Forecasting major winter storms is a critical function for all weather services. Conventional model-derived fields from numerical weather prediction models most frequently utilized by operational forecasters, such as pressure level geopotential height, temperature fields, quantitative precipitation forecasts, and model output statistics, are often insufficient to determine whether a winter storm represents a large departure from normal, or has the potential to produce significant snowfall. This paper presents a method, using normalized departures from climatology, to assist forecasters in identifying long-duration and potentially significant winter storms. The focus of this paper is on anomalous low- and upper-level wind anomalies associated with winter storms along the U.S. east coast.
Observed and forecast low-level (850 hPa) and upper-level (300 and 250 hPa) easterly wind anomalies are compared with a 30-yr (1961–90) reanalysis climatology. Anomalous easterly low-level winds are correlated with enhanced low-level forcing and frontogenesis. Strong low-level winds can also contribute to enhanced precipitation rates. Upper-level winds that are anomalously below normal, represented as easterly wind anomalies, are also correlated with systems that are cut off from the main belt of westerlies, which may result in slower movement of the system, leading to long-duration events. The proposed method of evaluating easterly wind anomalies is shown to assist in identifying potentially slow-moving storms with extended periods of enhanced precipitation.
To illustrate this method, winter storms on 25–26 December 2002 and 2–4 January 2003 will be compared with past historical winter storms. The results suggest that the low- and upper-level wind anomalies in the two recent snowstorms share common characteristics with several record snowstorms over the past 52 yr. Many of these storms were associated with easterly wind anomalies that departed significantly (2 or more standard deviations) from normal. The examination of climatic anomalies from model forecasts may assist forecasters in identifying significant winter storms in the short range (2–3 days) and potentially out to ranges as long as 7 days when ensemble forecast guidance is utilized.
East Coast winter storms (ECWSs; Hirsch et al. 2001) provide a wide variety of forecast challenges for all weather services. ECWSs are defined as storms that affect the United States east of the Appalachian Mountains, produce at least enough snow to require winter storm warnings from the National Oceanic Atmospheric Administration’s (NOAA’s) National Weather Service (NWS), and often produce coastal flooding. Winter storm warnings over the study area are issued when 2–3 in. of snow are expected in the Carolinas, 4–5 in. in Virginia and the Delmarva Peninsula, and 6–7 in. in the northeastern United States, including Pennsylvania, New York, and New England. Predicting snowfall amounts, areal extent, and areas of maxima are a challenge at all forecast time ranges, with decreasing forecast confidence in the longer time ranges, particularly beyond 48 h. It will be shown that utilizing information gleaned from analysis of upper- and low-level wind anomalies in numerical weather prediction (NWP) model forecasts can increase the confidence in forecasting the impact of ECWSs at all forecast time ranges, depending on the perceived NWP model accuracy.
There are two primary types of winter storms along the East Coast (Kocin and Uccellini 1990) including storms that move up and along the coast and secondary redevelopment. The Miller type-A (Miller 1946) storms, also referred to as “full coast storm,” typically develop in the vicinity of the Gulf of Mexico then strengthen and track northeastward affecting the U.S. east coast. These are relatively rare compared with the frequency of Miller type-B storms, as only four Miller type-A storms have occurred since 1993; the March 1993 Superstorm, the January 2000 Surprise Snowstorm, the 26 December 2004 storm, and the 27 February 2005 storm. Miller type-B (Miller 1946) storms are associated with a weakening primary cyclone in the Appalachian Mountains, as a new low pressure center redevelops and strengthens along the eastern U.S. coast.
