Accuweather calls its new 45-day weather forecast launched Monday “revolutionary.” I call it a joke.
By and large, weather forecasts beyond 7-10 days in the future simply do not have “skill.” That is, they provide no more accurate information than the historical average conditions in that time horizon.
AccuWeather insists it is innovating and pushing the envelope by providing these forecasts. In its press release, it boasts of being a leader in long-range forecasting.
In reality, it is simply peddling a useless product to people who don’t know better.
Let’s examine one of AccuWeather’s own scenarios demonstrating the supposed usefulness of the 45-day forecast, published in an article on its Web site:
If the 45-Day Forecast shows an expected high of 85 degrees and sunshine for Labor Day, the normal high for the date is 80 degrees and the record high for the date is 98 degrees, a lot of information can be gathered. Conclusions such as “an outdoor BBQ or swimming should be fair game” can be made.
The claim that the 45-day forecast is providing value-added information in this case is ludicrous. For one, you don’t need a 45-day forecast for Labor Day to know that Labor Day weather is usually warm and occasionally hot. All you have to do is look-up the averages and records. And, while occasionally long-range forecasts can give you a credible sense as to whether temperatures will be above or below normal, they cannot – under any circumstance – provide accurate information about whether it will rain or not.
Getting the timing specifics of storminess right even a few days into the future is incredibly challenging. 45 days into the future it’s impossible.
It is incredibly misleading for AccuWeather to imply its long-range forecasts can help planners – whether it be for a wedding or outdoor sporting event – determine whether and when to hold their event.
I’m not the only one who feels this way.
As a refresher, let me revisit some of the reactions from some of CWG’s senior meteorologists when AccuWeather launched the 25-day forecast in 2012:
Wes Junker, CWG’s winter weather expert, with over 30 years of forecasting experience said: “It gives people a false sense of where the science is. [In daily forecasts], there’s no accuracy beyond 7 days at all. You can’t tell the day the storm is going to be [beyond 7 days].”
Steve Tracton, who holds a Ph.D. in meteorology from MIT, said: “It undermines the credibility of the science of meteorology. There cannot be skill at those ranges – it goes back to chaos theory.”
When AccuWeather launched its 25-day forecasts in 2012, I tracked their accuracy in Washington, D.C. over a period of 30 days (April 28 – May 28, 2012). I found, not surprisingly, that its temperature forecasts 25 days into the future were no better historical averages – about 6 degrees off for both the high and low temperatures. On 6 of the 30 days, it had forecast errors exceeding 10 degrees for high temperatures.
Its rainfall forecasts were also problematic. It predicted rain on just 7 of the 14 days rain actually fell. The day it rained the most (1.22″ on May 15), it predicted no rain or “clouds and sun”.
To my knowledge, AccuWeather has never (in the last 10 years at least) published a rigorous analysis of how (in)accurate its forecasts are – which isn’t a particularly transparent approach for the company, given its name.
Instead, it serves up press release propaganda.
“The [45-day] forecast is the result of extensive scientific research and is backed by more than 110 meteorologists, providing the public with the Superior Accuracy they count on from AccuWeather,” says Steven Smith, AccuWeather’s Chief Digital Officer.
I fully appreciate AccuWeather’s desire to be on the cutting edge of weather forecasting, but doing so just because it can and because some of its customers may like it, doesn’t make it worth doing or valuable.
If AccuWeather wanted to innovate responsibly, rather than providing forecast details out to 45 days, it could provide some very general information about the direction of weather trends at this range. For example, it could list historical averages at long range complemented by colorful, easy-to-interpret graphical symbols to indicate whether temperatures and precipitation were favored to be above or below those averages. It could also provide confidence indicators, which would educate its customers about the lack of certainty in longer range forecasts.
But instead, AccuWeather has chosen to extend its scientifically-lacking approach, which is impossible for me to defend and will continue to be the subject of ridicule among meteorological peers.