7 August 2013

Is Roger Rabbit Running Accuweather??

Posted by Dan Satterfield

Screen Shot 2013-08-06 at 8.24.42 PMAccuWeather announced today that they are now producing “revolutionary” 45 day weather forecasts. Yes, you read that right, and while the public in general has a rather low understanding of science, I don’t think most people are that gullible. My first thoughts were that it might be a way of getting some free advertising, but perhaps they’re really serious.

Let me first be clear and tell you that synoptic weather forecasts are about 90% accurate out to 1 day, and are fairly accurate out to 5 days most of the time. Beyond 7 days the best forecast is to use is the 30 year averages. Not that there is no skill in long-range forecasting, but it’s more on the line of forecasting a region to be above or below normal, and wetter or drier than normal. Day by day, high and low forecasts (with rain chances) are just not reliable beyond 7 days.

If anyone tells you otherwise, they do not know what they are talking about.

Some comments on my FaceBook page today when i posted the news:

Screen Shot 2013-08-06 at 8.54.17 PM

There has already been a lot of reaction among meteorologists around the country, and laughter is probably the best way to describe it. Washington Post Meteorologist Jason Samenow has some quotes, and a similar take to mine. Worth reading.

I do not want to give you the idea that forecasting has not improved, because it is actually scientific success story. Check out this excerpt from the 2007 Policy Statement of the AMS on Weather Analysis and Forecasting:

Short-range Forecasts (12 hours to a few days)
Over the past several decades, improvements in observing systems, computer model physical processes, and assimilation of data into numerical weather prediction systems have produced steady improvement in the ability to predict the evolution of larger-scale weather systems as well as day-to-day variations in temperature, precipitation, cloudiness, and air quality.  Examples of these improvements in the short range include:

  • The 2006 average 48-hour forecast hurricane track error in the Atlantic basin was 111 miles, as compared with 336 miles in 1985.
  • 48-hour precipitation forecasts are now as accurate as 24-hour forecasts were a decade ago.
  • Winter storm watch lead time for the season ending in 2006 was 17 hours, an increase of 70% since 1999.

Significant room for improvement remains in forecasting high-impact weather such as hurricane intensity and winter storms.   For example, forecasting rapid intensification (or weakening) of hurricanes and differentiating between snow, sleet, freezing rain, and rain with narrow transition zones in winter storms remains a forecast challenge.  Advances in computer models and improved understanding and representation of cloud processes in the models will increase the ability to predict these small-scale features.  Together with advances in observations and methods of incorporating data, these steps will lead to enhanced forecasts of precipitation intensity and location. 

..and more from the statement on medium and long ranger forecasts:

Over the past three decades the skillful range of medium-range forecasts has been extended by roughly one day per decade. Major winter storms are now often forecast a week or more in advance, allowing road maintenance personnel and emergency managers time to prepare.  Examples of other advances in medium-range forecast skill include:

  • Three-day forecasts of 1 inch or more of precipitation are as accurate as two-day forecasts were in 1998.
  • The skill of five-day forecasts has more than doubled since the late 1970s. 
  • For major cyclonic storm location and intensity, five-day predictions are as skillful as three-day forecasts were in the early 1990s.
  • Surface temperature forecasts for the U.S. now show considerable skill on days three through five, with the skill decreasing to more marginal levels by day six.

Extended-range Forecasts (week 2 and beyond)

The current skill in forecasting daily weather conditions beyond eight days is relatively low.  However, products designed to highlight significant trends (e.g., warmer than normal, wetter than normal), such as 6–10 day and 8–14 day temperature and precipitation probability outlooks, often have useful skill.  The skill of operational forecasts of U.S. temperature and precipitation for an average of 6–10 days has more than doubled since the 1970s.  Precipitation forecasts are less skillful than those for temperature.  Forecasts for 8–14 days (week 2) have been issued since 2001.

AccuWeather has not only damaged their reputation, but also done a greatly disservice to the public.If you want to know the weather beyond seven days, and especially beyond 10 then look up the 30 year average. We meteorologists can sometimes tell you it may be a bit warmer or cooler than average but not more. Forecasting the high or the low temperature 40 days in the future is nothing more than a scam.