Tools & data

Background last update:23 Jul 2019

‘Technologies are only as good as their datasets’

Getting usable data and lowering data acquisition costs are the main challenges for the precision farming industry, says CEO and co-founder of Prospera, Daniel Koppel.

Israeli company Prospera was founded in 2014. The company specialises in analytics, and is on a mission to turn agricultural produce into a predictable and optimised asset. We talked to CEO and co-founder Daniel Koppel (34) about the opportunities and challenges in precision farming and the future of smart farming.

Why was Prospera founded? What triggered you to start a company that specialises in AI and data?

“As a data team, we were looking at using satellite data in agricultural applications, and we ended up spending a lot of time with growers in the fields. We started asking more detailed questions about the growing process and outcomes. We saw gaps in how technology was being applied to growing, and fell in love with the idea of not only of solving complex technical problems, but also applying our expertise to an industry that is responsible for feeding the world.”

At what level are we now when it comes to precision agriculture and remote sensing? Are we still just at the beginning, or is the industry already maturing and is it getting clearer in which way we are heading?

“It depends how you’re evaluating the industry. Some areas, particularly operations and mechanics, are really advanced – you have tractors driving by themselves through rows of crops based on precision GPS, for instance, and much more sophisticated spraying and irrigation capabilities.”

Daniel Koppel:

People still don’t really know what’s going on when and where in the field at a granular level

“The area with the biggest untapped potential is agronomic precision – being able to answer questions like how much should the grower spray, and where? That hasn’t happened yet in large part because the industry doesn’t have good sensing capabilities. People still don’t really know what’s going on when and where in the field at a granular level. And even if they do know what’s going on, they also need to be able to say, “Based on this information, this is the best decision to make.”

Precision farming technologies are being developed at an increasing pace, and are getting more affordable. Yet adoption by farmers is still slow. Why are farmers hesitating to invest in smart farming solutions? And what needs to change in order for more farmers to adopt precision ag technologies such as Prospera is developing?

“Farmers have a lot riding on their decisions, so it makes sense that they would be hesitant to adopt dramatically new approaches or technologies. That said, there is substantial initial traction with growers adopting the larger players’ platforms. A lot of digital technology adoption we’re seeing today is influenced by input companies (e.g. seed, chemical and ag retailers) that have the means to cover the cost. We’re not seeing a lot of individual farmers paying for precision technology.”

Text continues underneath image

Daniel Koppel:"The biggest challenges for precision agriculture are getting good, usable data and lowering data acquisition costs. Technologies like computer vision, machine learning and AI are only as good as their datasets." - Photos: Prospera
Daniel Koppel:"The biggest challenges for precision agriculture are getting good, usable data and lowering data acquisition costs. Technologies like computer vision, machine learning and AI are only as good as their datasets." - Photos: Prospera

“We need to prove to growers that new technologies work. We also need to be respectful of growers’ and agronomists’ experience and concerns, and approach technology deployment in a thoughtful way.”

“Optimisation – squeezing value out of technologies and transforming digital information into real value to growers – hasn’t happened at scale yet. The industry is in the initial phases of that. We think adoption will explode when we can provide significant value to growers, and that’s what we’re doing today.”

What is the biggest challenge when it comes to precision agriculture? And what is the biggest challenge for Prospera?

“The biggest challenges for precision agriculture are getting good, usable data and lowering data acquisition costs. Technologies like computer vision, machine learning and AI are only as good as their datasets. The biggest challenge for Prospera as a company is getting our technology deployed quickly and at scale.”

How will an arable farmer run his farm in 5 and in 10 years from now? What level of automation / AI will be involved?

“People across industries are concerned about the impact machine learning and artificial intelligence might have on their jobs and ways of life. There were similar concerns during the industrial revolution and the mechanical revolution in agriculture when suddenly, you could plant or harvest thousands of acres with just a few people instead of hundreds.

Daniel Koppel:

Ag technologies are less about replacing people and more about giving farmers and agronomists complementary technologies that help them do their jobs better

“I don’t think we’ll be replacing farmers or agronomists, but we will be teaching things like computer science and machine learning in agronomy programs. Ultimately, we believe ag technologies are less about replacing people and more about giving farmers and agronomists complementary technologies that help them do their jobs better.”

