Gunz and Rows – Agronomy Science at Granular
May 29, 2019
I have the privilege of leading our Agronomy Science work here at Granular and have been working in this space for more than 20 years. I hail from south central Iowa and am the 4th generation to farm there. In this post and several more to come, my goal is to provide some insight into agronomy data science at Granular and what we’re focusing on to help growers, both now and in the future.
To provide some context on my background, my home stomping grounds is not forgiving; it’s known for poorly drained silty clay loam soils that at times make it difficult to get crops planted in the spring and harvested in the fall due to excess moisture. The area receives on average 36 inches of precipitation, but it’s ranged from 27 to 60 inches in my 15 year farming career. It is possible to have drowned out crops in the spring and drought later in the same year. In a word, farming in this area is “unpredictable”.
Also, like many in ag, I have experienced the “Fog of Farming”, also known as 100 different things going on at once, making the right path forward very murky. Example: that forecasted one inch rain right after planting turns into four inches, and now you are trying to decide whether to wait and see what comes up or just tear it all up and try again. Or perhaps that 4 inch rain occurred later while the corn was emerged, but now it is an ugly yellow mess; should I hit it with another shot of nitrogen (which will cost more) or should I let it be? These big decisions are taking place along with 99 others, some of which will have equally major effects on the crop.
Now, I understand that our small area doesn’t have a monopoly on bad weather and less-than-ideal soils (can I get a high five, North Dakota?), but it gives some context to how I got started in Agronomy Science and what drives our efforts at Granular today. Trying to understand why my father’s and neighbors’ practices would work one year and then be an abject failure the next fueled my curiosity in Agronomy Science further. How could we learn from past seasons? What if we didn’t have to relive these difficulties? What if we could bypass the heartache and turmoil of a difficult season by leveraging observations we didn’t experience?
I see Agronomy Science as taking the best research in crop growth, soils, weather, fertilizers, management and more and building tools, in this case digital tools, to help growers make better decisions with their cropping plans and management. Let’s put the collective knowledge of agronomists, research scientists, and growers themselves to work in guiding the best path forward.
In this Digital Age, the technology we have available to us today, such as cloud computing, smart phones, machine learning and more, is expanding at a rate faster than we’ve ever seen. I believe there’s no reason why we in agriculture should not be a part of this growth.
But, I also understand the farmer’s need to know “why” and “how”. Farmers like being in control; they manage their own operations, call their own shots, and are pretty self-reliant. They take pride in knowing how things work, so when a word like “Machine Learning” comes along, where do you even begin to break that down? Some of the terms that have been thrown out in promotional materials are meant to impress, but sometimes they are downright confusing. When a bearing on a gearbox goes bad, you probably have a good idea on how to fix it, but when a Random Forest algorithm is incorrect, how would you even know it was wrong in the first place and then how to make it right again? Maybe take it out and leave it in some random forest, perhaps?
In coming posts, I want to share some of the terms we use in Agronomy Science and make them more understandable for everyone, because I think these are concepts you experience on a daily basis, but perhaps never had the words or structure to put it together. I also want to show you behind the scenes of how the calculations, algorithms, models and other decision tools are created.
Now, here’s something that may be surprising: every decision you make about your crops, whether it is to apply a particular herbicide or plant a certain brand of corn or perform tillage in a certain way, is part of a model. Every grower has a model that he or she uses to make decisions based on various sources of information, whether from previous experience, their trusted advisors, magazines, or just how Dad did it. This mental model evolves and grows with time; you learn about how new corn hybrids perform, and forget about how many quarts of Lasso should be applied per acre. But overall, growers have a framework for decision making that, if studied hard enough, could be broken down into a series of “if this, then that” or “for so much of this, do that” type of statements.
Computer models, regardless of the techniques used, are not that far off. They are based on observations, then synthesized into decision-making tools. They are only as good as the information and techniques used to train and build them. And, just like using a hammer to drive a wood screw into a 2 x 4, data science and statistics tools can be used improperly, but instead of a bent wood screw, you may end up with a faulty answer. But in the hands (and mind) of a trained individual, the results can be a masterpiece.
The end goal of this work is NOT, I repeat, NOT to replace the grower’s decision making ability, or their trusted advisors’. There are still many situations that cannot be answered by a computer, but are best handled by a wise person who can judge and weigh the evidence. Instead, these Agronomy Science tools are there to supplement the underlying knowledge growers have, perhaps giving them quantifiable insights, peaks around the corner, and head’s ups on what might be coming along.
At the heart of it all, Agronomy Science needs to make you a more productive grower. Our job isn’t to make your already busy situation even more so. Our job is to give you easy-to-use tools to make better decisions and have confidence in them.
Stay tuned for more posts that will cover the basics of data collection and why good data is necessary, how to make use of it, and what happens when we have many variables all pointing in different directions! Also feel free to check out my Gunz and Rows segment on the Digital Acre podcast, available on iTunes, Stitcher and all other podcast apps.
Enter your email to sign up for The Bottom Line, Granular’s monthly Newsletter