Photo Credit: Brian Mosley, CRA Policy Analyst

Well, that was a whopper!

The Washington, DC area received its first blizzard in many years this past weekend. For a city that normally only receives 15.4 inches of snow spread out over an entire season, the two plus feet of snow has been overwhelming for the city. DC is not equipped to handle this amount of snow.

But, we were warned. And, for the most part, the weather forecasters were right on target.

All thanks to amazing computer models that predicted the storms path and total accumulation with stellar accuracy 8 days before the storm hit the region.

This was highlighted by the Washington Post:

This forecast success would not have been possible without the state-of-the-art forecast modeling available to us thanks to decades of investment into computing resources by the U.S. and international weather enterprise (public, private and academic sectors), as well as advances in our theoretical understanding of these storms.

Not more than a week before the storm, the Washington Post reported that the National Weather Service (NWS) tripled their computing power to generate more accurate forecasts.

Currently the system has two supercomputers, from IBM and Cray, located in Reston, Virginia, and Orlando, Florida.

They are now running at 2.89 petaflops each for a new total of 5.78 petaflops of operational computing capacity, up from 776 teraflops of processing power last year, the NWS said in a press release.

This is all good news for the area of computing uncertainty. The Computing Community Consortium (CCC) held a workshop in 2014 on the Uncertainty in Computation. One of the talks was on the Uncertainty in Weather Forecasting by Eugenia Kalnay, from the University of Maryland. Kalnay noted that ensemble forecasting allowed the NWS to increase the public predictions from 3 to 7 days. See the workshop’s full report for additional information on her presentation and other topics in computational uncertainty.

The temperature was one aspect of the storm that the computing models did not predict correctly. Thankfully, the temperatures were mostly in the low-to-mid 20s rather than the predicted 25 to 30, which meant the snow was light and powdery rather than heavy and wet. As a result, power companies did not receive the number of power outages that they were anticipating. This was one prediction that many Washingtonians were very happy did not come true.

We’ve come a long way, but there’s always new innovations coming thanks to scientific research.