When to use satellites, when drones, when manned aircraft for remote sensing?

“Satellites and aerial imagery provide different levels of granularity, so we use them as complementary technologies. In our partnership with Valley, we’re tackling data acquisition in 3 layers. The first is cameras mounted on the pivots that circle in the fields 24/7, capturing tens of thousands of images each day. We complement that with a second layer of aerial imagery from planes and a third layer of satellite imagery. Each layer is helping with different functionalities to understand plant stress and pests and disease.”

Text continues underneath image

Machine vision technology by Prospera detects, analyses and reports blight on crops.
Machine vision technology by Prospera detects, analyses and reports blight on crops.

Will remote sensing by drones eventually be substituted by (open source / available) satellite imagery?

Probably not in the short term because of the level of granularity you need – for example, you can’t spot pest and disease on leaves with a satellite but you can with a drone.”

Are you also measuring soil moisture content?

“We don’t actively measure soil moisture, specifically; we focus more on using multispectral imagery to detect plant stress. We analyse many types of data to get a comprehensive picture of plant health.”

What is the next step (in precision farming) by Prospera?

“We’re doing some really interesting work on computer vision and machine learning, including patented approaches to detecting plant stress and pests and diseases. In terms of our business, we’re continuing to expand partnerships to distribute our technology.”

You recently signed a partnership with Valley. About the partnership with Valley Irrigation. Why?

“In order to truly impact agricultural productivity and efficiency on a global scale, we needed to apply our technology to open field environments, as well as greenhouses. We wanted to go to market quickly and at scale via a brand already trusted by growers, rather than trying to introduce our unknown brand. Besides the fact Valley is a recognisable brand, we really appreciated the way Valley treats growers and dealers, in terms of support for their products and providing growers with great service. Valley has a great culture and deep industry knowledge and understanding of irrigation.”

Text continues underneath image

A Prospera technician in the field with a pivot. The company aims to let AI-enabled center pivots gather information, analyse it and give recommendations to farmers and agronomists about what actions they should take to optimise their yield.
A Prospera technician in the field with a pivot. The company aims to let AI-enabled center pivots gather information, analyse it and give recommendations to farmers and agronomists about what actions they should take to optimise their yield.

How does your system work in the field? What hardware and software is being applied?

“We’re combining our software with Valley’s center pivots, 84K of which are already internet-connected. We expect the first stage of the project – anomaly detection – to cover 1 million acres by 2020. We’re starting with aerial imagery of commercial farms in Nebraska, Kansas and Washington, and we’re getting really positive feedback from the early users.

Daniel Koppel:

We’re trying to help the farmer make better decisions that will cost him less money and reward him with better yield and quality

The second stage of the project will focus on providing irrigation recommendations for farmers based on plant information, and the third and final stage of the project will add spraying and fertilisation capabilities to the pivot. Ultimately, the AI-enabled center pivots will be able to gather information, analyse it and give recommendations to farmers and agronomists about what actions they should take to optimise their yield.”

How does it benefit the farmer?

“Farmers benefit from much more granular insight into the real-time health of their crops than they could ever get with the naked eye. Ultimately, we’re trying to help the farmer make better decisions that will cost him less money and reward him with better yield and quality. For instance, with Valley, we’re aiming to optimise irrigation, which means the farmer can irrigate more precisely to the plants’ needs. By avoiding over- or under-irrigating, he can get better, more consistent yields and quality.”

Text continues underneath image

Prospera combines its software with Valley’s center pivots, 84K of which are already internet-connected.
Prospera combines its software with Valley’s center pivots, 84K of which are already internet-connected.

What level of investment for the farmer are we talking about?

“Like most of the industry, we’re doing a subscription model per acre or field, but it also depends on the functionality a grower wants (e.g. irrigation, spraying, fertilisation). Many farmers already have Valley irrigation equipment, so there’s no need to make a substantial investment in new infrastructure.”

How do you manage to deliver actionable insights with incomplete datasets?

“We collect a wide variety of data sets. Based on our large data sets and machine learning expertise, we can clean and interpret “messy” and partial data.”

Who owns the data?

“The farmer owns the data.”

Or register to be able to comment.