1. Introduction
Cloud-top generating cells (GCs) are small regions of locally high reflectivity at the top of stratiform clouds from which trails of hydrometeors originate (American Meteorological Society 2015). First observed in radar range–height indicator scans in the 1950s (Marshall 1953), GCs are typically 0.75–1.5 km wide and 1.0–2.0 km deep (Langleben 1956; Rosenow et al. 2014; Kumjian et al. 2014). Kinematically, they consist of updrafts from 0.75 to 3.00 m s−1 (Wexler 1955; Douglas et al. 1957; Wexler and Atlas 1959; Carbone and Bohne 1975; Rosenow et al. 2014; Kumjian et al. 2014) and adjacent downdrafts of similar magnitude. While updrafts of that magnitude are considered weak for summertime convection, in the ice-supersaturated environments common at the top of winter cyclones they can promote rapid depositional growth of ice crystals (Plummer et al. 2014, 2015). Trails of hydrometeors originating from GCs have been observed merging into bands of heavy precipitation or have been traced individually to the surface in winter cyclones (Evans et al. 2005; Cunningham and Yuter 2014; Plummer et al. 2014; Rauber et al. 2014a, 2015; Rosenow et al. 2014), demonstrating that GCs are a critical component of the precipitation process in extratropical cyclones. Accordingly, an understanding of the dynamics and thermodynamics that control their development and maintenance is needed.
Generating cells are typically observed in layers characterized by potential instability, the condition of equivalent potential temperature θe (or equivalent potential temperature with respect to ice θei when T < 0°C), decreasing with height (Wexler and Atlas 1959). Sharp vertical transitions from a moist (in cloud) to a dry (above cloud) atmosphere have been observed in the presence of upper-level fronts and cyclone dry slots (e.g., Wexler and Atlas 1959; Kreitzberg 1968; Hobbs et al. 1990; Grim et al. 2007). Large-scale ascent present in the warm-frontal and comma-head regions of extratropical cyclones favors the release of this instability (e.g., Wexler and Atlas 1959; Martin 1998). However, the sensitivity of GCs to varying cloud-top stability has not been established. Douglas et al. (1957) hypothesized that in the absence of cloud-top instability, GC updrafts could still exist as a result of latent heat released during depositional growth of ice crystals. More recently it has been suggested that longwave radiative cooling can decrease stability at cloud top, favoring maintenance of GCs (Kumjian et al. 2014; Rauber et al. 2014a,b). This claim is supported by an increase in GC coverage observed by Syrett et al. (1995) during a brief thinning of cirrus located above a deep cloud layer. The postulation that radiative cooling may be important suggests that the dynamics of GCs may have some similarities with that of stratocumulus, for which cloud-top longwave radiative cooling is the primary driver of convection (Wood 2012). Radiative cooling of 5–10 K h−1 is commonly observed in liquid-phase stratocumulus, with cooling concentrated in a shallow layer (as little as a few meters) having high liquid water content (Wood 2012). In clouds with lower liquid content, the peak cooling rate should decrease, and the depth over which it occurs should increase (Wood 2012).
In this paper, cloud-top diabatic forcing and the corresponding evolution of the vertical velocity field within generating cells are quantified using simulations representative of the comma-head region of an extratropical winter cyclone that was sampled on 14–15 February 2010, during the Profiling of Winter Storms (PLOWS) field campaign (Rosenow et al. 2014; Plummer et al. 2014, 2015). Specifically, the goal of this paper is to 1) quantify the temperature tendency at the GC level due to radiative and latent heating, 2) relate diabatic processes to upper-tropospheric stability and presence or absence of GCs, and 3) compare the structure and bulk microphysical characteristics of simulated GCs to data collected within the GCs during PLOWS. The impact of cloud-top longwave cooling, shortwave heating, and latent heating and cooling on generating cell structure and evolution are then assessed under thermodynamic conditions characteristic of this specific cyclone. This ultimately helps quantify the cloud-top radiative cooling in GCs and determines whether it is important to convective maintenance.
2. The Profiling of Winter Storms campaign
The PLOWS experiment had field phases from January–March 2009 and November 2009–March 2010, with the primary goal of improving the understanding of precipitation processes and the microphysical and kinematic structure of continental extratropical winter cyclones (Rauber et al. 2014b). In this paper, data are used from rawinsondes launched by the NCAR Mobile Integrated Sounding System (MISS) and from the Wyoming Cloud Radar (WCR; Wang et al. 2012) and microphysics probes installed on the NCAR/NSF C-130 aircraft. A detailed discussion of the quality control and processing procedures for the WCR reflectivity and vertical radial velocity data is included in Rosenow et al. (2014). Processing procedures for data from the microphysical probes are presented in Plummer et al. (2014, 2015). The very high resolution of the WCR dataset (~15 m), paired with a recent upgrade of the WCR that increased sensitivity to a minimum detectable equivalent reflectivity of −25 dBZe (Wang et al. 2012), enabled very-high-resolution observations of GC structure (Rosenow et al. 2014; Plummer et al. 2014; Rauber et al. 2014a, 2015).
3. The 14–15 February 2010 cyclone
The synoptic evolution of the cyclone and statistical analyses of the observed vertical radial velocity obtained from the Wyoming Cloud Radar at and below the generating cell level for the 14–15 February 2010 cyclone are discussed in Rosenow et al. (2014). Given that Alberta clippers are the most common cyclone type in the Great Lakes region (Bowie and Weightman 1914; Isard et al. 2000) and that the central pressure of the 14–15 February 2010 cyclone of 1008 hPa was near the peak in central pressure frequency in that region (Isard et al. 2000), this cyclone was representative of typical continental extratropical winter cyclones. The microphysical structure of precipitation within the cyclone is discussed in Plummer et al. (2014, 2015). Supercooled liquid water was not observed in the GCs between 6- and 8-km altitude in this cyclone since the temperature in their vicinity ranged from approximately −45° to −55°C (Plummer et al. 2014). On 14–15 February 2010, 10–25 cm of snowfall accumulated in southern Indiana as the warm-frontal and comma-head regions of an Alberta-clipper cyclone passed over the area. A composite radar image at 0535 UTC is shown in Fig. 1. Precipitation structure at the time of interest for this study (0300–0800 UTC 15 February) included a wide north–south band of moderate snowfall over southern Indiana that bifurcated into two shallower bands that curved cyclonically around the northwest quadrant of the low in central and southern Illinois.
WSR-88D base reflectivity composite (dBZ) at 0535 UTC 15 Feb 2010. Mean sea level pressure contours (hPa) are in yellow. The red dot in southwest Indiana corresponds to the location of the MISS rawinsonde launched at this time, and the white line corresponds to the C-130 flight leg flown repeatedly from 0352 to 1100 UTC 15 Feb.
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
The NCAR MISS was deployed northeast of Evansville, IN (red dot in Fig. 1), where rawinsondes were launched every 2 h from 2130 UTC 14 February to 1530 UTC 15 February. Multiple passes of the NCAR/NSF C-130 aircraft were flown over southern Indiana from west-southwest to east-northeast from 0352 to 1100 UTC 15 February along a track indicated by the white line in Fig. 1. These flight legs were flown back and forth at varying altitude over the ground operations, permitting analysis of microphysical structure directly within the GCs at cloud top (Plummer et al. 2014) and within the emanating fallstreaks (Plummer et al. 2015). WCR time–height cross sections of the broad precipitation area over southern Indiana (Fig. 2) show structures within the storm. These include a deep, precipitating nimbostratus cloud with cloud-top GCs, marked by areas of locally enhanced reflectivity and vertical radial velocity at the top of the cloud. High-resolution views of equivalent reflectivity factor and vertical radial velocity from the WCR are shown for a 10-km-long subset of a representative flight leg in Fig. 3. The GCs were approximately 1.0 km wide and 1.5–2.0 km deep, with echo tops extending to the tropopause (~7.7 km). The strongest updrafts (>2.00 m s−1) were typically located above 6.5 km, while the strongest downdrafts (magnitude > 2.00 m s−1) typically occurred between 6.0 and 7.0 km.
