Consider me highly skeptical about their value.
Nevermind, the skill in forecasts rapidly deteriorates beyond 5-7 days and is typically no better than judging what the weather will be based on climatology (historical averages). Today, AccuWeather delivers the forecast out to Saturday, April 28 to every zip code in the U.S.
For Washington, D.C. on 4/28, we should expect mostly cloudy skies with “showers around” and a high of 64 and low of 46. But don’t worry, if you buy this unrealistically precise forecast, just two hundredths of an inch of rain will fall - your outdoor picnic won’t be washed out.
Let’s get real for a second: I can hardly say with a lot of confidence whether 0.02” is a good rainfall prediction for today much less more than three weeks from now.
But the folks at AccuWeather are convinced there is value in these forecasts.
“People are really going to find this useful,” Accuweather senior meteorologist Henry Margusity told me. “If you’re doing planning, you’ll have a lot of information.”
Margusity stressed that these long-range forecasts will be complemented with additional data to provide context for the predictions, including normals (averages) and records. He admitted the accuracy will diminish with time, but emphasized users should focus on the trends.
The new forecasting scheme accurately predicted above normal temperatures during the recent Midwest heat wave according to Margusity.
“It caught the warm spell coming,” he said.
The 25-day forecast originates from a variety of datasets and models. Margusity would not divulge specifics about how the forecasts were generated, noting the methods are “proprietary.”
Margusity wasn’t surprised when I expressed my skepticism about these forecasts.
“They’re always doubters,” he said. “People are going to say it can’t be done, but it can be done… we have enough information.”
When I told Wes Junker, CWG’s winter weather expert, about the new 25-day forecast, he wasn’t impressed.
“It gives people a false sense of where the science is,” he said. “[In daily forecasts], there’s no accuracy beyond 7 days at all.”
Junker said there is some skill in providing a forecast for a multi-day period, and how it might compare to average but not in daily details.
“You can’t tell the day the storm is going to be [beyond 7 days],” he said.
CWG’s Steve Tracton, who worked on numerical weather prediction issues for three decades, was harsher in his assessment of the 25-day forecast concept.
“It undermines the credibility of the science of meteorology,” he said.
“There cannot be skill at those ranges - it goes back to chaos theory.”
Tracton wants AccuWeather to objectively demonstrate their forecasts have skill not only out to 25 days but even in the 7-15 day timeframe, something he says they haven’t done.
“Despite many inquiries, they have promised me, but never provided verification statistics,” he said.
He added: “If they can prove their 25 day forecasts have skill, that would be equivalent to demonstrating neutrinos can move faster than the speed of light, violating Einstein’s fundamental theory.”
All of this skepticism notwithstanding, I’m going to give these forecasts a chance.
For the next 25 days, I will keep track of the 25-day forecast for Washington, D.C. and then I’ll evaluate how well they performed compared to climatology (historical averages).
I’ll issue a post-mortem report in late May or early June. Stand by....