a. Data sources

The Unified Precipitation Dataset (UPD; Higgins et al. 1996) was utilized to identify cases exhibiting heavy precipitation in the northeast United States. This national dataset incorporates National Oceanic and Atmospheric Administration (NOAA) first-order station precipitation measurements, daily cooperative observer measurements, and data from NWS River Forecast Centers (RFCs), representing over 13 000 stations in the contiguous United States. Precipitation amounts represent 24-h accumulation ending at 1200 UTC, interpolated to a 0.25° latitude–longitude grid.

The NOAA Daily Weather Maps (DWMs) weekly series (NOAA 1995–2001) were used to corroborate the UPD measurements and determine the predominant precipitation type (liquid/frozen). This publication provides daily North American surface frontal analyses, surface snow cover, and 500-hPa geopotential height maps valid at 1200 UTC each day. In addition, temperature maxima and minima, and precipitation for the 24-h period ending 1200 UTC are plotted for over 120 stations in the United States.

Mosaic radar data used in this study were composed of reflectivity returns from the 0.5° elevation scan from each radar site, with 2-km spatial and 5-min temporal resolution. Archived data were obtained from the Cooperative Program for Operational Meteorology, Education and Training (COMET) and were available in April 1995 and then from October 1996 to April 2001, although a large portion of the 1996/97 winter season was missing.

The National Centers for Environmental Prediction– National Center for Atmospheric Research (NCEP– NCAR) reanalysis (Kalnay et al. 1996; Kistler et al. 2001) was utilized for the initial composite investigation. This global dataset has a 2.5° latitude–longitude spatial resolution and 6-h temporal resolution. Locally archived NCEP Eta Model (Black 1994) analyses and 6-h forecast fields served as a higher-resolution composite and case study dataset. Although the Eta Model resolution and physics changed during the course of the study period (see http://www.emc.ncep.noaa.gov/mmb/ research/eta.log.html for a list of changes), the Eta Model was chosen since it had a relatively fine resolution, was available for the entire study period, and is a current operational forecast model. Model grids were remapped to the CONUS 211 grid (Lambert conformal projection) with 80-km horizontal resolution and 50-hPa vertical resolution, providing gridded fields that yield smooth diagnostic calculations.

b. Methodology

2) Composites

Composite analyses were created for single- and multibanded events and nonbanded cases (Fig. 1). In order to develop composite analysis stratifications, a surface cyclone-relative composite of the single band distribution was developed. Surface cyclone positions at the analysis time (0000, 0600, 1200, and 1800 UTC) most representative of the banded structure were determined by utilizing the NCEP–NCAR reanalysis dataset 1000-hPa height fields (henceforth used as a surrogate for the sea level pressure field), and the observed band location at this time was determined from radar. Note that all events exhibited an identifiable surface low at their respective analysis time. For occurrences of secondary cyclogenesis, the surface low closest to the banded structure was used as the surface cyclone position.

The single-banded events were stratified into two composite classes according to the surface cyclone-relative distribution. Events in which the single band exhibited a majority of its length in the northwest quadrant relative to the surface cyclone were classified as the northwest class (Fig. 1). This class was subdivided into events where the bands exhibited a majority of their length within (northwest near) and beyond (northwest far) 500 km of the surface cyclone. Events in which the single band exhibited a majority of its length in the northeast or southeast quadrant relative to the surface cyclone were classified as the east class. One event could not be clearly identified as a northwest or east class event, and was included in both composite classes. Also, a few study cases exhibited bands in more than one quadrant of the cyclone, and therefore were included in more than one composite class. Study cases that did not exhibit any type of banding were defined as nonbanded cases, and were included in the nonbanded composite class (Fig. 1).

The initial analysis time (t = 0 h) for each banded event was defined as the closest NCEP–NCAR reanalysis time (0000, 0600, 1200, and 1800 UTC) to the onset of the event. The initial analysis time for the nonbanded cases was defined as the analysis time closest to the middle of the time period when the case was within the study domain. A program was developed to composite cases in each composite class. This program takes the reanalysis data within a 60° by 120° latitude–longitude box centered on the cyclone position at t = 0 h, and averages fields from all banded events and nonbanded cases included in each composite class. If more than one event occurred at the same time in the same composite class, only one analysis time was included in the composite. The composites were calculated at 6-h intervals from 12 h before (t = −12 h) to 12 h after (t = +12 h) the initial analysis time (t = 0 h).