(a) WCR equivalent reflectivity factor (dBZe) for the 0652–0718 UTC C-130 flight leg. The black box indicates the location for which data are shown in more detail in Fig. 3. (b) WCR vertical radial velocity for the same time period as in (a). The C-130 flight track is shown in gray. The thin line of high reflectivity and vertical radial velocity at 5.8 km is the result of ground reflection.
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
(a) Unthresholded WCR equivalent reflectivity (dBZe) between 5.0- and 8.5-km altitude for 0709:20–0710:45 UTC. The location of these data is shown as the black box in Fig. 2a. (b) Unthresholded WCR vertical radial velocity (m s−1) for the same time period and altitude range. The thin line of high reflectivity and vertical radial velocity at 5.8 km is the result of ground reflection.
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
4. Simulation methodology
Idealized numerical simulations were designed to assess the role of cloud-top radiative forcing on the dynamics of generating cells under stability and vertical wind shear conditions similar to those observed for the 14–15 February 2010 cyclone, so that comparisons to observed ice mass, kinematics, and structure obtained with the NCAR/NSF C-130 instrument platform could be made. For reasons discussed below, the initial conditions for the idealized simulations were obtained from a storm-scale simulation of the cyclone.
a. Storm-scale simulation
A WRF (version 3.3.1) simulation of the 14–15 February 2010 cyclone was performed to provide input data for the idealized generating cell simulations. The storm-scale simulation consisted of three domains nested at a 3:1 ratio, with 27-, 9-, and 3-km grid spacing and approximate dimensions of 4000 × 3500 km2, 1900 × 1800 km2, and 900 × 850 km2 (Fig. 4). The horizontal grid spacing for the storm-scale simulation is greater than the scale of the GCs, since the purpose of this simulation was to obtain a representation of the environment in which they developed, not to simulate them explicitly. A total of 178 vertical levels were spaced at uneven intervals, with model-level spacing starting at 150 m near the surface, decreasing gradually to 50 m by 4.0 km, and remaining at 50 m through the lower stratosphere (~8.5 km). Above this level, the spacing gradually increased to 250 m near the domain top of 100 hPa, with smooth transitions in level spacing, similar to the level spacing employed by Heymsfield et al. (2011) to simulate a shallow upper-tropospheric cloud layer. While most storm-scale simulations have fewer vertical levels (e.g., Molthan and Colle 2012), finer spacing was necessary here to reproduce the strong gradient in moisture that is typically present at cloud top in winter cyclones (Grim et al. 2007). Tests with coarser vertical spacing overly smoothed the profile of θei, decreasing the magnitude of potential instability that existed at the higher vertical resolution. A wave-damping layer in the top 5.0 km of the domain prevented wave reflection off the domain top using the Rayleigh-w damping method of Klemp et al. (2008). (See Table 1 for definitions of variables discussed regularly in this paper.)
Nested WRF grids for the simulation of the 14–15 Feb 2010 cyclone. Horizontal grid spacing for the outer, middle, and inner yellow boxes shown on the map are 27, 9, and 3 km, respectively. Source: WRF Domain Wizard, Lambert Conformal Conic projection.
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
Definition of variables discussed regularly in this paper. A horizontal bar over a given variable is used to denote a simulation domain-averaged field.
Rapid Update Cycle (RUC) hourly analyses provided the initial and lateral boundary conditions for the storm-scale simulation. The model was initialized at 0000 UTC 15 February 2010 and was integrated for 8 h with time steps of 20, 6.67, and 2.22 s for the three domains. Prior to carrying out the idealized simulations, experiments were conducted for the storm-scale simulation using a range of microphysical parameterizations available in WRF and the results were compared with the available data from the field campaign. The Thompson microphysical scheme (Thompson et al. 2008) produced results most comparable to the PLOWS microphysical data (Plummer et al. 2014). The Thompson microphysics scheme is well suited for this study since it permits supersaturation with respect to ice without automatic water vapor conversion to ice, unlike some other schemes. Automatic conversion is physically unrealistic, particularly when convective updrafts are present, as in the case of generating cells. Ice nucleation in the Thompson scheme is parameterized following a modified version of Cooper (1986) that permits nucleation after RHice exceeds 125% for heterogeneous nucleation of ice for T < 260 K. Results in a sensitivity study by Cintineo et al. (2014) further support the choice of microphysics scheme since, of the schemes they assessed, Thompson showed the least bias compared to observations of upper-level ice clouds. Thompson microphysics and Rapid Radiative Transfer Model for Global Climate Models (RRTMG; Iacono et al. 2000) radiation parameterizations were used for all domains. Ice crystal size is set by the RRTMG radiation scheme based on temperature within the cloud. The detailed comparisons we present in section 6 of predicted (in the idealized simulations) versus observed ice mass in the generating cells (Plummer et al. 2014) gave us further confidence in the use of the Thompson and RRTMG parameterizations.
The simulated cyclone’s horizontal and vertical cloud structure fully developed prior to the 0652–0718 UTC C-130 flight leg, for which the WCR data are shown in Fig. 2. A very tight gradient in relative humidity with respect to ice (RHice) was present at 7.0 km (Figs. 5c,d) in the simulation at 0705 UTC, with RHice less than 25% present in the western 50 km of the cross section between 4.0 and 6.0 km. This dry air undercut very moist air with RHice exceeding 130% just under the tropopause on the left side of the cross section (Fig. 5d), characteristic of an upper-level front (Reed 1955). The horizontal structure of the 2.0-km simulated reflectivity (Fig. 5a) shows the same general features in the observed radar reflectivity (Fig. 1), including the two bands over central and southern Illinois and the broader band of precipitation over central and southern Indiana. A cross section through the band in southern Indiana (Fig. 5b) shows shallow dry-slot precipitation to the west and deep stratiform precipitation to the east.
(a) The 2.0-km radar reflectivity (dBZ) and 10-m wind barbs (knots; 1 kt = 0.51 m s−1) from the WRF simulation at 0705 UTC, the center time of the flight leg shown in Fig. 2. The black line in southern Indiana is the location of the cross section shown in (b) and (d). The white dot corresponds to the location of the vertical profile of RH (liquid and ice) and θei shown in Figs. 6 and 8, respectively. The black box corresponds to the location over which the u and υ components of wind were averaged for Fig. 7. (b) Cross section of radar reflectivity from the WRF simulation at 0705 UTC with contours of θ shown every 1 K. (c) Relative humidity with respect to ice at 7.0 km from the WRF simulation at 0705 UTC. (d) Cross section [location shown in (a) and (c)] of RHice from the WRF simulation at 0705 UTC with contours of θ shown every 1 K.
