How well can you predict a storm? Accuracy of numerical weather models varies

In late October, a numerical weather model advertised the possibility of gale-force winds, rain, and high sea and swell on Friday through Sunday.

Sure enough, stormy weather developed. It was remarkable that a model could predict these unsettled conditions eight days in the future.

The European Centre for Medium-Range Weather Forecasts (ECMWF) model called for rain, while the National Oceanographic and Atmospheric Administration’s Global Forecast System (GFS) model did not until just a few days before this fall’s first rain event arrived.

This led to a few readers asking why do these numerical models differ so much in their predictions?

To start, the United States pioneered the groundbreaking science of computer numerical modeling. These numerical models are collections of mathematical formulas, usually run on powerful computers, which produce forecasts for a specific location over several time intervals.

These models are almost indispensable in giving guidance to forecasters.

The models perform billions of calculations to simulate the motion of weather patterns in the Earth’s chaotic atmosphere. This type of forecasting is possible because movements of the atmosphere follow natural laws expressed in mathematical equations.

British mathematician Lewis Fry Richardson first proposed numerical weather prediction in 1922. He wasn’t very successful because, without modern computers, it would take him nearly three months of performing calculations to produce a weather forecast that only predicted the weather 24 hours in the future.

But his work laid the foundation for the first real success by the U.S. Joint Numerical Weather Prediction Unit in 1955.

For the most part, weather models are either global, covering the entire Earth, or regional, covering only part of the planet.

As you may imagine, models become more accurate the closer to the predicted day because there’s less time for inaccurate oceanographic and atmospheric data to be amplified at model initialization.

Over the decades, the models have improved.

However, according to most in the professional weather community, the ECMWF model has proven more accurate than the GFS model. The ECMWF model predicted that Hurricane Sandy would swing toward the East Coast two days before the National Oceanographic and Atmospheric Administration (NOAA) model predicted it. The death toll would likely be much worse had it not been for the advance warning from the ECMWF forecast.

Cliff Mass has written extensively about this subject over the past several decades, including peer-reviewed literature; he’s also testified before Congress, and served on several NOAA/National Weather Service advisory committees and National Academy panels dealing with these issues.

“The U.S. numerical weather prediction, the cornerstone of all U.S. weather prediction, is behind other nations and far behind the state-of-the-art,” Mass wrote on his blog. “Our global model, the GFS, is usually third or fourth- ranked; behind the European Center and the U.K. Met Office, and often tied with the Canadians.” He wrote on his blog.

“To address this problem, Congress passed legislation (the National Integrated Drought Information System Reauthorization Act of 2018), which instructs NOAA to establish the Earth Prediction Innovation Center (EPIC) to accelerate community-developed scientific and technological enhancements into operational applications for numerical weather prediction (NWP).” Mass wrote.

Early this year, a Request for Proposals (RFP) for EPIC was issued, and offerers had until May 2020 to submit proposals. NOAA plans to make the award by the end of this year.

“Through EPIC, the United States has a unique opportunity to harness the talents of the most brilliant modelers in the world to advance operational global numerical weather prediction,” said Neil Jacobs, acting NOAA administrator. “Advancing our operational weather modeling capability will improve forecasts and lead to more resilient communities.”

In the coming years, it will be interesting to see how much further the models can improve.

How to get discounts on PG&E energy bills

Apply for a monthly discount through CARE or FERA, PG&E discount programs that help eligible customers pay their energy bills. More than 1.4 million customers are receiving a bill discount through these two programs. To learn more, please visit www.pge.com.

This story was originally published November 10, 2020 at 5:05 AM.