Snow, freezing rain, and shallow arctic Air 8-10 February 2015: NCEP HRRR success story By Richard H. Grumm National Weather Service State College, PA 1. Overview A short-wave (Fig. 1) moved over the strong ridge in the western United States over 7-8 February 2015 producing a winter precipitation event in the eastern United States. The strong and persistent ridge in the western United States was an interesting feature producing warm and areas of record warmth over much the western United States. This feature protected the eastern United States from the influx of warm Pacific air. The short-wave was associated with a surface wave (Fig. 2), which developed slowly along the boundary between warm air to the south and shallow arctic air to the north (Fig. 3). The projection used shows hints of the strong surface anticyclone eastern Canada. The 850 hpa temperatures show the -18 to -26ºC air over northern New England and Canada and the warmer air south of this boundary. The cyclone slowly slid south and eventually developed into an impressive surface cyclone on 10 February 2015. The initial stages of this impressive cyclone can be seen at 1200 UTC 10 February (Fig. 2e). It will be shown that the larger scale NCEP models did rather poorly with surface cyclone and shallow cold air evolution. As the short-wave moved east and the gradient between the warm and cold air tightened, the easterly winds on the cold side of the boundary increased (Fig. 4). In the strong easterly flow which slowly developed, a relatively long duration precipitation event (Fig. 5) and snow event began. Heavy snow was observed on the cold side of the boundary from eastern New York into eastern New England. The black dot for Boston shows strong easterly flow developed in eastern Massachusetts and persisted for at least 36 hours. This event contributed to the record snowfall observed in the Boston area during the record snowy February of 2015. The 16 to 24 mm or observed QPE (Fig. 5 upper) on 9 February fell as snow in eastern New England. The total QPE for the 3-day period was in the 24 to 32 mm range. The southern edges of the QPF shield fell as rain and freezing rain as the shallow arctic air slid southward. The NCEP HRRR 00-hour forecasts (Fig. 6) show the evolution of the shallow frontal boundary southward from New England (Fig. 6a) at 2100 UTC 8 February across Long Island into New Jersey and southward into northern Maryland (Fig. 6a-e). These data are in 4-hourly increments from hourly data. It is interesting to note how the warm air hung-up in the central Appalachians but drained southward in the coastal plain and eventually west of the mountains. Freezing rain was observed in Lancaster County and other locations in southeastern Pennsylvania. HRRR forecasts showed the evolution of the cold surge east side of the mountains and the transition of freezing rain.
This paper will document the pattern and anomalies associated with the ECWS of 8-10 February 2015. Section 3 will focus on the pattern and standardized anomalies to put the event into context. As with all high impact weather events, the forecasts and the communications of these forecasts are important. The 4 th section will examine the forecasts produced by the NCEP ensemble forecast systems and the HRRR. 2. METHOD AND DATA The large scale pattern was reconstructed using the Climate Forecasts System (CFS) as the first guess at the verifying pattern. The standardized anomalies were computed in Hart and Grumm (2001). All data were displayed using GrADS (Doty and Kinter 1995). For storm-scale details the 00-hour analysis from the hourly NCEP HRRR were used (see Figure 2). The precipitation was estimated using the Stage-IV precipitation data in 6-hour increments to produce estimates of precipitation during the even in 6, 12, 24 and 36 hour periods. Snowfall was retrieved from National Snow Analysis website. Snowfall data was obtained from both NWS public information statements and the National Snow site. The NCEP SREF were retrieved and examined in real-time and archived locally. These data helped identify the different predictability horizons of the forecast systems. The NCEP EFS data may not reflect public forecasts or perceptions of the forecasts. Many forecasters use a diverse set of forecast tools and often lean on the European Center model and post processed forecast data. 3. Forecast i) NCEP GEFS The NCEP GEFS and EC (not shown) were initially too cold and optimistic for heavy snow. This caused some early forecasts of 18 to 24 inches of snow fall across Pennsylvania and into western New England. The GEFS surface pattern (Fig. 7) showed considerable uncertainty with cyclone position on 9 February. Note the large spread north of the surface cyclone across New York and southern Canada (Figs. 7a-c). Longer range forecasts clearly showed large spread and thus low confidence in the forecasts. The area of the gradient and thus location of the frontal boundary had a significant impact on the GEFS QPFs (Fig. 8), which showed low probabilities for 16mm or more QPF. Confidence in heavy snow likely did not reflect the uncertainty contained in these forecasts. The probability of 25 mm was lower and is not shown. Clearly, widespread QPFs supporting heavy snow were limited and highly uncertain.
