How can AI help growers grow better and produce more? And what role do sensors play in this? Lotte Adema, Product Manager at Source.ag, joins us to explain how they approach this.
Need for greenhouses
"Worldwide, there is a great need for more greenhouses," Lotte explains. A greenhouse can use up to 20 times less water than outdoor crops and can also cope better with different climates. "More growers should be able to go from mid- and low-tech to high-tech," she says.
Scaling up, however, can be tricky. "Growers are basically superheroes who have to turn all kinds of knobs. Especially with the days getting longer, growers have to be available for 12 hours. If you have multiple greenhouses, that's actually not doable." Automation of irrigation management is therefore an important step.
AI around plant biology
Source.ag is building two models to help in that automation: a water uptake model and a nutrient model. "We build AI around the biology of the plant. By combining that with weather forecasts, we know exactly how much water and nutrients the plant will use in a day. Every night we create a new model, and we then build an autonomous strategy around that. And with new varieties, we learn how it works in 40 days - so you don't need a whole season for that to get used to a new crop."
Feed models with data
A nice concept, but for all those models you need data. "If we want to know exactly what the plant's water uptake is, we need irrigation data and drain data as well as substrate saturation. That's why we work with SenseNL's CARA MET substrate sensors. The data from that comes in directly wirelessly, and with that we train the irrigation models."
A great example of collaboration, Lotte thinks. "In this sector, you need different parties to achieve a result together: substrate suppliers, sensor companies, climate computers and AI. SenseNL was the first party to share their data directly with us. This allowed us to quickly create models that matched the incoming data."
This is useful because no matter how good a sensor is, you still have to interpret the data coming out of it. For growers, that can be tricky, but AI lends a hand. "A greenhouse usually has 10 CARA MET sensors. For that, we built an AI model that does anomaly detection on all that data. That way, we can see if oxygen has gotten in between somewhere at a sensor, then we signal the grower to put it back in the mat for a while." This is important, because if a sensor gives abnormal readings, the averages are no longer correct.
Sharing knowledge
The two companies complement each other well: SenseNL understands sensors, which measure the moisture content of the substrate mat. Source.ag is good at cleaning that data, interpreting and presenting it, and then converting that into action.
Source.ag is sensor agnostic, so it basically didn't matter which company they partnered with for the sensors. SenseNL was, however, one of the first parties to indicate it wanted to make use of the data collaboration opportunity offered by Source.ag.
"In doing so, SenseNL often puts 10 sensors in one greenhouse - we like to work with multiple in-slab sensors to get as much good data as possible. This is possible with them because the price of the sensors is favourable to the grower. What is also nice, they recognise that they are good at the hardware part and that we are good at the data cleaning part. So they were very open to sharing knowledge. Even when they develop new sensors, they ask us in what way we would like to receive the data to build products on that. They are also willing to put capacity into doing research together."
"Collaboration SenseNL and Source.ag is an example for the sector"
Lotte sees the collaboration as an example of where the sector needs to go. "We all want to do our own thing and we all have our own business to run. But we need to eventually move more towards covered cultivation, to use less water and land. To keep up the pace of that development, cooperation is crucial."
Winelis Kavelaars of SenseNL adds: "We provide affordable and reliable data and Source.ag is a reliable party that is doing good work. So this is a very nice combination. We are happy with that. Our sensors are sold รก 10 pieces per kit. Simply because the variations in a cabinet cannot be measured with one or two sensors. You once need several data points to pick up any deviations in the greenhouse, plants, climate, and/or sensors and then make the right decision on irrigation and/or nutrient management. But more data alone is not enough, data must be reliable and affordable. For that reason, we wanted to and were able to keep the CARA MET sensors affordable. We now dare to say that, at the moment, we have the most accurate and reliable substrate sensor on the market. Or actually, I don't say that myself. Source.ag, for instance, has indicated that SenseNL has a lot of traction among their customers. Parties like that have contact with multiple sensor manufacturers. So they are in contact with multiple parties. So if anyone can know, it's them. Sensors have to be reliable."
Winelis further indicates that SenseNL is open to multiple collaborations. "Our goal is to provide as many growers as possible with the tools to give more insight into their crop, thereby higher production with fewer resources."
For more information:
SenseNL
Tel: +31 (0)85 876 8909
www.cara met.com
[email protected]