The composite program was modified to incorporate the resolution and domain of the Eta Model grids. Because the archived CONUS 211 grid has an eastern boundary roughly along 55°W longitude, a new cyclone-centered data box approximately 2500 km by 2500 km in size was defined to accommodate the grid boundaries. If any portion of a particular event data box went off the Eta Model domain, the specific grid points included in that portion were not included in the composite calculation. This contingency was only an issue on the eastern edge of the composite, where the lack of data resulted in noisy composite fields. The NCEP– NCAR reanalysis composite data were used to guide subjective smoothing near the eastern edge of the model composite figures shown in this paper. The centroid position of each composite class at t = 0 h was used as geographical reference for displaying the resulting composite fields.

The General Meteorology Package (GEMPAK; desJardins et al. 1991) version 5.6 was used to calculate kinematic parameters and visualize the various gridded datasets. The simplified 2D form of the Miller (1948) frontogenesis equation,

View Expanded

was used to examine the frontogenetical effect of the horizontal wind. This simplification was made since the frontal zones included in this study were nearly saturated, so that the tilting and diabatic terms oppose one another in the updraft region of a thermally direct frontogenetical circulation.

a. Application of classification scheme

The band classification scheme (Table 1) was applied to the 88 study cases to document the observed mesoscale band type and frequency in the northeast United States, and to build a database of banded events to study. The results of this application are summarized in Table 2. Note that several cases exhibited a particular band type more than once, and/or more than one band type during the time period when the case was within the study domain. Since a given case can feature multiple events, the total number of events (162) is nearly double the total number of cases (88).

The most common event type observed in the northeast United States was the single (48 events), followed by the transitory (40), narrow cold frontal (36), and multi- (29). Nine events were classified as undefined. The large number of transitory events reveals that banded structure is often observed that does not maintain its identity for more than 2 h. Additionally, the predominance of single-banded events over narrow cold-frontal events reflects a relative lack of narrow cold-frontal events observed in the northeast United States during the cold season as compared with the Pacific Northwest (Houze et al. 1976) and the United Kingdom (Browning 1985). This finding also agrees with the relative proportion of single bands compared to narrow cold-frontal bands found in adjacent Atlantic Canada (Stewart 1991).

Not all cases exhibiting heavy precipitation in the northeast United States exhibit mesoscale banding. A total of 13 study cases were classified as nonbanded, representing ∼15% of the study cases. This subset of cases serves as an important dataset to be used to discriminate between banded and nonbanded cases.

b. Single-band characteristics

The geographical distribution of the 48 single-banded events is shown in Fig. 3a. Although only five cold seasons were studied, single-banded events were distributed relatively evenly over the northeast United States. The surface low positions corresponding to the analysis time of the 48 single-banded events are shown in Fig. 3b. Although the low centers represent a single time, the clustering of centers off the East Coast highlights the dominant coastal storm track, with evidence of a secondary storm track apparent in the Midwest.

The surface cyclone-relative single-band distribution is shown in Fig. 4. The predominance of bands in the northwest quadrant is readily evident. Thirty-nine of the 48 single-banded events (81%) exhibited a majority of their length in this quadrant. Also note that bands to the northeast of the surface cyclone exhibited a northwest– southeast mean orientation, while bands to the northwest exhibited a southwest–northeast mean orientation. These orientations are consistent with the idealized orientation of the thermal wind in a developing baroclinic system (an “S”-shaped thermal structure).

Analysis of the lifetimes of single-banded events (Table 3) shows that the events are generally short lived, as the frequency of events rapidly decreases with increasing event duration. Several events just barely met the 2-h minimum, consistent with the large number of transitory bands observed (Table 2). This outcome suggests that the conditions favoring single-band formation occur on limited time scales. However, a number of events persisted beyond 12 h, the longest of which lasted 22.5 h (4–5 February 1998).