ECWSs are quite common and tend to be more numerous during the warm phase of El Niño–Southern Oscillation (ENSO; Hirsh et al. 2001). The winters of 2002/03, 2003/04, and 2004/05 (hereafter winters of 2002–05) were characteristic of a moderate warm phase of ENSO; therefore, a higher probability of ECWSs could have been expected (Livezey et al. 1997). In addition to ENSO, the North Atlantic Oscillation (NAO; Lamb and Peppler 1987) was negative during the portions of the winters of 2002–05. The negative phase of the NAO is normally associated with colder than normal conditions over northeastern North America and western Europe. Several studies have shown that in addition to colder than normal conditions in the northeastern United States, there is a tendency for more snowfall during periods with a negative NAO (Hartley and Keables 1998; Bradbury et al. 2002). ENSO and the negative NAO may have produced conditions favoring ECWSs capable of producing heavy snow during the winters of 2002–05.
Two ECWSs produced heavy snow over interior sections of Pennsylvania northward into northern New England in late December 2002 and early January 2003. The first storm occurred on 25–26 December 2002 (hereafter Christmas 2002 storm), and the second storm occurred on 2–4 January 2003 (hereafter January 2003 storm). These two storms shared characteristics with other historical, long-duration ECWSs, including heavy snow, heavy rain, and coastal flooding. For this study, a historical ECWS is defined as one that produces widespread snow amounts >45 cm from Virginia through New York and New England, and >30 cm of snow from the Carolinas through Georgia. A key component of many historical ECWSs was the relatively long duration, ≥24 h, of precipitation in the region where the snow accumulation threshold was met during a single storm.
The Christmas 2002 and January 2003 storms were both Miller type-B events. In the mid-Atlantic region during Miller type-B events, an area of heavy snow often develops in the deformation zone between the two cyclone centers. It will be shown that upper- and low-level wind anomalies within this deformation zone were associated with the heavy snow from central Pennsylvania northward into upstate New York during the Christmas 2002 and January 2003 storms.
In this study, a comparison was made of the observed easterly wind anomalies, representing winds in the −U direction, during the Christmas 2002 and the January 2003 East Coast snowstorms. These data were compared with historical long-duration ECWSs associated with extreme snowfall and coastal flooding. The goal is to demonstrate how climatic anomalies may help identify potentially significant winter storms in the mid-Atlantic and northeastern United States when applied to NWP model forecasts. These data may assist forecasters in anticipating the severity and duration of the event, because frequently utilized NWP model-derived fields such as geopotential height, temperatures fields, quantitative precipitation forecasts, and model output statistics (MOS) often do not provide enough information to determine storm duration and severity. Regardless of the method used to evaluate severity and duration of ECWSs, lead times in anticipating the nature of the impact always vary and are dependant on the forecaster confidence in the NWP model guidance.
To derive departures from climatology for a specific event, a detailed and comprehensive climatology was developed. An overview of the datasets and definitions used in this approach are described below.
a. Gridded climatic datasets
The National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis dataset (Kalnay et al. 1996) was used for this analysis. The dataset has a 2.5° × 2.5° resolution at 17 pressure levels, extends from 1948 through January 2005, and is updated monthly. Many meteorological variables from that dataset were used, including mean sea level pressure (MSLP), geopotential heights, U and V (southerly) wind components, specific humidity, and temperatures at 12-h intervals. The latter two variables were available from 1000 to 300 hPa. The local climatology (for each 2.5° × 2.5° grid point) and 21-day centered means and standard deviations were based upon the fixed 30-yr period of record (POR) from 1961 to 1990. This 30-yr sample was chosen because of data availability and the National Climatic Data Center’s and NWS’s use of a 30-yr POR. Additionally, this is the most recent 30-yr period that includes the 1960s, which were characterized by exceptionally cold and snowy winters in the eastern United States. Further details on the climatic anomalies database and its applications can be found in Grumm and Hart (2001a,b).
The Grid Analysis and Display System (GrADS) software (Doty and Kinter 1995) was used to compute and display the means and standard deviations of the 500-, 700-, and 1000-hPa heights; 850-hPa temperatures; MSLP fields; and winds. The relatively coarse-resolution reanalysis data and smoothing of data by the GrADS software will likely smooth out the observed locally extreme values in both the 12-hourly reanalysis data and the 30-yr POR datasets. Conversely, model forecasts will tend to show stronger anomalies because of less smoothing of the data, typically only 0.5–1 standard deviation (SD), but up to 2 SDs in extreme cases, depending on how the absolute maximum (or minimum) wind values are resolved in the reanalysis and model analyses.
