On Sept. 12, USDA released its latest yield estimates for several major row crops, including a massive 50.6 bu. per acre prediction for soybeans. If realized, it will be the largest U.S. soybean crop on record.
But at this point in the season, actual yields are still up for grabs, and not everyone agrees with the 50.6 bu. prediction, including Descartes Labs, an ag tech startup that does its own proprietary analysis based on several components, including weather and satellite imagery. CEO Mark Johnson says USDA’s prediction is not necessarily a bad one, but Descartes Labs has a slightly different take.
“Our model agrees with the general consensus that 2016 is a year of record-high soybean yields and production in the U.S.,” he says. “The current release soybean model forecasts 12.75% of U.S. counties to have record-high yield, and 76% above-average yields.”
Still, the model doesn’t quite add up to 50.6 bu. per acre. It comes in slightly lower, at an estimated 48.3 bu. per acre. That’s a bin-busting crop but comes in about 5% lower than USDA estimates.
“The states that drive our national average yield below NASS are primarily Illinois, Missouri, Indiana, all major soybean producers,” he says. “We believe this is a result of the prolonged mixed weather patterns experienced by these states throughout August and now September, which to various degrees would impact not only seed development stages, but could also lead to the emergence of foliar disease and possibly premature leaf dropping.”
Descartes Labs uses a machine-learning algorithm that factors in this information, according to Johnson. The company will continue to post weekly corn and soybean yield estimates through the rest of the season to www.descarteslabs.com/forecast.html.