Radar animations of the single-banded structure show that the evolution may locally exhibit short-lived (0.5 h) finescale bands (widths < 20 km, intensities > 10 dBZ over the background reflectivity) that reach maturity in place, dissipate, but then are readily replaced by another finescale band in the same location. This observation is consistent with the theoretical work of Xu (1992), who showed through idealized numerical modeling that large-scale ascent can evolve into multiple smaller-scale updrafts when the geostrophic saturation equivalent potential vorticity (EPV^*_g) is reduced to negative values. In this environment, bands can exhibit finescale multibanded structures that stretch and merge into the dominant single band of moist ascent. In this case the single-banded structure may be considered a dominant band axis, or a band accumulation zone, rather than one single band. In fact, on 10 occasions there was interaction between single- and multibanded events in which individual bands in the region of multibanding merged with the single band for at least 2 h.

Radar animations of northwest class bands also revealed that many of the bands pivoted as they translated with the system. In fact, 24 of the 39 (62%) single-banded events in the northwest class were associated with an identifiable pivot point, with the band typically pivoting cyclonically. Radar animations of the east class bands showed a common motion approximating the velocity of the low- to midtropospheric wind (often toward the northeast).

a. Composite evolutions

The composite dataset allows examination of the cyclone evolution in each composite class through a 24-h period. For the purposes of this paper, only the northwest and nonbanded composite results will be shown at 12-h intervals. Results of the east, multibanded, northwest-near, and northwest-far composites are shown in Novak (2002, his section 3.2).

b. Composite cross sections

A comparison of the composite frontal environments can be made by cross-sectional analysis. Note that the lengths (∼2000 km) and frontal baroclinicity were nearly equal among the northwest and nonbanded class composite cross sections, facilitating comparisons of key features and frontal slope.

The northwest class composite cross section (orientation is shown in Fig. 6d) and nonbanded composite cross section (orientation is shown in Fig. 9d) at t = 0 h are presented in Figs. 11a,b and Figs. 11c,d, respectively. Moist symmetric stability is assessed in Figs. 11a,c by evaluation of the full wind saturation equivalent potential vorticity (EPV*). The full wind was used as opposed to the geostrophic wind in the calculation of EPV* since the full wind is likely more representative of the curved flow noted in the composites, although the applicability of the full wind to the assessment of moist symmetric stability in three-dimensional flow configurations continues to be debated (e.g., Schultz and Schumacher 1999, their section 3; Gray and Thorpe 2001; Clark et al. 2002). Note that the mean band position is centered in the cross section in Figs. 11a,b, whereas the center of the nonbanded cross section corresponds to the 750–650-hPa frontogenesis maximum.

The northwest class composite cross section shows that the mean band position is embedded within a sloping frontal zone (Fig. 11a). Examination of the saturation equivalent potential temperature (θ_es) and EPV* fields reveals that this front separates a conditionally unstable atmosphere on the equatorward side from a conditionally stable atmosphere on the poleward side. However, it is important to note the existence of a layer of small EPV* [i.e., values < 0.25 PVU (potential vorticity unit)] between 550 and 350 hPa in the center of the cross section, which is coincident with the mean band position. Although a large portion of this area is unsaturated (Fig. 11b), the portion directly above the band position exhibits relative humidity values near 80%, indicating the presence of small moist symmetric stability in this location. Frontogenesis is readily apparent in the cross section (Fig. 11a), with the maximum vertically aligned between 800 and 600 hPa and collocated with the mean band position under the area of small EPV*. As shown by Emanuel (1985), the frontogenetical response in the presence of small moist symmetric stability takes the form of a concentrated updraft on the equatorward flank of the frontal zone. Examination of the composite vertical velocity field (Fig. 11b) reveals a narrow ascent maximum with a slight tilt toward the cold air, consistent with Emanuel (1985). The band position is clearly evident in the relative humidity field (Fig. 11b), with a narrow relative humidity maximum correlated with the ascent maximum. The tilt of these two fields toward the cold air is suggestive of slantwise ascent.