Additional anomalies were computed using Short-Range Ensemble Forecasts (SREF; Tracton et al. 1998) and Medium-Range Ensemble Forecasts (MREF; Tracton and Kalnay 1993) using storms from the winters of 2002–05. The SREF and MREF tend to dampen out some of the extreme anomaly values from single model runs. However, the ensembles still adhere to the thresholds for historical storms determined from the reanalysis data. Upper-level anomaly data were analyzed at 300 hPa in the reanalysis data and 250 hPa in the model data. A comparison of 300- and 250-hPa anomalies resulted in no appreciable change in the conclusions of this study.
As in Grumm and Hart (2001b), all departures from normal are shown in standard deviations from normal. These departures are referred to as standardized anomalies. Throughout this paper, the term anomalous refers to fields that depart by more than 2.5 SDs from the 30-yr means. This value was arrived at based on the confidence limits determined using the Chebyshev theorem (Blaisdell 1993) as an upper limit and those of the normal distribution as a lower limit (see Grumm and Hart 2001b, Table 1). In an absolute sense, a departure of 2.5 SDs from normal implies that the anomalous field occurs between 5% and 16% of the time at any given location (Grumm and Hart 2001b). Based on the results of Hart and Grumm 2001, the actual confidence limits are probably closer to those of the normal distribution.
b. Case and event data
Labeling specific events with their impact and informal titles (e.g., the Ash Wednesday Storm) was accomplished using Storm Data (NOAA 1959–2003), Weatherwise, and the American Meteorological Society’s journals when case studies were available (Table 1). This information is presented simply to give the reader a reference for the date, type, and location of the event, not necessarily to directly connect the anomalies with the societal impacts. Such conclusions can only be made after detailed case studies and case comparisons that are beyond the scope of this paper. The analysis of events was limited to 25°–50°N and 95°–65°W, to focus on events impacting the eastern half of the United States.
Radar and satellite data in the Christmas 2002 and January 2003 storms, were archived in real time using the data from the Advanced Weather Interactive Processing System data archive software at the NWS Forecast Office in State College, Pennsylvania. These data were moved to the Weather Event Simulator (Magsig and Page 2003) for playback and image generation. Images from New York and New England were obtained from the NWS radar archives maintained at the NWS Eastern Region Headquarters in Bohemia, New York.
c. Gridded datasets
Model grids for diagnostics and image production were retrieved from the Internet. Gridded forecast data used for comparing operational weather prediction data with the climatology were obtained from the NCEP stepped-terrain Eta Model, the Regional Spectral Model from the Global Forecast System (GFS), MREF, and SREF guidance. The December 2002 and January 2003 cases were identified to show how these data could be used operationally to add value to real forecast problems. The reader should be aware of data resolution differences between the reanalysis data and the model data. As stated earlier, the finer-resolution model data will usually produce slightly larger departures from normal compared with the coarser reanalysis data used to compute the climatological means and standard deviations.
d. Case study selection
The selection of historic case studies was confined to events that affected the eastern United States over the period from 1948 to 2003. Storm Data (NOAA 1959–2003) and local climatological data were used to identify heavy snow events. Only winter events are presented in this study. For each case, the reanalysis data were compared with the 30-yr POR to determine if the event represented a substantial departure from normal.
