Some of the most studied farmland on Earth could help pave the way for the creation of an artificial intelligence assistant that could help producers make better decisions that will boost their profits.
As part of its HyperLayer project, Olds College of Agriculture & Technology collected about four terabytes of data from the fields on its Smart Farm. This amount of information is roughly equivalent to the combined data of 1.3 million smartphone photos taken at a typical size of three megabytes each.
It would take one person about 365 years to take that many photos if they snapped 10 per day. “A typical farmer wouldn’t be able to collect that much data because it wouldn’t be economically feasible for them,” said Dr. Felippe Karp, who is an instructor and researcher at the College’s Werklund School of Agriculture Technology.
Some of the data was collected using hyperspectral imaging by the International Space Station. A typical smartphone photo has three spectral bands of red, green and blue per pixel, but hyperspectral images contain more than 200 bands per pixel, including light that is invisible to the human eye.
These kinds of technologies allowed researchers to quickly study multiple fields in unprecedented detail, including the interaction between crops, soil and microclimates. For example, they were able to determine that even relatively small variations in height across a field can affect the severity of crop diseases in different locations by changing the humidity.
The $3.4-million HyperLayer project was launched in 2020 by Alex Melnitchouck, former chief technology officer of digital agriculture at Olds College. He coined the term, hyperlayer, to describe the many different layers or types of data that have been gathered at the Smart Farm, which is a living laboratory for crop and livestock research spread over 3,000 acres.
Although funding ended last year, Karp said research based on the project will continue at Olds College. The HyperLayer team built a digital platform to not only to share the project’s data with partners, but also to create a computer model that can make predictions for further research.
“Right now, I can go into that platform, access what I need to generate my prediction and click on a button. Less than a minute later, I can get my prediction. That’s how quick it is.”
Although it was trained using data from fields containing black soil at the Smart Farm in central Alberta, it has made realistic predictions about farmland in Saskatchewan operated by Olds College. The fields near Craik, Sask., north of Moose Jaw, contain different soil with less organic matter and a drier climate, said Karp.
“It’s quite impressive how the model did in terms of presenting the range of organic matter inside the fields. The model didn’t know, but it still tried. It definitely predicted a value that was slightly on the high end, but it knows about probability, which is to be expected because the model tends to use what it knows and predict values closer to it, while also trying to generalize the results.”
Karp said if similar or even partial sets of data as that of the HyperLayer project were collected from all the regions in Canada, researchers could use them to build an extremely powerful model for the entire country that could help the nation’s farmers.
He emphasized the HyperLayer project represents only the initial phase of what could potentially be exciting new technologies for agriculture. However, his vision is that efforts like these by researchers around the world will lead in a few years to something like an artificial intelligence assistant for farmers.
He pointed to ChatGPT’s use of artificial intelligence to engage in natural, human-like text responses and verbal conversations with people. “Imagine if you could pull up a personalized agriculture assistant on your smartphone,” he said.
“You get a message saying, ‘Hi. Let’s see. Based on soil moisture sensors and probability analysis, today’s the day for seeding,’ and you’re like, ‘Great. What should the seeding rate be?’ And it looks at the database you’ve built up over the years. It says, ‘OK, using the previous year’s sensor data and the forecast for this year, variable rate seeding is recommended. Here’s a prescription map for your fields. Can I upload it to the seeder?’”
Karp said it is important to understand that artificial intelligence is “just a tool” that works best in collaboration with human experts who can provide guidance, context and insights. He emphasized that farmers and agronomists will still have the final say as they work with the artificial intelligence assistant.
“You check that map and you’re like, ‘Hmm. You know what? I feel I could use a little bit more of a higher seeding rate. I’m feeling like taking a bit more risk. I’d like to go to 1.5 pounds more seed at the zones in my fields with the highest yield potential.’ The artificial intelligence assistant goes, ‘OK, sounds good. Can I now upload the revised map to the machine?’ And you say, ‘Sure, go ahead; now, I’m good with this.’”
The artificial intelligence assistant will ease the workloads of farmers and agronomists by acting like an operational manager that’s immediately available any time it’s needed, said Karp. “That’s how I see Olds College helping increase the capabilities of artificial intelligence to assist farmers through the data set that we’ve been building as part of the HyperLayer project.”
Karp acknowledged some producers may already be feeling overwhelmed by the rising flood of data about their farms from sources such as precision farm equipment. However, he said it is vital to collect good data, and store all of it, because it will pay off in a future where farmers must likely be ever more efficient due to rising input costs and the need to cut greenhouse gas emissions to reduce climate change.
“Perhaps you don’t have the knowledge or the time to process this data at the moment, but perhaps there will be tools in the future that can improve your farming practices and increase your profits. That’s more money in your pocket, plus it will help guarantee that our children and grandchildren will have the possibility of living in a world with the natural resources, such as healthy, productive soil, that we have right now.”