The nonbanded composite cross section (Figs. 11c,d) exhibits similar features to the northwest composite cross section, but differences are apparent. Although a region of midlevel conditional instability is found equatorward and above the frontal zone (Fig. 11c), the frontal zone has a shallower slope in the 800–500-hPa layer than in the northwest composite cross section [1:160 versus 1:90; based on considering the slope of the maximum θ_es gradient (not shown)]. For a given baroclinicity, the shallower sloped frontal zone would imply greater conditional stability within the frontal zone. Although small EPV* values are found above the frontal zone, relative humidities through the entire cross section are near or below 70% (Fig. 11d). Note also that the frontogenesis maximum is not as vertically aligned as in the northwest composite cross section, especially below 600 hPa. The maximum frontogenesis magnitude in the nonbanded cross section [0.4°–0.8°C (100 km)⁻¹ (3 h)⁻¹] is also weaker than the northwest composite [0.8°–1.6°C (100 km)⁻¹ (3 h)⁻¹]. Consistent with weaker frontogenesis and relatively larger conditional stability, the ascent accompanying this frontogenesis is relatively broad and weak, with magnitudes less than −4 × 10⁻³ hPa s⁻¹ (Fig. 11d).

a. 5–6 February 2001

On 5–6 February 2001, a major winter storm produced widespread snow accumulations greater than 30 cm throughout most of New England, with over 75 cm locally in New Hampshire. The large snowfall accumulations were primarily attributed to an intense single band that developed during the storm. Snowfall rates in excess of 10 cm h⁻¹ were observed in the extensive band, producing near white out conditions. Mosaic radar imagery shows the band at 0000 UTC 6 February 2001 (Fig. 2a; hereafter times will be abbreviated as “UTC/ day”).

A synoptic analysis of the event at 0000/6 is shown in Fig. 12. Rapid cyclogenesis occurred during the previous 6 h (700- and 1000-hPa height falls of 120 m) as the 500-hPa trough became negatively tilted (Fig. 12b). A closed circulation is noted from the surface to nearly 500 hPa (Figs. 12a–c), which results in a deep-layer deformation zone associated with a confluent asymptote northwest of the surface cyclone (Fig. 12c), just as in the t = 0 h northwest composite (Fig. 6c). This deformation acting on the thermal gradients results in midlevel frontogenesis, with maxima exceeding 2°C (100 km)⁻¹ (3 h)⁻¹ (Fig. 12c), found on the poleward side of a focused band of midlevel warm air advection (Fig. 12d). The observed band location (refer to Fig. 2a) closely corresponds to the axis of maximum midlevel frontogenesis (Fig. 12c).

b. 14–15 February 2000

On 14–15 February 2000 a weak cyclone moved from the Ohio Valley to the Gulf of St. Lawrence. Although not intense, the cyclone was still responsible for a large swath of over 25 mm of precipitation that fell as a mix of rain, sleet, and freezing rain over coastal locations, with over 20 cm of snowfall accumulation in interior New England. In contrast to the 5–6 February 2001 storm, no mesoscale banding was evident during the evolution of the 14–15 February 2000 storm (Fig. 2b).

Many of the same synoptic features noted in the nonbanded composite (Fig. 9) are present in the 14–15 February 2000 case. At 1200/14 the surface cyclone centered in Pennsylvania (Fig. 13a) was associated with a short-wave trough over the Maryland panhandle (Fig. 13b). This short wave was embedded within the confluent entrance region of a 70 m s⁻¹ jet (Fig. 13b). The 700-hPa flow featured a short-wave trough over the mid-Atlantic states, but in the absence of a closed circulation, the flow continues northeastward (Fig. 13c) instead of a portion of the flow turning cyclonically as in the 5–6 February 2001 storm. This flow pattern limits the development of deformation and warm air advection northwest of the surface cyclone, reducing the potential for frontogenesis in this sector of the cyclone (Fig. 13c). However, ahead (east) of the surface cyclone, midlevel warm air advection was found over a large portion of the northeast United States (Fig. 13d) and once again frontogenesis was found on the poleward side of this warm air advection (Figs. 13c,d).

c. Cross-section comparison

Cross sections through the midlevel frontogenesis maximum in each case show similar features to the respective composite cross sections. The 0000/6 cross section (Figs. 14a,b) exhibits intense frontogenesis within the frontal zone (Fig. 14a). Although the frontogenesis maximum and θ_es contours slope toward the colder air, the frontal zone is relatively upright, suggesting weak conditional stability. Conditional instability is noted south of the frontal boundary, but an area of small EPV* is found in the 600–400-hPa layer above the band position and frontogenesis maximum. The frontogenetical response is readily evident in Fig. 14b, with a narrow, sloping ascent maximum on the equatorward flank of the frontal zone.

