USask department is using AI to advance agricultural practices

[00:00:00]Artificial intelligence is being used by pretty much every industry and agriculture is no exception. Many new high-tech advancements will be highlighted at Ag in Motion this week.

[00:00:12]That's Western Canada's largest farm expo happening just 15 minutes northwest of here in Langham. Professor Steve Shirtliff is heading to Ag in Motion this morning. He is the director of the Nutrien Centre for Sustainable and Digital Agriculture and a professor of agronomy at the U of S. But first, he's with us here in studio. Good morning, Steve.

[00:00:31]>> Good morning, Jess.

[00:00:32]>> The Nutrien Centre for Sustainable and Digital Agriculture is still fairly new.

[00:00:36]How did it come about?

[00:00:38]>> Well, it it came about because of a gift from Nutrien. And I'll just there's we there we were actually rebranded before we launched as the Nutrien Digital Agriculture Centre. So, yeah, so we came about so Nutrien as part of the last fundraising effort Be What the World Needs campaign from the University of Saskatchewan. Nutrien, you know, very generous uh uh corporate Saskatchewan company gave the largest donation and one and a $15 to the University of Saskatchewan and a large chunk of that is going to the Nutrien to create a new centre, the Nutrien Digital Agriculture Centre. Our Dean uh in the College of Agriculture and Bioresources uh Angela Bedard Hawn, she it was this was really her brainchild. She had when she started being Dean about 5 years ago, she said she was going to move us into the area of digital agriculture and that's what she did. She negotiated this and then she tapped me on the shoulder and asked me if I want to be director and I was like, >> [laughter] >> "Yeah, put me in, coach." You know.

[00:01:35]>> Why is like why is it so important to have a centre like this here? Why do you digital and agriculture go hand in hand?

[00:01:42]>> You know, cuz I one of the things that's out there that that agriculture is a very data-rich industry, you know, in terms of how much data's out there. You know, there's you know, our our the fields are our factories and so there you know, so that so there's technologies like using satellites to monitor them as well. So and there's also large data sets. So you're dealing with these large data sets which are which you know, which can be difficult and imposing to analyze. So that's where you know, you mentioned AI earlier where it you know, that kind of meshes hand in hand is that we can use newer statistical techniques, you know, to analyze them and you know, including AI.

[00:02:22]>> I want to talk more about that because at the center you are using AI. You mentioned satellites and drones to model all of the farmland in Western Canada.

[00:02:29]Can you tell us about that project?

[00:02:31]>> Yeah, well it really stems out of some work that we've been doing for about a dozen years at the University of Saskatchewan where there's been not just myself, but others in soil science and agricultural economics as well as colleagues of mine in computer science and engineering and geography. We've been working together in the area of digital agriculture.

[00:02:50]It's very much a collaborative a collaborative area. We started off we started off with a perk grant about you know, a bit over 10 years ago and that's where we started phenotyping plants using drones using UAVs. So quadcopters we'd image them and working with the plant breeders and geneticists at the University of Saskatchewan to develop phenotypes and phenotypes are essentially just character things we can characterize from the crop, you know, whether it be the amount of flowers, the intensity of flowers, the number of heads, the reflection off the plot. So this allowed us to work with plant breeders and to speed up the breeding process by that to give them more accurate because the drone can actually measure things instead of just saying that oh this plot is kind of laying down. It's somewhat lodged. We could quantify exactly how much or we could say that this one this variety of canola flowered earlier and flowered longer. So we're able to do that and be able to give breeders plant breeders tools to increase their selection efficiency to get good varieties. And so with that we're partnering with a lot with the mostly breeders in the Crop Development Centre at the university, another centre at the University of Saskatchewan.

[00:04:00]>> How will all of this help farmers? Like, how can they access some of this this data from the centre?

[00:04:05]>> Yeah, well, so the centre is we're really prefaced on three research platforms. And so, there's I mentioned the phenotyping already.

[00:04:14]Uh the the the the one the other one that has a lot to do with satellites and larger data, and that's where we use a lot of uh machine learning and AI is what we call the gauge, and I think you mentioned it originally. It's a the geospatial agroecosystem inference generator, and it's not pronounced gag, gauge.

[00:04:31]Uh and uh and so, what that with the idea behind that is that we're using is that we're essentially creating a digital twin of every field on every farm in all of Western Canada at a 10-m pixel level. So, at every 10 m is where we have one piece of information using historical satellite information, as well as information about the soil, about the soil topography, and about the weather, etc. We'll be able to understand that. And this started from a project where this started from a project where farmers have shared yield data with us. So, we can use that to train an AI model. So, the farmers' yield maps, yield maps modern combines can produce maps of where the yields are higher and lower in a field. And I should say that the one of the the the main driver behind this is to really understand within field spatial variability. As we drive around Saskatchewan, we look at fields, we know this year we're pretty lucky. There's most fields are quite even, but there's some fields that are drowned out in the low spots. In a lot of years, in most years in fact, those lower spots are often the higher yielding spot higher yielding areas of the fields because of there's more moisture. And we're usually moisture limited. This year is a of course, you know, everybody swatting mosquitoes knows whether or not we're not moisture limited. So, do understand that and we can So, with the satellite information over time, we can understand the variability in space within a field and in time. And so, by putting this model together, we're looking at ways of managing the land better so we can grow more with less.

[00:06:05]>> Your center obviously doing a lot of research using AI, but are a lot of farmers using it in their day-to-day work?

[00:06:12]>> You know, I just I I read some stats recently and agriculture is one of the industries where AI is not been implemented as much as others. So, it is it is There are people are looking for ways. I think most people think of AI as being the large language models of the chat GPTs and the clouds of this world.

[00:06:29]You know, we we often use We don't use it that way. We use it in terms of you know, we used You know, large language models are based on transformers and we can and we use those to analyze data over time. Uh and similarly, we use a lot of machine learning, which is a form of AI as well. So, we tend to use those more, but agriculture, interestingly enough, if there was recently There was recently an AI strategy released for Canada from from the federal government and it identified agriculture as one of the key industries that can benefit from AI in Canada and actually singled out singled out a Saskatchewan company Crop Tuff Mistic as being one that's implementing AI already. And so, there's I know we visited We had the deputy the deputy parliamentary secretary visit our lab. And so, they're well aware of that.

[00:07:19]And they see and that's really exciting for us because the federal government sees a lot of potential of agriculture being a data-rich environment and being very important, a fundamental industry in Canada to So, they feel we're poised to benefit from it.

[00:07:33]>> Steve, just before I let you go, you're on your way soon to Ag in Motion. What are you going to be doing there today?

[00:07:38]>> So, I'm going to be I'm going to be at the uh the College of Agriculture and Bioresources Uh uh booth there. People can come by, see it. We have some UAVs outside. We also have uh We also have some displays inside. There'll be people from uh the other people from the center there as well. There's also always popular There's always a There's a soil pit next door. They have a hole in the ground where people can go in and learn about the soils. That's a key part of all this as well, you know, the digital soil mapping and understanding what's beneath us, you know, our you know, Saskatchewan has had 10,000 years uh since the last glaciation to build up this great soil we have and to understand more about that really affects the sustainability of agriculture in Saskatchewan.

[00:08:21]>> Steve, really good to talk to you this morning. Have fun at Ag in Motion later today.

[00:08:24]>> Well, thank you very much.

[00:08:26]>> That is Steve Shirtliff, director of the Nutrien Digital Agriculture Centre. He's also a professor of agronomy at the U of S, and Ag in Motion is on until tomorrow.