Computers Confused Over Snow Predictions? - NBC 10 Philadelphia

Computers Confused Over Snow Predictions?



    Computer Models Go Haywire About Potential Snowstorm

    It's going to be a close call for snow the day after Christmas after U.S. and European models disagree. But officials said there are "numerous problems" with American prediction models. (Published Friday, Dec. 24, 2010)

    This is a weird one -- something I don't remember seeing before.

    The forecasts from the U.S. computer models are DRASTICALLY different from the ones from other parts of the world. Now, we usually get some differences between computer models, but as we get closer to the event in question, they get closer together, and the forecast becomes clearer.

    Not Friday!

    The U.S. models (known as the GFS and NAM, among others) on Friday tracked the storm much closer to the tri-state region and strengthen the storm much more than they did on Thursday.

    EarthWatch Forecast: Computer Confusion Over Snow Totals

    [PHI] EarthWatch Forecast: Computer Confusion Over Snow Totals
    Christmas day should remain dry for much of the area but as for Sunday it isn't clear after American computer models went haywire Friday.
    (Published Friday, Dec. 24, 2010)

    If we took the NAM literally, we would get a few inches in the Philly area and more than six inches at the shore.

    The GFS is even wilder. It suggests 6 to 10 inches in Philadelphia and over a foot at the shore.

    When I came into work Christmas Eve and saw those solutions, I was ready to freak out! At the same time, the foreign models (European, Canadian, and UKMET, among others) stayed with Thursday's solution of having the storm stay far enough east to give us just a little snow -- if any.

    But after looking further into this, it started making more sense.

    A statement from the National Weather Service in Washington -- where the computer models originate -- said: "Initialization errors in numerous diagnostic quantities… the specific predictions by all deterministic guidance are in question." What that means in English is that some bad data got into the basic "initialization" of the current weather.

    If we don't have accurate information worldwide on the current conditions being fed into the models, we can't trust their outcome. "Garbage in, garbage out," as the saying goes.

    So, Thursday's basic thinking held as of late Friday afternoon, perhaps with a bit of a west shift that would increase a bit the chance of at least a little snow Sunday afternoon and night.

    Be sure that I will be carefully watching new U.S. computer guidance through Friday night, and if it doesn't shift more toward the international models, I will get very little sleep worrying about what will happen.

    Fan, follow and download: Get the latest from anytime, anywhere. Follow NBC10 Earthwatch on Facebook. Sign up for our weather newsletter. And, get weather forecasts delivered right to your mobile phone -- just text PHIWEATHER to 639710 to sign up. (Message and data rates may apply.)