Abstract

Environments that accompany mesoscale snowbands in extratropical cyclones feature strong midlevel frontogenesis and weak symmetric stability, conditions conducive to vigorous ascent. Prior observational and numerical studies document the occurrence of upward vertical velocities in excess of 1 m s−1 near the comma head of winter cyclones. These values roughly correspond to the terminal fall velocity of snow; snow lofting has been measured directly with vertically pointing radars. Here, we investigate the occurrence of lower-tropospheric snow lofting near mesoscale bands and its contribution to snowfall heterogeneity. We test the hypothesis that hydrometeor lofting substantially influences snowfall distributions by analyzing the vertical snow flux in case-study simulations, by computing snow trajectories, and by testing sensitivity of simulated snowbands to parameterized snow terminal fall velocity and advection. These experiments confirm the presence of upward snow flux in the lower troposphere upstream of simulated mesoscale snowbands for two events (27 January 2015 and 2 February 2016). The band of lower-tropospheric lofting played a more important role in the January 2015 case relative to the February 2016 event. Lofting enhances the horizontal advection of snow by increasing hydrometeor residence time aloft, influencing the surface snowfall distribution. Experimental simulations illustrate that while lofting and advection influence the snow distribution, these processes reduce snowfall heterogeneity, contrary to our initial hypothesis. Our findings indicate that considerable horizontal displacement can occur between the locations of strongest ascent and heaviest surface snowfall. Numerical forecasts of snowbands are sensitive to formulations of terminal fall velocity of snow in microphysical parameterizations due to this lofting and transport process.

1. Introduction

Previous theoretical and observational studies have identified mechanisms leading to banded precipitation in extratropical cyclones (e.g., Bennetts and Hoskins 1979; Emanuel 1983; Thorpe and Emanuel 1985; Sanders and Bosart 1985; Xu 1989). These include conditional symmetric instability (CSI; Schultz and Schumacher 1999) and lower-tropospheric frontogenesis in the presence of small or negative equivalent potential vorticity (EPV; e.g., McCann 1995; Glinton et al. 2017; Ganetis and Cole 2015; Ganetis et al. 2018). Especially when temperatures are sufficiently cold for precipitation to reach the surface in the form of snow, mesoscale bands of heavy precipitation can impart significant societal impacts, and are difficult to forecast (e.g., Wiesmueller and Zubrick 1998; Nicosia and Grumm 1999; Jurewicz and Evans 2004; Novak et al. 2006; Evans and Jurewicz 2009; Novak et al. 2010; Novak and Colle 2012). Results from these and other studies provide operational forecasters with useful guidance concerning the synoptic-scale environments conducive to mesoscale banding. The advent of high-resolution numerical model forecasts provides an opportunity for explicit prediction of the bands, with such numerical guidance complementing the use of ingredients-based environmental parameter methods in the forecast process for banded snowfall. High-resolution model simulations also provide an opportunity to elucidate the mesoscale dynamics of the bands and their associated underlying physical and microphysical processes.

Observational studies of winter cyclones document the occurrence of upward vertical air velocities in excess of 1 m s−1 in several storm-relative regions (e.g., Cronce et al. 2007; Rosenow et al. 2014; Rauber et al. 2017). While the factors that determine the fall speed of snow are complex and difficult to model or measure, a representative value for the approximate terminal fall velocity of snow is ~1 m s−1 (e.g., Barthazy and Schefold 2006, their Fig. 1; Garrett and Yuter 2014, their Fig. 2; Molthan et al. 2016, their Fig. 9a). Thus, the preceding observations suggest that snow crystals may be lofted when these conditions are present in winter cyclones. Indeed, snow lofting has been measured directly with vertically pointing radars, including with airborne W-band Doppler radar as reported by Rauber et al. (2017). These prior studies document lofting in elevated convection, including cloud-top generating cells (e.g., Stark et al. 2013; Keeler et al. 2016a,b), and in regions of gravity wave activity in highly sheared frontal regions. However, to our knowledge, observations of snow lofting in the lower troposphere in association with mesoscale snowbands are few. More generally, detailed observations of air and hydrometeor motions in and near mesoscale snowbands are somewhat limited (e.g., Novak et al. 2008; Stark et al. 2013).

Studies of mesoscale snowbands document the enhanced response to frontogenetical forcing in the presence of small moist symmetric stability (e.g., Novak et al. 2004; Novak et al. 2008). Specifically, Novak et al. (2004, their Fig. 16) and Novak et al. (2008, their Fig. 12c) demonstrate upward air motions approaching the fall velocity of snow, suggesting the possibility of hydrometeor lofting, or at least of extended hydrometeor residence time. Inspired by these studies, we hypothesize that lofting can occur in the lower troposphere near or immediately upwind of mesoscale snowbands, and that this mechanism can contribute substantially to the banded nature of snowfall (Fig. 1). Because a significant fraction of snow cannot fall through the updraft in these situations, it may be lofted and/or suspended, advected horizontally, and concentrated in a zone of hydrometeor flux convergence at the point where the updraft velocity no longer exceeds the mass-weighted fall velocity of snow particles (Fig. 1a). This location could be dictated in part by hydrometeor loading (which could limit the updraft strength), and is sensitive to microphysical details that affect fall velocity such as crystal habit, size distribution, riming, and aggregation. The idealized schematic provided in Fig. 1 was inspired by the prior work of Novak et al. (2004, their Fig. 16) and Novak et al. (2008, their Fig. 12c). While suggestive of lofting, these previous studies were generally working with lower-resolution data that may not have fully captured narrow ascent maxima in these cases. Observations with vertically pointing radar document hydrometeor lofting in winter storms, but the relation of such regions to mesoscale snowbands, and the precipitation distribution, remains unclear.

Fig. 1.

(a) Preliminary hypothesized snow mixing ratio and vertical motion fields near mesoscale snowbands: Idealized schematic cross section of vertical velocity (dark red contour and shading, labeled, shade and contour interval 1 m s−1) and snow mixing ratio (blue shading). (b) As in (a), but in a situation where the upward vertical velocity does not exceed the terminal fall velocity of snow, resulting in a relatively uniform snowfall distribution. Labeled arrows denote approximate hypothetical vertical and horizontal length scales.

Fig. 1.

(a) Preliminary hypothesized snow mixing ratio and vertical motion fields near mesoscale snowbands: Idealized schematic cross section of vertical velocity (dark red contour and shading, labeled, shade and contour interval 1 m s−1) and snow mixing ratio (blue shading). (b) As in (a), but in a situation where the upward vertical velocity does not exceed the terminal fall velocity of snow, resulting in a relatively uniform snowfall distribution. Labeled arrows denote approximate hypothetical vertical and horizontal length scales.

The hypothesized impact of lower-tropospheric snow lofting on the surface snowfall distribution is that it accentuates snowfall heterogeneity relative to cases in which the vertical velocity is not sufficient to produce lofting (Fig. 1). There are related implications for operational forecasting, numerical model resolution requirements, the formulation of snow terminal fall velocity in microphysical parameterization schemes, and treatment of hydrometeor advection in numerical weather prediction (NWP) models. Lofting of snow enhances horizontal transport by advection, both by increasing hydrometeor residence time, and by elevation of snow mass into zones of stronger horizontal wind speed. The combination of enhanced snow production accompanying strong ascent, horizontal transport of snow, and snow lofting could affect the location of heavy surface snow accumulation, or perhaps influence the location of snowfall gradients near the edge of a cyclone’s precipitation shield.

The importance of hydrometeor lofting in lake-effect snowbands was demonstrated by Reeves and Dawson (2013), and lofting has been documented in studies of strongly forced orographic precipitation (e.g., Garvert et al. 2005), and in cloud-top generating cells (e.g., Rosenow et al. 2014). To the best of the authors’ knowledge, the importance of hydrometeor lofting in association with single-banded snow events accompanying winter cyclones has not been extensively investigated. As mentioned above, many previous studies have documented vertical air motions approaching or exceeding 1 m s−1 in winter storms, including in elevated convection, in cloud-top generating cells, and in association with frontal regions (e.g., Geerts and Hobbs 1991; Rosenow et al. 2014, 2018; Keeler et al. 2016a,b, 2017). Physical mechanisms contributing to strong ascent include release of potential instability (e.g., Cronce et al. 2007; Kumjian et al. 2014; Rosenow et al. 2018) and cloud-top radiative cooling (e.g., Kumjian et al. 2014; Keeler et al. 2016a,b, 2017). There is evidence that hydrometeors originating in generating cells in the comma head region of extratropical cyclones can seed and enhance precipitation within stratiform cloud layers beneath, contributing to banded precipitation (e.g., Stark et al. 2013; Keeler et al. 2016a; Plummer et al. 2014, 2015).

NWP models calculate a plethora of useful atmospheric fields, many of which are either not output from the model, or are not routinely examined by users of model output. This is certainly true of model microphysics schemes, which compute hydrometeor size distributions, terminal fall velocities, and other microphysical quantities that are not typically written to output files. Here, we alter the Weather Research and Forecasting (WRF) Model (Skamarock et al. 2008) and the Thompson et al. (2004, 2008) microphysics scheme to compute and output the vertical snow flux (hereafter VSNF), along with other microphysical properties. Thus, we are able to directly test our hypothesis concerning hydrometeor lofting: We predict the presence of an upward snow flux in the vicinity (upwind) of strong mesoscale snowbands. We do not expect that lofting is solely responsible for mesoscale snowbands, and we recognize the importance of previously identified environmental factors leading to banded precipitation, such as lower-tropospheric frontogenesis in the presence of weak moist symmetric stability. We analyze simulations of two geographically diverse cases featuring prominent snowbands, and perform model sensitivity experiments to test our hypothesis.