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
b. Idealized generating cell simulations
The input fields for the idealized GC simulations consisted of a vertical profile of potential temperature θ, water vapor mixing ratio qυ, and the u and υ wind components from the storm-scale simulation at the mean time of the C-130 flight leg, 0705 UTC. The profiles of θ and qυ were taken from the center point of the C-130 flight leg, indicated by the dots in Figs. 5a and 5c and vertical bars in Figs. 5b and 5d. The vertical profiles of observed (MISS sounding, 0535 UTC, red dot in Fig. 1) and simulated (0705 UTC, white dot in Figs. 5a and 5c) relative humidity with respect to liquid and ice are in general agreement, except for a difference of about 10% in RH and 20% in RHice at the GC level (Fig. 6). The idealized simulations that follow show that RHice varies substantially in the horizontal at the GC level within the convective updrafts and downdrafts of the GCs. A single rawinsonde has the potential to pass through updrafts (high-RHice regions) or downdrafts (low-RHice regions). There is no way to assess whether a sonde passed within or between GCs. Using a single sonde therefore can introduce a possible humidity bias in the upper troposphere. To avoid this, the storm-scale simulation was used for the input fields instead of the observed sounding. The horizontal grid spacing of 3.0 km in the storm-scale simulation is not sufficient to represent strong horizontal gradients in RHice at the GC level, ensuring that the RHice profile does not introduce the same possible humidity bias. The presence of an upper-tropospheric RHice peak in the storm-scale simulation is consistent with soundings from PLOWS-observed cyclones, as well as aircraft spirals through a winter cyclone in Ontario, Canada, during the Canadian CloudSat/Cloud–Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) Validation Project (C3VP) [Molthan and Colle (2012), their Fig. 8a], with RHice values over 140% in a layer just under the tropopause. Based on these data, the input sounding herein has reasonable and representative moisture and thermal characteristics.
Vertical profiles of relative humidity with respect to liquid water (gray: observed; black: simulated) and ice (light blue: observed; dark blue: simulated). The observed profiles are from the MISS rawinsonde launched at 0535 UTC 15 Feb 2010 from the location indicated by the red dot in Fig. 1. The simulated profiles are from the location indicated by the white dot in Figs. 5a and 5c.
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
Wind components shown in Fig. 7 are the mean profiles for the area indicated by the black square in Figs. 5a and 5c. It was necessary to average over this area rather than using a point sounding because of strong vertical wind shear present in a narrow corridor ahead of the upper-level front in western Indiana. This wind profile is representative of the area in which GCs were observed by the WCR on 15 February 2010. The frontal slope was sufficiently large that horizontal averages of θ and qυ would increase θei near 7 km and decrease θei near 6.0 km, thus reducing the cloud-top potential instability. For this reason, a single location profile of θ and qυ was used. The mid-to-upper-tropospheric profile of θei for the input data (Fig. 8) consists of two layers of potential instability. Maxima in CAPE with respect to ice-saturated ascent are 20.2 and 9.5 J kg−1 when air is lifted from 4.9 and 5.9 km, respectively.
Vertical profiles of the u wind component (gray: observed; black: simulated) and υ wind component (light blue: observed; dark blue: simulated). The observed profiles are from the MISS rawinsonde launched at 0535 UTC 15 Feb 2010 from the location indicated by the red dot in Fig. 1. The simulated profiles are average values for the location indicated by the black box in Figs. 5a and 5c.
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
Profile of θei from the WRF simulation for the location indicated by the white dot in Figs. 5a and 5c. Maxima in CAPE with respect to ice are indicated at 4.9 and 5.9 km.
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
The idealized WRF (version 3.3.1) simulations were performed on a single 50.1 × 50.1 × 15.0 km3 grid with double-periodic boundary conditions. This sufficiently large domain size was required so as not to enforce an artificial spatial scale on the GCs. The relatively high wavenumber evident in both horizontal directions associated with the GCs (shown later in Fig. 21) is evidence that the domain size was adequate for these simulations. Vertical levels were set to the same model coordinates as in the storm-scale simulation, with an additional 20 levels added in the lower stratosphere to smooth the transition from fine to coarser spacing, and thus reduce the chance of spurious wave development; this yielded a total of 199 vertical levels.
Because of the fine horizontal scale of generating cells (~0.75–1.5 km), fully capturing the spectrum of their vertical velocities and microphysical structure required horizontal grid spacing at a scale of 100 m. Skamarock (2004) evaluated WRF kinetic energy spectra versus that from theory and observations and recommended that departure of model spectra from expected values at high wavenumber be an indication of that model’s effective resolution. For WRF, he found this to be ~7∆x. With 100-m horizontal grid spacing, this suggests that the model is best for simulating phenomena at or greater than 700 m in width. Rosenow et al. (2014) found generating cells to be 0.75–1.5 km wide; on this basis, we believe the model resolution used in the idealized simulations to be adequate for all but the smallest generating cells. The agreement in vertical velocity between model GCs and observed GCs discussed in section 6 supports the use of 100-m grid spacing. Bryan et al. (2003) argue that a grid spacing of order 100 m is required for traditional large-eddy simulation (LES) closures to perform appropriately for their design. This was in the context of simulating deep convection. In their analysis of LESs, they discuss the relationship among three spatial scales: the “large” energy-containing eddies of scale l, the model’s filter scale (approximated with the grid spacing ∆), and the small dissipative eddies of scale η, related by 
A dynamical time step of 0.5 s ensured numerical stability at this grid spacing. As in the case study simulation, the Thompson microphysics and RRTMG radiation parameterizations were used. Radiation was called every 30 s since sensitivity tests showed essentially no change in simulated vertical velocity with more frequent radiation calls. Convection was simulated explicitly. A wave-damping layer was employed in the top 7.0 km of the grid as in Klemp et al. (2008). Surface fluxes of momentum, heat, and moisture were turned off in the model, and the surface type was set to snow.
Representation of the large-scale ascent dynamically consistent with the thermal and moisture profile at the initial time was necessary to maintain the deep nimbostratus cloud deck located below the GCs in the 14–15 February cyclone. Large-scale ascent and subgrid turbulence were parameterized using the WRF-LES module described in Yamaguchi and Feingold (2012). This module parameterizes the effect of large-scale vertical air motion 
Large-scale forcing (prescribed ascent) for WRF-LES. Ascent prescribed is the mean w 
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
5. Simulated generating cell structure and evolution
Three idealized simulations were carried out, all of which used the input fields described above. The first is fully representative of the 14–15 February 2010 cyclone, in that the simulation time, like the observations, was at night (0700–1000 UTC, 0200–0500 LT) when longwave radiative cooling is not offset by shortwave warming at cloud top. The second simulation was set in the afternoon (1800–2100 UTC, 1300–1600 LT), so that the effect of shortwave radiation could be assessed. The final simulation was conducted with radiation turned off in the model to test the hypothesis that radiative cooling is necessary to maintain cloud-top instability and, thus, the presence of GCs in this cyclone environment. Integration for 180 min was sufficient to determine whether GCs would persist. The 180-min time limit also avoided continuing the simulations through either dawn or dusk, when radiative forcing would no longer be consistent with the experimental design. Last, the background parameterized large-scale vertical velocity eventually led to sufficient compression of the potential temperature gradient at the tropopause after approximately 180 min such that nonrealistic gravity waves began to form in the no-radiation simulation, where GCs were not present to mix out the shear.
a. Bulk properties of simulated generating cells
Statistical analyses of the nighttime, daytime, and no-radiation simulations were performed to compare the general evolution of cloud-top convection and the corresponding diabatic heating rates. Two types of analyses are presented: contoured frequency by time diagrams (CFTDs) of w (Fig. 10) and Hovmöller-like domain-averaged time–height displays of the net (
CFTD analysis of the vertical air velocity data at the GC level (6.0–8.0 km) every 5 min with 0.10 m s−1 bin width for the (a) nighttime, (b) daytime, and (c) no-radiation simulations. Contours show the 99th (top contour), 95th, 90th, 75th, 50th, 25th, 10th, 5th, and 1st (bottom contour) percentiles of vertical air velocity data.
