MIT Tech Review: Crystal Ball for Corn Crop Yields Will Revolutionize Commodity Trading.

August 10, 2017

Today at noon EST, the USDA released its August Crop Production Report, which included forecasts for end of season US corn and soy yield. This report is a major event for the corn and soy markets every year, and this season was no different!


Today’s report provided a 2017 US corn yield forecast of 169.5 bu/ac, a 1.2 bu/ac decrease (0.7%) compared to the July WASDE report. This revision is much smaller than the average change of 4 bu/ac between July and August reports since 2004 (excluding 2012, when the USDA forecast was revised downward by 23 bu/ac).


For soy, which is earlier in its growing season, the production report provided an end-of-year yield forecast of 49.5 bu/ac, a 1.5 bu/acincrease (3.1%) from the July WASDE figure of 48.0 bu/ac. This revision islarger than the average of 1.1 bu/ac since 2004, and more importantly, the adjustment to a higher yield forecast comes as a surprise to most observers.

How does this compare to the TellusLabs view?

The 10-day average of our corn outlook is 168.0 bu/ac, and that outlook has been consistentlybelow the USDA July forecast of 170.7 bu/ac since July 22nd (thought not as far as what we’ve seen from some commentators).  For US corn, our 10-day average (across two models) is 52.5 bu/ac, and that outlook has been consistentlyabove the July forecast of 48.0 bu/ac since July 13th.

We were (in all candor) surprised to see the USDA maintain their high July forecasts in the face of complex weather patterns and its own unpromising crop progress reports.  However, our models, which are trained on end of season outcomes at county scales (as opposed to these monthly within-season reports) have pointed to strong yields for both crops, and above what most observers were forecasting before today’s report.

The conversation on ag-twitter certainly reflects the surprise experienced by the grain markets at the news today, and we are pleased that our models were directionally correct relative to the market consensus.

Diverging directions for corn and soy?

At first glance it seems unusual that the USDA would be improving its outlook on one crop (in this case, soy), while tempering its outlook for another (here, for corn). The last time this happened for the July-August USDA revision was in 2009.  

In fact we’ve seen a similar oppositional pattern in our daily Kernel forecasts since the middle of July, with the median daily change in our preferred model for soy at plus 0.1 bu/ac while the same measure for corn has been minus 0.1 bu/ac.  Given that soy is earlier in its growing season and is more tolerant to dry conditions early in the season, these trends are plausible.  It is also important to note that weather patterns in the center of the country have shifted over the last two weeks that favor stronger yields.

A word on TellusLabs

TellusLabs is a Boston-based satellite imagery and machine learning company. In 2016, our agricultural intelligence product, Kernelbeat all publicly available forecasts for end of season US soy and corn yield. We've been featured in MIT Tech Review, raised $3m in institutional funding (TechCrunch profile), and have our first wave of Kernel customers onboard this season.

Please reach out with any questions, thoughts, views!


Also in News

TellusLabs' Summer Adventures
TellusLabs' Summer Adventures

September 22, 2017

Recap of TellusLabs' summer adventures, from San Francisco to Nova Scotia to Grand Island, Nebraska.
TellusLabs new Kernel models show a 2017 US growing season back on track
TellusLabs new Kernel models show a 2017 US growing season back on track

June 28, 2017

Three themes we like about FinTech today
Three themes we like about FinTech today

May 25, 2017