Northeastern United States Snowstorm of 9 February 2017 By Richard H. Grumm and Charles Ross National Weather Service State College, PA 1. Overview A strong shortwave produced a stripe of precipitation from the western Plains to East Coast on 8-9 February 2017 (Fig. 1). Most of the precipitation along and north of the track of the surface cyclone (Fig 2) fell as snow (Fig. 3). The highest snowfall totals were observed from northern New Jersey across Long Island and into New England. Though not shown, radar and lightning data showed an intense band of snow and a prolonged period of thunder-snow in the band from southern Maine into Massachusetts. Areas in this band received over 16 inches of snowfall. This storm system came on the heels of system which brought warm weather to much of the east and rain to the Mid-Atlantic region. The developing storm carved out a deep trough over the eastern United States and adjacent western Atlantic by 0000 UTC 10 February 2017 (Fig. 4f). The previous system had brought a surge of deep moisture into the eastern United States (Fig. 5c-d) and left a lingering area of above normal precipitable water in the Ohio Valley (Fig. 5e). Hints of the first surface cyclone are visible on the northern limbs of Figure 2a. The surface cyclone developed along the frontal boundary of the leading cyclone in the southern Plains and was over Arkansas at 1800 UTC 8 February (Fig. 2a). This surface system raced northwestward across the Mid-Atlantic region and rapidly deepened east of the Appalachian Mountains (Figs. 2b-e). The 850 hpa (Fig. 6) temperatures were well above normal over much of the eastern United States on 8 February. However, cold air and below normal 850 hpa temperature air accompanied the deepening upper level trough depicted in the 500 hpa fields (Fig. 4). Many areas which received heavy snow were experiencing unseasonably warm weather just hours before the arrival of snow. At 850 hpa (Fig. 7) a strong frontal circulation developed over central Pennsylvania around 0600 UTC 9 February (Fig. 7c) as a strong 850 hpa northeasterly jet developed. The u-wind anomalies in the 850 hpa jet were around -4 below normal over southern New England (Fig. 7d-e), in close proximity to where the heaviest snow was observed (Fig. 1). The NCEP 3km HRRR will be used to show how warm it was over much of the region before the onset of the precipitation and how fast the cold air came in to produce the snow and heavy snow. The rapidly developing cyclone east of the mountains, the surge of cold air, and the development of the strong easterly flow produced heavy snow in central Pennsylvania between approximately 0000 and 1200 UTC 9 February (Fig. 8a) and as the cyclone move off the coast, heavy snowfall was observed from Long Island into New England (Fig 8b). An examination of the observed
liquid equivalent precipitation (QPE not shown) implied most of the snow in western Pennsylvania fell before 1200 UTC and most of the snow fell after 1200 UTC in New England. Radar data showed a similar timing. This significant winter storm had a short predictability horizon with forecasts of deep cyclone and the snow potential having only 2-3 days of predictability. Forecasts of the cyclone and the QPF from the NCEP models are examined in the following section. Forecasts of precipitation type and precipitation transitions were critical but are not addressed here. 2. NCEP Forecast Guidance of i. The Global Ensemble forecast system (GEFS) The GEFS surface cyclone forecasts from 6 forecast cycles are shown in Figure 9. These data show no signal of the strong surface cyclone from forecasts produced at 1200 UTC 4 February 2017. The 5 February GEFS showed a potential cyclone development in the frontal trough (Fig. 9b) and successive GEFS runs showed a stronger and more defined surface cyclone (Fig. 9b-f). The GEFS 850 hpa winds and u-wind anomalies showed a similar evolution with a strong 850 hpa u-wind developing north of the cyclone with about 1.5 to 2 days of lead-time (Fig. 10). The GEFS QPF forecasts (Figs. 11-12) showed an increasing threat for over 12.5 mm of QPF as the forecast length decreased. The largest shift of the QPF shield to the northwest was observed between forecasts issued at 1200 UTC 4 to 5 February 2017. After 1200 UTC 5 February the probabilities of 12.5 mm or more QPF increased but the general region of the QPF remained relatively constant. The ensemble mean QPF showed in gradual increase in the QPF amounts. Note all forecasts had a 25 mm contour from 1 more ensemble members at all initialization times. However, in the mean the 25 mm contour did not appear until forecasts produced at 0000 UTC 8 February and as the forecast length decreased the 25 mm contour expanded to include portions of the higher terrain in southwestern Pennsylvania (Fig 12f). ii. The Global Forecast System (GFS) The GFS had predictability issues with the surface cyclone and the QPF shield were similar to those displayed by the GEFS and the focus here is on the QPF in the 2 key 12 forecast periods ending at 1200 UTC 9 and 0000 UTC 10 February 2017. We do not address the critical precipitation type issues which affected much of the region from southern Pennsylvania to Long Island and southern New England. But due to the warm air ahead of the system most locations started as rain.