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
1) Evolution of GC vertical velocities
In all simulations the w spectrum at 6.0–8.0 km (Fig. 10) visibly diverges from 0 m s−1 at t = 15 min. The nighttime-simulation w spectrum range (as measured by the difference between the 99th and 1st percentiles of w; Fig. 10a) broadens rapidly through t = 50 min owing to release of potential instability (Fig. 8), with 99th and 1st percentiles of w equal to 1.70 and −1.55 m s−1, respectively. The w spectrum range then decreases by approximately half by t = 90–95 min with 99th and 1st percentiles of w equal to 0.85 and −0.80 m s−1. This is followed by a second increase in convection with the maximum 99th percentile of w reaching 1.80 m s−1 at t = 135 min and the minimum 1st percentile of w reaching −1.60 m s−1 at t = 140 min. A slight decrease in the w spectrum range occurs through the end of the simulation, with 99th and 1st percentiles of w equal to 1.40 and −1.35 m s−1 at t = 180 min.
As in the nighttime simulation, the daytime simulation w spectrum range broadens beginning at t = 15 min (Fig. 10b), reaching a maximum extent at t = 50 min, with 99th and 1st percentiles of w equal to 1.75 and −1.60 m s−1. After that time, the w spectrum range decreases more rapidly than during the nighttime simulation, with 99th and 1st percentiles of w equal to 0.75 and −0.75 m s−1 at t = 90 min. The w spectrum range continues to decrease until it reaches approximately one-third of its maximum extent at t = 115 min, with 99th and 1st percentiles of w equal to 0.60 and −0.50 m s−1. Convection then increases from t = 115 to 155 min. After t = 155 min the 99th percentile of w was steady at 1.15 m s−1, while the 1st percentile of w decreased to −1.05 m s−1 by t = 180 min.
The w spectrum is notably different in the no-radiation simulation (Fig. 10c). The 1st percentile of w reaches a minimum of −1.35 m s−1 at t = 45 min, followed by the maximum 99th percentile of w of 1.35 m s−1 at t = 50 min. For the majority of the remainder of the simulation, the w spectrum range narrows, with a 99th and 1st percentile of w equal to 0.20 m s−1 and −0.20 m s−1, respectively, at t = 155 min. The slight increase in the w spectrum range in the last 15 min of the simulation corresponds to development of waves on the tropopause that do not produce precipitation. Similar waves do not develop in the other simulations, since turbulence associated with the GCs reduces wind shear in the vicinity of the tropopause and prevents shear instability from developing.
2) Evolution of longwave cooling
Early in the nighttime and daytime simulations, a deep layer of domain-averaged longwave cooling 
Time evolution of the mean longwave heating rate (K h−1) for (a) nighttime and (b) daytime simulations. The mean value is calculated for the entire horizontal model domain for data interpolated to 50-m height intervals. As in Fig. 10, contours show the 99th (top contour), 95th, 90th, 75th, 50th, 25th, 10th, 5th, and 1st (bottom contour) percentiles of vertical air velocity data from the simulation corresponding to individual panels.
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
3) Evolution of shortwave heating
The domain-averaged shortwave heating 
Time evolution of the mean shortwave heating rate (K h−1) for the daytime simulation. The mean value is calculated for the entire horizontal model domain for data interpolated to 50-m height intervals. Contours show the 99th (top contour), 95th, 90th, 75th, 50th, 25th, 10th, 5th, and 1st (bottom contour) percentiles of vertical air velocity data from the daytime simulation.
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
4) Evolution of latent heating
The evolution of the domain-averaged latent heating 
Time evolution of the latent heating rate (K h−1) for (a) nighttime, (b) daytime, and (c) no-radiation simulations. The mean value is calculated for the entire horizontal model domain for data interpolated to 50-m height intervals. As in Fig. 10, contours show the 99th (top contour), 95th, 90th, 75th, 50th, 25th, 10th, 5th, and 1st (bottom contour) percentiles of vertical air velocity data from the simulation corresponding to individual panels.
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
5) Net diabatic heating
The nighttime simulation net diabatic heating rate is the sum of longwave cooling and latent heating (Fig. 14a). In general, cooling was confined to higher altitudes and heating lower altitudes, together leading to destabilization for the duration of the simulation. Within the higher altitude layer of cooling, three distinct maxima in cooling occurred at t = 35, 90, and 175–180 min. These maxima were interspersed by weaker net cooling or weak net warming corresponding to the maxima in latent heating due to stronger updrafts at t ~ 50 and 135 min. Low levels were dominated by latent heating, with only a slight offset by longwave cooling.
Time evolution of the mean net diabatic heating rate (K h−1) for (a) nighttime (longwave radiation and latent heat) and (b) daytime (longwave and shortwave radiation and latent heat) simulations. The mean value is calculated for the entire horizontal model domain for data interpolated to 50-m height intervals. As in Fig. 10, contours show the 99th (top contour), 95th, 90th, 75th, 50th, 25th, 10th, 5th, and 1st (bottom contour) percentiles of vertical air velocity data from the simulation corresponding to individual panels.
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
In the daytime simulation, the inclusion of shortwave warming notably reduced the cooling near cloud top (Fig. 14b) and, thus, the rate of destabilization. However, weak domain-averaged destabilization did occur for the duration of the simulation. Domain-averaged cooling maxima were only −0.10–0.00 K h−1 at t = 30–40, 80–145, and 170–180 min, with weak net warming between. The collocation of shortwave warming and latent heating maxima at t = 50 min explains the large increase in net heating at ~6.2–7.7 km relative to the nighttime simulation. A reduction in latent heating due to weaker convection late in the simulation and a deeper layer of longwave cooling due to increased spacing of GCs reduced the extent net warming > 0.10 K h−1 from t ~ 120 to 150 min. While shortwave heating of cloud top partially offset destabilization due to longwave cooling, the magnitude of shortwave heating was lower than that for longwave cooling, and continuous weak destabilization occurred in the daytime simulation.
b. Development and evolution of convection
The evolution of GCs on small scales for the nighttime, daytime, and no-radiation simulations is presented in this section. Figures 15 and 16 show 10-km-long cross sections along the shear vector (west-southwest to east-northeast) at the level of the GCs at 45, 90, and 135 min into the simulation. This orientation was chosen so that precipitation fallstreaks emanating from GCs were more likely to remain in the cross section. As was shown in section 5a, these times correspond to maxima (t = 45 and 135 min) and minima (t = 90 min) in the vertical air velocity spectrum range at the GC level and, thus, illustrate distinct time periods in the evolution of the simulations.
(a)–(c) Vertical air velocity (m s−1), (d)–(f) relative humidity with respect to ice (%), and (g)–(i) ice precipitation mixing ratio (g kg−1) for (left)–(right) t = 45, 90, and 135 min for the nighttime radiation simulation. All cross sections are along the wind shear vector.
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
(a)–(c) Longwave and (d)–(e) latent heating rates (K h−1) for (left)–(right) t = 45, 90, and 135 min for the nighttime radiation simulation.
