High-Throughput Phenotyping: Sensing Technologies and Data Mining

For the layman, the topic of sensing technologies and data mining for high-throughput (HT) phenotyping can be more than a little confusing. For Dr. Sindhuja Sankaran of Washington State University (WSU), however, it’s day-to-day stuff. In the invitation to her seminar Applying Advanced Sensing Technologies and Data Mining for High-throughput, In-field and Post-harvest Crop Phenotyping, she describes it this way: “Pastoral environments of agricultural crop fields are being invaded by armies of robots and unstaffed drones collecting essential data from plants—things like growth rates, size and shape, measuring photosynthesis, and sniffing volatile chemicals emitted by plants subjected to stresses like drought, heat, wind, and mechanical abrasions. The question becomes: What do you do with this huge amount of data? How do you put it to best use so farmers can treat crops to improve production and lower the assault of plant disease agents?”

Sankaran, an expert in agricultural and biosystems engineering who has toiled and taught in those fields domestically as well as in India and Thailand, answered her own question. She told her audience, “The definition of high-throughput (HT) phenotyping differs depending on whom you talk to, but basically it’s a non-destructive, highly accurate, sometimes automated, and generally less-expensive method for breeders to get a real-time picture of their crops. And early warning signs prompting a need for early decisions can make a big difference.” In other words, it’s an important aspect of breeding programs designed to produce better crops.

Sensor Technologies Being Developed to Monitor Crops

Conventional phenotyping is traditionally a visual operation. Disease symptoms and severity are picked out using standard, labor-intensive techniques like caliper measurements. However, working with both row and field crops, Sankaran is developing and integrating sensor technologies into the monitoring of crop phenotyping to support plant breeding, plant research, and applications for precision agriculture.

Rather than just leaf-level measurements, she’s using of a range of sensors that can also measure product characteristics and use those to quantify crop stressors and nutrients. It is making a difference in predictions and probabilities.

“In Washington state, we have 300 agricultural commodities from buckwheat to blueberries, so agriculture is a billion-dollar business that requires working closely with breeders in disease prevention and detection and crop yield potential,” she says. “We evaluate many different things, from disease resistance at various stages of development to a plant’s architecture and its cold tolerance. Sensors do an unbiased rating of the fruit, eliminating any subjectivity or personal favoritism in evaluation.”

Dr. Murat Kacira at the University of Arizona’s Controlled Environment Agriculture Center has collaborated on some of her crop monitoring research and calls Sankaran a “rising research star focusing on sensor technologies,” collecting essential crop data in terms of growth rates, canopy architectures, and stress factors to support crop improvement studies and help growers looking at precision agriculture applications.

“This is a field where the objective is to quantify crop characteristics as a response to a given environment or treatment or stress to further understand crop behavior,” Kacira says. “Once you understand that behavior, you can use it to either improve the crops through genetics applications or through environmental controls to get the most out of the plant within given stress or environmental conditions.”

Now to synthesize all the big words and morph them into regular-person speak: HT phenotyping quantifies things like stress factors, crop growth rates, even overall crop health status in a timely way with information used to take corrective action in a timely fashion.

“You can look at this data to perhaps revise your irrigation schedule or fertilization program,” Kacira says. “We’re able to analyze anomalies early on.”

Take a drone, for example. “Drone technology can detect a specific aberration in the field,” says Sankaran. “It relays the message that ‘something is wrong here’ and it GPS tags it. Then, you send in another sensor or platform to gather more data like leaf samples for a biochemical assay. If you’re just relying on a visual inspection, you’re missing what’s inside that can be discovered by other means like CT imagery scans.”

Looking To The Future

Because Sankaran and colleagues work under the mantra of improving the world through technology, they dream of an agricultural future without waste accomplished via technology integration. “The future of sensing is headed in that direction,” she says. “Not just phenomics but sensing in general. The concept is here. The technology is following closely behind and all the robotic initiatives in phenotyping are moving towards that direction.

“The ideal future would be that you don’t have to go out and scout your farm, you would get data input to your cellphones or computers in your home and, if there is a crop problem, you’d be quickly warned about it and you could just as quickly respond to change watering or disease prevention or harvesting decisions. That day is coming.”