In contrast, the 1200/14 cross section (Figs. 14c,d) exhibits a shallow sloped frontal zone (Figs. 14c), with two disconnected and weak frontogenesis maxima at 1000 and 700 hPa (Fig. 14c). Although a region of conditional instability is found equatorward and above the frontal zone, the shallow slope of the front and weak disconnected frontogenetical forcing limits the frontal circulation as is evident in Fig. 14d. Note that the ascent is less than half the magnitude at the corresponding time in the 5–6 February 2001 storm (Fig. 14b) and occurs over a broader region, consistent with the comparison between the banded and nonbanded composite cross sections (Figs. 11b,d).

a. Climatology

Despite the rather strict criteria used to define mesoscale precipitation bands, 162 banded events were documented in the northeast U.S. study region during the study period. This number testifies to the frequency with which operational forecasters are challenged by mesoscale band formation during the cold season in the northeast United States. However, the occurrence of 13 nonbanded cases during the study serves as a reminder that not all heavy precipitation systems exhibit mesoscale banding as defined in this paper.

The predominance of single-banded events in the northwest quadrant relative to the surface cyclone is consistent with case studies conducted in the eastern half of the United States (e.g., Sanders and Bosart 1985; Nicosia and Grumm 1999; Martin 1998a,b); however, it differs from the predominant single-band location (ahead of the warm front) documented in the Pacific Northwest (Houze et al. 1976) and in the United Kingdom (Browning 1985). Note that these latter two studies were located in the exit region of the Pacific and Atlantic storm tracks (Pacific Northwest, United Kingdom), which may influence the storm-relative location of single-band formation. This hypothesis is supported by the results of this study and Nicosia and Grumm (1999), which suggest that single-banded events in the northeast United States are often associated with rapid cyclogenesis, which is more frequent in the western ocean basins (Whittaker and Horn 1984).

The tendency for single bands to be short lived and the large number of transitory events suggest that the conditions favoring single-band development occur on limited time scales, although in rare cases single bands persisted for more than 12 h. Also of interest is the common pivoting motion of the observed single bands, since the pivot point experiences heavy precipitation for a sustained period. Operational quantitative precipitation forecasts (QPF) could benefit from understanding the dynamics associated with this pivoting motion.

b. Composites and case examples

Composites calculated from the Eta Model analyses and 6-h forecast fields for banded and nonbanded cases highlight the role deformation, frontogenesis, and moist symmetric stability play in mesoscale band formation, and places these elements into the context of the synoptic and mesoscale flow evolution.

In the northwest composite the midlevel deformation zone forms northwest of the surface cyclone when a closed midlevel circulation develops and strengthens as the cyclone deepens. Consistent with the case studies of Nicosia and Grumm (1999), the deformation is found in diffluent flow north of the 700-hPa low center. This deformation acting on the temperature gradient produces differential temperature advections and associated frontogenesis. The midlevel frontogenesis develops in response to increasing warm air advection and associated warm air advection gradient, and is focused along the confluent asymptote of the deformation zone through a deep layer, providing forcing for the single-banded events.

The northwest composite frontal zone was marked by deep-layer frontogenesis coincident with small moist symmetric stability. Application of the semigeostrophic form of the Sawyer–Eliassen equation (Sawyer 1956; Eliassen 1962) would predict that the direct circulation induced by the frontogenesis in the presence of small moist symmetric stability would lead to an intense, narrow sloping updraft on the equatorward flank of the frontal zone. The northwest composite vertical velocity fields exhibited this very feature. In fact, all single-banded composite frontal zones (northwest near, northwest far, and east) exhibited the above frontal features (see Novak 2002, his section 3.2), suggesting that deep-layer frontogenesis in the presence of small moist symmetric stability is the primary forcing mechanism for single bands in both the northwest and east quadrants.