The forecasts of high snow amounts likely resulted from use of single models, viewing the models as what will happen and not what may happen, and applying snow liquid rations (SLRs) on poor QPFs. One thing that did favor the snow and snow farther south was the GEFS and EC (not shown) 850 hpa temperature forecasts, which were relatively cold (Fig. 9). As cold as these forecasts were, the longer range forecasts showed spread values north of the implied frontal boundary on the order of 3 to 8ºC. These forecasts suggested low confidence in position of the boundary and thus the location of the 850 hpa isotherm as a first cut rain-snow demarcation. ii) SREF The SREF showed relatively high spread with the cyclone position and the gradient between the cyclone and anticyclone. This led to differences in the location of the maximum QPF in the SREF from cycle to cycle. Six SREF QPF of 16mm or more QPF (Fig. 10) confined to eastern New York and New England. The SREF also showed the 850 hpa 0ºC line at 2100 UTC 8 February (not shown) and 0000 UTC 9 February (Fig. 11) likely to penetrate into southern New York. The SREF was generally colder than the GEFS and GFS (not shown). The colder solutions likely led to more optimistic snow forecasts in northern and eastern Pennsylvania than verified. The SREF 850 hpa temperatures did show considerable uncertainty in the high spread, mainly along and north of the frontal boundary at 850 hpa (Fig. 11). iii) HRRR The HRRR analysis was used to show the slow and steady intrusion of shallow, arctic air on 9 February, which produced freezing rain from southern New York and Long Island down into southeastern Pennsylvania. The forecasts of HRRR 2m temperatures (Fig. 12) show the intrusion of shallow cold air down the coast. These forecasts were relatively consistent in keeping the warm air over the central mountains of Pennsylvania and allowing shallow cold air to slide down the coastal plain into northern Maryland. The HRRR precipitation type data and 2m temperatures from 3 HRRR cycles (Fig. 13) show the colder solutions with freezing rain slowly moving farther down the coastal plain. As these forecasts updated they provide better guidance as to the potential for freezing rain farther south. Only a small sample of images was presented here at one valid time. Latter valid times, showed the potential for freezing rain in later forecasts down to the Maryland border by 1400 UTC 9 February 2015. These data imply some potential value in using the HRRR for more precise short-range forecasts. 4. Conclusions
A strong baroclinic zone in the eastern United States separated arctic air from relatively warm air to the south. Easterly flow on the colder side of the boundary produced a long duration snow event from about 7-10 February. A short-wave moved this baroclinic zone 8-9 February producing snow and areas of heavy snow deep in the cold air and a region of freezing rain as the shallow arctic air drifted southward on 8-9 February. Heavy snow was observed in east-central New York and eastward to Boston. This system had a relatively low predictability horizon (PH) and forecasts from 4-6 February had considerable spread and thus low confidence in where the higher QPF amounts would occur. The GEFS showed large spread in the baroclinic zone in the temperature and mean sea-level pressure fields. Despite the low confidence rather robust forecasts for heavy snow were presented at times from central Pennsylvania to New England. The NCEP GEFS (Figs. 7-9) showed the low confidence at longer ranges. However, the GEFS did correctly focus the higher probability for higher QPF amounts near the region where the higher snow amounts were observed. The NCEP SREF showed warmer 850 hpa temperatures and the potential for the rain snow line to be farther north than the GEFS and EC (not shown). The SREF also showed the potential for mixed precipitation. As the forecast horizon decreased, the spread decreased and the forecasts converged toward the higher QPF and thus snow from central New York into eastern Massachusetts. The SREF and GEFS both indicated considerable uncertainty that may or may not always be utilized by forecasters. Late on 8 and 9 February, as the cyclone began to develop, the precipitation increased. The higher observed QPE amounts (Fig. 5) were observed from 0000 UTC 9-10 February. During this period of time the HRRR showed