The following section discusses case selection, data sources, model configurations, and experimental design. Section 3 presents two case studies, including snow flux diagnostics and experimental simulations, followed by conclusions, implications, and suggestions for future research in section 4.

2. Data and methods

In selecting case studies, we worked with collaborators in the U.S. National Weather Service (NWS) and Central Michigan University to identify candidate cases of banded and nonbanded snowfall. We based our case selection in part on the perceived forecast difficulty by operational NWS forecasters, recent occurrence for data availability, and geographical diversity. To determine the presence or absence of mesoscale snowbands, we utilized an objective band-detection algorithm from Radford (2019). This algorithm is designed to capture single meso-β-scale bands, rather than multibands of the type documented by Ganetis et al. (2018). We initially selected heavy snowfall events from 27 January 2015, 2 February 2016, and 7 March 2018 as examples of banded events, and an event from 20 December 2008 as an example of a significant snowfall event that did not feature prominent, persistent banding. These events took place in either the Midwest or northeastern United States. We omitted the 7 March 2018 event owing to difficulty in obtaining satisfactory model simulations, and in order to allow more in-depth treatment of the banded cases, we focus exclusively on the two remaining banded cases, eliminating the nonbanded case from 2008.

Each of these events was simulated using a customized version of the WRF Model (WRF-ARW V3.9), modified to output three-dimensional fields of mass-weighted terminal fall velocity for snow (υts), Earth-relative snowfall velocity (υtsw), vertical snow flux qs(υtsw), depositional snow tendency, and total snow tendency. By computing the vertical snow flux internally within the Thompson microphysics scheme, we avoid any ambiguity introduced by spatial interpolation. Within the Thompson scheme, we compute the depositional snow tendency as the sum of three terms: large cloud ice converting to snow, deposition of vapor onto snow, and deposition of vapor onto large ice. The total snow tendency is the sum of 9 terms, including sink terms such as melting. See Table 1 for a complete listing of model configuration choices. A variety of initial and boundary condition data were used in the case-study simulations, including the North American Mesoscale Forecast System (NAM; Janjić 2003), and the Rapid Refresh (RAP; Benjamin et al. 2016).

Table 1.

Configuration choices for WRF Model case-study simulations. All simulations utilized 50 vertical levels, a model top of 50 hPa, the diabatic filter initialization (DFI), and the adaptive time step feature. Unless otherwise stated, all simulations utilized Kain–Fritsch convective parameterization (Kain et al. 2006) on domains with a grid length larger than 4 km, the YSU PBL scheme, RRTMG radiation scheme, and the Noah land surface model.

Configuration choices for WRF Model case-study simulations. All simulations utilized 50 vertical levels, a model top of 50 hPa, the diabatic filter initialization (DFI), and the adaptive time step feature. Unless otherwise stated, all simulations utilized Kain–Fritsch convective parameterization (Kain et al. 2006) on domains with a grid length larger than 4 km, the YSU PBL scheme, RRTMG radiation scheme, and the Noah land surface model.
Configuration choices for WRF Model case-study simulations. All simulations utilized 50 vertical levels, a model top of 50 hPa, the diabatic filter initialization (DFI), and the adaptive time step feature. Unless otherwise stated, all simulations utilized Kain–Fritsch convective parameterization (Kain et al. 2006) on domains with a grid length larger than 4 km, the YSU PBL scheme, RRTMG radiation scheme, and the Noah land surface model.

In addition to control simulations of each event, we conducted experimental simulations of the 27 January 2015 and 2 February 2016 banded snow events. One experiment is designed test the sensitivity of banding to the computation of the terminal snowfall velocity in the Thompson microphysics scheme (Thompson et al. 2004, 2008). In this scheme, the fall velocity of snow, υts, is given by

 
υts=aυsDbυsexp(Dfυs),
(1)

where D is particle diameter, and s, s, and s are constants with default settings of 40 s−1, 0.55, and 100.0 m−1, respectively, in recent and current versions of the Thompson scheme in the WRF Model code (WRF V3.9–V4.0). An experiment designed to minimize the effects of snow lofting was run with s set to 160 s−1, which produces terminal fall velocity values for snow that are roughly 4 times larger than those in the control. Displays of the mass-weighted terminal fall velocity confirm that the experimental simulation exhibited the expected increase in fall speed (not shown).

The importance of horizontal snow advection is of potential significance, especially with strong horizontal flow and vigorous upward vertical air motions (relative to typical synoptic-scale values), which can result in greatly increased hydrometeor airborne residence time and enhanced horizontal advective transport (e.g., Colle et al. 2005). To quantify the importance of this process, we modified the WRF Model code in a second experiment to omit advective and diffusive tendencies of snow, while retaining full microphysical treatments and advective and diffusive tendencies of other quantities. In addition to its relation to lofting and horizontal transport, this experimental configuration allows us to answer the more fundamental question of how much spatial separation may occur between regions where snow crystals form and grow aloft, relative to where they reach the surface.

Approximate hydrometeor trajectories were computed for snow by using the mass-weighted terminal snowfall velocity output from the Thompson microphysics scheme along with the vertical air motion. In computing these trajectories, we restrict trajectory computations to locations where the snow mixing ratio exceeded 0.1 g kg−1, and assumed that the ground-relative fall speed for snow is the difference between the air vertical motion and the terminal fall velocity. Three-dimensional trajectories were computed using the Unidata Integrated Data Viewer (IDV) software with the masked zonal, meridional, and ground-relative fall velocities. Owing to relatively coarse temporal resolution, these trajectories are rather highly approximated, but serve to illustrate snow behavior in and near mesoscale snowbands. We reran a simulation of the January 2015 case with an output interval of 15 min, and found that trajectory results were not significantly different from those computed from hourly output (not shown).

3. Results

In the interest of brevity, we will not present in-depth case studies for the two events analyzed. Instead, we present radar imagery and analysis fields for a small subset of times for each event, followed by comparisons to WRF simulations, and vertical snow flux diagnostics.

a. 26–27 January 2015 Northeast U.S. snowband event

On 26 and 27 January 2015, a nor’easter brought heavy snow to parts of downstate New York and southern New England, with storm-total accumulations in some locations approaching 100 cm (WPC 2015; Greybush et al. 2017). At 0000 UTC 26 January, a strong short-wave trough at the 500-hPa level was centered over the Ohio Valley, with confluent flow over the northeastern United States and southeastern Canada; a surface cyclone was centered immediately to the east of the upper trough, while a strengthening surface anticyclone was located over southern Ontario and western Quebec (Fig. 2a). By 1200 UTC 27 January, a new cyclone center had formed and rapidly strengthened to the south of New England with an analyzed central pressure below 984 hPa at this time (Fig. 2b). The trough had also intensified at the 500-hPa level, with a closed 528-dam contour nearly collocated with the surface cyclone. Pronounced warm advection is evident to the north of the surface low center, manifest as strong and sharply veering geostrophic flow between sea level and the 500-hPa level. A short-wave ridge at the 500-hPa level was located to the north of the cyclone, and strengthening of the associated surface anticyclone contributed to a strong sea level pressure gradient over New England (Fig. 2b).

Fig. 2.

Observational and analysis fields for January 2015 banded snowfall case: (a) NAM analysis (40-km grid spacing) of 500-hPa height (black contours, interval 6 dam), absolute vorticity (×10−5 s−1, shaded as in legend), wind barbs, and sea level pressure (red contours, interval 4 hPa) at 0000 UTC 26 Jan 2015. (b) As in (a), but valid at 1200 UTC 27 Jan; (c) observed composite radar reflectivity at 0900 UTC 27 Jan, radar data obtained from Iowa State University; and (d) NOAA NOHRSC 24-h snowfall accumulation (cm, shaded as in legend) ending 1200 UTC 27 Jan 2015, data obtained from https://www.nohrsc.noaa.gov/snowfall/.

Fig. 2.

Observational and analysis fields for January 2015 banded snowfall case: (a) NAM analysis (40-km grid spacing) of 500-hPa height (black contours, interval 6 dam), absolute vorticity (×10−5 s−1, shaded as in legend), wind barbs, and sea level pressure (red contours, interval 4 hPa) at 0000 UTC 26 Jan 2015. (b) As in (a), but valid at 1200 UTC 27 Jan; (c) observed composite radar reflectivity at 0900 UTC 27 Jan, radar data obtained from Iowa State University; and (d) NOAA NOHRSC 24-h snowfall accumulation (cm, shaded as in legend) ending 1200 UTC 27 Jan 2015, data obtained from https://www.nohrsc.noaa.gov/snowfall/.

Composite radar imagery valid ~0900 UTC 27 January indicates the presence of a curved band of intense snowfall embedded within a broader shield of light and moderate snow (Fig. 2c). The band of heaviest snow extended from coastal Maine, over extreme southeastern New Hampshire, across Massachusetts, Connecticut, and Long Island, New York, at this time. A sequence of radar images indicates that several such bands formed and propagated slowly westward between 0300 and 1200 UTC 27 January (not shown). Snowfall accumulation exhibited a pronounced gradient along the western edge of the precipitation region, with 24-h accumulations ranging from 30 to over 60 cm in eastern New England to <10 cm over western Massachusetts and Connecticut (Fig. 2d).

