Can predictions be so accurate!

A Startup That Defy USDA

Mark Johnson brings a revolution in the agriculture sector by envisaging America’s crop yield more accurately. Instead of setting out numerous workers to analyze thousands of ranches a month forth October corn harvest, America’s biggest crop, Johnson’s Descartes Labs, a startup, launched in 2014 is using a machine learning algorithm and 4 petabytes of satellite imagery to survey the corn crop from space. Johnson is challenging the USDA corn predictions, with just 20 employees, who never leave the Los Alamos, New Mexico office.

All the concerned persons including ethanol producers, grain elevator operators, hedge funds, insurance companies, commodities traders, and even the farmers await for the corn to sprout out in August so that they can get the USDA’s August crop report. As corn yield prediction is a huge business in the US, so billions of dollars are at risk every year. Now, Descartes crop yield estimate is ahead of the crop report from August last year.

As per Johnson, founder of Descartes, their model constantly beats the crop prediction accuracy of the USDA’s 2015 August by a percentage point. And their algorithms have gotten more precise with time, with an average margin of 2.5 percent of error, when to go through the historical backtests.

Why Agriculture?

According to Johnson and his partners, there are some specific reasons behind choosing agriculture tracking. One of the major issues or reasons is food scarceness and global climate change. Second, the data set of the year already exists from the images taken by MODIS and Landsat that could be needed to train their machine erudition models. Third, as corn grows at the slower pace so, farmers can take advantage by observing it with additional spectral bands such as infrared that can be recorded by both newer and older generations of the imaging satellites.

To calculate the chlorophyll levels, Descartes makes use of spectral information which is not visible to human eye. According to Descartes, it measures satellite data of the entire farm individually on a daily basis in the US and updates corn yield prediction in every two days, whereas USDA updates its prediction once a month only.

Big Plans of Descartes for Future

As stated by Johnson, with the help of better access to data, better satellites, and more powerful algorithms, his models are improving for better. Descartes plans to move on to the other crops, once its models get ready for it. They have already started tracking soybean yield and other locations such as Argentina and Brazil, the Black Sea region, EU and the China.

Descartes vision is to understand the natural resources, how those natural resources move around, and how humans change this planet for better tomorrow.

Does Descartes Model Prove Its Accuracy in Future? Please share your thoughts in the comments section below.

Subscribe with us to get your dose of interesting news, research & opinions in the startup segment. Fill the form below:

Name

Email*