Case study
Unlocking Economic Gains with Beyond Weather's New US Yield Forecaster
Download this case study to access it later offline
At Beyond Weather, scientists have developed the first operational soy and corn yield forecasting system that delivers accurate predictions up to 8 months ahead, i.e. before seeds hit the soil. Powered by AI, our Yield Forecaster searches Big climate data for predictable signals, trained to forecast yields and key meteorological variables like temperature and soil moisture. Economic analyses reveal the significant competitive advantage our forecasts offer to agricultural stakeholders and commodity traders, setting new standards in the industry.
The importance of accurate yield predictions
The importance of breadbaskets for global food production and prices cannot be overstated. The United States breadbasket alone provides 30% of the global production of both soy and corn. Weather variability, especially extreme events like droughts, can drastically impact harvests. The 2012 US drought, for instance, caused dramatic price spikes in both regional and global markets. To stay ahead of such fluctuations, commodity traders and agricultural stakeholders must anticipate these changes to mitigate risks and optimize trading strategies.
Beyond Weather's Yield Forecaster
Earlier work by Beyond Weather scientists identified the dominant climatic processes driving weather variability throughout the US growing season. They’ve also developed and trained advanced AI methods to directly forecast yields. Building on this fundamental science, Beyond Weather has now fine-tuned and launched its operational US Yield Forecast system. “This is a landmark moment for our young start-up company”, says CEO Jannes van Ingen. “We not only demonstrate that we can provide accurate forecasts but also that they provide real economic value, providing a clear use case for agricultural stakeholders.”
"We not only demonstrate that we can provide accurate forecasts but also that they provide real economic value, providing a clear use case for agricultural stakeholders."
Agricultural stakeholders typically rely on the USDA's yield outlooks, which are based on crop planting data but lack weather insights. The USDA releases its first US crop outlook in May, updating them monthly until a final production report in December. The August report is regarded as a fairly good estimate of the production, since the crucial crop growing period takes place in July and August. In contrast, Beyond Weather’s soy forecast is accurate as early as February. The forecasting system is especially good at forecasting in those years where we see big off-trend yield swings. These years are called medium (outer 50% of the distribution) and high (outer 30% of the distribution) confidence years. In medium or high confidence years, the error in Beyond Weather’s February forecast is respectively 35% and 42% smaller than USDA’s August outlook. Looking at all years combined, Beyond Weather’s February forecast still outperforms USDA’s August outlook by 10%. Thus, Beyond Weather is more accurate and nails it at a much earlier stage.
Mitigating financial risks
Financial analyses using basic trading strategies demonstrate the proficiency for agricultural stakeholders to mitigate financial weather risks when using Beyond Weather’s Yield Forecaster. The returns from these strategies show that these forecasts are not only an interesting insight, but give actionable and reliable results. Climate change is already making weather extremes, and in particular drought, more intense, a trend that will continue in the years ahead. Agricultural stakeholders will thus be exposed to ever growing risks. Effective forecast systems like Beyond Weather’s Yield Forecaster will thus become indispensable tools for mitigating risks and optimizing trading strategies.
“An important aspect of our Yield Forecaster is that we can assign confidence levels to individual forecasts” says Dr. Sem Vijverberg, Chief Technical Officer at Beyond Weather. “In some years, the climate system has better predictability than in others. We know this when we do our forecast, enabling us to assign confidence levels”. The yield forecast for 2024 has a medium-to-high confidence level, making it valuable for trading purposes.
"In some years, the climate system has better predictability than in others. We know this when we do our forecast, enabling us to assign confidence levels."
Expanding globally
Currently Beyond Weather is shifting its efforts towards the South American breadbasket. “Both Argentina and Brazil are equally important soy producing regions as the US” says Vijverberg. ”Training our AI methods for these regions is giving promising results and we plan to extend our operational Yield Forecaster to include South America later this year”. In the longer term, Beyond Weather is dedicated to bring the best AI-powered forecasts to regions around the globe, with a focus on the Global South. Van Ingen concludes: “We are a team of climate scientists and entrepreneurs who care about the escalating impacts from climate change, hitting vulnerable nations in the Global South particularly hard. It is our mission to save lives, money and resources by contributing to effective early warning systems here”.