Snowstorms that impacted the East Coast were readily identified (Kocin and Uccellini 1990). A summary of the climatic anomalies associated with these events and more recent snow events was shown by Grumm and Hart (2001b). Events such as the Ash Wednesday Storm (6–8 March 1962), Blizzard of 1978 (6–8 February 1978), Blizzard of 1996 (6–8 January 1996), and Blizzard of January 2000 were analyzed to identify storms that had a substantial impact on populated areas and to determine if the Christmas 2002 and January 2003 storms were characterized by similar wind anomalies. Table 1 depicts the MSLP, and 850-, 500-, and 300-hPa height anomalies that existed in these historical storms. There is a lack of a clear signal for historical snowstorm activity in the height anomalies. These extreme events provide the forecaster with a quantitative measure of the “range” of atmospheric extremity. The examples in this study emphasize snowstorms because such events have been examined in the published literature. However, the forecast approach presented here for identifying extreme weather events need not be limited to snowstorms (Grumm and Hart 2001a).
a. Historical ECWSs
Wind anomalies at 850 and 300 hPa were analyzed for 92 ECWSs between 1948 and 2005. Four storms were described by Stuart and Grumm (2004) to illustrate typical 850- and 300-hPa wind anomalies in historical ECWSs that affected separate regions of the East Coast, including the Southeast (December 1989), mid-Atlantic (Ash Wednesday Storm of 7 March 1962), the Northeast (Blizzard of 1978), and the entire East Coast (January 2000 Surprise Snowstorm). These four storms also represented both Miller type-A (January 2000) and type-B (December 1989, March 1962, and February 1978) storms. These storms were chosen to illustrate that the low- and upper-level wind anomaly thresholds were consistent for all historical ECWSs, regardless of the region affected or storm type.
For this study, departures from the 1960–90 climatology in the easterly component of the wind are U-wind anomalies in units of negative standard deviations, because the westerly wind component represent positive U-wind anomalies. Easterly winds represent winds that can advect moisture from the Atlantic Ocean and provide forcing in the form of convergence, promoting vertical motions and precipitation production. Table 2 lists a subset of ECWSs used in this study, from the Ash Wednesday Storm of March 1962, through the winters of 2002–05. In all of the storms prior to the President’s Day storm of 2003, the reanalysis 850-hPa U-wind anomalies peaked between −4 and −5.5 SDs, while the 300-hPa U-wind anomalies peaked between −2 and −4.5 SD. The values of −4 SDs at 850 hPa and −2.5 SDs at 300 hPa were determined to be the threshold values. The anomalies indicated in NWP model output for the storms during the winters of 2002–05 peaked between −4 and −6.3 SDs, while the 250-hPa U-wind anomalies peaked between −2.5 and −3.5 SDs.
b. Overview of the Christmas 2002 and January 2003 storms
1) Christmas 2002 storm
The observed snowfall during the Christmas 2002 storm is shown in Fig. 1. In this event, the heaviest snow was observed in northeast Pennsylvania, upstate New York, and New England. The event produced heavy snow over a very large area, with some locations receiving nearly 100 cm of snow. The hardest hit areas were between Binghamton and Albany, New York. Initially, the major metropolitan areas of the northeastern United States were spared the heavy snow; however, late in the event, the rain changed to snow in both the New York, New York and Boston, Massachusetts metropolitan areas, where 15–30 cm of snowfall was observed. In addition to the heavy snow inland, strong low-level easterly winds brought flooding problems to coastal regions of the northeastern United States.
The Christmas 2002 storm moved out of the midwestern United States producing snowfall over portions of the Midwest, Tennessee Valley, and Ohio Valley (not shown). Prior to the storm, a large surface anticyclone was located over eastern Canada and extended into New England (not shown). The storm redeveloped off the Maryland coast, and deepened as it moved east off Cape Cod, reaching 965 hPa southeast of Montauk, New York. The MSLP for the Christmas 2002 storm was more than −4 SDs from the 30-yr norms (Table 3). At 700 hPa (not shown), there were two dominant anomalies over North America: below normal heights over Canada and slightly below normal heights over the northern Gulf states associated with the trough. Some confluence in the flow was evident along the East Coast and the adjacent western Atlantic Ocean. As the upper cutoff low developed over the Great Lakes, there were below normal westerlies at 300 hPa (−1.5 SDs) to the north of the upper low and anomalous southwesterly winds (+2 SDs) to the east. This latter feature implied that an anomalously strong upper jet was present over eastern North America. The southwesterly jet at 300 hPa peaked on the order of +3 SDs above the seasonal normal values. However, below normal 300-hPa U-wind anomalies north of the cutoff low, in the Eta and GFS models, as well as in the reanalysis data, peaked at −3.0 to −3.5 SDs, implying a slow-moving storm cut off from the predominant westerlies.
