To paraphrase an old saying: Everybody talks about the impact of extreme weather, but nobody does anything about it. 

That’s changing. By rethinking traditional predictive inputs and applying the latest advances in AI and machine learning technology, scientists can improve the accuracy of subseasonal-to-seasonal (S2S) forecasts to levels that were previously unattainable.

Case in point: In March of 2021 our company, Salient Predictions, used a proprietary forecast model to predict what proved to be a historic drought in the Upper Midwest — North Dakota and a portion of South Dakota in particular — when it was still several months away.

It’s not our intention to boast about this accomplishment. Rather, we’re simply illustrating what’s possible when you take a different approach to S2S forecasting. Instead of relying on legacy modeling approaches, which are derived from rapidly changing atmospheric conditions, our model looks for long-range indicators such as sea surface salinity, ocean-driven oscillations, and other teleconnections.  Machine learning technology then enables us to detect weather patterns that were previously unseen.

Many stakeholders across the agriculture industry could benefit from these more accurate S2S forecasts:

• Farmers. The traditional lack of reliable S2S forecasting has forced farmers to take a reactive approach to extreme weather events such as flooding or drought. Affordable crop insurance in the US, made possible through government subsidies, has proven to be a lifesaver at times. But more accurate S2S forecasts could also enable farmers to implement climate-smart farming practices that prevent or reduce losses in the first place. In addition, climate-smart farming could mitigate impacts not always, or not often, covered by insurance, such as diminished crop quality or increased costs of production. Climate-smart farming is essential in countries where crop insurance is more expensive or not available. And it lessens negative environmental impacts everywhere by helping farmers adjust irrigation practices and application of fertilizer and pest-control products in response to more accurate seasonal forecasts. In short, climate-smart farming and better forecast models go hand in hand.

• Insurers. For both private insurers and the federal government, the unpredictability of climate change and extreme weather events will make it difficult to accurately measure risk. This could potentially lead to financial instability in the insurance market. The better the S2S forecasting data, the better the odds of keeping crop insurance programs viable. Of particular value, machine learning can enable more precise forecast models that predict extreme weather events such as seasonal drought on a regional scale. That allows for more informed risk management and portfolio construction. Insurers could also offer farmers incentives to use climate-smart farming practices tied to more accurate forecast models.

• Ancillary Industries. Beyond farmers and insurers, there’s a whole range of agriculture-related businesses that are impacted by disruptive weather events. From suppliers that provide seeds and crop-protection products, to food-and-beverage producers that depend on predictable crop yields to meet their projections, a large sector of the national economy depends on a stable agricultural industry. As such, they also have a significant incentive to support research into improved S2S forecast models.

Politicians also need to be part of this equation. We encourage elected officials in the US and across the world who represent regions that depend heavily on agriculture to support increased funding for research on climate science and weather forecasting. This includes not only traditional research grants, but also funding for nontraditional programs such as the “forecast rodeo” conducted by the federal government’s Bureau of Reclamation.

Salient Predictions owes its existence to that program. In 2017-2018, our founder, climate scientist Ray Schmitt, teamed with his sons, machine learning engineers Eric and Stephen, to win the Bureau’s S2S Forecast Rodeo using a novel approach that tied water evaporation from the ocean to rainfall in the West.

Our experience has shown what’s possible. To meet the demands of our rapidly changing climate, we need new approaches in everything from forecasts to funding.

‍

Anthony Atlas, VP Business Development at Salient Predictions, is a cofounder of Stanford Alumni in Food & Agriculture.

‍