the surge of shallow arctic air southward along the coast (Fig. 11) and the southward progression of the freezing rain southward with time. The HRRR correctly forecast the shallow arctic air down into northern Maryland by 1200-1400 UTC 9 February into northern Maryland and into western Pennsylvania (Fig. 12) while keeping the relatively warm air anchored in the higher terrain of the Appalachian mountain into central Pennsylvania. These data imply that the high resolution, rapidly updated HRRR (Fig. 6,11, & 12) can simulate shallow cold air masses down the coastal plain and thus provide clues to potential freezing rains. The high resolution of the HRRR and its high resolution terrain provide it with a critical advantage over coarser Regional and Global models. In this case, the HRRR did remarkably well with both the southward advance of the shallow cold air and the persistence of the relatively warmer air in the central Appalachians. 5. Acknowledgements:
Thanks for editing and comments from Samantha Ballard. 6. References Doty, B.E. and J.L. Kinter III, 1995: Geophysical Data Analysis and Visualization using GrADS. Visualization Techniques in Space and Atmospheric Sciences, eds. E.P. Szuszczewicz and J.H. Bredekamp, NASA, Washington, D.C., 209-219. DeGaetano, A. T., M. E. Hirsch, and S. J. Colucci. 2002. Statistical prediction of seasonal East Coast winter storm frequency. Journal of Climate 15:1101 17. Kahneman, D, 2011: Thinking Fast Thinking Slow. Farrar,Straus, and Giroux, NY,NY. 511pp. Kalnay, Eugenia, Stephen J. Lord, Ronald D. McPherson, 1998: Maturity of Operational Numerical Weather Prediction: Medium Range. Bull. Amer. Meteor. Soc., 79, 2753 2769. Roebber, P.J., M.R. Butt, S.J. Reinke and T.J. Grafenauer, 2007: Real-time forecasting of snowfall using a neural network. Wea. Forecasting, 22, 676-684.
Figure 1. CFSR 500 hpa heights (m) and standardized anomalies in 12 hour increments from a) 1200 UTC 7 February through f) 0000 UTC 10 February 2015. Return to text.
Figure 2. As in Figure 1 except for mean sea-level pressure (hpa) and standardized anomalies in 12 hour increments from a) 0000 UTC 8 February through f) 1200 UTC 10 February 2015. Return to text.
Figure 3. As in Figure 1 except for 850 hpa temperatures (C ) and standardized anomalies in 12 hour increments from a) 0000 UTC 8 February through f) 1200 UTC 10 February 2015. Return to text.
Figure 4. As in Figure 3 except for 850 hpa winds and and u-wind standardized anomalies in 12 hour increments from a) 0000 UTC 8 February through f) 1200 UTC 10 February 2015. Return to text.
Figure 5. Stage IV QPE for the period of 0000 UTC 9-10 February and the period of the longer duration event from 0000 UTC 7 to 10 February 2015. Units are mm and plotted using Pygrib. Return to text.
Figure 6. NCEP HRRR forecasts of 1000 hpa temperatures (C) in 2C intervals. Shading shows values below 2C for 4 hour windows from a) 2100 UTC 8 February through f) 1700 UTC 9 February 2015. Return to text.
Figure 7. GEFS mean sea-level pressure (mb) and spread (mb) valid at 0000 UTC 9 February 2015 from GEFS forecasts issued at a) 1200 UTC f 04 February, b. 1200 UTC 5 February, c. 1200 UTC 6 February, d. 1200 UTC 7 February, e) 0000 UTC 8 February, and f) 0600 UTC 8 February. Return to text.
Figure 8. As in Figure 7 except for 16mm of QPF. Return to text.
Figure 9. As in Figure 8 except for GEFS 850 hpa ensemble mean temperatures and spread about the ensemble mean. Return to text.
Figure 10. As in Figure 9 except for SREF mean QPF (mm) and the probability of 16mm or more QPF. Data valid for the period ending at 0000 UTC 10 February 2015 from SREF initialized at a) 2100 UTC 8 February, b) 1500 UTC 8 February, c) 0900 UTC 8 February, d) 0300 UTC 8 February, e) 2100 UTC 7 February, and e) 1500 UTC 7 February. Python based images. Return to text.
Figure 11. As in Figure 10 except for SREF 850 hpa temperatures valid at 0000 UTC 09 February. Return to text.
Figure 12. HRRR forecasts of 2m temperatures in 5C intervals shaded to show values below 0C. All forecasts valid at 1200 UTC 9 February from HRRR intialized at a) 10000 UTC 9 Feb, 06000 UTC 9 Feb, c) 0300 UTC 9 Feb, d) 0000 UTC 9 Feb, e) 2300 UTC 8 Feb, and f) 2100 UTC 8 Feb 2014. Return to text.
Figure 13. As in Figure 12 except for HRRR 2m temperatures and precipitation type as indicated by the color bar by type and intensity. Return to text.