The upper- and low-level wind anomalies peaked at −3.2 and −3.7 SDs, respectively, in the reanalysis data. The 80-km Eta Model derived low- and upper-level anomalies that peaked at −6.0 and −3.3 SDs (Figs. 2a and 2c), respectively, were likely due to the finer spatial resolution. These upper- and low-level wind anomaly values can be compared to historical storms such as the Blizzard of 1978 (Fig. 3) and the Ash Wednesday Storm of 1962 (Fig. 4), two of the most extreme storms in terms of duration and snowfall amounts.
Just north of the developing surface cyclone, a strong low-level (850 hPa) southeasterly jet developed over the Delmarva Peninsula and into the mid-Atlantic region (not shown). Winds within the jet were on the order of 30 m s−1 and were on the order of 2–3 SDs below normal. Several snowbands developed over the mid-Atlantic and northeast United States, particularly over northeastern Pennsylvania along and near the New York border. The bands developed just north and west of the intensifying easterly jet at 850 hPa.
2) January 2003 storm
The observed snowfall during the January 2003 storm is shown in Fig. 5. The heavy snow fell in the same general area that received heavy snow during the Christmas 2002 storm. Some of these areas received nearly 150 cm of snow from the combination of both the Christmas 2002 and January 2003 storms. The hardest hit areas were again between Binghamton and Albany. Similar to the Christmas 2002 storm, heavy snow was observed inland and the strong low-level easterly winds brought coastal flooding problems to the northeastern United States. Unlike the Christmas 2002 storm, the January 2003 storm MSLP pressure anomalies near the surface cyclone center deviated much less from the 1960–90 climatology. The central pressure, analyzed by the Eta, was on the order of 1–2 SDs below normal, reaching a pressure of about 992 hPa at around 1800 UTC 4 January 2003 (not shown).
From an anomaly perspective, the height, MSLP, and temperature anomalies associated with the January 2003 storm were on the order of 1–2 SDs below normal (not shown). However, the wind anomalies with this event deviated much more from the 1960–90 climatology. Integrated from 1000 to 100 hPa, the absolute values of the wind anomalies were on the order of 4 SDs from normal in the Eta Model at 1800 UTC 4 January 2003 (Fig. 6), a value indicative of a rare to historical storm, with an average return period of around 5 yr for a storm with an identical integrated anomaly value. (Grumm and Hart 2001a).
Similar to the Christmas 2002 storm, a strong low-level (850 hPa) southeasterly jet developed over the Delmarva area and into the mid-Atlantic region. The winds and anomalies within the jet were on the order of 30 m s−1, corresponding to 2–3 SDs below normal (Fig. 7a). Multiple snowbands developed just north and west of the intensifying easterly jet at 850 hPa over the mid-Atlantic and the northeast United States, particularly over northeastern Pennsylvania along and near the New York border. This 850-hPa jet was on the order of −4.5 SDs from normal from Connecticut to northeastern Pennsylvania in the Eta Model’s 0-h forecasts initialized at 0000 UTC 4 January 2003 (Fig. 7b). This jet intensified to over −5 SDs below normal as it moved northward into Massachusetts around 0600 UTC 4 January 2003 (Fig. 7c). This −5 to −5.8 SDs anomalous low-level jet slowly lifted to the northeast and was focused over eastern Maine and southern New Hampshire by 1200 UTC 4 January 2003 (Fig. 7d).
