Identification and Significance of a Problem or Opportunity

The problem we propose to address is the worldwide, systemic inefficient application of crop inputs on farm fields. A principal focus will be on the chronic mismanagement of nitrogen fertilizer. Our society manipulates the nitrogen (N) cycle to great benefit, but chronic inefficient use of nitrogen fertilizer has led to the hypoxic “dead zone” in the Gulf of Mexico and the leaching of nitrates into groundwater. The National Academy of Engineering (2012) has declared managing the nitrogen cycle a “Grand Challenge,” and the N cycle has been labeled a “planetary boundary that has been transgressed” (Rockström, et al. 2009). But currently agricultural science is far from understanding the processes determining crop yields, and therefore from discovering how efficient input management practices vary over time and space. Reimer, et al. (2017, p. 6A) described the problem well:

The incredible complexities of the biophysical systems alone are still not well understood by researchers, advisors, or farmers. There are a myriad of variables involved in a decision about how much N to apply in a given growing season … This makes providing reliable recommendations inherently difficult, both for researchers trying to distill complex science into useable information and for advisors trying to craft recommendations for individual farms and fields. (Reimer, et al., 2017, p. 6A)

Unless scientists can gain a better understanding of these complexities, farmers will continue applying fertilizer in ways that send chemicals that everyone wants to go into crops actually go down rivers. The objective of the proposed project is to develop a research infrastructure and encourage the development of a commercial infrastructure that will generate the data needed for this better understanding. If successful, the research will increase farm income and enhance the quality of the nations’ waters.

Of late, great hope has been placed in using the components of digital agriculture, including precision agriculturetechnology,  agricultural Big Data, remote and proximal sensing, and on-farm experimentation to increase farm management efficiency. Schlam (2019) offered a typical quote:

 The use of analytics will be crucial to enabling smarter agriculture, with farmers leveraging data on soil, local diseases and pests, climate, and other environmental factors to optimize yields and seed selection. ... Soil sensors, aerial drones, GPS-enabled tractors, and more will generate the data upon which analytics solutions will rely … .

But, however impressive these technologies and techniques are individually, and despite public and commercial enthusiasm about them, academic agriculturists looking to the future are expressing frustration with the current state of their use, and have recognized a need to bring them together systematically:

Digital agriculture … has been trying to attract customers before the ecosystem has been properly constructed. What we believe is missing is a standardized way to gather and interpret data, and then translate actionable insights to commercial users—insights which then, in turn, can deliver value to growers. (Zuckerberg, 2017)

In the following, we propose an integrated project to drive the creation of the system Zuckerberg calls for above. We will argue that the key impediment to the development of digital agriculture is lack of information about crop responses to factors of production, but that components of digital agriculture can be used to generate that same information.

A Conceptual Framework and the Fundamental Question

We conceptualize the problem of crop input management in terms of four types of variables. The first variable, y, is crop yield. The second type we represent by a multi-element vector c, of unmanaged, spatially distributed “field characteristics.”  The third is a multi-element vector z, of unmanaged and temporally stochastic variables (principally, weather). The fourth is a multi-element vector x of “managed input variables,” (e.g., fertilizer application rate, seeding rate). We conceptualize yield, y, as resulting from a natural process described by a function f, which depends on farmer choices, field characteristics, and weather: y = f(x, c, z).

The fundamental research question for crop input management is, “What is f?”  Agricultural scientists have been attempting to generate data to estimate ffor major row crops for almost two hundred years (Odell, et al. 1984). Indeed, estimating f was motivated the advance of modern statistical theory, as R.A. Fischer (1935) developed his pioneering statistical research on randomization in experimental design to address estimating f using data from agronomic experiments (Antle 2019). The extraordinary efforts to learn about f have been made because f reveals how crop yields respond to producer choices, and how yield responses change with growing conditions. Better estimates of f are key to better scientific support of farm management and all the social benefits that follow. Figure 1 provides the simplest of illustrations. Versions of figure 1 and explanations of the implications for producer management choices appear in most introductory microeconomics textbooks (e.g., Pindyck and Rubinfeld 2013, pp. 204-208), and are typically presented in the first few weeks of introductory microeconomics courses. It is assumed in figure 1 that there is only one managed input, x. The characteristics of a site A on which the crop is grown are c^A. For simplicity, it is assumed that when a value of x is chosen, weather is assumed known and given as the vector of constant values z²⁰¹⁹. The input price and output price are assumed constant at levels w and p. Elementary calculus shows that the profit-maximizing input application rate, x^*_A,19(p, w), is that value at which the slope of the yield response curve is equal to the price ratio, w/p. When multiple inputs are chosen, similar mathematical rules apply, optimizing over more than two dimensions. Space limitations prevent us from illustrating more complicated situations.