Simulations of this event initialized at 1200 UTC 26 January were able to capture the main synoptic and mesoscale features of the nor’easter. At hour 21 of the simulation, the 12-km outer domain of the WRF simulation depicted a cyclone with central pressure near 980 hPa to the south of New England at 0900 UTC 27 January (Fig. 3a). The simulated cyclone location was close to that analyzed, while the central pressure was ~3–4 hPa lower than analyzed (not shown, but see Fig. 2b, valid 3 h later). Strong frontogenesis in the presence of near-zero or negative EPV is evident at the 700-hPa level in a band extending from the Gulf of Maine southwestward (Fig. 3b), indicating environmental conditions favorable for banded snowfall.

Fig. 3.

WRF Model simulation of January 2015 event, initialized at 1200 UTC 26 Jan. (a) As in Fig. 2a, but for domain 1 (12-km grid spacing) of WRF 24-h simulation valid 1200 UTC 27 Jan. (b) Simulated 700-hPa height (black solid contours, interval 3 dam), equivalent potential vorticity (red dashed contours, interval 0.5 PVU for regions of negative EPV; 1 PVU = 10−6 K kg−1 m2 s−1), and frontogenesis [K (100 km)−1 h−1, shaded as in legend] for 21-h WRF simulation valid 0900 UTC 27 Jan; (c) domain 2 (4-km grid spacing) WRF simulated composite reflectivity (dBZ, shaded as in legend) valid 0900 UTC 27 Jan; and (d) domain 2 24-h snowfall accumulation computed as change in physical snow depth (cm, shaded as in legend) ending 1200 UTC 27 Jan.

Fig. 3.

WRF Model simulation of January 2015 event, initialized at 1200 UTC 26 Jan. (a) As in Fig. 2a, but for domain 1 (12-km grid spacing) of WRF 24-h simulation valid 1200 UTC 27 Jan. (b) Simulated 700-hPa height (black solid contours, interval 3 dam), equivalent potential vorticity (red dashed contours, interval 0.5 PVU for regions of negative EPV; 1 PVU = 10−6 K kg−1 m2 s−1), and frontogenesis [K (100 km)−1 h−1, shaded as in legend] for 21-h WRF simulation valid 0900 UTC 27 Jan; (c) domain 2 (4-km grid spacing) WRF simulated composite reflectivity (dBZ, shaded as in legend) valid 0900 UTC 27 Jan; and (d) domain 2 24-h snowfall accumulation computed as change in physical snow depth (cm, shaded as in legend) ending 1200 UTC 27 Jan.

Model simulated composite reflectivity from domain 2 of the WRF control simulation (4-km grid spacing) valid 0900 UTC 27 January indicates heavy snow, especially over eastern Massachusetts, Connecticut, and Long Island (Fig. 3c). Some banding is evident in the reflectivity field, with reflectivity values ~35 dBZ in the areas of heaviest simulated snowfall, comparable to that observed (Figs. 2c and 3c). Accumulated snow from the 4-km domain, computed as the difference between the snow depth at the end of the simulation and that at the beginning1, indicates a large swath of >30 cm accumulation extending across southern New England and eastern Long Island. Simulated maximum accumulations are lower than observed, although the region of strong snowfall gradient generally compares favorably between model simulation and observation (Figs. 2d and 3d); an exception is the New York City area and eastern New Jersey, where the model simulation produced greater than observed snow accumulation.

An elongated region of upward snow flux (VSNF > 0) is evident in the 1.33-km domain upwind of the band over eastern Long Island at 1100 UTC 27 January, depicted as a white shaded isosurface (Fig. 4). The region of upward snow flux is centered between 800 and 600 hPa (Fig. 4c), and is displaced upwind and aloft from the band of largest reflectivity; we provide a 3D view in Fig. 4b in order to offer an additional perspective and to give a sense of the altitude and vertical extent of this feature. An elongated region of weaker simulated reflectivity is present beneath and immediately downwind of the band of upward snow flux (Figs. 4a–c), which we speculate could correspond to the region of reduced snow mixing ratio located beneath the strongest updraft presented in Fig. 1a. Cross sections of VSNF, snow mixing ratio, and vertical velocity support this interpretation, although the zone of ascent and lofting is upright, not sloped as in the hypothesized schematic (Figs. 1a and 4c). In cross-sectional view, there are several regions of snow lofting, with the westernmost one centered between 2- and 4-km altitude. The simulated band of lower-tropospheric snow lofting is present in both the 4- and 1.33-km domains, extending parallel to the band of heavy snow, and present in the same band-relative location for several hours (not shown). Additional areas of upward snow flux are evident in convective towers to the south and east of the mesoscale snowband (Fig. 4b), consistent with previous studies of lofting in elevated convection and cloud-top generating cells (e.g., Rauber et al. 2014); these features appear to be ubiquitous in high-resolution model simulations of events with convective updrafts.

Fig. 4.

(a) Simulated 900-hPa reflectivity (shaded as in legend), 600-hPa wind barbs, and upward vertical snow flux isosurface (light gray shading) for hour 23 of domain 3 (1.333-km grid spacing) WRF simulation valid 1100 UTC 27 Jan 2015; black line marks cross-sectional location. (b) As in (a), but for three-dimensional (3D) perspective view. (c) Cross section in 3D perspective: snow mixing ratio (kg kg−1, shaded as in legend at top), positive vertical snow flux (red contours, interval 1 × 10−4 m s−1, beginning at +1 × 10−6 m s−1 so that outermost contour represents a small upward value), and upward vertical velocity (white contours, every 0.25 m s−1 beginning 0.75 m s−1). Gray horizontal contours show altitude at 2-km intervals. Section orientation shown in (a); in (c), gray dashed latitude–longitude lines are shown at 1° intervals for scale.

Fig. 4.

(a) Simulated 900-hPa reflectivity (shaded as in legend), 600-hPa wind barbs, and upward vertical snow flux isosurface (light gray shading) for hour 23 of domain 3 (1.333-km grid spacing) WRF simulation valid 1100 UTC 27 Jan 2015; black line marks cross-sectional location. (b) As in (a), but for three-dimensional (3D) perspective view. (c) Cross section in 3D perspective: snow mixing ratio (kg kg−1, shaded as in legend at top), positive vertical snow flux (red contours, interval 1 × 10−4 m s−1, beginning at +1 × 10−6 m s−1 so that outermost contour represents a small upward value), and upward vertical velocity (white contours, every 0.25 m s−1 beginning 0.75 m s−1). Gray horizontal contours show altitude at 2-km intervals. Section orientation shown in (a); in (c), gray dashed latitude–longitude lines are shown at 1° intervals for scale.

A band of strong lower-tropospheric frontogenesis, evident in Fig. 3b, is shown in three dimensions in Fig. 5a. The 1 m s−1 vertical motion isosurface, shown in red in Fig. 5b, is centered within and directly above the frontogenesis maximum. Finally, superimposing the isosurface of upward snow flux (white, Fig. 5c) reveals that lofting is favored at higher altitudes, where snow particle sizes are smaller and exhibit smaller terminal fall velocities; weaker ascent is sufficient to loft these lighter snow crystals (Fig. 5c). The zones of upward air motion greater than 1 m s−1 and snow lofting shown in Fig. 5 are in the lower troposphere; the uppermost portion of the lofting region in Fig. 5c is located below the 600-hPa level. To demonstrate that snow is actually traversing the lofting region and being suspended there, we computed backward hydrometeor trajectories for snow as described in section 2. Snow reaching the surface over eastern Long Island at 1100 UTC 27 January can be traced to the east and northeast of this location several hours prior; there is a clear inflection in the trajectories as they pass through the zone of strong ascent and lofting (Fig. 6). The trajectories descend rapidly from above the 5-km level until they approach the zone of strong ascent; there, the trajectories level out, with many ascending, as they pass through the zone of upward snow flux (Fig. 6). Thus, it is clear that the snow particles reaching the surface in the simulated snowband were previously lofted, and delayed in their descent, prior to reaching the surface. The prolonged residence time may also enhance particle growth through deposition (not shown).

Fig. 5.

Perspective view of simulated snowband properties for hour 23 of domain 3 (1.333-km grid spacing) WRF simulation valid 1100 UTC 27 Jan 2015: simulated composite reflectivity (base image) superimposed with isosurfaces of (a) frontogenesis (blue, 1 × 10−5 K m−1 s−1), (b) vertical air velocity (red, 1 m s−1), and (c) upward vertical snow flux (white, upward).

Fig. 5.

Perspective view of simulated snowband properties for hour 23 of domain 3 (1.333-km grid spacing) WRF simulation valid 1100 UTC 27 Jan 2015: simulated composite reflectivity (base image) superimposed with isosurfaces of (a) frontogenesis (blue, 1 × 10−5 K m−1 s−1), (b) vertical air velocity (red, 1 m s−1), and (c) upward vertical snow flux (white, upward).

Fig. 6.

Perspective view of snow trajectories from 1.333-km WRF simulation ending 1100 UTC 27 Jan looking southwest, superimposed on 950-hPa snow mixing ratio (kg kg−1, shaded). Backward snow trajectory ribbons colored by altitude, red shaded above 5 km. Trajectories computed from 1.333-km WRF domain with 15-min frequency. White shaded isosurface depicts upward snow flux. (a) View from higher altitude perspective, and (b) reveals the region beneath snow trajectories, indicating an upward bend in the trajectories near the region of lofting. Two-letter state abbreviations and directional arrows are provided for orientation.

Fig. 6.