At 300 hPa, the U-wind anomalies peaked at −2.0 SDs (not shown), similar to the Christmas 2002 storm, implying a slow-moving storm cut off from the predominant westerlies. Eta Model 250-hPa U-wind anomalies peaked at −3.0 SDs for the January 2003 storm (not shown). The anomalies were generally less in the reanalysis data because of the resolution differences between the operational models and the coarser climatic data (see Table 2).
c. Radar and Eta Model frontogenesis
Strong 850-hPa U-wind anomalies can have a profound effect on 850-hPa frontogenesis by increasing the potential temperature gradient, which may impact processes such as conditional symmetric instability (CSI), possibly leading to bands of heavy snow (Grumm and Nicosia 1997). The bands of heavy snow were well depicted during both the Christmas 2002 and January 2003 storms in the radar data (Figs. 8a, 8b, 9a–d, and 10) Areas of strong frontogenesis in both storms were aligned near the axis of the anomalous easterly 850-hPa jet (Figs. 11 and 12), and the snowbands were oriented along the frontogenetic zones (Clark et al. 2002). Reflectivity data over upstate New York from the Christmas 2002 (Figs. 8a and 8b) and January 2003 storms (Figs. 9a, 9d and 10) show the well-defined bands during a particularly intense stage of development. The strongest bands in the Christmas 2002 storm were oriented from southwest to northeast, while the bands in the January 2003 storm were oriented west to east. The bands in both storms were shifted slightly north of the frontogenesis in the Eta Model forecasts (Figs. 11 and 12 for the Christmas 2002 and January 2003 storms, respectively), but were in close proximity to the 850-hPa U-wind anomaly (Figs. 3 and 4, respectively). So while the frontogenesis was relatively well forecasted by the Eta Model, the locations of the strongest frontogenesis can often be best evaluated in real time by radar analysis of bands of snow exhibiting enhanced reflectivity (Nicosia and Grumm 1999).
Snowfall rates of 2.5–10.0 cm h−1 (1–4 in. h−1) were reported within the snowbands. These snowbands are consistent with the areas of heavy snow near Binghamton eastward to Albany and across Massachusetts in Figs. 1 and 2. Some locally heavy snow amounts (>50 cm) were observed around Binghamton, Albany, and southern New Hampshire.
Both the Christmas 2002 and January 2003 storms produced 30–90 cm of snow in bands across Pennsylvania, New York, and New England. The extremely heavy snow was observed in the strong low-level easterly flow north and west of the rapidly developing surface cyclone. This tendency for heavy snow to fall north and west of the track of the surface and 850-hPa surface cyclone was documented by Brown and Younkin (1970) in a statistical study of heavy snow. Independently, Stuart and Grumm (2004) showed that strong low-level easterly winds were a common occurrence in major winter snowstorms. Similarly, strong easterly winds, on the order of 3–5 SDs below normal at 850 hPa, were well correlated with heavy rainfall events in the eastern United States (Grumm and Hart 2001b) and in ECWSs (Stuart and Grumm 2004). The two storms analyzed in this study had low-level (850 hPa) U-wind anomalies 3.5–5 SDs below the 1960–90 climatological normals. Eta Model–derived 850-hPa wind anomalies peaked at near −6.0 SDs in both storms (Figs. 3 and 4) because of the finer resolution of the model data compared with the relatively coarse reanalysis data (−3.7 SDs in the Christmas 2002 storm and −4.5 SDs in the January 2003 storm). These Eta Model wind anomalies are comparable to historical storms that were also characterized by relatively strong upper- and low-level U-wind anomalies (Table 2). The North American Regional Reanalysis (Mesinger et al. 2006) dataset, with finer resolution on the order of 32 km, would improve the utility of climatic anomalies as a forecast tool using mesoscale models.
These two storms shared similar characteristics, including secondary cyclogenesis along the East Coast as the primary low weakened over the Appalachian Mountains and an anomalously strong low-level easterly jet. Although not shown, a modest band of snowfall was observed along and to the left of the primary 850-hPa cyclone track, consistent with the previous work of (Brown and Younkin 1970). The second band of heavy snow was also observed north and to the left of the 850-hPa cyclone track. The area to the north and left of the 850-hPa cyclone track is an area where strong −U winds are observed and where conditions are favorable for frontogenesis. In both cases, the anomalous 850-hPa U winds were present in areas where it was sufficiently cold to receive extremely heavy snowfall. The 850-hPa wind anomalies implied there could be potentially strong forcing and may serve as a proxy in identifying areas of potentially strong frontogenesis. Frontogenesis calculations using these winds and thermal fields, showed strong frontogenesis where the anomalous winds were producing strong forcing. In the January 2003 storm, the lack of significant height, MSLP, and thermal anomalies, in this case, show the power of unidirectional shear and anomalous easterly flow for big precipitation events along the East Coast.