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
1) Nighttime simulation
The evolution of vertical air velocity w, RHice, and ice precipitation mixing ratio qi is shown in the top, middle, and bottom rows of Fig. 15, respectively. Fields are shown between 5.0- and 8.5-km altitude to illustrate their evolution in the vicinity of GCs. The ice precipitation mixing ratio is the sum of the cloud-ice and snow-mixing ratios. By t = 45 min, updrafts and downdrafts with magnitudes in excess of 1.00 m s−1 (~95th percentile of w; Fig. 10a) were ubiquitous between 6.5 and 7.5 km, with extreme w values in excess of ±2.00 m s−1. RHice generally increased with height in the vicinity of updrafts, with maxima frequently greater than 150% between ~6.75 and 7.5 km. Maxima in qi in excess of 0.09 g kg−1 were common in areas of high ice supersaturation, with some precipitation cores in excess of 0.12 g kg−1. The impact of the GC updrafts was to increase qi at higher altitudes by triggering the formation of additional precipitation at these altitudes. This was also obvious in Hovmöller-like domain-averaged time–height displays of qi when compared to the daytime and no-radiation simulations (not shown). Diabatic heating in the vicinity of GCs (Fig. 16) consists of longwave cooling QL in excess of 0.25 K h−1 at ~6.0–7.5 km, with maxima of ~0.75–1.00 K h−1 near GC tops at 6.7–7.3 km. Latent heating rates QH of 0.75–4.00 K h−1 were present in areas of ice supersaturation due to ice deposition, with cooling of 0.25–0.50 K h−1 within local downdrafts where lower RHice was present and ice sublimation occurred.
At t = 90 min, when the w spectrum range had decreased notably (Fig. 10a), two layers of RHice, qi, and w maxima were centered near 6.0 and 7.5 km (Figs. 15e,h,b), with substantially decreased coverage of strong vertical circulations and reduced magnitudes of w and RHice. The value of QH also decreased substantially at that time (<2.50 K h−1 at the lower level and <1.00 K h−1 at the upper level). The rates of QL had similar coverage and magnitude at t = 45 and 90 min (Figs. 16a,b).
The GCs reinvigorated substantially by t = 135 min, as evidenced by w maxima in excess of 2.00 m s−1 (~99th percentile of w in Fig. 10a) and in extreme cases 3.00 m s−1 (Fig. 15c). These stronger updrafts supported higher values of RHice and corresponding qi maxima. In Figs. 15c, 15f, and 15i, w maxima of 3.00–3.50 m s−1 correspond to RHice and qi in excess of 170% and 0.24 g kg−1, respectively. Diabatic heating and cooling rates were much stronger at this time with longwave cooling up to 3.50 K h−1 near cloud top (~7.5–8.0 km; Fig. 16c) and latent heating of up to 4.00 K h−1 in ice-supersaturated locations (Fig. 16f). Latent cooling had also increased by this time, with maximum cooling of 0.50–0.75 K h−1 between GCs. Fallstreaks with qi > 0.12 g kg−1 were easily identified emanating from the GCs (Fig. 15i).
2) Daytime simulation
Fields of w, RHice, and qi are shown t = 45, 90, and 160 min in Fig. 17 along the same cross section shown in Fig. 15. At t = 45 min, there was little difference between the fields in the nighttime and daytime simulation (Figs. 17a,d,g). Accordingly, QL and QH also showed little difference (Figs. 18a,g). Absorption of solar radiation near cloud top corresponded to shortwave heating rates QS of 0.25–0.50 K h−1 between 6.0 and 7.5 km, with maxima in excess of 0.50 K h−1 near 7.0 km (Fig. 18d). Typical QS magnitudes at cloud top were weaker than QL, which favored persistent destabilization of the cloud top and maintenance of GCs.
(a)–(c) Vertical air velocity (m s−1), (d)–(f) relative humidity with respect to ice (%), and (g)–(i) ice precipitation mixing ratio (g kg−1) for (left)–(right) t = 45, 90, and 160 min for the daytime radiation simulation.
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
(a)–(c) Longwave, (d)–(f) shortwave, and (g)–(i) latent heating rates (K h−1) for (left)–(right) t = 45, 90, and 160 min for the daytime radiation simulation.
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
Convection again weakened substantially by t = 90 min, as in the nighttime simulation. Updrafts and downdrafts were less pronounced, with small (~100–200 m diameter) extrema ranging from −1.50 to 1.00 m s−1 embedded within weaker ascent/descent (Fig. 17b). Two layers of RHice, qi, and QH maxima near 6.0 and 7.5 km were also present at t = 90 min for the daytime simulation. While the upper layer had similar characteristics to the nighttime simulation, the lower layer was much less pronounced with RHice, qi, and QH maxima near 115%, 0.12 g kg−1, and 0.75–1.00 K h−1, respectively (Figs. 17e,h; Fig. 18h). Longwave cooling was slightly weaker than the nighttime simulation, with broad cooling in excess of 0.25 K h−1 from 6.0 to 7.5 km and maxima near cloud top from 0.75 to 1.00 K h−1 (Fig. 18b). Shortwave warming also decreased, with less coverage of QS in excess of 0.25 K h−1 and a decrease in maxima to near 0.50 K h−1 (Fig. 18e).
Unlike the nighttime simulation, convection had yet to reinvigorate by t = 135 min in the daytime simulation (Fig. 10b). Instead, the vertical velocity spectrum range increased to a maximum near t = 160 min, which persisted through the end of the simulation at t = 180 min. Thus, data from t = 160 min are shown in the third column of Figs. 17 and 18. GCs had peak w of 1.00–1.50 m s−1 at that time (Fig. 17c). RHice maxima in the GCs were near 160% (Fig. 17f), slightly lower than with the stronger convection present in the nighttime simulation. Accordingly, it is not surprising that qi values were also lower, with typical maxima near 0.18 g kg−1 (Fig. 17i). The latent heating rates had maxima and minima of only 1.50 and −0.25 K h−1, respectively (Fig. 18i). While the maximum values of longwave cooling were only slightly lower in the daytime simulation (1.50 K h−1; Fig. 18c), the areal coverage of the highest values was much less relative to the nighttime simulation (Fig. 16c). Shortwave warming rates were similar to those earlier in the simulation, with maxima in excess of 0.50 K h−1 near cloud top (Fig. 18f). Despite weaker maxima in qi relative to the nighttime simulation, fallstreaks were still evident emerging from the bases of the GCs (Fig. 17i).
3) No-radiation simulation
At t = 45 min in the no-radiation simulation, the simulated w, RHice, qi, and QH fields (Figs. 19 and 20) were very similar to that of the nighttime and daytime simulations since potential instability was present in the initial conditions of all simulations, regardless of the inclusion or exclusion of radiative forcing. Note that Fig. 20 includes only the latent heating rate, since radiative forcing was not included in this simulation.
As in Fig. 15, but for the no-radiation simulation.
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
Latent heating rate (K h−1) for t = (a) 45, (b) 90, and (c) 135 min for the no-radiation simulation.
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
Without the destabilizing influence of radiative forcing, convection was drastically weaker by t = 90 min, with |w| maxima of only ~0.50 m s−1 near 7.5 km (Fig. 19b). Two layers of RHice maxima were present at this time, albeit much less pronounced compared to the other simulations, with isolated RHice maxima of 145% in the upper layer (~7.5 km; Fig. 19f). RHice maxima in the lower layer (~6.0 km) were near 125%. The precipitation content of GCs was only slightly reduced, with typical qi maxima of 0.06–0.09 g kg−1 and isolated maxima near 0.12 g kg−1 in the upper layer (Fig. 19h). Precipitation content was higher in the lower level, with typical and isolated qi maxima of 0.12 and 0.15 g kg−1, respectively. Latent heating in the upper layer was notably weaker than for the other simulations, with maxima near 0.50 K h−1 and minima from 0.00 to −0.25 K h−1 (Fig. 20b).