The first 12 hour period (Fig. 13) that the GFS had the QPF axis in a similar area as the GEFS with an axis of 24 mm or more QPF from West Virginia to Long Island. It is unclear why the 0000 UTC 7 January forecasts were so dry relative to previous and latter forecasts. Relative to verification, most of these forecasts were too wet over southern Pennsylvania and Maryland. The second QPF (Fig. 14) window covered most of the heavy snowfall over LI and southern New England. The GFS forecasts had a swath of 12 to 24 mm of QPF over most of the affected region. Again the 0000 UTC 7 February forecast cycle showed a south and east shift to the QPF shield. The verifying 12 hour QPE for these two periods (Fig. 15) showed. The GFS was too wet in the first period over a large portion of southern Pennsylvania. The second 12 hour period had better QPF and a smaller error field (not shown). iii. The North American Mesoscale Forecast System (NAM) The NAM is shown using forecasts produced within about 48 hours of the onset the QPF in western Pennsylvania. Again only the QPFs are shown here focused on the Mid-Atlantic snowfall and the heavy snow in southern New York and New England. The NAM forecasts for the first 12 hour period (Fig. 16) were similar to those produced by the GFS. The one significant difference was that the NAM had higher QPF values along the northern edge of the precipitation shield, which in this case provided useful guidance for the band of heavier snow to the north. The NAM forecasts for the second phase include the 0600 to 1200 UTC period to cover the QPF in New Jersey and Long Island slightly better (Fig. 17). These forecasts show the NAM had over 24 mm of QPF over the region. This was forecast to mainly be snow or rain to snow over some areas. iv. The High resolution Rapid Refresh (HRRR) model The HRRR is used here to show that it too had too much QPF over portions of Pennsylvania (Fig. 18) and did not capture the relative QPE minimum observed. The HRRR is also useful in showing the issues related to the warm air which had to be displaced before the precipitation could change to snow. HRRR 00-hour 2m temperatures are shown in Figure 19. These data show that 2m temperatures were greater than 16C over Virginia and Maryland at 0500 UTC 8 February and
rose to over 18C during the day on 8 February. Cold air with 2m temperature below normal was present over New England B (blue shading) and cold air was observed moving in from the northwest between 1400 and 2000 UTC 8 February 2017. The cold air came in rapidly between 0000 and 1500 UTC on 9 February (Fig. 20). These 00-hour forecasts were quite similar to the forecast temperature changes in the HRRR and other models. Simulating the cold air though not shown, was handled quite well by most of the NCEP guidance. In addition to the 2m temperatures, the HRRR 850hPa temperatures showed the rapid intrusion of cold air at 850 hpa (Fig. 21). Many first cut precipitation type forecast use 850 hpa temperatures. From southern Pennsylvania to southern New England, the intrusion of sub-freezing air at 850 hpa preceded the onset of the QPF. 3. Summary A strong shortwave produced a stripe of precipitation from the western Plains to East Coast on 8-9 February 2017. This short wave interacted with pre-existing moisture and unseasonably warm along the East Coast. The result was rapid cyclogenesis and a rapid change over from rain to snow and ultimately from Pennsylvania, across New Jersey, southern New York and into New England for a significant snow storm. Over 12 inches of snow was observed in the mountains of central Pennsylvania, Long Island and portions of eastern New England. The 00-hour HRRR hourly analysis showed how rapidly the 850 hpa and 2m temperatures fell to below 0C as the precipitation shield blossomed and rapidly moved eastward. Snow fall was observed in areas where temperatures range from 16 to 20C only 12 hours before the onset of measureable snow fall. This rapidly developing and moving system was relatively well predicted by the NCEP model suite with about 2-3 days of lead-time. The GEFS was used to show forecasts in the 4-5 day range. These longer range forecast showed some indications of a storm system but it was not until 2-3 days prior to the event that a potentially significant winter storm began to emerge in these forecast. A single GFS run on the 7 February actually decreased the total QPF over the region. The event was separated into two twelve hour windows. The first window covered most of the period of heavy snow in central Pennsylvania and the latter covered the heavy snowfall over southern New York and New England. These data showed that the NCEP models over forecast the QPF over much of Pennsylvania and underestimated the QPF over north-central Pennsylvania. In fact the axis of higher observed QPE was well north of most of the NCEP forecast guidance. The NAM focused the higher QPF farther south but also had a more QPF farther north than the other NCEP forecast systems.
