An Examination of the 30 January 2018 Long Island Snowfall Forecast Bust

During the overnight hours of 30 January 2018, a surface cyclone passed approximately 500 km to the southeast of Long Island. About 6-12 hours before the event the National Weather Service at Upton, NY forecast 2.5-5.0 cm (1.0-2.0 in) of snow accumulation on the eastern half of Long Island, and less than 2.5 cm elsewhere. Convection-allowing models like the 3-km NAM and High Resolution Rapid Refresh (HRRR) were simulating occasional snow showers across Long Island and into Connecticut overnight with very light accumulations, supporting this forecast.

A broad stratiform region of snowfall developed after approximately 0400 UTC over all of Long Island. Within this stratiform area were numerous finescale snow bands. A larger single snow band also formed over the eastern half of the island between 1000 and 1100 UTC and persisted for several hours. Snowfall ended from west to east from about 0900 to 1400 UTC. By this time, snowfall had accumulated to 15.2 cm (6.0 in) at Islip Airport and 22.1 cm (8.7 in) at the NWS office. The unexpected snowfall occurring overnight and continuing into the morning rush hour was responsible for widespread school delays and cancellations. Police reported at least 147 automobile accidents during the morning commute, including one that proved fatal.

This study examines what synoptic and mesoscale features may have been responsible for this unexpectedly heavy snow missed by the NWP models and human forecasters. Preliminary results show that the HRRR struggled to adequately predict the precipitation even at forecast lead times as short as 2-3 hours. Although the forecast of the surface cyclone was fairly well predicted, the model misplaced the region of maximum upper-level divergence associated with the coupling of equatorward and poleward jet streaks. The HRRR also consistently under-forecast the strength and southwestward extent of a northeasterly lower level jet (LLJ) that stretched from Cape Cod to south of Long Island, which likely led to errors in low-level frontogenetical forcing for precipitation banding. The PV in the 700-500 hPa layer early in the forecast was also underforecast and likely contributed to upper-level and LLJ jet errors and the growing precipitation errors. This presentation will combine observations with the HRRR, RAP, and NAM to highlight the evolution of this event and associated model errors at these short lead times.