Stuart and Grumm (2004) also concluded that storms with the longest precipitation durations (24 h or more) tended to have U-wind anomalies at upper levels (250–300 hPa) on the order of −2.5 to −5 SDs. In both storms, reanalysis upper U-wind anomalies at 300 hPa were between −2.5 and −3.5 SDs below normal, suggesting that the storms were cut off from the predominant westerlies. Upper-level wind anomalies within this range may indicate an upper-level system that is nearly or completely cut off from the main belt of westerlies. Therefore, these systems will typically move more slowly and −U anomalies in this range at upper levels may provide a signal to anticipate a slower-moving, longer-duration precipitation event, resulting in a historical ECWS. The Christmas 2002 and January 2003 storms met these criteria.
During both events, heavy snow was observed along and to the left of the track of the secondary surface cyclones in a region of anomalously strong low-level easterly flow. Anomalously strong low-level easterly flow may also contribute to enhanced frontogenesis, increasing the potential for CSI banding, thus contributing to increased precipitation rates (Grumm and Nicosia 1997). These processes contributed to the presence of several intense snowbands as the secondary lows strengthened and became the primary low pressure centers.
The radar imagery clearly shows that the areas of strongest frontogenesis were closely aligned with the strongest bands on radar (Figs. 8a, 8b, 10a–c and 11). The areas of strongest frontogenesis were also aligned along and to the west of the nose of the extremely anomalous low-level jet (Figs. 3 and 4), along and just left of the 850-hPa cyclone track. The strong 850-hPa U-wind anomalies indicated a strong jet circulation, and from this, one could infer the frontogenesis resulting from this flow had the potential to be extraordinary too.
Historical ECWSs (included in Tables 1 and 2) and false alarm events (not shown) were evaluated during the winters of 2002–05, and some consistent signals were found in the individual and ensemble model output regarding the occurrence and nonoccurrence of historical ECWSs. These cases (not shown) included The President’s Day Storm of 16–18 February 2003, the New England storm of 5–7 December 2003, the northern New England storm of 15–16 December 2003, the Canada Great Maritime Blizzard of 18–19 February 2004, the Carolinas storm of 26–27 February 2004, the New England blizzard of 22–24 January 2005, and the northern New England storm of 10–11 February 2005.
There were no false alarms during the study period of 1948 through the winter of 2004/05. However, there were at least two recent storms that produced historical amounts of snowfall that were characterized by low- and upper-level wind anomalies below the thresholds outlined in this study, namely the April Fool’s Day 1997 Storm that affected New England, and the Millennium Storm of 30 December 2000 that affected interior southern New York and New Jersey. These ECWSs exhibited low- and upper-level wind anomalies 0.5–1 SD from the thresholds outlined in this study. The effects of extreme frontogenesis are being investigated for future studies to possibly explain the existence of these historical ECWSs, which did not exhibit the wind anomalies outlined in this study.
In numerous cases, when one individual source of long- or short-range guidance forecasted the 850- and 250-hPa wind anomalies to meet or exceed the thresholds, but not multiple sources of forecast guidance, a historical ECWS was not realized. Conversely, ensemble-based forecast output provided superior guidance, especially the SREF guidance. In every case when SREF output forecasted the wind anomalies to meet or exceed the anomaly thresholds within 60 h, a historical ECWS occurred (the snowfall criteria were met in the respective region and the duration of precipitation exceeded 24 h). Additionally, no one individual NWP model was superior in forecasting a historical ECWS. However, when the operational Eta and GFS models both forecasted the wind anomalies to meet or exceed the thresholds in the 60–84-h time range, a historical ECWS resulted.
