Business Impact

Crystal Ball for Corn Crop Yields Will Revolutionize Commodity Trading

TellusLabs is using NASA imagery, machine learning, and expert knowledge about vegetation to deliver accurate, in-season agricultural yield estimates.

Deriving financial insights from satellite images isn’t a new idea, but TellusLabs is putting a twist on it. The Boston startup analyzes satellite imagery from NASA as well as weather data from the National Oceanic and Atmospheric Administration and seasonal, crop-growing information from the U.S. Department of Agriculture. It then uses machine-learning algorithms to generate intelligence about natural resources, such as predicting agricultural yields.

The strategy might sound similar to that of other satellite imagery analysis companies like Descartes Labs and Orbital Insight. However, TellusLabs plans to differentiate itself by applying scientific expertise in vegetation and climatology to its analysis, maintaining a narrow focus on natural resources, and quickly rolling out new products. Its goal is to be “a Bloomberg terminal for Earth signals.” “There’s a broad base of people who have to make tough decisions around natural resources, and we want to give them quality data, quickly,” says TellusLabs CEO and cofounder David Potere. 

The company’s first foray into the market is Kernel, an agricultural commodities forecast modeling tool that recently entered a publicly accessible open-beta phase. The free, beta version of Kernel has limited features, but the full-fledged product is an interactive, online dashboard that shows a map of the main corn-growing regions in the U.S.—across 18 states—and key financial indicators, such as predicted yield, harvested area, and total production. Users can view data at a state, agricultural district, or county level and look at historical yield data sourced from the USDA. The dashboard also features an indicator arrow—analogous to a stock-market ticker—that denotes the average change in corn yield estimates, week over week. TellusLabs will update the forecasts daily.

The dashboard of Kernel, TellusLabs’s agricultural commodities forecasting tool.

Like a Bloomberg terminal, Kernel is designed to be a nexus for fast, reliable financial data, which people can utilize multiple ways. A commodities trader could use the information to make money off trades in the futures market. An ethanol plant operator could consult Kernel to gauge whether its contracted farmers will be able to supply enough corn to keep it running. An agribusiness company like John Deere could license the data feed and integrate it into a smart pump that automatically adjusts how much water it gives crops. 

The natural resources know-how behind Kernel comes mostly from TellusLabs’s other cofounder and chief technology officer, Mark Friedl. Friedl is a professor in Boston University’s earth and environment department and leads the school’s land cover and surface climate research group, which studies continental-scale vegetation mapping and monitoring. TellusLabs also has a handful of science advisors who are proficient in remote sensing of agriculture, forests, and bodies of water. One advisor works for NASA’s Land Science team and another at Woods Hole Research Center, a Massachusetts-based environmental research institute. Potere, the TellusLabs CEO, holds a master’s degree in satellite remote sensing and a PhD in geo-demography and helped establish and lead the data science team at Boston Consulting Group.

Accuracy and speed could also give TellusLabs an edge in this market. The company says a recent, internal test showed it is able to project the end-of-year yield for U.S. corn more accurately than the government. In the test, TellusLabs ran publicly available, historical USDA corn yield data from 2004 to 2014 through its algorithms and made predictions about year-end numbers. In that 10-year period, the startup’s estimates beat the government’s 69 percent of the time during August and September, which are the key trading months for corn.

TellusLabs also recently released a corn-yield estimate for the 2016 growing season, so prospective customers can compare its prediction with the USDA’s forecast, which the agency will release later this month.

Quandl, a Toronto-based aggregator of financial, economic, and alternative data like satellite imagery analysis, is currently testing Kernel to decide whether to resell it on its platform, which is used by hedge funds, asset managers, pension funds, and investment banks. Early signs are encouraging. “You can make money by being more accurate than the market, or by being faster; TellusLabs is both,” says Quandl chief data officer Abraham Thomas. “A 70 percent ‘beat’ is not in itself hugely compelling, but couple that with a speed advantage and it becomes compelling.”

TellusLabs plans to introduce a forecast model for soybeans to Kernel in September. It eventually aims to release an outlook model for wheat; to extend its corn yield data to Argentina, Brazil, and China; and to monitor forests and large, fresh-water reservoirs by satellite. “We have a pipeline of ideas,” says Potere. “There is a whole bunch of interesting, global-scale, geospatial questions that haven’t been asked yet.”

Learn more about artificial intelligence at EmTech Digital 2017.
Register now

Uh oh–you've read all five of your free articles for this month.

Insider Premium

$179.95/yr US PRICE

More from Business Impact

How technology advances are changing the economy and providing new opportunities in many industries.

Want more award-winning journalism? Subscribe to Insider Basic.

  • Insider Basic {! insider.prices.basic !}*

    {! insider.display.menuOptionsLabel !}

    Six issues of our award winning magazine and daily delivery of The Download, our newsletter of what’s important in technology and innovation.

    See details+

    What's Included

    Bimonthly home delivery and unlimited 24/7 access to MIT Technology Review’s website.

    The Download. Our daily newsletter of what's important in technology and innovation.

You've read of free articles this month.