South American grain exchanges are projecting an Argentina soybean crop that exceeds 3 metric tons per hectare (MT/ha). But an agtech startup company specializing in satellite imagery analysis is predicting significantly lower yields.
According to beta forecasting from TellusLabs, Argentina soybean production could land closer to an average end-of-season yield of 2.75 MT/ha. All TellusLabs modeling from the past month have ranged between 2.69 to 2.89 MT/ha. Compare that to the Rosario Exchange (3.07 MT/ha) and USDA estimates (2.92 MT/ha).
But severe flooding in Argentina means there may be pressure to lower current estimates. TellusLabs is the first to assert that prediction, according to David Potere, CEO and founder.
“Everyone’s holding their breath right now to see who revises downward,” he says. “We’re showing 5% to 10% lower end-of-season yields than the consensus estimates.”
To determine its projections, TellusLabs analyzed tens of thousands of satellite images across the past decade. The company trained its predictive models against a decade’s worth of yields reported from about 300 corn and soybean growing districts across Argentina. Potere says flooding in January and February has been well above normal.
Potere notes that soybeans are a resilient crop, but at this point in the season, “there’s only so much catch-up you can do. It’s still a heck of a year, but it’s disappointing based on the output of the last five or six years.”
Last fall, TellusLabs predicted U.S. corn and soybean yields to within 1% of USDA estimates. The company is working on predictive models for several other key crops and geographies, including Argentina corn, Brazil corn and soybeans, U.S. winter wheat and China soybeans. Potere says the Argentina soybean model was released first because it has “cleared the bar on accuracy,” adding that TellusLabs will share the results of other models after they are sufficiently fine-tuned.
For more information, visit www.telluslabs.com/.