Technical Feasibility of the Work

The focus of the proposed project is on-farm precision experimentation (OFPE) and the analysis of OFPE data, which are illustrated in figure 2. The first panel of figure 2 shows the randomized design of a nitrogen rate OFPE on corn. In form, the design is like those of traditional small-plot trials, but covers a much wider area. The field in figure 2 is 37 ha in size. The second panel shows the trial being “put in the ground” using variable input rate application is accomplished by using GPS-based computer software to pre-program a variable application rate “plan” into a computer aboard farm machinery. That program “instructs” application equipment to apply inputs at the planned rates on the designated plots as the farmer just drives the farm equipment through the field in the usual manner. The final figure shows the resultant data. Led by PI David Bullock, the Data-Intensive Farm Management (DIFM) project has now worked with participating farmers to conduct over one hundred large-scale OFPEs in eight US states, Argentina, Brazil, and South Africa. Trials have been run on cotton, corn, soybeans, and wheat. DIFM has collected, processed, and analyzed the data, communicated the analytical results to the participating farmers, and published journal articles describing the techniques and results. The basic elements of OFPE research, then, are clearly feasible. Those techniques inexpensively provide excellent data, and one very legitimate aim of the proposed multistate project will simply be to pass the proven techniques to scientists currently unfamiliar with them, so that they might work with farmers to conduct OFPEs. But the project goals are much more ambitious than simply teaching other researchers OFPE techniques. The aim is to build a research infrastructure and encourage the development of a commercial infrastructure that will permit tens of thousands of OFPEs to be designed and conducted farmers and their crop consultants annually, as well as the handling, processing, and analysis of the resulting data. The system would annually generate vastly more field trial data than has been generated since the first agronomic field trials were conducted in the first half of the nineteenth century.

The main elements of the strategy to scale up data generation involve the creation of a cloud-based, “on-farm precision experiment design” software system. That system will allow crop consultants, who have received training from Extension personnel or others, to upload basic information about a farmer’s field, such as a geo-referenced file of the field’s border, and the sizes of the farmer’s machinery, and then, with some “pointing and clicking,” design a statistically legitimate agronomic experiment over the farmer’s entire field. All data would be transferred wireless to and from farm machinery. Writers of this proposal have already developed parts of this automated trial design system, and are confident that developing the rest will take hard work, but is feasible.

The principal element of scaling up OFPE data analysis will be an automated “analytical engine,” which can import OFPE data, and then with minimal “human-in-the-loop” effort, employ econometric analysis and a variety of machine learning methods to develop management recommendations. For example, methods based on reinforcement learning can be used to optimize prescriptions for seeding or applying fertilizers or herbicides. Learning algorithms such as those based on deep learning (e.g., convolutional networks or long short-term memories), locally (spatial or temporal) random forests, or Gaussian processes can be used to capture spatial and temporal properties in fields. Other techniques based on a variety of heuristics (local search) and metaheuristics (evolutionary and swarm-based search) in constraint satisfaction, Markov random fields, and even natural language models to describe agricultural processes can be applied to derive data-based farm management support.

The statistical and machine learning models will be accompanied by responsive, interactive statistical visualizations to explain which data have the most impact on the predictions (the “why” of the statistical models), helping build the trust of farmers and consultants in the “black box” models used for prediction. In addition, OFPE data will be used to improve the calibrations of existing crop growth models, which in turn can be used to model yield response and estimate optimal management strategies. Automating OFPE data analysis will be a major challenge. We envision a future in which significant parts of crop science and agricultural economics academia devote themselves to this very task. But it will be feasible, and fascinating, to make a convincing start to this research endeavor in the five years of the proposed project.

Automating the communication of the implications of a farm’s OFPE data will involve the creation of a cloud-based “decision tool” software system, which displays in user-friendly, visual ways, OFPE data results and implications. Several such “decision tools” are currently available on commercial markets. Several of the 135 interviews of farmers, professional crop consultants, and extension personnel conducted by two of the writers of this proposal (Montesdeoca, et al. 2018) made clear that many farmers and consultants have limited faith in their effectiveness. The crop science and agricultural economics literature impels skepticism of current commercial decision tool software packages. They are based on long debunked yield-based input application algorithms (Rodriguez, et al. 2019) and crop growth models with parameters that in general cannot be well enough calibrated to allow the models to provide adequate management advice (Antle 2018).

The DIFM project has proved that conducting trials and analyzing the data is feasible. We believe that the scaling-up of that work is also feasible within the five-year timeline. But, certainly, the project faces risks. Data analysis is difficult, and to develop automated techniques for it that work well enough to provide farmers with profitable management recommendations will be a challenge. But this is why a multistate project is needed--we plan to put dozens of very skilled, highly talented researchers to the tasks at hand, for five years. We are confident that we can develop a workable infrastructure by then, and that the next generation of agronomic science and farm management rely upon, and continue to advance that infrastructure.

The Importance of the Work, and Consequences of It not Being Done

“Site-specific” agriculture technology appeared on the market in the 1990s. But, as Reimer describes in the quote above, very little about the yield response process is known. As a result, little is known about how to use precision agriculture technology well. Farmers complain that they do not know what to do with their data. In truth, in many ways neither does anyone else. If the proposed work is not done, the ignorance that Reimer describes will continue for longer than necessary, and farmers will continue having to make input application decisions that depress their incomes and are bad for the environment.