Perspective view of snow trajectories from 1.333-km WRF simulation ending 1100 UTC 27 Jan looking southwest, superimposed on 950-hPa snow mixing ratio (kg kg−1, shaded). Backward snow trajectory ribbons colored by altitude, red shaded above 5 km. Trajectories computed from 1.333-km WRF domain with 15-min frequency. White shaded isosurface depicts upward snow flux. (a) View from higher altitude perspective, and (b) reveals the region beneath snow trajectories, indicating an upward bend in the trajectories near the region of lofting. Two-letter state abbreviations and directional arrows are provided for orientation.

To further isolate the role of snow lofting, we conducted two experimental simulations to reduce or eliminate lofting. The fast-snow experiment features increased terminal snow fall velocity by a factor of ~4, while the no-snow-advection simulation eliminates snow advection and diffusion. Comparisons of composite reflectivity between the control, fast-snow, and no-snow-advection simulations reveal that reflectivity is generally lower over southern New England in the fast-snow simulation relative to the other two (Figs. 7a,b). This is likely attributable to a greatly reduced snow mixing ratio in the fast-snow experiment, as snow material is quickly removed due to fallout, and thus cannot contribute to the simulated reflectivity (not shown). Despite the absence of hydrometeor lofting in this experimental simulation, a band of relatively high simulated reflectivity is seen in the fast-snow experiment extending from southeastern Connecticut south-southwestward across eastern Long Island (Fig. 7b); this band is more narrow, and located to the east of the band in the control simulation. This is consistent with the interpretation that snow is falling directly beneath the strongest updraft in this simulation, where as it is being lofted and advected laterally downwind of the strongest ascent in the control experiment (Fig. 4c).

Fig. 7.

Reflectivity comparison at 900-hPa level from hour 21 of domain 3 (1.333-km grid length) WRF simulation valid 0900 UTC 27 Jan 2015: (a) control simulation, (b) fast-snow simulation, and (c) no-snow-advection simulation.

Fig. 7.

Reflectivity comparison at 900-hPa level from hour 21 of domain 3 (1.333-km grid length) WRF simulation valid 0900 UTC 27 Jan 2015: (a) control simulation, (b) fast-snow simulation, and (c) no-snow-advection simulation.

In the no-snow-advection experiment, regions of larger reflectivity are generally confined to southern and eastern portions of the snowfall region, as snow is essentially required to fall directly beneath regions where it forms. As in the fast-snow experiment, the band of heaviest snow is located to the east of its location in the control experiment. Given that the bands in this case are oriented largely parallel to the flow in the lower and midtroposphere, the reflectivity patterns between the control and no-snow-advection experiment are generally similar (Figs. 7a,c). Cross-sectional plots similar to Fig. 4c, but for the experimental simulations, are consistent with these interpretations (not shown). Computation of snow trajectories for the fast-snow experiment show what we would expect: That the snow falls nearly directly downward and with greatly reduced residence time (not shown).

Interpretation of differences in lower-tropospheric reflectivity between the control and experimental runs is complicated by differences in snow mixing ratio and may not correspond to actual differences in surface snow accumulation. Therefore, we now compare the water-equivalent snowfall between the control and experiments through hour 21 of the simulations, valid 0900 UTC 27 January (Fig. 8). Note that there may not be consistency between the snow accumulation determined in this way and the increase in physical snow depth as presented in Fig. 3d (which shows 24-h accumulations); however, this calculation provides greater spatial continuity by including overwater regions. The control simulation produced liquid-equivalent snow depths greater than 20 mm across southeastern Massachusetts, Rhode Island, and much of Long Island (Fig. 8a). The fast-snow simulation produced a more highly structured field that again featured a band of heavier accumulation extending from eastern Massachusetts southward across eastern Connecticut and Long Island (Fig. 8b). A difference plot reveals that the aforementioned band in the fast-snow simulation resulted in snowfall accumulations that exceeded those in the control simulation in the band from eastern Massachusetts to eastern Long Island (Fig. 8c). Snowfall differences between the fast-snow and control simulations indicate regions of both positive and negative difference, but there are two prominent areas of positive difference, one in the aforementioned band, and the other over western Long Island (Fig. 8c). These differences arose earlier in the model simulation (not shown). The generally more detailed snow accumulation pattern in the fast-snow simulation relative to the control suggests that reduced residence time for snow results in a less diffuse surface snowfall distribution (Figs. 8a,b).

Fig. 8.

Domain 3 (1.333-km grid spacing) accumulated snowfall comparison (mm of liquid equivalent snowfall, shaded as in legend) through 0900 UTC 27 Jan: (a) control simulation, (b) fast-snow simulation, (c) difference field, fast-snow minus control, (d) no-snow-advection simulation, and (e) difference field, no-snow-advection minus control.

Fig. 8.

Domain 3 (1.333-km grid spacing) accumulated snowfall comparison (mm of liquid equivalent snowfall, shaded as in legend) through 0900 UTC 27 Jan: (a) control simulation, (b) fast-snow simulation, (c) difference field, fast-snow minus control, (d) no-snow-advection simulation, and (e) difference field, no-snow-advection minus control.

The no-snow-advection simulation also produced a swath of slightly heavier snow over eastern MA, but with peak values only reaching 15 mm over eastern Massachusetts relative to 25 mm in the fast-snow simulation (Figs. 8b,d). As in the fast-snow simulation, the no-snow-advection run also produced a zone of enhanced snow over parts of Long Island. A similar pattern of difference from the control simulation is evident in the no-snow-advection experiment as for the fast-snow experiment, although the magnitude of the positive differences is generally smaller than that between the fast-snow and control simulations (Fig. 8e). The reasons for the overall reduction in snow accumulation in the no-snow-advection experiment are not immediately obvious, but further investigation of this is beyond the scope of this study. We offer the speculative explanation that in both of the experimental simulations, reduced depositional snow growth may result from reduced residence time and vapor availability; efficiency of the seeder–feeder mechanism would be reduced in these experiments.

The results of the experimental simulations illustrate the sensitivity of banding and snow accumulation to terminal snow fall velocity and advection, but are not generally consistent with our initial hypothesis. The liquid-equivalent snowfall pattern appears to be more homogeneous in the control simulation relative to the fast-snow experiment, indicating that lofting and lateral advection may serve to disperse snow in this case, perhaps due to the combined effects of strong directional wind shear, horizontal advection, and differential snow fallout over and downwind of the snow-formation zone. Shear-induced dispersion of snow has been documented in studies of cloud-top generating cells (e.g., Rosenow et al. 2014), and a similar mechanism may be at work here. With faster snow terminal velocities, and when advection is eliminated, greater snow accumulation is seen over the New York City area, suggesting that advection and lofting may have served to limit snow accumulation in this region in the control simulation. The similarity between regions of positive and negative difference in the two experimental runs is understandable, given that in each of these experimental simulations, snow falls more directly beneath formation regions, relative to the control simulation. Overall, the conclusions drawn from the experimental simulations are that (i) these processes (lofting and advection) indeed exert a considerable influence on the snow distribution, but (ii) their impacts, while complex, do not lead in this case to the hypothesized increase in snowfall heterogeneity.

b. 2 February 2016 Midwest snowband event

An intense upper-level disturbance moved over the southwestern United States and northern Mexico on 1 February 2016, and was centered over Colorado and Northern New Mexico by 0000 UTC 2 February (Fig. 9a). A strong jet was located to the south of the trough axis, with diffluent flow and warm advection evident in a veering geostrophic wind profile over much of the central Great Plains at this time. By 1200 UTC 2 February, an associated surface cyclone was centered over Kansas (Fig. 9b), and a swath of heavy precipitation was occurring to the north of the low center in a region of warm advection. As the strongest portion of the jet shifted to the east of the upper trough axis by 0000 UTC 3 February, the upper-level system began to lift, and the associated surface cyclone weakened somewhat (not shown).

Fig. 9.

Observational and analysis fields for February 2016 banded snowfall case: (a) NAM analysis (40-km grid spacing) of 500-hPa height (black contours, interval 6 dam), absolute vorticity (×10−5 s−1, shaded as in legend), wind barbs, and sea level pressure (red contours, interval 4 hPa) for 0000 UTC 2 Feb 2016; (b) as in (a), but valid 1200 UTC 2 Feb; (c) observed composite radar reflectivity for 1400 UTC 2 Feb, data obtained from Iowa State University; and (d) NOAA NOHRSC 48-h snowfall accumulation (cm, shaded as in legend) ending 1200 UTC 3 Feb 2016, data obtained from https://www.nohrsc.noaa.gov/snowfall/.

Fig. 9.

Observational and analysis fields for February 2016 banded snowfall case: (a) NAM analysis (40-km grid spacing) of 500-hPa height (black contours, interval 6 dam), absolute vorticity (×10−5 s−1, shaded as in legend), wind barbs, and sea level pressure (red contours, interval 4 hPa) for 0000 UTC 2 Feb 2016; (b) as in (a), but valid 1200 UTC 2 Feb; (c) observed composite radar reflectivity for 1400 UTC 2 Feb, data obtained from Iowa State University; and (d) NOAA NOHRSC 48-h snowfall accumulation (cm, shaded as in legend) ending 1200 UTC 3 Feb 2016, data obtained from https://www.nohrsc.noaa.gov/snowfall/.

Composite radar reflectivity indicates the presence of a region of moderate to heavy snow across much of eastern Nebraska and northwestern Iowa at 1400 UTC 2 February (Fig. 9c). There is some evidence for mesoscale snowbands extending from northwestern Iowa southwestward across Nebraska and into extreme northern Kansas at this time. Banded features first appear in radar imagery around 1200 UTC 2 February and persist until just after 0000 UTC 3 February (not shown). The reflectivity field over southeastern Iowa and northern Missouri was more convective in character, including cellular features approaching 50 dBZ (Fig. 9c). A swath of analyzed 48-h snowfall accumulation in excess of 30 cm extends from northwestern Iowa westward into eastern Colorado and northwestern Kansas (Fig. 9d), with maximum accumulations approaching 50 cm in central Nebraska.

For simulations of this case, we used a single WRF domain featuring 3-km grid spacing (Table 1). This simulation deepened the surface cyclone ~4 hPa more than analyzed, although the location was close to that in the analysis over central Kansas (Figs. 9b and 10a). Examination of 700-hPa frontogenesis and EPV valid 1400 UTC 2 February indicates the presence of a narrow band of strong frontogenesis extending from southwest to northeast across eastern Nebraska (Fig. 10b). Although negative values of EPV were not collocated with the zone of strong frontogenesis at this level and time, cross sections taken perpendicular to the frontal zone confirm the presence of small or negative EPV in the near vicinity of the frontogenesis region (not shown). Thus, environmental conditions were generally conducive to mesoscale snowbands.

Fig. 10.

WRF Model simulation of February 2016 event, initialized 1800 UTC 1 Feb: (a) as in Fig. 9a, but for WRF 18-h simulation valid 1200 UTC 2 Feb 2016. (b) Simulated 700-hPa height (black solid contours, interval 3 dam), equivalent potential vorticity (red dashed contours, interval 0.5 PVU, only for regions of negative EPV), and frontogenesis [K (100 km)−1 h−1, shaded as in legend] for 20-h WRF simulation valid 1400 UTC 2 Feb. (c) Simulated composite reflectivity (dBZ, shaded as in legend) for 20-h WRF simulation valid 1400 UTC 2 Feb; (d) WRF 36-h accumulated snowfall (cm, shaded as in legend).

Fig. 10.

WRF Model simulation of February 2016 event, initialized 1800 UTC 1 Feb: (a) as in Fig. 9a, but for WRF 18-h simulation valid 1200 UTC 2 Feb 2016. (b) Simulated 700-hPa height (black solid contours, interval 3 dam), equivalent potential vorticity (red dashed contours, interval 0.5 PVU, only for regions of negative EPV), and frontogenesis [K (100 km)−1 h−1, shaded as in legend] for 20-h WRF simulation valid 1400 UTC 2 Feb. (c) Simulated composite reflectivity (dBZ, shaded as in legend) for 20-h WRF simulation valid 1400 UTC 2 Feb; (d) WRF 36-h accumulated snowfall (cm, shaded as in legend).

Simulated composite reflectivity exhibited many similar characteristics to that observed, although simulated precipitation coverage was greater over northern and central Missouri than observed (Figs. 9c and 10c). A band of enhanced simulated composite reflectivity is seen near the frontogenesis feature mentioned previously, extending southwestward across Nebraska (Figs. 10b,c). Accumulated snowfall values were similar to those observed (Figs. 9d and 10d), although because this is a 36-h simulation, accumulations over Iowa were reduced relative to the analyzed 48-h totals shown in Fig. 9d.

During a period of prominent banding, 1400 UTC 2 February, a region of upward snow flux is located upwind (east) and aloft of the band (Fig. 11). Upward snow flux was not observed over or near the western portions of the band over central Nebraska. The strong easterly wind component in the lower and middle troposphere may have advected lofted snow westward along the axis of higher reflectivity, across central Nebraska and northwestern Kansas; we will test this speculation with hydrometeor trajectories and with experimental simulations, described below. Additional regions of upward snow flux are evident in association with convective cells over southeastern Iowa, northeastern Missouri, and western Illinois (Figs. 11a,b).

Fig. 11.

Simulated reflectivity (shaded as in legend), 700-hPa wind barbs (black barbs), and isosurface of upward snow flux (gray isosurface) for 20-h WRF simulation valid 1400 UTC 2 Feb 2016: (a) top view, and (b) as in (a), but for 3D view looking northward.

Fig. 11.

Simulated reflectivity (shaded as in legend), 700-hPa wind barbs (black barbs), and isosurface of upward snow flux (gray isosurface) for 20-h WRF simulation valid 1400 UTC 2 Feb 2016: (a) top view, and (b) as in (a), but for 3D view looking northward.

As in the January 2015 case, superposition of the 1 m s−1 vertical motion isosurface with the upward snow flux isosurface reveals that upward snow flux is widespread in regions of convection, and is favored aloft where snow crystals are smaller and lighter (Fig. 12). At lower altitudes, especially in regions of strong updraft, the snow has likely experienced riming, and has a fall velocity greater than 1 m s−1 (not shown). Despite the widespread presence of snow lofting to the south and east of the band across Nebraska, there was less lower-tropospheric lofting in this event relative to the January 2015 case. At times, there were similar bands of lofted snow, but primarily to the east, and in this case, the midlevel flow was more parallel to the band relative to that in the January 2015 event. Backward snow trajectories (Fig. 13), taken at several points along the band across Nebraska, indicate that the snow in the band can be traced back to regions of lofting in convection to the south of the primary snowband (Figs. 13b,c), in contrast to what was found for the 2015 case. Three back-trajectory regions along the western portion of the band indicate origins in convective turrets to the south (Figs. 13a–c). At least some of this convection appears to be surface based, while those cells farther east and north appear to be elevated. The latter are consistent with cloud-top generating cells (e.g., Rosenow et al. 2014; Keeler et al. 2016a,b).

Fig. 12.

Perspective view of upward vertical velocity greater than 1 m s−1 (red isosurface, 1 m s−1), upward snow flux (gray isosurface), and 700-hPa wind vectors (vector arrows, m s−1, shaded by magnitude as in legend at top) for 20-h WRF simulation valid 1400 UTC 2 Feb 2016. View is looking west-southwestward; direction arrow and 2-letter state abbreviations added for orientation.

Fig. 12.

Perspective view of upward vertical velocity greater than 1 m s−1 (red isosurface, 1 m s−1), upward snow flux (gray isosurface), and 700-hPa wind vectors (vector arrows, m s−1, shaded by magnitude as in legend at top) for 20-h WRF simulation valid 1400 UTC 2 Feb 2016. View is looking west-southwestward; direction arrow and 2-letter state abbreviations added for orientation.

Fig. 13.

Snow trajectories superimposed on simulated 800-hPa simulated reflectivity for 2 Feb 2016 event. Backward snow trajectory ribbons colored by altitude, red shaded above 5 km, valid 1400 UTC 2 Feb 2016. (a) Backward snow trajectories for westernmost portion of band, (b) snow trajectory cluster farther east along band, (c) snow trajectory cluster east of band center, and (d) snow trajectory cluster from easternmost portion of band, along with isosurface of upward slow flux (white). View in (c) is toward the southwest; 2-letter state abbreviations, and directional arrow in (d) are provided for orientation.

Fig. 13.

Snow trajectories superimposed on simulated 800-hPa simulated reflectivity for 2 Feb 2016 event. Backward snow trajectory ribbons colored by altitude, red shaded above 5 km, valid 1400 UTC 2 Feb 2016. (a) Backward snow trajectories for westernmost portion of band, (b) snow trajectory cluster farther east along band, (c) snow trajectory cluster east of band center, and (d) snow trajectory cluster from easternmost portion of band, along with isosurface of upward slow flux (white). View in (c) is toward the southwest; 2-letter state abbreviations, and directional arrow in (d) are provided for orientation.

As for the January 2015 case, in order to examine the roles of lofting and snow advection, we again present two experimental simulations, one in which the fall velocity of snow was increased by a factor of ~4 (see section 2), and another in which the advection of snow was omitted. The hypothesized outcome of each of the two experimental simulations is that banding would be reduced or eliminated, and that the location of the snowfall region would shift eastward and southward, owing to a reduction in westward and northward horizontal transport. In this case, the band of lower-tropospheric lofting is less important, but these same conclusions would apply to lofted snow in convection to the south and east of the snowfall region.

Examination of the simulated composite reflectivity at 1400 UTC 2 February indicates the presence of a snowband extending across central Nebraska in the control simulation (Figs. 14 and 15a), featuring reflectivity values of over 30 dBZ. The corresponding display from the fast-snow experimental simulation reveals that the snowband was largely absent (Fig. 14b), with some indication of enhanced reflectivity immediately to the cold side of the 0°C 2-m temperature isotherm (this feature was also present in the control simulation). The only difference between these two simulations is an increase in the value of s as discussed in section 2, resulting in an increase of the terminal fall velocity of snow. Similarly, the no-snow-advection experiment also showed much less prominent banding, although reflectivity values were closer to those in the control simulation across central Nebraska (Fig. 14c). In the no-snow-advection experiment, there is some indication of a band extending from central Iowa into east-central Nebraska (Fig. 14c). Owing to the reduced mixing ratio and reflectivity in the fast-snow experiment, we must again examine accumulated snowfall in order to ascertain the impact of enhanced fall velocity on the surface snowfall distribution.

Fig. 14.

Composite reflectivity comparison for 20-h WRF simulation valid 1400 UTC 2 Feb 2016: (a) control simulation, (b) fast-snow simulation, and (c) no-snow-advection simulation. The 0°C isotherm at the 2-m level is included for reference (solid black line).

Fig. 14.

Composite reflectivity comparison for 20-h WRF simulation valid 1400 UTC 2 Feb 2016: (a) control simulation, (b) fast-snow simulation, and (c) no-snow-advection simulation. The 0°C isotherm at the 2-m level is included for reference (solid black line).

Fig. 15.

Accumulated snowfall comparison. (a) Accumulated snow (cm, shaded as in legend) for 36-h period ending 0600 UTC 3 Feb 2016. (b) as in (a), but for fast-snow experiment, (c) difference between control and fast-snow [(b) minus (a), fast-snow minus control], (d) as in (a), but for no-snow-advection experiment, and (e) difference between control and no-snow-advection [(d) minus (a), no-snow-advection minus control].

Fig. 15.

Accumulated snowfall comparison. (a) Accumulated snow (cm, shaded as in legend) for 36-h period ending 0600 UTC 3 Feb 2016. (b) as in (a), but for fast-snow experiment, (c) difference between control and fast-snow [(b) minus (a), fast-snow minus control], (d) as in (a), but for no-snow-advection experiment, and (e) difference between control and no-snow-advection [(d) minus (a), no-snow-advection minus control].

The total snowfall accumulation from the control simulation was maximized in a swath extending southwest to northeast across Nebraska and into northwestern Iowa (Figs. 10d and 15a). Accumulation in the fast-snow experiment featured more small-scale variability, with the axis of maximum accumulation shifted southward and eastward relative to that in the control simulation (Figs. 15a,b). Differences in accumulation between these two simulations were substantial, ranging up to 30 cm (Fig. 15c). Snow accumulation in the no-snow-advection experiment was strikingly similar to that in the fast-snow experiment (Fig. 15d), also featuring southward and eastward shifts relative to the control simulation (Fig. 15e). This is not surprising, because the physical effect of faster snow fall velocity is to reduce residence time and thus horizontal advection. The southerly and easterly flow evident at the 700-hPa level in Fig. 11 is consistent with snow advection to the north and west; therefore, reducing or omitting this process is consistent with the southward and eastward shift in the region of heavy snow accumulation evident in Fig. 15. Consistent with results from the January 2015 experiments, lofting and advection appear again in this case to produce a smoother snowfall distribution.

The shift in snow accumulation between the control simulation and that in the two experimental simulations demonstrates that there can be a significant spatial offset between the physical locations where snow forms and grows aloft, and where it reaches the surface. These simulations highlight the importance of lofting and horizontal transport in determining surface snowfall accumulation. In the simulation with increased terminal fall velocity, the mesoscale snowband did not form. The experimental results from this case differ from those of the 27 January 2015 event, suggesting that in this case, the band was more strongly linked to lofting and advection originating in convective towers to the south and east of the snowband. However, as in the 27 January 2015 case, the experimental simulations again exhibited more spatial variability than in the control simulation, contrary to our initial hypothesis.

4. Conclusions

Earlier studies of mesoscale snowbands have identified the environmental conditions that are conducive to banding, namely forcing for ascent (frontogenesis) in an environment of weak stability, favoring bands of strong ascent (e.g., Wiesmueller and Zubrick 1998; Nicosia and Grumm 1999; Jurewicz and Evans 2004; Novak et al. 2004; Novak et al. 2008). Both observational and modeling studies demonstrate that the strength of ascending air motions in the comma head region of such winter storms can approach or exceed 1 m s−1, a value that roughly corresponds to the fall velocity of snow. Ascent of sufficient strength to loft snow in winter cyclones has been documented in several distinct locations, including in elevated convection and in cloud-top generating cells (e.g., Cronce et al. 2007; Stark et al. 2013; Rosenow et al. 2014; Keeler et al. 2016a,b). However, these studies did not document lofting of snow in the lower troposphere in association with mesoscale snowbands, although results from several prior studies strongly suggested this possibility as discussed in section 1 (e.g., Novak et al. 2004; Novak et al. 2008). These studies led us to investigate whether lofting was happening in numerically simulated winter cyclones that featured banded snowfall, and once documented, to explore its importance. Here we present analyses of simulated air and hydrometeor motions in the lower and midtroposphere near snowbands in simulations of two winter cyclone events in differing geographical regions.

There are several potentially important implications for snow lofting in the lower and midtroposphere. First, because snow cannot fall through an updraft that exceeds its fall velocity, lofting could generate or enhance the spatial offset between the region of strongest ascent and the location where the snow actually reaches the surface. This means that lofting increases the horizontal advection of snow, potentially leading to large displacement of surface accumulation from where the strongest forcing for ascent is taking place. Operational forecasters who are examining environmental predictions from mesoscale NWP models would need to recognize and account for this offset in their forecast process. For a similar synoptic situation taking place with a considerably higher freezing level, such an offset would not occur or be greatly diminished because the fall velocity of rain is roughly an order of magnitude larger than that of snow. In other words, it is much more common for upward motion in winter storms to exceed the fall velocity of snow than for rain. Advective transport of snow could also potentially affect other operational products, such as radar-derived quantitative precipitation estimation. Second, the presence of lofting and the associated increase in horizontal advective transport must be accounted for in model microphysical parameterizations. This importance was documented for lake-effect snowbands by Reeves and Dawson (2013), and is potentially important for mesoscale snowbands as well. In general, this indicates that the parameterization of the fall velocity for snow may be more critically important than for other hydrometeors in these situations because of lofting and advection. Third, understanding the role of lofting and transport to surface snowfall heterogeneity is a basic science question that could best be answered through a combination of observational and modeling studies. Here, our goal is to provide initial documentation of this process in numerical simulations and to present an initial analysis to assist in understanding and assessing its importance.

We have simulated several cases of banded and nonbanded snowfall; for brevity, we restrict our presentation here to two cases. Our method involves outputting several fields from the Thompson microphysics scheme that are not routinely output, including the terminal snow fall velocity and an internally computed vertical snow flux. We also conducted experimental simulations in which the snow fall velocity was increased by a factor of 4, and in which advection of snow was omitted from the model. With these modifications, we were able to test our hypothesis directly. In addition, outputting the terminal snow fall velocity enabled us to compute approximate snow trajectories, which were a useful complement to Eulerian diagnostic fields.

In both of the simulated banded cases, we identify regions of lower- and midtropospheric snow lofting near and upwind of the simulated snowbands. Though we were inspired by previous studies of snowbands, this region of snow lofting has not, to the authors’ knowledge, been previously documented either in model simulations or in observations. The regions of lofting did not always accompany the bands, but were present during times when intense bands were occurring in the simulations. When present, this lofting feature took the form of an elongated band located between approximately 2 and 5 km in altitude, upwind of the near-surface snowbands. The zone of lofting was the result of a band of upward air motion exceeding the parameterized fall velocity of snow; beneath this feature was a region of strong frontogenesis in the lower troposphere. In the 27 January 2015 event, a region of relatively weak surface snowfall rate (simulated reflectivity) was located directly beneath the zone of lofting and strongest ascent, with heavy snowfall displaced downstream. This spatial offset was much less evident in the 2 February 2016 event, which also featured a more band-parallel flow orientation in the lower troposphere and a less organized band of lofting. In the 2 February 2016 case, lower-tropospheric lofting was primarily confined to regions well to the east of the band, while in the 27 January 2015 case, lofting was in closer proximity to the region of heaviest snowfall. In the February 2016 case, snow trajectories indicate that lofting in convection, well to the south and east of the primary snowband, was important. Examination of the parameterized depositional snow growth demonstrates that the snow crystals grew substantially during their descent (not shown). In addition, we also noted the presence of previously identified regions of lofting, including cloud-top generating cells and elevated convection.

Key findings from our analysis of these simulations and experiments include:

  • The relation of lofting regions to bands of heavy snowfall differed between the two cases studied. Lofting was more directly associated with banding in the 27 January 2015 case.

  • Beneath regions of snow lofting and in locations where the surface precipitation type was snow, a zone of reduced reflectivity and snow mixing ratio was present beneath the location of strongest ascent (Figs. 8 and 12), consistent with our initial hypothesis (Fig. 1a).

  • Contrary to our initial hypothesis, in the two cases presented the zones of snow lofting tended to be more upright rather than sloped, consistent with convective instability more so than a sloping or slantwise circulation.

  • Also contrary to our initial hypothesis, accelerating the parameterized fall velocity of snow and omitting horizontal advection of snow increased snowfall heterogeneity. These experiments also demonstrate the importance of these processes to the overall snowfall distribution, which changed significantly in their absence.

  • Snow trajectories, computed using only winds and Earth-relative fall velocities for snow, are consistent with the importance of lofting to the snowband in the 27 January 2015 East Coast cyclone event. In the February 2016 Midwest event, snow trajectories in the central portion of the band originated in convection or elevated convection to the south of the snowband, though snow growth took place within the band as well.

