This paper describes the evolution of an intense precipitation band associated with a relatively weak warm front observed during the Global Precipitation Measurement (GPM) Mission Cold Season Precipitation Experiment (GCPEx) over southern Ontario, Canada, on 18 February 2012. The warm frontal precipitation band went through genesis, maturity, and decay over a 5–6-h period. The Weather Research and Forecasting (WRF) Model nested down to 1-km grid spacing was able to realistically predict the precipitation band evolution, albeit somewhat weaker and slightly farther south than observed. Band genesis began in an area of precipitation with embedded convection to the north of the warm front in a region of weak frontogenetical forcing at low levels and a weakly positive to slightly negative moist potential vorticity (MPV*) from 900 to 650 hPa. A midlevel dry intrusion helped reduce the midlevel stability, while the precipitation band intensified as the low-level frontogenesis intensified in a sloping layer with the warm front. Aggregates of unrimed snow occurred within the band during early maturity, while more supercooled water and graupel occurred as the upward motion increased because of the frontogenetical circulation. As the low-level cyclone moved east, the low-level deformation decreased and the column stabilized for vertical and slantwise ascent, and the warm frontal band weakened. A WRF experiment turning off latent heating resulted in limited precipitation band development and a weaker warm front, while turning off latent cooling only intensified the frontal precipitation band as additional midlevel instability compensated for the small decrease in frontogenetical forcing.
Warm fronts and their associated precipitation areas have been studied since the Norwegian School (Bjerknes 1919; Bjerknes and Solberg 1922), but they have generally received less attention than cold fronts. During the 1970s and 1980s, observational studies used radar, aircraft, and conventional surface and upper-air data to document some of the surface frontal wind, temperature, and precipitation structures. Houze et al. (1976) found a tendency for precipitation bands to be oriented approximately parallel to surface fronts and developed a six-category band classification scheme according to their location with respect to surface fronts. Several studies noted the presence of precipitation bands associated with a surface warm front (Browning et al. 1973; Houze et al. 1981; Austin and Houze 1972; Roach and Hardman 1975; Heymsfield 1979; Houze and Hobbs 1982). Warm frontal bands are primarily stratiform, develop either along or ahead of the surface warm front, and have been attributed to low-level warm advection forcing (Houze and Hobbs 1982).
The intensification of warm frontal rainbands has been attributed to the “seeder–feeder” mechanism (Houze and Hobbs 1982; Matejka et al. 1980; Houze et al. 1981; Rutledge and Hobbs 1983), in which ice particles nucleated in generating cells aloft grow by vapor deposition or accretion and aggregation through a feeder cloud as they fall. Generating cells are typically associated with potentially unstable layers, as noted by many other studies (e.g., Marshall 1953; Gunn et al. 1954; Houze et al. 1981; Browning 1983; McFarquhar et al. 2011; Cunningham and Yuter 2014; Rauber et al. 2014; Rosenow et al. 2014), and the seeder–feeder process has been documented in several other studies (Wexler et al. 1967; Herzegh and Hobbs 1980; Matejka et al. 1980; Houze et al. 1981; Browning 1983; Syrett et al. 1995; Trapp et al. 2001; Schultz et al. 2004; Stark et al. 2013; Cunningham and Yuter 2014). The Profiling of Winter Storms (PLOWS) field program focused on precipitation structures over the central United States. (Rauber et al. 2014). Plummer et al. (2015) showed using PLOWS data the ubiquitous presence of cloud-top generating cells and fall streaks in the stratiform region ahead of the surface warm front, which helps promote ice nucleation and growth in the upper part of the cloud.
Precipitation bands in the comma head have been shown to develop within a region of midlevel frontogenesis and weak moist symmetric stability (Emanuel 1985; Thorpe and Emanuel 1985; Xu 1989; Banacos 2003; Moore et al. 2005; Nicosia and Grumm 1999; Novak et al. 2009; Novak et al. 2010). The vertical motion and narrowing of the region of ascent with fronts has been attributed to latent heating through cloud condensation and other microphysical processes (Thorpe and Emanuel 1985; Chan and Cho 1991; Han et al. 2007). In addition, the ageostrophic circulation can be further modified by the positive (negative) potential vorticity anomalies that develop below (above) the region of maximum heating (Chan and Cho 1991). For example, Novak et al. (2009) showed that band formation and dissipation could occur in conjunction with changes in frontogenesis induced by diabatic PV anomalies resulting from latent heat release.
In addition to the seeder–feeder mechanism, other studies have linked smaller-scale precipitation enhancements along the front to convective cores and gravity waves. For example, Locatelli and Hobbs (1987) showed that the vertical structure of the warm front is not continuous; rather, the frontal zone can have a “staircase” pattern. Neiman et al. (1993) also saw this pattern and referred to the more vertical sections of the front as the “elevator” parts, while the more gradual sloping areas were the “escalator” sections. The heavier convective cores embedded within the warm frontal boundary were associated with the elevator parts of the front. Lindzen and Tung (1976) suggested that ducted gravity waves, with an unstable or neutral layer above or below a stable layer, can create vertical motions that may favor banded precipitation. Wakimoto and Bosart (2001) related precipitation variability along the warm frontal to Kelvin–Helmholtz waves.
