arena.
Can Precision Ag Meet
the Top Farmer Challenge?
In 2004 yield monitor data analysis at the Top Farmer Workshop will be an experiment. Can yield monitor data analysis provide enough payback to participants to justify spending time on it during the two and a half day workshop? Based on experience with linear programming, the expectation is that while not all Top Farmer participants will be interested in the yield monitor analysis, some will find it to be the compelling reason to participate.
The initial focus will be on planned comparisons in which the grower changed something and wants to know the effect on yields and profits. Methods for “data mining” are in the early research stage and not ready for farm use. The term “data mining” is used to describe pattern recognition analysis of data collected from functioning systems without any treatments or planned comparisons.
The Top Farmer yield data analysis will use statistical methods drawn from geostatistics and spatial econometrics. Part of the challenge will be to communicate those statistical results in ways that farmers can use. The Top Farmer team expects that analysis will be particularly useful for cases in which simpler visual or descriptive statistic methods do not reveal clear differences.
In 2004 most of the data will probably be from split planter, strip or block comparisons. The team has some experimental methods that might be used to analyze split fields or paired field comparisons. Australian researchers have suggested designs which place relatively small plots at strategic locations in the landscape, but few US farmers have the equipment or the background to implement this kind of comparison. Because many farmers use yield monitors without a Global Positioning System (GPS), the Top Farmer team will work with both GPS and “load” data. Load data would be using strip or block average yields and/or moistures as stored by the yield monitor.
Economic comparisons will be a key part of the analysis. The Top Farmer team will help work with the grower to budget out the costs of each practice or technology, and the benefits given estimated yield differences. When appropriate these budgets will be used to create return surfaces to help growers understand how returns to a new technology would vary across the landscape.
Over 30 years experience with linear programming indicates that a good analysis requires a close working relationship between the farmer and the Purdue team. A good analysis will probably require several iterations. The initial analysis will be simple. Subsequent analyses may add layers (e.g. soil type, topography, slope, previous management) or scouting information (e.g. location of pest and plant disease problems, stand differences). For both the Purdue staff and the farmer, each analysis is expected to be an experience in learning what data is important. Farmers interested in participating in having their yield monitor data analyzed are encouraged to contact the Top Farmer Team before the workshop to began initial analysis.
The purpose of the yield data analysis will be to answer the grower’s questions about the technology tested. It will not be to teach growers how to do this type of analysis. Experience with linear programming suggests that while a few farmers learned to do the analysis themselves, the priority was on obtaining answers that could be used in making better decisions. The Top Farmer Team expects that eventually crop consultants and other farm advisors will offer this kind of yield monitor data analysis.
In 2004 participation in the yield monitor data analysis will be limited to 10 farms. Until now yield monitor data analysis has been handled on an individual basis, field-by-field. A researcher might spend days or weeks working on data from a single field. Scaling up from one-by-one to 10 farms in a three-day workshop is a big step. Participation will be first-come-first-serve. Workshop organizers expect to scale up the yield data analysis in subsequent years.
If you are interested in yield monitor data analysis, please contact the Top Farmer Team to reserve a slot. The members of the Top Farmer team responsible for yield monitor data analysis are: Jess Lowenberg-DeBoer, phone: (765) 494-4230, email: lowenbej@purdue.edu, and Terry Griffin, phone: (765) 494-4257, email: twgriffi@purdue.edu.
For More Information
Brouder, Sylvie, and Robert Nielsen, 2000 “On-Farm Research,” in Precision Farming Profitability, J. Lowenberg-DeBoer and K. Erickson, eds., Purdue University Agricultural Research Programs, p. 103-112.
Finck, Charlene, 1998, “Precision Can Pay Its Way.” Farm Journal, Mid-January, p. 10-12.
Finck, Charlene, 2001. “The Root of All Yields”, Farm Journal, July/August, p. 14-18. http://www.flexharrow.com/content/whatsnew/stlkchpprartcle.html
Kessler, Mark and J. Lowenberg-DeBoer, “Regression Analysis of Yield Monitor Data and Its Use in Fine Tuning Crop Decisions,” Proceedings of the 4th International Conference on Precision Agriculture, SSSA, Madison, WI, 1998, pp. 821-829.
Lowenberg-DeBoer, J. and Anthony Aghib, “Average Returns and Risk Characteristics of Site Specific P & K Management: Eastern Corn Belt On-Farm Trial Results,” Journal of Production Agriculture 12 (1999), pp. 276-282.