The Challenge of a 7 Day Forecast

Every morning and every night we provide a 7 day forecast to our viewers. When I started at WVIT in 2005 it was a 5 day outlook. Why did we switch? Because viewers wanted the heads up!

Unfortunately, a day 6 or 7 forecast can be wildly inaccurate at times while perfectly spot on others. So why is this? The answer comes from West Hartford native Ed Lorenz – the famous meteorologist who discovered “chaos theory”.  Here is a quote from his landmark work Deterministic Nonperiodic Flow from the Journal of Atmospheric Science.

Two states differing by imperceptible amounts may eventually evolve into two considerably different states … If, then, there is any error whatever in observing the present state — and in any real system such errors seem inevitable — an acceptable prediction of an instantaneous state in the distant future may well be impossible….In view of the inevitable inaccuracy and incompleteness of weather observations, precise very-long-range forecasting would seem to be nonexistent.


And therein lies the problem. While we can do a very good job at initializing what the atmosphere looks like (from remote sensing, weather balloons, surface observations, etc.) that initial representation will never be perfect. The less than perfect initialization will grow into larger and larger errors as time moves on in our computer models solving equations of motion for the atmosphere.

Some storms are more susceptible to those small initialization errors on the models and some areas of the earth are relatively data-poor and initialization errors are larger there just by a function of geography.

Lorenz’s work is something that everyone sees time and again with weather forecasts. A day one forecast is more accurate than a day 4, 5, or 6 forecast almost all the time. The reason why is that the errors in our computer models grow at a non-linear rate with time.

About 7 days before this Wednesday’s storm (on March 5th) here’s what I posted on Facebook.



Now is that post of any use to someone? I’m not sure. Some say yes and some say no. I view a day 6 or 7 forecast as a “head’s up!” or a “stay tuned” kind of deal. No hype involved just telling people that there’s a possibility of something. Some people commented with “Then why say anything at all?” or “When an accurate prediction of what the weather will be in 2 days can start happening then maybe I’ll believe a prediction 10 days out”

I understand both of those comments! Personally, I would rather post about a storm potential and be honest about the uncertainty associated with that forecast.

As it turned out we did get a storm. The storm was never really hyped (by us, not sure about the others) and it turned out to be rain. It still amazes me that 7 days out we can sometimes have a pretty good idea that yes there will be a storm or no there won’t be one. In that case we were only about 150 miles away from having a damn good snowstorm.

One thing that all meteorologists need to be better at is expressing uncertainty in our forecast and when we don’t know being able to say we don’t know. One way we can do this is through ensemble modeling which runs a given computer models many different times while each time the initial conditions are tweaked. That will tell us how susceptible a given storm is to model initialization. If all 50 tweaks on the Euro Ensembles still produce 45 “snow storm hits” then we can say a snow storm is a good bet! If 50 tweaks produce 50 dramatically different results than we need to have a spine to say “we just don’t know!”

That brings me to Monday. Our computer models over the last 48 hours have been all over the place. The operational run of the European model has very little snow on Monday though it has come substantially farther north since its overnight run.



The European ensemble members as a whole, however, are substantially more “juiced” and farther north than the main operational run. Of the 50 ensemble members 40% of the produce >0.25″ liquid in 24 hours by Monday evening. The ensemble mean (take the average of all 50 members) is nearly 0.55″ of liquid in the 36 hours ending Tuesday morning.

This tells us that there’s a lot of spread in possible solutions! Many members showing no storm, some showing a big snowstorm, and others showing a glancing blow. The best thing we can do now is watch for trends and watch future runs. How will this storm evolve? Nothing to hype now other than say there COULD be a snowstorm on Monday. If we had more information we’d pass it along but at this point that’s the best I can do.


3 thoughts on “The Challenge of a 7 Day Forecast

  1. Ryan – another excellent post! Knowing how much can go wrong, I am still amazed whenever the forecast actually verifies. When someone asks me how accurate I am, I usually say I’m close to 100% accurate, but sometimes I can only give you 5 minutes notice.

    • Thanks Bob, I appreciate it. I feel like we’ve become our own worst enemy. We make forecasts so convenient and easy for people to access (web, mobile, Facebook, Twitter) that we’re giving people forecasts without context. a 7 day “snow flake” icon on a mobile app or a “model snow forecast” for 5 days out that gets shared without the accompanying caption/context is just killer for us. Research shows most people will stop watching a web video after 45 seconds so even there it’s hard to provide context.

  2. I can understand the sensitivities that on-air and public Meteorologists sense they face when generating and sharing a long range forecast with the public. Since I have a basic understanding of Meteorology, having started in a Met program at LSC in 78 (along with Nick Gregory), I appreciate the advancement made in modeling and longer range forecasting. Lorenz’s principles basically went viral outside of the meteorology community, with their overall application to non linear precesses (Chaos theory) – in fields such as engineering, an occupation that I settled in.
    With my continued interest in Meteorology as a layman, I do take issue with the Meteorology community, and the AMS in criticizing the use of models by “amateurs” as a vehicle to spread hype. That is nonsense. All possibilities in the spread of potential solutions should be considered in light of freedom of speech and expression. It’s all part of the sharing of knowledge, within the realm of possible outcomes.

Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s