Using post-processed short-range ensemble products to enhance precipitation type forecasts: the 1 December 2007 Upper Midwest storm

A fast-moving, higher predictability winter storm struck the Upper Mississippi River Valley on 1 December 2007. This was the first storm during the 2007-2008 Winter to impact the public with freezing rain, sleet, and heavy snow. Across most of Iowa, 5-12 mm of ice accumulation occurred, mainly from freezing rain, while further north across central Minnesota and northern Wisconsin, heavy snow fell (100-175 mm). Geographically situated between those two regimes, a period of rapidly accumulating snow occurred (75-80 mm/3 hr), followed by a 5-h period of pure sleet (25-50 mm), then freezing rain (3-6 mm), then rain and drizzle - all within a 15-h period.

Predictability for this winter event was higher than average with a strong signal in the deterministic and probabilistic forecast models. This above-average predictability derived from similar and consistent deterministic model solutions led to higher confidence forecasts. National Weather Service winter warning lead times were in excess of 24 hours for some locations affected by the storm. Forecast information from the National Center for Environmental Prediction's Short-range Ensemble Forecast (SREF) system provided operational forecasters an improved visualization of the threat and its timing, and further increased forecaster confidence in the solution. Post-processed SREF data, those that provide unique guidance aside from the standard ensemble means and spreads, are available from the Storm Prediction Center web site (Bright et. al., 2007) and also from a collaborative web site from the Pennsylvania State University and the National Weather Service in State College, PA (Grumm, 2006).

This presentation will focus on applications of post-processed SREF forecast data for the December 1 2007 winter storm, with emphasis on improved visualization of the forecast outcome. These SREF data include the probability of freezing rain lasting three or more hours, the probability of freezing rain changing to rain, the probability of new snow or ice on roadways, and precipitation type “plume” visualizations. In particular, forecasts will be examined in their run-to-run framework verifying during the winter storm's peak. SREF signals for mainly freezing rain across Iowa, sleet across southern Minnesota and central Wisconsin, and heavy snow over central Minnesota and northern Wisconsin remained consistent beginning at 48-51 hours prior to the event. Within 24 hours, many SREF probabilities tended to step-increase in magnitude and shift to the verifying location. The SREF probabilities for post-processed fields requiring multiple conditions be satisfied over a time duration (e.g., the probability of freezing rain lasting three or more hours) remained lower (10-40%) than those with fewer conditions (e.g., such as the probability of measurable freezing rain). For these multiple requirement diagnostics, the location of relative maxima better indicated possible threats to the public. SREF precipitation type and amount “plumes” provided excellent visualization of the complex scenario for this winter storm. A few comparative remarks on the SREF post-processed data performance for the January 26-28 2009 mega ice storm across the central United States will also be made.