Big data helps growers track crop stress and yield
By Peter Mitham
Robotic scans pick up 400,000 data points on plant health and development.
By Peter Mitham
The human eye feeds 10 million bits per second to the brain, but when workers walk a greenhouse to check plants there’s only so much info they can take in, let alone absorb. Often, they’ll conduct spot checks and make decisions based on what they see.
But a sensor developed by a North Vancouver company is giving growers the potential to collect and analyze data from every plant and make more accurate management decisions across the entire greenhouse. Known as ‘SABER’ – a Spanish verb meaning “to know,” or “to be aware” (distinct from Miresmailli’s own name, the Persian word for ‘patient’) – this sensor is at the core of their first product, a trolley-mounted robot named ‘IRIS!’ that runs autonomously on a greenhouse pipe rail.
Built with Dutch partners Metazet FormFlex and Micothon (creators of the mobility platform), IRIS! received the 2018 GreenTech Innovation Concept Award at the GreenTech trade show in the Netherlands this past June.
“The philosophy of the technology is very simple. If you have good quality information, you can make better decisions,” says Saber Miresmailli, a former greenhouse grower who started Ecoation Innovative Solutions Inc. with his wife Maryam, an engineer, in his living room in 2010. “Usually, human scouts or workers select every other row, or one row per bay, to monitor. Our robot essentially goes [down] every single row, and it monitors and scans every single post in that row.”
In fact, the robot can automatically switch between rows and scan crops at two different heights – low to mid as well as mid to high – without human intervention. The result is a wealth of data rooted not just in observation but measurements of plant health based on subtle shifts in plant chemistry. Robotic scans of less than 50 milliseconds pick up 400,000 data points that indicate plant health, development and – when correlated with a library of stress indicators – problems.
“We believe the most accurate source of information in any greenhouse is the plant itself,” Miresmailli explains. “The technology that we have assesses the chemical composition of the plant.”
Scans occur every four to five days, ensuring regular data collection. Data analysis takes place within 12 hours, accelerating grower response times when problems start to manifest.
“The pests and the disease don’t stop spreading during this time,” he says. “[So] we wanted to make sure the moment this information is collected, it’s available to the decision-maker.”
How it works
Scanning plants to determine health is nothing new, but the systems available today are a step beyond older technologies such as NDVI (Normalized Difference Vegetation Index). NDVI examines the absorption and reflection of solar radiation as an indicator of plant health, but Ecoation’s system focuses on plant chemistry. Its proprietary process uses various light-based chemical analytical techniques that allow it to assess the displacement of multiple nutrients and their bioavailability.
While other technologies simply measure NDVI or other indices, we create fingerprints of defensive secondary metabolites in plants that can help us gain a clearer understanding of the crop’s status and health, he explains. We discovered – by accident – that there is a small time delay between the two types of plant responses, and we focus some of our specific measurements within that window, providing a clear snapshot of nutrient and defensive signal status, he says. “We get the true response of the plant.”
Combined with climatic information, such as temperature, moisture and carbon dioxide, the system is able to identify plants that are stressed as well as those that are thriving.
“The beauty of this particular approach and the wealth of the data that we collect is that it can actually be used beyond detecting stress,” Miresmailli says. “You can train this robot to detect the plants that are hurting, so you can go and treat them, or you can design this robot to identify your star plants, and then … try and elevate the rest of the plants to this level.”
The system’s capacity to assess plant health increases with each scan. This has allowed it to create models of how plants respond under specific conditions, allowing it to work on not only analyzing existing plants, but anticipating how environmental conditions may affect plants in the future.
“In a virtual greenhouse, we can create multiple seasons, multiple angles of sun, multiple lights, multiple elevations and locations,” he says, along with using images and data from real plants. “This is the exact same technology that self-driving cars are using.”
The technology has started to roll out in locations from Ontario to Mexico, in collaboration with operators such as Houweling’s and NatureFresh Farms. Other than IRIS!, the SABER sensor has also been incorporated into their recently unveiled scissor lift known as OKO. While growers use the lift for other tasks, the sensor conducts automated pest and disease scouting, yield measurements, pipe alignment and environmental monitoring of temperature, relative humidity, carbon dioxide and photosynthetically active radiation. The data is then fed into their interactive software, which will also allow for task management, labour tracking and storage of historical data on yields and crop health per square meter.
“Our software is like the Google street view of greenhouses,” says Miresmailli. “You can virtually walk inside your greenhouse, all the issues will pop on the screen, and you can then go and get more details about it.”
While autonomous greenhouses may be some ways off, the potential exists. Saber’s data can be fed to fertigation, lighting and other systems so that they can work in tandem with the plants to boost health and yields.
Big steps forward
Greenhouse industry veteran Peter Cummings, who recently retired as COO of Houweling’s, says Ecoation has the potential to change how growers manage greenhouse crops.
On the one hand, it’s giving them the jump on disease by picking up on cues within the plants that aren’t visible to the naked eye.
“Some disease states in plants have been notoriously hard to detect for human scouts, so if you’re able to isolate a disease situation in a plant before it’s visible, it allows the grower to respond rapidly,” he says.
The faster the response, the more limited it is, too, because the issue is smaller and will typically require fewer materials to address. This saves growers money on labour, replanting, and boosts productivity, which adds revenue. But what really fascinates Cummings is the potential to address issues not just in Canada but also in California.
While greenhouses seem like closed systems, they’re not isolated from the surrounding environment. A greenhouse in a complex growing area – one with a wide variety of crops, something not uncommon in California – faces a greater range of pest and disease risks.
“You get down on that Oxnard plain where there’s so many more different crops out there, you’ve just got a lot more different insects to deal with,” he says. “You’re constantly being presented with this barrage of unknowns, so you get these disease states that are hard to detect.”
This means growers need to be aware of what’s going on, something the Saber sensor helps them do quickly.
Cummings says the next big step for the system will be yield forecasting. Regular scans of each plant opens the door to practices from initial blossom counts to optical analysis of ripening fruit. An accurate read on crop load and maturation date will help growers know what to promise buyers and how to price it.
“If there’s too much, you have to drop the price to clear it out. If you don’t have enough then you’ve got upset customers and you have to end up sourcing elsewhere to try and fill the gap,” Cummings says. “So the sales guys are always looking for more certainty on what’s in-bound, plus or minus.”
Miresmailli, meanwhile, is continuing to forge ahead. The company now employs approximately 20 people full-time, recently completed a second round of financing, and is one of the United Nations’ partners in promoting their Sustainable Development Goals.
Ecoation’s high-profile investors include artificial intelligence entrepreneur Barney Pell, Jeff Mallett, the driving force behind Yahoo!, as well as Howard Balloch, former Canadian Ambassador in China. Associate co-founder of Singularity University and former R&D area manager at NASA, Pell sits on Ecoation’s board, and has connected the company with key figures at Amazon and other companies breaking ground in the area of machine learning. Their equally qualified advisory board consists of Casey Houweling of Houweling’s Tomatoes, Peter Quiring of NatureFresh Farms, Vijay Mital of Microsoft’s AI Architecture & Strategy, and Aaron Bickell of Viscon Group. A research partnership with Netherlands-based Delphy will provide independent third-party evaluation of the technology.
Technology should keep growers in touch with their crop, Miresmailli says, not divorce them from what’s going on in the greenhouse. He believes the technology will enhance traceability, and give growers the information needed to tell buyers exactly how a crop was grown.
“The very knowledgeable, expert grower cannot be everywhere at once,” he says. “If we can bridge the gap, and if we can actually provide information on the status of all the plants in a way that’s as if he or she were walking the plants, they will be able to execute their knowledge or expertise over a broader area.”