Evidence for Stakeholder Identification

Evidence for stakeholder identification of the need for the proposed project’s outputs was provided in abundance during the 135 interviews of farmers, professional crop consultants, and extension personnel, mentioned above. That study concluded that many farmers are already experimenters. Certified Crop Advisors (CCAs) and farmers told us repeatedly that they were placing various types of informal “test plots” in their fields, to see how alternative management strategies fared against their actual strategies.  But CCAs also told us that carrying out field trials was a great deal of work, and that they realized that the statistical designs of their trials were technically flawed.

Another principal conclusion from the interviews is that many farmers are frustrated that they now possess multitudes of data that they do not know how to analyze. A few representative quotes follow:

“Once you get all this data, you need to hire someone to analyze it. If you can trust them. …  Lots of people are just trying to sell you something.”  -western Illinois corn grower 

“Our seed rates?  Pretty much a shot in the dark.”  -three south Texas cotton growers 

“It would be wonderful if my decision-making were more analytical instead of just off-the-hip SWAG (‘Scientific Wild-A** Guesses’)!” –northern Illinois corn grower

The interviews also showed that neither farmers nor their CCAs want “black-box” management recommendations from decision tool. They want intuitive agronomic explanations of how field characteristics determine the responses of yields to managed variables.

The Advantages of Doing the Work as a Multistate Effort

The proposed project’s collaborators come from all over the US. Including multiple geographically dispersed universities allows trials in crops from different regions: maize, soybeans, white winter wheat, red spring wheat, barley, cotton, canola, sugarcane, and sorghum, under varied topography, soil characteristics, and climates. The multiple-university team of researchers and extension personnel is comprised of personnel with significant experience already working together, and its professionals are capable of providing the considerable needed expertise.

Likely Impacts of Successfully Completing the Work

The proposed project’s principal goal is ambitious:  we want to revolutionize agronomic research and how the information from that research is transferred to those who actually make farming decisions. It will make farmers central to the research process. The data will allow crop consultants to provide farmers management advice derived from data from the same fields for which that advice needs to be applied, instead of data conducted by researchers somewhere else. Farmers will improve their input management strategies, basing them on data, not speculation. Farm income will be increased, and environmental damage from agricultural chemicals will decrease. In addition, the commercial crop consultancy industry will change, providing advantages to those who understand experimental design and statistics. It will attract highly skilled workers into rural areas, which will have multiple positive social and economic impacts in those regions.

Proposed Project Members

Bullock, David
dsbulloc@illinois.edu

Boerngen, Maria
maboern@ilstu.edu

Hawkins, Elizabeth
hawkins.301@osu.edu

Griffin, Terry
twgriffin@ksu.edu 

Vanderplas, Susan
svanderplas2@unl.edu 

Jung, Jinha
jinha@purdue.edu 

Miao, Yuxin
ymiao@umn.edu 

Chowdhary, Girish
girishc@illinois.edu 

Miguez, Fernando
femiguez@iastate.edu 

Mitchell, Paul D
pdmitchell@wisc.edu 

Brorsen, Wade
wade.brorsen@okstate.edu 

Ashworth, Amanda
Amanda.Ashworth@ars.usda.gov 

Kharel, Tulsi
Tulsi.Kharel@usda.gov 

Li, Xiaofei
xiaofei.li@msstate.edu 

Ketterings, Quirine
qmk2@cornell.edu 

Sheppard, John
john.sheppard@montana.edu 

Daccache, Andre
adaccache@ucdavis.edu 

Sun, Xin
xin.sun@ndsu.edu 

Dhillon, Jagmandeep
jagman.dhillon@msstate.edu

Allen, Cody
allencm@illinois.edu 

Baath, Gurjinder
gurjinder.baath@ag.tamu.edu

Reed, Vaughn
vr401@msstate.edu 

Guo, Wenxuan
wenxuan.guo@ag.tamu.edu 

Tao, Haiying
haiying.tao@uconn.edu 

Froes de Borja Reis, Andre
areis@missouri.edu 

Mieno, Taro
tmieno2@unl.edu 

Longchamps, Louis
ll928@cornell.edu

Jha, Gaurav
gjha@ksu.edu 

Ransom, Curtis
Curtis.Ransom@usda.gov 

Enciso, Juan
Juan.Enciso@ag.tamu.edu 

Shajahan, Sunoj
sunoj@illinois.edu 

Nugent, Paul
paul.nugent@montana.edu 

Sellars, Sarah
sarah.sellars@sdstate.edu 

Kaur, Upinder
kauru@purdue.edu 

Bagavathiannan, Muthukumar
muthu@tamu.edu 

Proulx, Rob
rob.proulx@ndsu.edu 

Mizuta, Katsutoshi
toshi.m@uky.edu 

Pinto, Ricardo
ricardo.pinto@montana.edu 

Yost, Matt
matt.yost@usu.edu 

Ghatrehsamani, Shirin
spg5994@psu.edu 

Becker, Talon
tbecker2@illinois.edu 

Diss, Meagan
mcdiss@illinois.edu

Jones, John D.
jones86@illinois.edu 

Kumar, Hemendra
hemendra@umd.edu 

Mendes Bastos, Leonardo
lmbastos@uga.edu