BlogCornWheatSoybeansSupplyReleases and Updates
27 January 2023
In a time of unprecedented supply and demand shocks, early and accurate yield forecasts and planted acreage estimates are indispensable.
Gro’s highly accurate Global Yield Forecast and Acreage Models offer a holistic view of global food supply up to 10 months in advance of final government reporting. These models are a key indicator for supply, prices, and farmer profitability and can be used to make decisions around product pricing, sourcing, trading, or policymaking.
Gro’s suite of machine-learning models currently cover 80% of global production for corn, soy, and wheat. They provide an unmatched combination of accuracy, global coverage, granularity down to the district level, and daily in-season updates.
Gro’s 2022 Models Accurately Predicted Yields and Planted Acreage Up To 10 Months Before Final Government Reporting
In our latest performance report, we look back on the performance of our Global Yield Forecast Models across corn, soy, and wheat, as well as our US Planting Intentions Model and US Prevent Plant Model for corn, soy, wheat, and cotton, in 2022.
This is the only report currently available in the market that compares model performance against ground truth data in a comprehensive, transparent way. Our analysis shows that Gro’s models accurately predicted yields and planted acreage, averaging within 95%-98% of final government reporting up to 10 months in advance.
Key takeaways of our 2022 report include:
- Historically, for the past seven years, Gro’s US Corn Yield Forecast Model has been on average within 98.3% of the USDA final January report by September (i.e., four months in advance). Our US Soybean Yield Forecast Model has been on average within 99% of the USDA’s final number by September (i.e., four months in advance) for the last four years.
- This remained true during the 2022 US growing season. By September, Gro’s forecasts for corn and soy were already within 98% of the USDA’s final January reporting. Gro’s models thus provided an earlier view of final numbers, allowing customers to move ahead of the market.
During the 2022 growing season, Gro’s US Corn Yield Forecast Model was showing yields consistently below USDA estimates starting in June, providing early insight into lower-than-expected production and potentially lower exports. The model dropped mid-season due to extreme high temperatures, recovering a bit in August as conditions improved. The forecasts stabilized in September and were already within 98% of USDA final January reporting.
- In critical global regions, where the current marketplace does not have many machine learning-based predictive models, such as Ukraine, Brazil, and China, Gro’s 2022 Global Yield Forecast Models were on average 95% accurate up to 10 months ahead of final government reporting.
- Gro’s US Planting Intentions Model for corn and soybeans did not initially account for the surge in fertilizer prices in early 2022. As a result, our model’s estimates, published early February, were higher than the USDA’s Prospective Plantings report, published March 31, 2022. However, after we enhanced our Planting Intentions Model by incorporating fertilizer prices, our estimates for corn and soybeans planting intentions showed significant improvement and were within 98% of the USDA’s March 31, 2022 report.
The Unique Power of the Gro Platform
Gro Intelligence is an AI-driven data and analytics company helping businesses, financial institutions, and governments make better decisions on food security and climate risk.
Gro has over 2 million unique, proprietary data series that address a range of questions across crop yield and production, supply and demand, growing conditions, and climate scenarios. Gro’s data series are built by our team of 50+ domain experts and scaled through AI , generating predictive models and real-time applications. Our products provide accurate forecasts and holistic views into the most pressing challenges across climate, agriculture, and the economy.
Gro’s Global Yield Forecast Models leverage high-resolution satellite data that capture every field in every district or province, not just random fields as with traditional crop survey methods, making the data more precise. Our Yield Forecast and Acreage Models are significantly more accurate, and available earlier, because of the blending of multiple unique skill sets across research and data science, comprehensive global data sources and variables, geospatial expertise, and machine and human learning.
Contact our team for a demo or to learn more about our predictive models.
Download the full 2022 performance report here.