The over forecast QPF in southern Pennsylvania was present in the GEFS, GFS, NAM, and short-range HRRR forecasts. The mesoscale QPE minimum over portions of south and southcentral Pennsylvania (Fig. 1) was not forecast by the NCEP guidance. Though not shown, GOES water vapor imagery implied two mesoscale dry slots moved across this region which may have limited the QPE and thus snowfall totals. The significant winter storm of 9 February had a short predictability horizon with forecasts of deep cyclone and the snow potential having at best 1.5 to 3 days of predictability. It was an intense storm with many potential avenues of study. What caused the relative QPE minimum in south-central Pennsylvania? Why did the models fail to forecast this feature? And based on observations how unique or rare was the thunder snow over eastern Long Island and southeastern Massachusetts? 4. Acknowledgements
Figure 1. Total estimated precipitation (mm) for the period of 0000 UTC 8 to 0600 UTC 10 February 2017. Data was determined using the 6-hour Stage-IV QPE data set in grib format. Return to text.
Figure 2. GFS 00-hour forecasts of mean sea level pressure (hpa) in 6-hour increments from a) 1800 UTC 8 February through f) 0000 UTC 10 February 2017. Shading shows the standardized anomalies and isobars are every 4 hpa. Return to text.
Figure 3. National Snow Analysis data using 12-hour grib files to produce a two-day snow total in inches. There is like bad data in northwestern Connecticut. Return to text.
Figure 4. As in Figure 2 except for 500 hpa heights (m) and height anomalies (shading) in 24 hour increments from a) 0000 UTC 5 February through f) 0000 UTC 10 February 2017. Return to text.
Figure 5. As in Figure 4 except for the precipitable water (mm). Return to text.
Figure 6. As in Figure 5 except for 850 hpa temperatures (C ) in 6 hour increments from a) 0600 UTC 8 February through f) 1200 UTC 12 February 2017. Isotherms every 2C. Return to text.
Figure 7. As in Figure 6 except for 850 hpa winds and u-wind anomalies. Winds in ms-1. Return to text.
Figure 8. As in Figure 3 except for the 24 hour periods ending at a) 1200 UTC 9 and b) 1200 UTC 10 February 2017. Return to text.
Figure 9. NCEP GEFS forecasts of mean sea level pressure and pressure anomalies valid at 0600 UTC 9 February 2017 from GEFS forecast initialized at a) 1200 UTC 4 February, b) 1200 UTC 5 February, c) 1200 UTC 6 February, d) 1200 UTC 7 February, e) 0000 UTC 8 February, and a) 1200 UTC 8 February 2017. Return to text.
Figure 10. As in Figure 9 except for GEFS 850 hpa winds (ms-1) and u-wind anomalies valid at 1200 UTC 9 February 2017. Return to text.
Figure 11. As in Figure 9 except for the probability of 12.5 mm of QPF for the 24 hour period ending at 0000 UTC 10 February 2017. Shading is in percent, solid black contour is the mean 12.5 mm contour and the black dot is the location of State College where 6 to 7 inches of snow was observed. Return to text.
Figure 12. text. As in Figure 11 except for the ensemble mean QPF (shaded) and each member s 25 mm contour if forecast. Return to
Figure 13. NCEP GFS forecast total QPF for the period from 0000-1200 UTC 9 February 2017 from 6 GFS runs initialized at a) 0000 UTC 6 February, b) 0000 UTC 7 February, c) 1200 UTC 7 February, d) 0000 UTC 8 February, e) 0600 UTC 8 February, and f) 1200 UTC 8 February Return to text.
Figure 14. As in Figure 12 except for the 12 hour period ending at 0000 UTC 10 February 2017. Return to text.
Figure 15. As in Figure 1 except for QPE totals for the 2 12 hour periods ending at a) 1200 UTC 9 and b) 0000 UTC 10 February 2017. Return to text.
Figure 16. As in Figure 13 except for the NCEP NAM 12 hour accumulated QPF ending at 1200 UTC 9 February from NAM forecasts initialized at a) 0000 UTC 7 February, b) 1200 UTC 7 February, c) 0000 UTC 8 February, d) 0600 UTC 8 February, e) 1200 UTC 8 February, and f) 1800 UTC 08 February. Return to text.
Figure 17. As in Figure 16 except for the 18 hour forecast window ending at 0000 UTC 10 February 2017. Unlike the GFS forecasts, this time frame was chosen to better cover the full period of rain and snow over New York and Long Island. Return to text.
Figure 18. As in Figure 16 except for HRRR forecasts of the QPF for the period of 0300 to 1200 UTC 9 February 2017 from forecasts produced at a) 2100, and b) 2300 UTC 8 February and c) 0000 UTC, d) 01000 UTC, e) 0200 UTC and f)0300 UTC 9 February 2017. Return to text.
Figure 19. Select HRRR 00-hour 2m temperature forecasts valid in 3-hour intervals at a) 0500 UTC through f) 2000 UTC 8 February 2017. Contours every 2C, shading as in the color bar. Return to text.
Figure 20. As in Figure 19 except for HRRR 00-hour forecasts valid every 3 hours from a) 0000 UTC through f) 1500 UTC 09 February 2017. Return to text.
Figure 21. As in Figure 20 except for HRRR 850 hpa temperatures. Return to text.