84th AMS Annual Meeting

Tuesday, 13 January 2004
An experiment in probabilistic quantitative snowfall forecasting
Room 4AB
Alan M. Cope, NOAA/NWSFO, Mount Holly, NJ; and M. P. DeLisi
Poster PDF (125.0 kB)
For the past two winter seasons, beginning in December 2001, the National Weather Service at Mount Holly, New Jersey (PHI) has issued experimental probabilistic quantitative snowfall forecasts (PQSF) for potential winter storms in its forecast area. The intent behind these forecasts is to convey the uncertainty associated with forecasts of snowfall amount, and to help focus attention on the most likely snowfall amounts, rather than the highest possible totals.

The PQSF forecasts are issued as tabular text products which give the probability that storm-total snowfall amounts will exceed any of seven thresholds, ranging from 2 inches to 24 inches. The seven probabilities are listed for five sites in the PHI forecast area: Philadelphia, Allentown and Mount Pocono in Pennsylvania; Atlantic City, New Jersey; and Georgetown, Delaware. To prepare a PQSF product, the forecaster refers to a table of pre-computed probability values, according to the mid-range of the storm-total snowfall amount expected at each location. These pre-set probabilities were derived from a statistical analysis of observations and forecasts for past winter storms. Forecasters are free to modify these values based on their confidence in the storm track and intensity; however, anything beyond modest changes to the probabilities is discouraged.

The first winter of the experiment, 2001/02, was unusually warm and dry in the mid-Atlantic region and testing was very limited. Winter 2002/03 was much more active, and PQSFs were issued for 11 snow events. We have verified the PQSF values by comparing them with snow reports from numerous cooperative observers in the region around each of the forecast sites. For most storms, the percentage of coop reports exceeding the mid-range forecast snow amount for a given site is in good agreement with the corresponding PQSF probability value. An exception was the Presidents' Day storm of 2003, when snow amounts were significantly underestimated.

At the conference, we will present a detailed explanation of the development of our PQSF technique, including some background information and the motivation behind the PQSF. We will show examples of its use during the winter 2002/03. We will explain our verification methodology and present the full verification results for the past two winters. Finally, we will outline our plans to promote greater use of the PQSF product for winter 2003/04.

Supplementary URL: