Tuesday, February 4, 2014

"The sky is falling! The sky is falling!"

This has been a rough winter for much of the central and eastern U.S. There has been persistent cold and storm after storm.  It has been a few years since this much of the country has experienced a winter of this magnitude.

There have been a couple of occasions this winter when unofficial  forecasts (i.e. not by the National Weather Service) for snow were released and spread like wildfire through social media and in some cases conventional media. One recent example is for the storm that is currently affecting the central and eastern portions of the country. Last Friday, January 31, a forecast of up to 2 feet of snow for Iowa, northern Illinois, and southern Wisconsin for today received more than a little attention in social media.

A forecast through Wednesday, February 5 issued on January 31.

So what was this all about?  This forecast for five days out was someone's interpretation of the results from some computer model. I don't know who produced this forecast nor what they used to produce it.   The bottom line is that the capability to accurately forecast snow amounts and location this far in advance just doesn't exist.

The atmosphere is very complex, as are the computer models that simulate the atmosphere. The accuracy of these simulations depends a lot on the initial conditions the model start with. These initial conditions are usually weather observations, but the density of weather observations varies and assumptions or estimations  have to be made where there is sparse data. If these assumptions are off or incorrect then the performance of the model suffers, and the result (the forecast) gets worse the farther out in time you go.  This is just one aspect of forecasting snow. Snow amounts are tough to forecast. We tend to think of snow as a generally uniform blanket, but snow amounts can vary in much the same way rainfall varies in the spring. That's the reason snowfall forecasts are a range of amounts, for example, "6 to 10 inches". Snow storms can have convective elements, i.e. "thundersnow" that produce intense snowfall rates over relatively small area of the storm's footprint. Bands of snow may sit over the same area producing higher amounts of snow than areas around them.

The Chicago office of the National Weather Service put together an excellent video describing the complexities in forecasting snowfall amounts, including the uncertainty in the computer models and the accuracy of snowfall forecasts made from one to six days out. It's well worth the time to watch.


  1. Great blog. Thanks for the video. It was a real eye-opener when it showed how far off the models forecasted the changes in the jet streams. Thank you!

  2. Very informative. Thank you! I only wish the general public could see this so they wouldn't be so hard on weather forecasters! :)