Each historical storm during the winters of 2002–05 produced snowfall for 24–48 h, and met or exceeded the snowfall thresholds for their respective regions (Carolinas to New England and southeast Canada). The areal extent of the historical snow accumulations was observed within the nose of the 850-hPa wind core (forward half of the region defined by anomalies ≤ −3 SDs), labeled the “threat zone.” Forecasters must still determine exactly where in the threat zone the extreme snowfall amounts will occur. Low- and upper-level anomalies for these historical ECWSs can be found in Table 2. The difference in the reanalysis and Eta Model anomalies further illustrate the differences in anomaly values that can occur depending on resolution, an important consideration for forecasters. As forecast model guidance continues to evolve, such as through the development and implementation of the Weather Research Forecast (WRF; Michalakes et al. 1998) model system, lead times for anticipating historical ECWSs will continue to increase.
The low- and upper-level wind anomalies serve as indicators for assessing the potential impact of ECWSs. Knowledge of these wind anomalies can aid forecasters in anticipating the severity and duration of an ECWS, because frequently utilized NWP model–derived fields such as geopotential height, temperatures fields, quantitative precipitation forecasts, and model output statistics (MOS) often do not provide enough information to determine storm duration and severity. The utility of low- and upper-level wind anomalies analyses were illustrated by an analysis of two major snowstorms that affected the interior regions of the mid-Atlantic and northeastern United States within a span of 2 weeks, 25–26 December 2002 and 2–4 January 2003. These two storms shared many similarities with each other, with respect to the duration, locations of heavy snowfall, and the snowbands on radar relative to the low-level wind anomalies. These wind anomalies may imply potential regions of strong frontogenesis and potentially long-duration events. Unlike the Christmas 2002 storm, the January 2003 storm had only a modest surface cyclone and never underwent a period of rapid cyclogenesis during the period of heavy snow over the eastern United States, yet the observed wind anomalies were very similar.
Preliminary data during the 2002–05 winter seasons suggest that lead times in forecasting historical ECWSs vary, depending on the NWP model resolution of low- and upper-level winds. Long-range guidance did not indicate anomalies that met the historical storm criteria for the majority of the storms, which may be due in part to the coarse spatial resolution. In the MREF output and some of the individual forecast models [GFS long range, European Centre for Medium-Range Weather Forecasts model, Navy Operational Global Atmospheric Prediction System Model, and Canadian Meteorological Centre Model], anomalies were within 1 SD of the thresholds 3–5 days in advance of many of the storms, but consensus by all the models was not observed in these time ranges. The anomaly thresholds were well resolved and met in the short-range guidance (Eta, GFS short range, and SREF) within 60 h of nearly all the storms, and 84 h for some of the storms. Preliminary data also suggest that the historical snow amounts were observed within the nose of the 850-hPa wind core, specifically the forward half of the region defined by anomalies ≤ −3 SDs, but not in the entire area.
Future work will include evaluating the areal extent of the historical snow accumulations and further evaluation of the SREF and MREF model output, WRF model output, frontogenesis, low-level temperature anomalies, low-level moisture, upper heights, and surface pressure anomalies. More data will be evaluated during future winter seasons to determine if significant weather events provided a characteristic signal relative to the climatology that could be used to anticipate the potential intensity of the event, and to evaluate potential lead times for predicting historical ECWSs.
The authors thank Robert Hart of The Florida State University for anomaly graphics and frontogenetic routines. Thanks also to John LaCorte for producing the snowfall graphics. Wayne Albright also added additional insight into the analysis of anomalies. Peter P. Neilley of UCAR provided online WSR-88D NEXRAD images and NEXRAD composite imagery.
* Current affiliation: NOAA/National Weather Service, Albany, New York
Corresponding author address: Neil Stuart, NOAA/National Weather Service, 251 Fuller Rd., Albany, NY 12203. Email: email@example.com