The importance of radiative forcing in the maintenance of GCs under initially unstable conditions is clearly indicated by the near lack of variation in the w field by t = 135 min, relative to the other simulations, with |w| well under 0.50 m s−1 (Fig. 19c). The RHice field is characterized by a 100-m layer of 125% at 8.0–8.1 km and isolated maxima of the same magnitude near 7.5 km (Fig. 19f). There were no closed contours in the qi cross section; however, remnant fallstreaks with qi ~ 0.12 g kg−1 were present below 6.5 km (Fig. 19i). Latent cooling was not present in the cross section (Fig. 20c); maximum QH below 6.5 km had values typical of fallstreaks in the other simulations (0.50–0.75 K h−1).
6. Comparison of simulated generating cells to PLOWS observations
In this section, the kinematic and bulk microphysical properties of the nighttime simulation of the GCs are compared to PLOWS observations, since the observations were made from 0515 to 0819 UTC (0015–0319 LT) 15 February 2010.
a. Vertical velocity spectrum of generating cells
Statistical analysis of WCR vertical radial velocity W from Rosenow et al. (2014) is compared to the vertical air velocity spectrum w at t = 135 min (when GCs had reinvigorated; see Fig. 10a) in the nighttime simulation in Table 2. The value of W is the sum of w and the reflectivity weighted terminal velocity of hydrometeors in the pulse volume υt. The mean 50th percentile of W for the 6.0–8.0-km layer in Rosenow et al. is −0.29 m s−1, while the 50th percentile of w for the nighttime simulation at t = 135 min is −0.01 m s−1 (Fig. 16a). This difference of 0.28 m s−1 falls within the range of expected υt velocities (Rosenow et al. 2014, their Fig. 2) corresponding to the median mass diameters observed within GCs in this cyclone (0.66–0.78 mm; see Plummer et al. 2014). When the Rosenow et al. W percentile values shown in Table 2 are adjusted by 0.28 m s−1 to account for the plausible 0.28 m s−1 fallout of the hydrometeors, there is very strong agreement with the simulated nighttime w spectrum, with a Pearson’s linear correlation coefficient of 0.9994. The GC vertical velocity is also within the range observed by many others including Wexler (1955), Douglas et al. (1957), Wexler and Atlas (1959), Carbone and Bohne (1975), Rauber et al. (2014a,b), and Kumjian et al. (2014).
Comparison of the 6.0–8.0-km vertical air velocity data from the nighttime simulation at t = 135 min to vertical radial velocity data from the 15 Feb 2010 cyclone (Rosenow et al. 2014). (right column) The difference between Rosenow adjusted and the nighttime simulation.
b. Precipitation mass
The simulated bulk cloud microphysical properties are compared with in situ observations from 14 to 15 February 2010 derived by Plummer et al. (2014) in Table 3. The ice water content (IWC) derived by Plummer et al. at 6.7 and 7.3 km has been converted to ice precipitation mixing ratio qi using the air density at the appropriate level in nighttime idealized simulation. The bottom row of Table 3 includes corresponding qi percentiles from 6.7 to 7.7 km during the t = 135–150-min time period in the nighttime simulation. These data from the simulation are from a slightly higher altitude than those measured in situ since the simulated GCs were at that level. There is very close agreement between the 6.7–7.7-km simulated qi and qi observed at 6.7 km, with the 5th-, 25th-, 50th-, and 75th-percentile values within 0.01 g kg−1 of each other. The difference between 6.7–7.7-km simulated qi and 6.7-km observed IWC increases slightly to 0.02 g kg−1 at the 95th percentile. The simulated qi is lower than the observed 7.3-km IWC, particularly at the 75th and 95th percentiles, where qi is 0.07 and 0.17 g kg−1 lower than observed, respectively. At the 5th, 25th, and 50th percentiles, the simulation is within 0.02 g kg−1 of the 7.3-km observations. It is worth noting that while the 95th percentile of qi was lower than that observed by Plummer et al. (2014) at 7.3 km (0.18 and 0.35 g kg−1, respectively), isolated instances of ice precipitation mixing ratio of 0.27–0.33 g kg−1 did occur in the nighttime simulation (Figs. 15i and 21a–d).
(top),(middle) Percentiles of ice precipitation mixing ratio (g kg−1) at 6.7 and 7.3 km as derived using microphysical probes (Plummer et al. 2014). (bottom) Mean percentiles of ice precipitation mixing ratio for t = 135–150 min in the 6.7–7.7-km layer for the nighttime simulation.
Differences between observed and simulated precipitation content are consistent with the differences between observed and simulated vertical air velocity, in that the observations agree very strongly except at high percentiles. As shown in Fig. 5, an upper-level front was present at the western edge of GCs in the 14–15 February 2010 cyclone. It is probable that this frontal feature enhanced upward vertical air motion in the western portion of the flight legs and locally enhanced ice crystal growth. Given the horizontal homogeneity in the initial conditions of the idealized simulations (aside from the very small thermal and moisture perturbations at 6.45–6.55 km), the local enhancement in vertical motion and precipitation due to the frontal feature is not represented in the idealized WRF simulations.
7. Discussion
The 6.0–8.0-km vertical air velocity spectrum range (Fig. 10) broadens early in all three idealized simulations. This increase in the vertical air velocity spectrum range primarily represents a release of the instability present in the model initial conditions (Fig. 8). However, there is some influence of radiative forcing since the simulations with radiation have a broader vertical air velocity spectrum at t = 45 min, with the 99th-percentile maxima of 1.70, 1.75, and 1.35 m s−1 in the nighttime, daytime, and no-radiation simulations. The model solutions diverge substantially following the initial release of instability. In the nighttime simulation, the reduction in the vertical velocity spectrum range is less pronounced and shorter lived than in the daytime simulation, with peaks occurring at t = 50 min and again at t = 135–140 min, compared to t = 50 and 180 min in the daytime simulation. This indicates that when cloud-top longwave cooling is offset by shortwave warming there is both a delay in destabilization and a reduction in the magnitude of instability attained through radiative forcing. The values of 
Plan views of qi and QL at 7.5 and 6.5 km are shown for the nighttime and daytime simulations at t = 150 min in Fig. 21. Coverage of precipitation at 7.5 km is much greater in the nighttime simulation (Fig. 21a), which focuses much of the longwave cooling at 7.5 km (Fig. 21e). In the daytime simulation there is much less coverage of qi > 0.03 g kg−1 at 7.5 km (Fig. 21c). Areas at 6.5 km below gaps in the 7.5-km ice precipitation mixing ratio field exhibit greater longwave cooling than for the nighttime simulation (Fig. 21h). Thus, the larger spacing between GCs late in the daytime simulation prevents the longwave cooling from focusing near 7.6 km as in the nighttime simulation (Fig. 11), possibly indicating a negative feedback between instability and GC spacing. This relationship will be explored further by varying the cloud-top stability in simulations to be discussed in Keeler et al. (2016, hereafter Part II).
(top) Ice precipitation mixing ratio (g kg−1) for the full 50 × 50 km2 model grid at t = 150 min for (a) nighttime simulation at 7.5 km, (b) nighttime simulation at 6.5 km, (c) daytime simulation at 7.5 km, and (d) daytime simulation at 6.5 km. (bottom) Longwave heating (K h−1) for the full 50 × 50 km2 model grid at t = 150 min for (e) nighttime simulation at 7.5 km, (f) nighttime simulation at 6.5 km, (g) daytime simulation at 7.5 km, and (h) daytime simulation at 6.5 km.
