You collect data on crop performance and yields, including data from on-farm trials. Ag companies and researchers at universities collect tons of data from large-scale and small-plot trials. The more technology develops, the more data people collect. How can all this data be turned into a tool that helps you make more informed crop management decisions?
Ron Baruchi, CEO of Agmatix, and Sagi Katz, chief agronomist for the Israel-based startup, believe the service they provide unlocks answers. One of their primary functions is using computer software and machine learning to combine various sets of data from all over the world and put it into a format that makes the results meaningful. At the same time, they use computer modeling to help sort out common factors behind a disease or nutrient trend lurking within the reams of research. The modeling they develop makes predictions about things as diverse as whether sudden death syndrome will appear in your soybeans or how much nitrogen you can convert into grain during the season.
“Researchers set up experiments and field trials using different techniques,” Baruchi explains. “You need a way to get data into a common format so you can tease out conclusions and common threads. That’s what we’re doing — making mountains of data useful. We provide interpretation for it.”
Here are two examples that illustrate how powerful this concept could be. Both projects are ongoing, so more information will likely be gleaned later.
SDS and soybean cyst nematode. Five Midwestern universities plus the Ontario Ministry of Agriculture are studying SDS outbreaks to learn how to predict where they will occur more accurately. The universities are Purdue, Iowa State, Michigan State, Illinois and Wisconsin. So far, they have data from 90 SDS trials conducted in the six regions over five years. The challenge, Baruchi and Katz say, is turning all that research information into something you can use as a farmer.
Cooperating with researchers, Agmatix broke the data into three classes — no disease, moderate disease and severe outbreaks. The model developed predicts the correct class of disease that would occur in the real world with 78% accuracy.
“We also discovered that the severity of SDS is affected by several factors, including soybean genetics, timing and type of fungicide applied, and amount of rainfall received during specific soybean growth stages,” Katz says.
Nitrogen in corn. Wageningen University in the Netherlands and the International Fertilizer Association are cooperating with Agmatix in an ongoing project to build a first-of-its-kind database of nutrients in crops under an array of environmental conditions. More than 30 researchers from 50 countries shared data for the project, covering corn, rice, soybeans and wheat.
What information exists in that data for just one crop, corn, and just one nutrient, nitrogen, that might help you better understand how to manage it?
Using artificial intelligence technology, Agmatix developed a protocol to standardize and harmonize the data. That led to Agmatix developing powerful models, Katz says.
Using 5,377 observations collected in the U.S., China and Nigeria, Agmatix developed a decision tree to predict how much nitrogen would wind up in grain. So far, the model is very accurate, with a prediction error of 7.2%. Plus, the model uncovered that hybrid maturity, nitrogen input and soil organic matter have the greatest influence on nitrogen availability.