flownaksala - Fotolia
Improvements in agriculture have predominantly relied on better chemicals over the last several decades. Now though, researchers and businesses are starting to develop a new generation of agricultural oriented applications to bring further gains to plant production performance. At the Silicon Valley Robotics Agriculture Summit, experts talked about the future of information technology in agriculture.
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
"There are several reasons agriculture is ready for technology," said Jim Ostrowski, VP of engineering at Blue River Technology, which is making lettuce thinning and weeding robots. "One is the need to increase ag production. Some stats suggest we need to increase production by 30% over the next 10 years. The land required to do this is not increasing. Also, there are huge labor shortages in agriculture." In addition, the effectiveness of some of the most common weed killers is starting to wane with the rapid evolution of herbicide tolerant weeds.
Ostrowski said the three most important areas of agricultural applications include:
Better plant care. This involves in developing systems to automate processes like weeding and identifying practices that bring better results.
Better plant production performance through resistance to drought and disease. This requires a new generation of plants, sensors, and seed data management systems on the back end to quickly identify the best breeds.
Improving the data and providing farmers applications to make better decisions.
Improving plant production performance
One important use case is the growth of agricultural robotics. This can help address the labor issues and bring new efficiencies. To address this need, Blue River is developing a line of robots that can dramatically reduce the time and costs of thinning and weeding large fields using hot oil or targeted application of weed killer chemicals. This requires the development of better machine vision techniques and machine learning techniques to recognize plants from weeds.
There are big upsides to improve plant care. For example, average corn yields are about 171 bushels per acre. But some farmers can grow about 500 bushels per acre. "There is a huge gap between the average and what it could be," said Ostrowski. "Most farming is one size fits all. We can use cameras and technology to understand what is going on at the individual plant level." He believes that more precision in applying herbicides and fertilizer could eliminate 90% of agricultural chemicals and save the US $45 billion a year.
Building apps for better plants
A key component of better agriculture lies breeding seeds with the best properties for a given environment. Large scale seed breeding operations require an inordinate amount of data collected from seed growing labs. Until recently the expense of doing this was limited to larger companies like Monsanto, Pioneer, Syngenta, and Bayer focused on commercial crops like corn, soy, and wheat.
Benson Hill Biosystems CropOS is a new cloud application that is that brings the same capabilities to much smaller seed growing operations. The service applies robust analytics and machine learning to better predict which crop varieties will perform better.
Benson Hill CEO Matthew Crisp, said CropOS is the first cloud based platform to empower the plant science community to develop novel plant varieties that can be customized for different environmental conditions and geographies. This data includes breeder notes, IoT sensors, drone imagery, and the molecular characteristics of the plants. Crisp said, "Plant genes might be 99% the same, but little variations can make significant changes in plant behavior and performance."
It’s only been recently where smaller researchers could access the quantities of data and computational horsepower to look at what these smaller genetic changes meant for different plant varieties. Crisp said, "Its only through cloud computing that we can mobilize that data and reduce that complexity that allows organizations of any size to innovate and produce seeds that are competitive with much larger players."
Plant production performance and crop data
Researchers are also looking at how to collect richer data from plants in the field using a wide variety of sensors. Richard Slaughter at UC Davis said they have created a lab with IoT sensors that can aggregate information about plant health so that farmers can make more targeted decisions.
This can be used to reduce by aggregating data from individual sensors moisture and plant stress sensors. Slaughter said, "We had a respite from the drought, but agricultural in general needs to be more efficient with water."
There is additional work on building sensors and machine learning systems to automatically identify and address plant health issues including nutrient analysis, weeds, and pests. This requires fusing data from cameras, chemical sensors, and other IoT devices to cloud analytics platforms.
The real money requires a focus on the main crops grown in the US, said Robert Morris, CEO of TerrAvion. They developed an aerial imaging plant monitoring service using piloted airplanes for specialty crops. This data is fed into a variety of farming applications that help guide decision processes such as Land Magic and Pasture Map.
The service languished until TerrAvion decided to focus front and center on corn, soy, and wheat. Farmers spend $4 per acre for a multispectral image of their crops and can see an ROI of over 56x. Morris said, "The entire artichoke crop in California is smaller than one larger farm in Nebraska." He does not believe that drone imaging will ever have achieve the efficiencies of scale to provide enough of a benefit to commercial farmers. He believes it is more cost effective today to give farmers data that can help guide decisions at key moment of the farming cycle, rather than deluging them with continuous data.
Learn how to fix performance issues before they become a problem
Learn about how componentization architectures help applications perform better
How to design for performance