Ensemble Sensitivity of Precipitation Type to Initial Conditions for a Major Freezing Rain Event in Montreal

The predictability of precipitation type in a January 2017 winter storm over the northeastern United States and southeastern Canada is examined using a convective-scale initial-condition ensemble with the Weather Research and Forecasting (WRF) Model. Real-time forecasts of the event by Environment and Climate Change Canada predicted 15-25 cm of snow accumulation in Montreal, Quebec, Canada. However, the initial 4 h of the event had 5-8 mmof freezing rain instead, followed by 7 cm of snow. While the total liquid-equivalent precipitation was consistent with the forecast, the unexpected freezing rain caused significant disruption in the Montreal region. The fraction of freezing precipitation (freezing rain and/or ice pellets) over the initial 4 h in Montreal varied greatly across the ensemble, with some members producing nearly all snow and others producing nearly all freezing precipitation. In members with larger fractions of freezing precipitation (as opposed to snow), the cyclone's midlevel trough was displaced slightly to the northwest, and its downstream (eastern) edge was narrower, the latter of which was traced back to model initialization. These differences increased the midlevel southerly flow into southern Quebec, which both enhanced the horizontal warm advection and decreased the vertical cold advection leading up to the event. The consequent midlevel warming over Montreal in these members produced an above-zero layer that melted falling precipitation, leading to freezing upon contact with the ground. This case study highlights the value of convective-scale ensembles for identifying mechanisms by which initial synoptic-scale uncertainties lead to high-impact localized errors in precipitation type.