U.S. model could drastically improve winter weather predictions

Canadians could have a better idea of how cold winter will be and how much snow they'll be shovelling if weather experts in this country adopt a new U.S. methodology that promises more accurate long-term winter forecasts.

Scientists at North Carolina State University say they've created a new way to forecast temperatures and to estimate precipitation that will increase accuracy by 20 per cent.

The new ability would also be important for governments and utilities, since it gives key information for use in predicting energy consumption and water availability, they say.

"Predicting winter precipitation is extremely useful because winter is the most important season in terms of recharging water supplies . . . and ensuring water will be available in the summer," says lead researcher Sankar Arumugam, a climatology expert.

Environment Canada senior climatologist David Phillips says the new method — which factors in temperatures and conditions in the Pacific Ocean on top of the usual sea surface temperatures and past ice and snow conditions — could mean more accurate and timely forecasts.

"I think it's really the future. It's the way we have to go. There is hope that there is credibility, timeliness and accuracy to this," said Phillips.

As for whether Environment Canada might consider adopting the new methodology, he was noncommital.

Meteorologists currently use multiple general climate forecast models (GCMs), which are composed of sea surface temperatures and past ice and snow conditions, to create forecasts, but Arumugam says the approach can be improved by more closely considering Pacific weather factors as well.

The current model used by Environment Canada to predict its three-month long-term winter outlook monitors GCMs twice a day for five days before Dec. 1, but doesn't closely consider weather patterns in the tropical Pacific Ocean.

In the U.S., weather predictions are currently about 40 per cent accurate, Arumugam said, but by taking into account both GCMs and patterns in the tropical Pacific between 1951 and 1996, Arumugam's researchers found they were able to reduce uncertainty in winter predictions in the U.S. by 20 per cent.

False alarms — when predictions are far off — are significantly reduced when following the new method, according to Arumugam.

Phillips likes Arumugam's theory.

"Our models never tell us why it's saying it's going to be warmer than normal or colder than normal. It just spews out of the computer, but hey, it could be the opposite and that's been embarrassing in the past," Phillips said.

He believes new ideas on improving predictions will steadily increase as experts continue to monitor patterns.

"It's a growing kind of understanding and there is hope. It's more than a Farmer's Almanac and it's more than the using-a-groundhog now," he said.

Arumugam's findings were published in the latest issue of Monthly Weather Revi