In hindsight, our initial hypothesis as presented in Fig. 1 was oversimplified, and failed to account for complexities such as veering wind profiles and nearby convection. Lofting reduces snowfall heterogeneity due to the lateral spreading of snowfall via advection, which evidently ultimately imparts a smoothing effect on the overall snow distribution. Having now documented the occurrence of snow lofting in simulations of banded snowfall, it is important to continue this work with both observational and numerical studies. Observations from field programs would be valuable in determining if the simulated lofting regions upwind of mesoscale bands are actually present. If present, such observations would aid in understanding the importance of this process in different settings, and in providing observational comparisons for numerical model simulations of the type presented here. It would be especially beneficial to compare parameterized terminal fall velocity values for snow, which may be especially important in improving quantitative prediction of surface snowfall, to observations, as has been done for other regions of winter storms (e.g., Keeler et al. 2016a; Molthan et al. 2016). Another approach to quantify the numerical sensitivity of surface snow accumulation to parameterization of snow terminal fall speed would be to run an ensemble, varying only the fall speed parameters, and compute the resulting ensemble spread in surface snowfall accumulation. Finally, we note that outside of banded snowfall, upward snow flux was evident in convection, often found aloft in regions where precipitation reached the ground in the form of rain. Owing to the ubiquitous presence of cold-cloud processes even when the surface precipitation type is in the form of rain, the influence of the processes discussed here on banded rainfall may also be worthy of future investigation.

Acknowledgments

This research was supported by NOAA Grant NA16NWS4680003, awarded to North Carolina State University. The Weather Research and Forecasting Model is made available by the National Center for Atmospheric Research (NCAR), funded by the National Science Foundation. We greatly appreciate the helpful review of Prof. Bob Rauber and two anonymous reviewers on earlier versions of this manuscript. We acknowledge the assistance of Dr. Martin Baxter of Central Michigan University for input on case selection and helpful comments, and Dr. Jimy Dudhia of NCAR for assistance with the no-snow-advection experiments. The gridded analyses used as boundary conditions for our simulations, along with the Stage IV precipitation analyses, are made available by the National Centers for Environmental Prediction and the National Centers for Environmental Information (NCEI) NOMADs system. The ERA-Interim reanalysis data used in this study were obtained from NCAR Computational Information Systems Laboratory (CISL). The Iowa State University meteorology archive is acknowledged for the composite radar data shown in Figs. 2c and 9c, along with other gridded NWP model datasets.

