In March of 2021 Salient’s proprietary forecast model detected the possibility of a significant drought in the Upper Midwest -- including North Dakota and a portion of South Dakota in particular -- during the upcoming growing season. Our forecast, represented by a 30-day rolling average of daily precipitation, predicted below-average precipitation from mid-April to August 1. (We released an updated forecast each week, which remained consistent throughout the period.)

How does a retrospective, or “hindcast,” of our 2021 forecast compare to actual conditions? Our model predicted what proved to be an historic drought when it was still several months away.¹ That’s something unique to Salient -- and the implications of this novel sub-seasonal-to-seasonal (S2S) forecasting technology for the agriculture industry are far-reaching.

Why Did We Focus on the Dakotas?

Because the Dakotas produce significant yields of oats and barley, the repercussions of a drought in that region could be significant to anyone whose livelihood depends even indirectly on those yields. The ability to foresee drought conditions across an entire growing season and prepare for diminished yields would be of enormous benefit to three primary groups of stakeholders:

1. The food and beverage industry. This ranges from brewers to cereal and snack producers -- any company that sources oats and barley for their products. If there’s a shortage of oats and barley due to drought or other extreme weather conditions, this can lead to higher prices on the commodities market. Producers risk buying grains at higher prices, which causes their own prices to go up. If you can’t make up the shortfall even using the spot market, you’ll lose sales. Or you might have to transfer ingredients from other regions. And even if the supply of grain is adequate, there may be quality issues. Maybe the grain has higher protein content under drought conditions, creating a need to mix in a lower-protein source to hit target specifications for a particular recipe.
2. Farmers. If they knew a drought was coming, they could mitigate the impact with informed decisions. Maybe they could sow their crop at a different time and avoid the worst of the heat and the drought at a key period of the crop’s development. That would give it a better chance of reaching maturity and producing a higher yield. Or they could adjust the planting rate to create a lower density and improve soil moisture availability per plant. Or they could choose cover crops that might help retain moisture going into the season. They could also take ancillary measures such as not applying nitrogen preparations that are ineffective in drought conditions. This would not only save money but also reduce groundwater contamination. 
3. ‍Agriculture-input companies. These companies provide seeds that have different traits, crop-protection products, nutrients, and so on to produce the greatest yield for different crops under different conditions. In addition to helping determine when to ramp up production of particular strains for anticipated conditions, accurate S2S forecast data could help input companies provide better advisory services. And because our forecasts are made at grid point resolution (25km), companies can fine-tune advisories to a local level. If you know the drought conditions will be severe in a fairly confined area, there’s no reason to change your approach across the board and supply drought-resistant strains in areas where they won’t be needed. That helps build trust with growers.

How Did We Get the Forecast Right?

Conventional wisdom holds that it is impossible to forecast weather conditions beyond two weeks with meaningful accuracy. But that conventional wisdom is rooted in a conventional approach. Government-run dynamical models tend to focus on atmospheric conditions, which is most effective in the short-range (<2 weeks). However, insight into long-range weather requires more emphasis on ocean data -- oceans have high heat capacity and inertia, making them a key contributor to seasonal weather patterns. Salient’s proprietary S2S forecasting models incorporate these stable, consistent inputs, offering us greater accuracy and reliability on much longer lead times. As a result, companies have an opportunity to start thinking differently about business planning than they have in the past.

And different thinking is necessary. Weather volatility is increasing worldwide. Companies that have created their supply-and-demand curves based on trailing or historical weather averages for the most recent 5-10 years often find that their assumptions don’t match what actually occurs. Extreme fluctuations from year to year are becoming more common.

The ability to predict those fluctuations is especially critical to those on the front lines of climate change. This is particularly true of drought and extreme heat, which a growing number of scientists agree are directly linked to global warming.

Farmers along the 100^th meridian will be among the first to feel the impact. Historically, there’s been more precipitation east of the 100^th meridian and drier conditions to the west. But as that demarcation shifts eastward, the people on the leading edge will have to be very agile in adapting their processes. If you’re in an area like Iowa or Illinois, where the soils have a ton of yield capacity and high water retention, you might not feel the impact of drought conditions yet — but you will soon. Scientists have a higher degree of confidence that climate change is a contributing factor in the severity and frequency of droughts, as compared to other extreme event types.

Rapidly changing climate conditions will impact not only agriculture, but also a range of other industries, including fisheries and energy production. We all need to start planning for the climate of the future — and Salient’s novel S2S forecasts have made that possible.

-----

¹ Salient's mission is to put scientific knowledge to work to help people and organizations better understand and prepare for the increasing volatility of climate change. Climate and weather forecasting is complex and dynamic. The blog content accurately represents how Salientʼs forecasting models combine research-based data insights into our machine learning algorithms. However, due to the inherent uncertainty in long range weather and climate forecasting, measuring performance over a longer time scale rather than specific events provides a much more representative and accurate view. In addition to calibrated probabilistic forecasts, Salient offers a suite of metrics to assess the forecasts' quality and benchmark Salient's skill against the best publicly available models. Additionally, Salient models are back-tested over an extended period of 20 to 30 years to verify performance.