The Challenge of Forecasting Snow

Some of DC’s largest snowstorms have occurred during the second half of January, including the infamous “Knickerbocker” Storm of 1922 and, more recently, DC’s fourth largest snowstorm just two years ago.  A high degree of skill is required when forecasting snow in the DC Metro Area given the complexity of winter storms and the unique geography of the Mid-Atlantic Region.  

Several of the record snow events in the Mid-Atlantic Region over the last 20 years were well-forecast in advance.  For example, the January 22-23, 2016 event that tied for fourth largest snowfall in the Nation’s Capital and was first mentioned in forecasts the previous weekend.  Also, area meteorologists, including my colleagues and I on the WUSA9 weather team, gave area residents several days advance notice for high impact storms in December 2009, February 2003 and January 1996.  

These storms were all such blockbuster events that the computer models meteorologists use were able to identify the seeds that became the storms days in advance.  The farther out from an event you are, there is a higher degree of fluctuation between computer model runs.  That leads to an old weather adage in the weather community that “the trend is your friend.”  If several models show a particular storm or event over several model runs, then meteorologists have more confidence that said event might occur.

There has also been an incredible amount of improvement in the forecasting models as computers have become more powerful and the mathematical equations used have been refined to account for bias.  But there still remains an inherent amount of uncertainty in weather forecasting, with each model having its own bias.  That’s why meteorologists are so important since they recognize the bias associated with each model and can adjust their forecasts accordingly.

Despite the increased accuracy in forecasts due to advances in numerical weather prediction, forecasting “busts” do occur on occasion when an anticipated winter storm either doesn’t materialize or more snow than originally expected occurs.  For example, some computer models were suggesting on March 2, 2001 that a major winter storm with possible blizzard conditions would develop in the DC Metro Area on March 5-7.  While snow totals of 12”/+ did eventually occur, they were further north in New York and New England.  Many DC Area school districts had closed in anticipation of a significant winter storm, but only 0.2” fell at National Airport on March 5, 2001.

Ironically, today is also the 18^th anniversary of another numerical weather prediction failure.  On January 23, 2000, a storm was expected to be too far east to have any significant impact on the DC Metro Area.  However, by 11:00 pm on January 24, local meteorologists recognized that it would be a high impact storm for the Nation’s Capital.  Most people had already gone to bed and woke up to a surprise snow day on January 25.  Snow totals of 9” to 18” were common from I-81 to the Chesapeake Bay.  Many residents held local media outlets and meteorologists responsible for the unwelcome surprise.  This remains one of the more significant numerical weather prediction failures in recent memory for the Mid-Atlantic Region.

Some take-away lessons are that while weather forecasting continues to improve, forecasting “busts” do happen.  People need to pay attention to the latest forecasts for any changes.  Weather forecasting is an evolving process and it’s vital to get the latest updates from reputable sources.