In contrast, the nonbanded composite cyclone evolution was dominated by its position in the confluent entrance region of an upper-level jet, limiting the development of a closed midlevel circulation and an associated deformation zone northwest of the surface cyclone. Although the cases in the nonbanded composite did not exhibit mesoscale banding, it is important to note that midlevel deformation and frontogenesis were evident, associated with confluence ahead of the cyclone. In this regard, the existence of midlevel frontogenesis alone does not ensure mesoscale band formation. Composite and case study cross sections through respective frontogenesis maxima showed that compared to their banded counterparts, the frontogenesis was weaker (especially near the surface in the nonbanded composite) and lacked vertical continuity (as in the 14– 15 February 2000 case). Additionally the nonbanded composite and the 14–15 February 2000 case exhibited a relatively shallow slope (see Figs. 11c and Fig. 14c), suggesting greater conditional stability. The combination of weaker frontogenesis and greater conditional stability likely limited the frontal circulation and subsequent band development.

The stark contrast in the degree of development between the northwest and nonbanded composite and case study evolutions supports the assertion made by Nicosia and Grumm (1999) that banded structure is favored during cyclogenesis, when temperature advections are enhanced (contributing to frontogenesis) and differential moisture advection can reduce moist symmetric stability. Also, the results from the present study suggest that single-band formation in the comma-head portion of cyclones occurs during favored periods when midlevel deformation and associated frontogenesis are maximized as the midlevel circulation forms, and when the moist symmetric stability is minimized. The northwest composite also shows that as the midlevel center deepens and broadens during the occlusion process (see Fig. 7), the primary deformation zone and associated frontogenesis shift northeastward, leading to band dissipation in the northwest quadrant. This evolution may explain the limited lifetime of single-banded events noted in Table 3, at least with respect to single-banded events observed in the northwest quadrant.

c. Conceptual models

Conceptual models of single-banded events and nonbanded cases are presented in Fig. 15. The single-banded event (Fig. 15a) exhibits a double-jet structure, placing the surface cyclone under upper-level divergence, which promotes cyclogenesis. The existence of a closed midlevel circulation creates a deformation zone associated with a confluent asymptote northwest of the surface cyclone. This deformation zone supports frontogenesis, which provides forcing for band development northwest of the surface cyclone. Diffluent flow ahead of the midlevel disturbance also contributes to deformation and associated frontogenesis, supporting band development ahead of the warm front. Note that this conceptual model is consistent with the findings of Nicosia and Grumm (1999, see their Fig. 17). The nonbanded case (Fig. 15b) highlights the dominance of the upper-level jet. A weak surface low associated with a weak midlevel trough is found in the equatorward entrance region of the jet. Confluent flow ahead of this midlevel trough contributes to deformation and associated frontogenesis. The conceptual models of single-banded events and nonbanded cases are also consistent with the combined observational and idealized modeling study of Schultz et al. (1998), who found that cyclones in diffluent flow tend to exhibit frontogenesis in the northwest quadrant of the cyclone (as seen in the single-banded conceptual model; Fig. 15a), whereas cyclones in confluent flow tend to feature frontogenesis in the northeast quadrant (as seen in the nonbanded conceptual model; Fig. 15b).

A schematic of the characteristic single-banded and nonbanded cross-sectional environments is shown in Fig. 16. The banded frontal zone is marked by a sloping region of frontogenesis extending through the midtroposphere, coincident with weak conditional stability. The direct circulation induced by the deep-layer frontogenesis in the presence of small moist symmetric stability results in a narrow, intense sloping updraft on the equatorward flank of the frontal zone (Fig. 16a). The nonbanded frontal zone (Fig. 16b) has similar features as the banded cross section, except that the frontogenesis is weaker and the associated ascent is weaker and broader. In addition to the weaker frontogenetical forcing, the slope of the frontal zone is shallower than in the banded cross section. In this situation, the Sawyer–Eliassen equation would predict that the ascent forced by the frontogenesis would be weaker and broader, as depicted in the vertical velocity field (Fig. 16b).