Why It's Disruptive
Satellite images can tell us a lot about Earth; Descartes Labs is taking that knowledge to the next level. The company, which spun out of the Los Alamos National Lab, analyzes photos of the globe and compares them, pixel by pixel, to make predictions about anything from crop outputs to the effects of climate change. In 2016, Descartes projected U.S. corn yields--which number in the billions of bushels--to within 1 percent accuracy, beating the USDA's estimate, according to co-founder and CEO Mark Johnson. "We're able to get to an astonishing margin of error," Johnson says, "without ever seeing an ear of corn."
Johnson says the startup's clients include a number of agriculture companies, scientists, and government agencies. Descartes recently signed a contract with Darpa to monitor crop outputs in the Middle East to warn of food shortages. Another client tracks stagnant water to predict the flow of mosquito-borne diseases like Zika, and another is trying to use Descartes's data to estimate worldwide pollen counts.
The company's biggest asset is also its biggest challenge: data. Johnson says that Descartes's supercomputer processes so much data that the first time the company did a test run, Google thought its cloud services were falling victim to a denial of service attack. The startup processes quadrillions of pixels of data in a single analysis, which can take hours. Downloading millions of satellite images from the web takes time too--so much that instead of downloading them over the web, the company has considered having them shipped via FedEx. Descartes's growing number of partnerships with satellite companies has helped address the problem, since the data sets those companies provide are often more manageable--and better suited to the startup's unique needs--than the massive files it downloads from NASA's servers. --Kevin J. Ryan