Citation: Journal of the Atmospheric Sciences 73, 4; 10.1175/JAS-D-15-0126.1
The extent of the vertical velocity spectra late in these simulations represents what is sustainable given the radiative forcing present. The evolution of the no-radiation simulation clearly demonstrates that without radiative forcing, GCs are not sustainable for the shear and thermodynamic conditions present in the 14–15 February 2010 cyclone. The small increase in the vertical velocity spectrum range late in the no-radiation simulation (Fig. 10) is due to development of waves on the tropopause (not shown) that do not produce precipitation. Recall that Douglas et al. (1957) suggested that under initially stable conditions, GCs develop as a result of latent heat released during depositional growth of ice crystals. In the no-radiation simulation, GCs do not persist despite an initially unstable layer and persistent 
We note that the results presented herein may be sensitive to uncertainties in the treatment of ice microphysics and ice–radiation interactions that are calculated within the parameterizations selected for these model runs. We also note that one other mechanism, differential thermal advection, not explored in these simulations, could lead to sufficient destabilization to continuously maintain GCs. For the cyclone under consideration, the storm-scale simulation was assessed to determine the potential impact of differential thermal advection on destabilization of the GC layer. Analysis of the storm-scale simulation showed that differential thermal advection had a stabilizing effect on the upper troposphere in the location where the idealized simulation thermal and moisture profiles were obtained. Weak warm-air advection of up to 0.7 K h−1 occurred at the same level as the θei minimum in that location, while neutral thermal advection occurred at the base of the GC layer. Farther east, where GCs were also observed in the 14–15 February 2010 cyclone, weak cold-air advection of approximately 0.5 K h−1 occurred throughout the GC layer, so there was no differential thermal advection to destabilize the layer. Although differential thermal advection did not contribute to destabilization in this cyclone, we cannot conclude that in all cyclones this is true. In general, this mechanism may also contribute to the maintenance of GCs.
8. Summary
Cloud-top precipitation generating cells (GCs) were simulated in an idealized environment representative of the comma-head region of the 14–15 February 2010 cyclone, which produced 10–25 cm of snow in southern Indiana. Wyoming Cloud Radar (WCR) observations of this cyclone (Rosenow et al. 2014; Plummer et al. 2014) showed the presence of 1.5–2.0-km-deep GCs atop deep nimbostratus cloud deck in which precipitation fallstreaks emanating from the GCs merged into bands of snowfall as they fell toward the surface. Consistency of these GC and fallstreak observations with those of other cyclones observed during the Profiling of Winter Storms campaign (PLOWS; Rauber et al. 2014a,b; Rosenow et al. 2014; Plummer et al. 2014) is strong evidence not only of the ubiquity of GCs in extratropical winter cyclones but for the critical role of GCs in the precipitation process therein. The favorable comparison of the bulk vertical velocity and precipitation mass in the nighttime radiation idealized simulation compared to PLOWS observations (Tables 2 and 3) provides evidence that the simulations can be used to assess the sensitivity of GCs to various parameters. In this paper, the roles of radiative forcing and latent heating on GC dynamics were examined, and the following conclusions were obtained:
Under the thermodynamic and shear conditions present in the 14–15 February 2010 cyclone, longwave radiative cooling at cloud top is critical to the maintenance of GCs. Latent heat release due to depositional growth of ice crystals was not sufficient to maintain convection at cloud top in the absence of radiative forcing.
Domain-averaged longwave cooling rates > 0.60 K h−1 centered at 7.6 km in the nighttime simulation (with cooling > 3.00 K h−1 atop some individual GCs) supported maintenance of GCs and a broad vertical velocity spectrum, with isolated updrafts near 3.00 m s−1 and a 99th percentile of w = 1.80 m s−1 in the 6.0–8.0-km layer.
Domain-averaged longwave cooling (~0.30 K h−1) was more diffuse late in the daytime simulation because of an increased distance between GCs or possibly a reduction in optical depth in individual generating cells relative to nighttime simulations. Shortwave warming, while weaker by comparison (0.10–0.20 K h−1 and maxima of 0.50 K h−1 atop GCs), was not sufficient to stabilize the cloud top to the point where GCs do not persist. That said, the daytime vertical velocity spectrum range is notably narrower late in the simulation, with a 99th percentile of w in the 6.0–8.0-km layer of 1.20 m s−1, compared to 1.80 m s−1 for the nighttime simulation.
When present, GCs are characterized by high ice supersaturation with RH ice > 150%, suggesting that rapid depositional growth of ice crystals should occur. Rapid depositional growth in GCs has been confirmed in the observations reported by Plummer et al. (2014). Ice precipitation mixing ratio maxima of >0.15 g kg−1 were common among GCs, particularly in the nighttime simulation when some maxima > 0.30 g kg−1 developed. Fallstreaks were clearly evident extending from the base of GCs into the underlying nimbostratus in both the simulations and observations.
This is the first paper of a three-part series investigating the mechanisms that control the development, maintenance, and organization of GCs atop winter cyclones. Subsequent papers will assess the influence of varying cloud-top stability and wind shear.
Acknowledgments
The authors thank the staff at the National Center for Atmospheric Research Environmental Observing Laboratory, particularly Alan Schanot, Jorgen Jensen, and the Research Aviation Facility staff for their efforts with the C-130 and the staff of the University of Wyoming King Air facility for their support of the WCR deployment. We thank Major Donald K. Carpenter and the U.S. Air Force Peoria National Guard for housing the C-130 during the project. Collaborations with National Center for Atmospheric Research coauthors were made possible through an NCAR Advanced Study Program Graduate Visitor Program fellowship received by the lead author. We thank David Plummer and Andrew Rosenow for providing the processed cloud probe and WCR data. Code for the WRF-LES module was provided by Takanobu Yamaguchi of NOAA/ESRL. The composite radar data shown in Fig. 1 were provided by the Iowa Environmental Mesonet, which is maintained by the Iowa State University Department of Agronomy. Rapid Update Cycle data for the storm-scale WRF simulation boundary and initial conditions were provided by the National Climatic Data Center’s NOMADS. Funding for this research was provided by NSF Grants ATM-0833828 and AGS-1247404 to the University of Illinois. All simulations were run on the Stampede supercomputer with support from XSEDE Grant TG-ATM050014N.
REFERENCES
American Meteorological Society, 2015: Generating cell. Glossary of Meteorology. [Available online at http://glossary.ametsoc.org/wiki/Generating_cell.]
Bowie, E. H., and R. H. Weightman, 1914: Types of storms of the United States and their average movements. Mon. Wea. Rev., 42 (Suppl. 1), 3–37.
Bryan, G. H., J. C. Wyngaard, and J. M. Fritsch, 2003: Resolution requirements for the simulation of deep moist convection. Mon. Wea. Rev., 131, 2394–2416, doi:10.1175/1520-0493(2003)131<2394:RRFTSO>2.0.CO;2.
Carbone, R. E., and A. R. Bohne, 1975: Cellular snow generation—A Doppler radar study. J. Atmos. Sci., 32, 1384–1394, doi:10.1175/1520-0469(1975)032<1384:CSGDRS>2.0.CO;2.
Cintineo, R., J. A. Otkin, M. Xue, and F. Kong, 2014: Evaluating the performance of planetary boundary layer and cloud microphysical parameterization schemes in convection-permitting ensemble forecasts using synthetic GOES-13 satellite observations. Mon. Wea. Rev., 142, 163–182, doi:10.1175/MWR-D-13-00143.1.
Cooper, W. A., 1986: Ice initiation in natural clouds. Precipitation Enhancement—A Scientific Challenge, Meteor. Monogr., No. 43, Amer. Meteor. Soc., 29–32.
Cunningham, J. G., and S. E. Yuter, 2014: Instability characteristics of radar-derived mesoscale organization modes within cool-season precipitation near Portland, Oregon. Mon. Wea. Rev., 142, 1738–1757, doi:10.1175/MWR-D-13-00133.1.