REFERENCES

REFERENCES
Barthazy
,
E.
, and
R.
Schefold
,
2006
:
Fall velocity of snowflakes of different riming degree and crystal types
.
Atmos. Res.
,
82
,
391
398
, https://doi.org/10.1016/j.atmosres.2005.12.009.
Benjamin
,
S. G.
, and Coauthors
,
2016
:
A North American hourly assimilation and model forecast cycle: The Rapid Refresh
.
Mon. Wea. Rev.
,
144
,
1669
1694
, https://doi.org/10.1175/MWR-D-15-0242.1.
Bennetts
,
D. A.
, and
B. J.
Hoskins
,
1979
:
Conditional symmetric instability—A possible explanation for frontal rainbands
.
Quart. J. Roy. Meteor. Soc.
,
105
,
945
962
, https://doi.org/10.1002/qj.49710544615.
Chen
,
F.
, and
J.
Dudhia
,
2001
:
Coupling an advanced land surface-hydrology model with the Penn State–NCAR MM5 modeling system. Part I: Model implementation and sensitivity
.
Mon. Wea. Rev.
,
129
,
569
585
, https://doi.org/10.1175/1520-0493(2001)129<0569:CAALSH>2.0.CO;2.
Colle
,
B. A.
,
M. F.
Garvert
,
J. B.
Wolfe
,
C. F.
Mass
, and
C. P.
Woods
,
2005
:
The 13–14 December 2001 IMPROVE-2 event. Part III: Simulated microphysical budgets and sensitivity studies
.
J. Atmos. Sci.
,
62
,
3535
3558
, https://doi.org/10.1175/JAS3552.1.
Cronce
,
M.
,
R. M.
Rauber
,
K. R.
Knupp
,
B. F.
Jewett
,
J. T.
Walters
, and
D.
Phillips
,
2007
:
Vertical motions in precipitation bands in three winter cyclones
.
J. Appl. Meteor. Climatol.
,
46
,
1523
1543
, https://doi.org/10.1175/JAM2533.1.
Ek
,
M. B.
,
K. E.
Mitchell
,
Y.
Lin
,
E.
Rogers
,
P.
Grunmann
,
V.
Koren
,
G.
Gayno
, and
J. D.
Tarpley
,
2003
:
Implementation of Noah land surface model advances in the National Centers for Environmental Prediction operational mesoscale Eta model
.
J. Geophys. Res.
,
108
,
8851
, https://doi.org/10.1029/2002JD003296.
Emanuel
,
K. A.
,
1983
:
The Lagrangian parcel dynamics of moist symmetric instability
.
J. Atmos. Sci.
,
40
,
2368
2376
, https://doi.org/10.1175/1520-0469(1983)040<2368:TLPDOM>2.0.CO;2.
Evans
,
M.
, and
M. L.
Jurewicz
,
2009
:
Correlations between analyses and forecasts of banded heavy snow ingredients and observed snowfall
.
Wea. Forecasting
,
24
,
337
350
, https://doi.org/10.1175/2008WAF2007105.1.
Ganetis
,
S. A.
, and
B. A.
Colle
,
2015
:
The thermodynamic and microphysical evolution of an intense snowband during the Northeast U.S. Blizzard of 8–9 February 2013
.
Mon. Wea. Rev.
,
143
,
4104
4125
, https://doi.org/10.1175/MWR-D-14-00407.1.
Ganetis
,
S. A.
,
B. A.
Colle
,
S. E.
Yuter
, and
N. P.
Hoban
,
2018
:
Environmental conditions associated with observed snowband structures within Northeast U.S. winter storms
.
Mon. Wea. Rev.
,
146
,
3675
3690
, https://doi.org/10.1175/MWR-D-18-0054.1.
Garrett
,
T. J.
, and
S. E.
Yuter
,
2014
:
Observed influence of riming, temperature, and turbulence on the fallspeed of solid precipitation
.
Geophys. Res. Lett.
,
41
,
6515
6522
, https://doi.org/10.1002/2014GL061016.
Garvert
,
M. F.
,
C. P.
Woods
,
B. A.
Colle
,
C. F.
Mass
,
P. V.
Hobbs
,
M. T.
Stoelinga
, and
J. B.
Wolfe
,
2005
:
The 13–14 December 2001 IMPROVE-2 Event. Part II: Comparisons of MM5 model simulations of clouds and precipitation with observations
.
J. Atmos. Sci.
,
62
,
3520
3534
, https://doi.org/10.1175/JAS3551.1.
Geerts
,
B.
, and
P. V.
Hobbs
,
1991
:
Organization and structure of clouds and precipitation on the mid-Atlantic coast of the United States. Part IV: Retrieval of the thermodynamic and cloud microphysical structures of a frontal rainband from Doppler radar data
.
J. Atmos. Sci.
,
48
,
1287
1305
, https://doi.org/10.1175/1520-0469(1991)048<1287:OASOCA>2.0.CO;2.
Glinton
,
M. R.
,
S. L.
Gray
,
J. M.
Chagnon
, and
C. J.
Morcrette
,
2017
:
Modulation of precipitation by conditional symmetric instability release
.
Atmos. Res.
,
185
,
186
201
, https://doi.org/10.1016/j.atmosres.2016.10.013.
Greybush
,
S. J.
,
S.
Saslo
, and
R.
Grumm
,
2017
:
Assessing the ensemble predictability of precipitation forecasts for the January 2015 and 2016 East Coast winter storms
.
Wea. Forecasting
,
32
,
1057
1077
, https://doi.org/10.1175/WAF-D-16-0153.1.
Janjić
,
Z. I.
,
2003
:
A nonhydrostatic model based on a new approach
.
Meteor. Atmos. Phys.
,
82
,
271
285
, https://doi.org/10.1007/s00703-001-0587-6.
Jurewicz
,
M. L.
, and
M. S.
Evans
,
2004
:
A comparison of two banded, heavy snowstorms with very different synoptic settings
.
Wea. Forecasting
,
19
,
1011
1028
, https://doi.org/10.1175/WAF-823.1.
Kain
,
J. S.
,
S. J.
Weiss
,
J. J.
Levit
,
M. E.
Baldwin
, and
D. R.
Bright
,
2006
:
Examination of convection-allowing configurations of the WRF Model for the prediction of severe convective weather: The SPC/NSSL Spring Program 2004
.
Wea. Forecasting
,
21
,
167
181
, https://doi.org/10.1175/WAF906.1.
Keeler
,
J. M.
,
B. F.
Jewett
,
R. M.
Rauber
,
G. M.
McFarquhar
,
R. M.
Rasmussen
,
L.
Xue
,
C.
Liu
, and
G.
Thompson
,
2016a
:
Dynamics of cloud-top generating cells in winter cyclones. Part I: Idealized simulations in the context of field observations
.
J. Atmos. Sci.
,
73
,
1507
1527
, https://doi.org/10.1175/JAS-D-15-0126.1.
Keeler
,
J. M.
,
B. F.
Jewett
,
R. M.
Rauber
,
G. M.
McFarquhar
,
R. M.
Rasmussen
,
L.
Xue
,
C.
Liu
, and
G.
Thompson
,
2016b
:
Dynamics of cloud-top generating cells in winter cyclones. Part II: Radiative and instability forcing
.
J. Atmos. Sci.
,
73
,
1529
1553
, https://doi.org/10.1175/JAS-D-15-0127.1.
Keeler
,
J. M.
,
B. F.
Jewett
,
R. M.
Rauber
,
G. M.
McFarquhar
,
R. M.
Rasmussen
,
L.
Xue
,
C.
Liu
, and
G.
Thompson
,
2017
:
Dynamics of cloud-top generating cells in winter cyclones. Part III: Shear and convective organization
.
J. Atmos. Sci.
,
74
,
2879
2897
, https://doi.org/10.1175/JAS-D-16-0314.1.
Kumjian
,
M. R.
,
S. A.
Rutledge
,
R. M.
Rasmussen
,
P. C.
Kennedy
, and M. Dixon,
2014
:
High-resolution polarimetric radar observations of snow-generating cells
.
J. Appl. Meteor. Climatol.
,
53
,
1636
1658
, https://doi.org/10.1175/JAMC-D-13-0312.1.
McCann
,
D. W.
,
1995
:
Three-dimensional computations of equivalent potential vorticity
.
Wea. Forecasting
,
10
,
798
802
, https://doi.org/10.1175/1520-0434(1995)010<0798:TDCOEP>2.0.CO;2.
Molthan
,
A. L.
,
B. A.
Colle
,
S. E.
Yuter
, and
D.
Stark
,
2016
:
Comparisons of modeled and observed reflectivities and fall speeds for snowfall of varied riming degrees during winter storms on Long Island, New York
.
Mon. Wea. Rev.
,
144
,
4327
4347
, https://doi.org/10.1175/MWR-D-15-0397.1.
Nicosia
,
D. J.
, and
R. H.
Grumm
,
1999
:
Mesoscale band formation in three major northeastern United States snowstorms
.
Wea. Forecasting
,
14
,
346
368
, https://doi.org/10.1175/1520-0434(1999)014<0346:MBFITM>2.0.CO;2.
Novak
,
D. R.
, and
B. A.
Colle
,
2012
:
Diagnosing snowband predictability using a multimodel ensemble system
.
Wea. Forecasting
,
27
,
565
585
, https://doi.org/10.1175/WAF-D-11-00047.1.
Novak
,
D. R.
,
L. F.
Bosart
,
D.
Keyser
, and
J. S.
Waldstreicher
,
2004
:
An observational study of cold season–banded precipitation in northeast U.S. cyclones
.
Wea. Forecasting
,
19
,
993
1010
, https://doi.org/10.1175/815.1.
Novak
,
D. R.
,
J. S.
Waldstreicher
,
D.
Keyser
, and
L. F.
Bosart
,
2006
:
A forecast strategy for anticipating cold season mesoscale band formation within eastern U.S. cyclones
.
Wea. Forecasting
,
21
,
3
23
, https://doi.org/10.1175/WAF907.1.
Novak
,
D. R.
,
B. A.
Colle
, and
S. E.
Yuter
,
2008
:
High-resolution observations and model simulations of the life cycle of an intense mesoscale snowband over the northeastern United States
.
Mon. Wea. Rev.
,
136
,
1433
1456
, https://doi.org/10.1175/2007MWR2233.1.
Novak
,
D. R.
,
B. A.
Colle
, and
A. R.
Aiyyer
,
2010
:
Evolution of mesoscale precipitation band environments within the comma head of northeast U.S. cyclones
.
Mon. Wea. Rev.
,
138
,
2354
2374
, https://doi.org/10.1175/2010MWR3219.1.
Plummer
,
D. M.
,
G. M.
McFarquhar
,
R. M.
Rauber
,
B. F.
Jewett
, and
D. C.
Leon
,
2014
:
Structure and statistical analysis of the microphysical properties of generating cells in the comma-head region of continental winter cyclones
.
J. Atmos. Sci.
,
71
,
4181
4203
, https://doi.org/10.1175/JAS-D-14-0100.1.
Plummer
,
D. M.
,
G. M.
McFarquhar
,
R. M.
Rauber
,
B. F.
Jewett
, and
D. C.
Leon
,
2015
:
Microphysical properties of convectively generated fall streaks in the comma-head region of continental winter cyclones
.
J. Atmos. Sci.
,
72
,
2465
2483
, https://doi.org/10.1175/JAS-D-14-0354.1.
Radford
,
J. T.
,
2019
:
Verification of high-resolution banded snowfall forecasts. M.S. thesis, Department of Marine, Earth, and Atmospheric Sciences, North Carolina State University, 105 pp., http://www.lib.ncsu.edu/resolver/1840.20/36330
.
Rauber
,
R. M.
,
M. K.
Macomber
,
D. M.
Plummer
,
A. A.
Rosenow
,
G. M.
McFarquhar
,
B. F.
Jewett
,
D.
Leon
, and
J. M.
Keeler
,
2014
:
Finescale radar and airmass structure of the comma head of a continental winter cyclone: The role of three airstreams
.
Mon. Wea. Rev.
,
142
,
4207
4229
, https://doi.org/10.1175/MWR-D-14-00057.1.
Rauber
,
R. M.
,
S. M.
Ellis
,
J.
Vivekanandan
,
J.
Stith
,
W.-C.
Lee
,
G. M.
McFarquhar
,
B. F.
Jewett
, and
A.
Janiszeski
,
2017
:
Finescale structures of a snowstorm over the northeastern United States: A first look at high-resolution HIAPER Cloud Radar observations
.
Bull. Amer. Meteor. Soc.
,
98
,
253
269
, https://doi.org/10.1175/BAMS-D-15-00180.1.
Reeves
,
H. D.
, and
D. T.
Dawson
II
,
2013
:
The dependence of QPF on the choice of microphysical parameterization for lake-effect snowstorms
.
J. Appl. Meteor. Climatol.
,
52
,
363
377
, https://doi.org/10.1175/JAMC-D-12-019.1.
Rosenow
,
A. A.
,
D. M.
Plummer
,
R. M.
Rauber
,
G. M.
McFarquhar
,
B. F.
Jewett
, and
D.
Leon
,
2014
:
Vertical velocity and physical structure of generating cells and convection in the comma head region of continental winter cyclones
.
J. Atmos. Sci.
,
71
,
1538
1558
, https://doi.org/10.1175/JAS-D-13-0249.1.
Rosenow
,
A. A.
,
R. M.
Rauber
,
B. F.
Jewett
,
G. M.
McFarquhar
, and
J. M.
Keeler
,
2018
:
Elevated potential instability in the comma head: Distribution and development
.
Mon. Wea. Rev.
,
146
,
1259
1278
, https://doi.org/10.1175/MWR-D-17-0283.1.
Sanders
,
F.
, and
L. F.
Bosart
,
1985
:
Mesoscale structure in the megalopolitian snowstorm of 11–12 February 1983: Part I: Frontogenetical forcing and symmetric instability
.
J. Atmos. Sci.
,
42
,
1050
1061
, https://doi.org/10.1175/1520-0469(1985)042<1050:MSITMS>2.0.CO;2.
Schultz
,
D. M.
, and
P. N.
Schumacher
,
1999
:
The use and misuse of conditional symmetric instability
.
Mon. Wea. Rev.
,
127
,
2709
2732
, https://doi.org/10.1175/1520-0493(1999)127<2709:TUAMOC>2.0.CO;2; Corrigendum, 128, 1573, https://doi.org/10.1175/1520-0493(1999)127<1573:CORRIG>2.0.CO;2.
Skamarock
,
W. C.
, and Coauthors
,
2008
:
description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp., https://doi.org/10.5065/D68S4MVH
.
Stark
,
D.
,
B. A.
Colle
, and
S. E.
Yuter
,
2013
:
Observed microphysical evolution for two East Coast winter storms and the associated snow bands
.
Mon. Wea. Rev.
,
141
,
2037
2057
, https://doi.org/10.1175/MWR-D-12-00276.1.
Thompson
,
G.
,
R. M.
Rasmussen
, and
K.
Manning
,
2004
:
Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. Part I: Description and sensitivity analysis
.
Mon. Wea. Rev.
,
132
,
519
542
, https://doi.org/10.1175/1520-0493(2004)132<0519:EFOWPU>2.0.CO;2.
Thompson
,
G.
,
P. R.
Field
,
R. M.
Rasmussen
, and
W. D.
Hall
,
2008
:
Explicit forecasts of winter precipitation using an improved bulk microphysics scheme. Part II: Implementation of a new snow parameterization
.
Mon. Wea. Rev.
,
136
,
5095
5115
, https://doi.org/10.1175/2008MWR2387.1.
Thorpe
,
A. J.
, and
K. A.
Emanuel
,
1985
:
Frontogenesis in the presence of small stability to slantwise convection
.
J. Atmos. Sci.
,
42
,
1809
1824
, https://doi.org/10.1175/1520-0469(1985)042<1809:FITPOS>2.0.CO;2.
WPC
,
2015
:
Mid-Atlantic to northeast major winter storm, 26-28 January 2015. Weather Prediction Center (WPC), 3 pp., https://www.wpc.ncep.noaa.gov/winter_storm_summaries/event_reviews/2015/MidAtlantic_Northeast_Major_WinterStorm_Jan2015.pdf
.
Wiesmueller
,
J. L.
, and
S. M.
Zubrick
,
1998
:
Evaluation and application of conditional symmetric instability, equivalent potential vorticity, and frontogenetic forcing in an operational forecasting environment
.
Wea. Forecasting
,
13
,
84
101
, https://doi.org/10.1175/1520-0434(1998)013<0084:EAAOCS>2.0.CO;2.
Xu
,
Q.
,
1989
:
Frontal circulations in the presence of small viscous moist symmetric stability and weak forcing
.
Quart. J. Roy. Meteor. Soc.
,
115
,
1325
1353
, https://doi.org/10.1002/qj.49711549008.

Footnotes

© 2019 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).

1

Snow depth is computed by the Noah land surface model; see Chen and Dudhia (2001) and Ek et al. (2003) for details.