Douglas, R. H., K. L. S. Gunn, and J. S. Marshall, 1957: Pattern in the vertical of snow generation. J. Meteor., 14, 95–114, doi:10.1175/1520-0469(1957)014<0095:PITVOS>2.0.CO;2.
Evans, A. G., J. D. Locatelli, M. T. Stoelinga, and P. V. Hobbs, 2005: The IMPROVE-1 storm of 1–2 February 2001. Part II: Cloud structures and the growth of precipitation. J. Atmos. Sci., 62, 3456–3473, doi:10.1175/JAS3547.1.
Grim, J. A., R. M. Rauber, M. K. Ramamurthy, B. F. Jewett, and M. Han, 2007: High-resolution observations of the trowal–warm-frontal region of two continental winter cyclones. Mon. Wea. Rev., 135, 1629–1646, doi:10.1175/MWR3378.1.
Heymsfield, A. J., G. Thompson, H. Morrison, A. Bansemer, R. M. Rasmussen, P. Minnis, Z. Wang, and D. Zhang, 2011: Formation and spread of aircraft-induced holes in clouds. Science, 333, 77–81, doi:10.1126/science.1202851.
Hobbs, P. V., J. D. Locatelli, and J. E. Martin, 1990: Cold fronts aloft and the forecasting of precipitation and severe weather east of the Rocky Mountains. Wea. Forecasting, 5, 613–626, doi:10.1175/1520-0434(1990)005<0613:CFAATF>2.0.CO;2.
Iacono, M. J., E. J. Mlawer, S. A. Clough, and J.-J. Morcrette, 2000: Impact of an improved longwave radiation model, RRTM, on the energy budget and thermodynamic properties of the NCAR community climate mode, CCM3. J. Geophys. Res., 105, 14 873–14 890, doi:10.1029/2000JD900091.
Isard, S. A., J. R. Angel, and G. T. VanDyke, 2000: Zones of origin for Great Lakes cyclones in North America, 1899–1996. Mon. Wea. Rev., 128, 474–485, doi:10.1175/1520-0493(2000)128<0474:ZOOFGL>2.0.CO;2.
Keeler, J. M., B. F. Jewett, R. M. Rauber, G. M. McFarquhar, R. M. Rasmussen, L. Xue, C. Liu, and G. Thompson, 2016: Dynamics of cloud-top generating cells in winter cyclones. Part II: Radiative and instability forcing. J. Atmos. Sci., 73, 1529–1553, doi:10.1175/JAS-D-15-0127.1.
Klemp, J. B., J. Dudhia, and A. D. Hassiotis, 2008: An upper gravity-wave absorbing layer for NWP applications. Mon. Wea. Rev., 136, 3987–4004, doi:10.1175/2008MWR2596.1.
Kreitzberg, C. W., 1968: The mesoscale wind field in an occlusion. J. Appl. Meteor., 7, 53–67, doi:10.1175/1520-0450(1968)007<0053:TMWFIA>2.0.CO;2.
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, doi:10.1175/JAMC-D-13-0312.1.
Langleben, M. P., 1956: The plan pattern of snow echoes at the generating level. J. Meteor., 13, 554–560, doi:10.1175/1520-0469(1956)013<0554:TPPOSE>2.0.CO;2.
Marshall, J. S., 1953: Precipitation trajectories and patterns. J. Atmos. Sci., 10, 25–29, doi:10.1175/1520-0469(1953)010<0025:PTAP>2.0.CO;2.
Martin, J. E., 1998: The structure and evolution of a continental winter cyclone. Part II: Frontal forcing of an extreme snow event. Mon. Wea. Rev., 126, 329–348, doi:10.1175/1520-0493(1998)126<0329:TSAEOA>2.0.CO;2.
Molthan, A. L., and B. A. Colle, 2012: Comparisons of single- and double-moment microphysics schemes in the simulation of a synoptic-scale snowfall event. Mon. Wea. Rev., 140, 2982–3002, doi:10.1175/MWR-D-11-00292.1.
Naylor, J., and M. S. Gilmore, 2012: Convective initiation in an idealized cloud model using an updraft nudging technique. Mon. Wea. Rev., 140, 3699–3705, doi:10.1175/MWR-D-12-00163.1.
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, doi: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, doi:10.1175/JAS-D-14-0354.1.
Rauber, R. M., M. K. Macomber, D. M. Plummer, A. A. Rosenow, G. M. McFarquhar, B. F. Jewett, D. Leon, and J. M. Keeler, 2014a: 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, doi:10.1175/MWR-D-14-00057.1.
Rauber, R. M., and Coauthors, 2014b: Stability and charging characteristics of the comma head region of continental winter cyclones. J. Atmos. Sci., 71, 1559–1582, doi:10.1175/JAS-D-13-0253.1.
Rauber, R. M., and Coauthors, 2015: The role of cloud-top generating cells and boundary-layer circulations in the fine-scale radar structure of a winter cyclone over the Great Lakes. Mon. Wea. Rev., 143, 2291–2318, doi:10.1175/MWR-D-14-00350.1.
Reed, R., 1955: A study of a characteristic type of upper-level frontogenesis. J. Meteor., 12, 226–237, doi:10.1175/1520-0469(1955)012<0226:ASOACT>2.0.CO;2.
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, doi:10.1175/JAS-D-13-0249.1.
Skamarock, W. C., 2004: Evaluating mesoscale NWP models using kinetic energy spectra. Mon. Wea. Rev., 132, 3019–3032, doi:10.1175/MWR2830.1.
Syrett, W. J., B. A. Albrecht, and E. E. Clothiaux, 1995: Vertical cloud structure in a midlatitude cyclone from a 94-GHz radar. Mon. Wea. Rev., 123, 3393–3407, doi:10.1175/1520-0493(1995)123<3393:VCSIAM>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, doi:10.1175/2008MWR2387.1.
Wang, Z., and Coauthors, 2012: Single aircraft integration of remote sensing and in-situ sampling for the study of cloud microphysics and dynamics. Bull. Amer. Meteor. Soc., 93, 653–668, doi:10.1175/BAMS-D-11-00044.1.
Weisman, M. L., and J. B. Klemp, 1982: The dependence of numerically simulated convective storms on vertical wind shear and buoyancy. Mon. Wea. Rev., 110, 504–520, doi:10.1175/1520-0493(1982)110<0504:TDONSC>2.0.CO;2.
Wexler, R., 1955: Radar analysis of precipitation streamers observed 25 February 1954. J. Atmos. Sci., 12, 391–393, doi:10.1175/1520-0469(1955)012<0391:RAOPSO>2.0.CO;2.
Wexler, R., and D. Atlas, 1959: Precipitation generating cells. J. Atmos. Sci., 16, 327–332, doi:10.1175/1520-0469(1959)016<0327:PGC>2.0.CO;2.
Wood, R., 2012: Stratocumulus clouds. Mon. Wea. Rev., 140, 2373–2423, doi:10.1175/MWR-D-11-00121.1.
Yamaguchi, T., and G. Feingold, 2012: Technical note: Large-eddy simulation of cloudy boundary layer with the Advanced Research WRF model. J. Adv. Model. Earth Syst., 4, M09003, doi:10.1029/2012MS000164.
Yuter, S. E., and R. A. Houze Jr., 1995: Three-dimensional kinematic and microphysical evolution of Florida cumulonimbus. Part II: Frequency distributions of vertical velocity, reflectivity, and differential reflectivity. Mon. Wea. Rev., 123, 1941–1963, doi:10.1175/1520-0493(1995)123<1941:TDKAME>2.0.CO;2.