The Global Precipitation Measurement (GPM) Mission Cold Season Precipitation Experiment (GCPEx; Skofronick-Jackson et al. 2015) provided in situ and aircraft observations of lake-effect and synoptic-scale snowfall events over southern Ontario, Canada, during January–February 2012 in order to calibrate satellite precipitation retrievals and validate bulk microphysical parameterizations. Most of the events studied within GCPEx involved unorganized precipitation structures, but the 18 February 2012 event provides an opportunity to investigate snowbands within the cyclone comma head. On 18 February 2012, there was an intense warm frontal precipitation band that crossed the GCPEx study domain (Fig. 1), which will be the focus of this paper.
Figure 2 shows the radar evolution of this band over the GCPEx study domain on 18 February 2012 that rapidly evolved from genesis, maturity, and decay from 0730 to 1334 UTC 18 February 2012. In this paper the band region is defined by the outline of the 24-dBZ threshold. During its peak intensity, the band stretched for at least a few hundred kilometers and was associated with snowfall rates of 2–4 cm h−1. Narrow and intense bands have been well documented for cold fronts (e.g., Houze and Hobbs 1982), but not for warm fronts. The conceptual model of a warm frontal band is fairly broad (~50 km wide), with a strong connection to generating cells aloft [cf. Fig. 5 in Houze and Hobbs (1982)]. Kenyon (2013) illustrated that narrow bands just north of the warm front can develop in a region of low-level frontogenesis parallel to the axis of dilatation. The microphysical variability across a warm frontal precipitation band has not been explored. Stark et al. (2013) showed that snowbands in the cyclone comma head have more intense vertical motions and riming on the warm side of a band forced by frontogenesis. The factors leading to this narrow and relatively intense warm frontal band in Fig. 2 and associated microphysical variations need to be investigated in order to separate the potential role of forcing for the band, generating (seeder) cells aloft, and other processes (convection, gravity waves, riming, etc.).
As noted above, diabatic heating can be important in enhancing the vertical motion and band intensity in a region of frontogenesis. We hypothesize that dry dynamics alone are not sufficient to produce the well-defined band in Fig. 2. It is also not clear where these bands will develop along the sloping warm frontal zone extending from the surface to midlevels. Novak et al. (2008, 2010) showed that a single primary snowband develops to the north and northwest of the surface cyclone with frontogenesis along a midlevel trough extending to the north of the parent cyclone. Finally, most studies have focused on band development, but few studies have documented the band life cycle from genesis to decay. Novak et al. (2010) showed that for bands northwest of the cyclone center there was an increase in conditional stability and reduction in frontogenesis when the PV anomalies east of the band modified the kinematic field.
High-resolution numerical output is used to complement the surface, radar, and aircraft in situ data during the 18 February 2012 GCPEx case to help address the following questions:
What processes led to the rapid intensification and subsequent weakening of the warm frontal precipitation band?
What is the role of latent heating and cooling on the band evolution?
How did the ice and water amounts change within the band as it evolved?
How well can a mesoscale model predict this warm frontal band evolution?
2. Data and methods
This study used GPM field campaign measurements acquired during GCPEx (Skofronick-Jackson et al. 2015), which featured a broad suite of ground instrumentation and intensive observation periods that included aircraft data (Fig. 1). Nearly all the observations for this study were taken at the Centre for Atmospheric Research Experiments (CARE) site (Fig. 1). Surface data included a ground-based millimeter microwave Doppler radar to estimate hydrometeor fall speeds; dual-polarization precipitation radars at Ka-, Ku-, and C-band to sample microphysical characteristics; numerous video and optical disdrometers; and traditional surface and upper-air weather observations. The University of North Dakota (UND) Citation aircraft was equipped with the King probe, a two-dimensional optical array cloud probe (2D-C), and a high-volume precipitation spectrometer (HVPS). Liquid water content (LWC) was derived from the King probe measurements while particle number concentrations for 38 size bins ranging from 0.05 to 30 mm were derived from merging 2D-C and HVPS measurements. The particle number concentrations were used to derive ice water content (IWC) based on the methodology in Heymsfield et al. (2004), which is applicable for snow aggregates. Thus, for other ice habits, such as rimed ice particles, the derived IWC is highly uncertain. The temperature calculations from the UND Citation temperature probes likely have a random error of 0.2°C in clear air (Field et al. 2006). It appears that for our case study the temperature measurements were biased low by about 0.2 K. We verified this by analyzing areas of the aircraft profiles where the King probe LWC was greater than 0.06 g m−3, which should correspond to regions of relative humidity with respect to water of approximately 100%. When directly using the aircraft temperature measurements in our RH calculations, the mean relative humidity was 101.7%. By adding 0.2 K to the temperature measurements, the mean humidity was approximately 100%.
The GCPEx campaign also used an airborne GPM satellite simulator equipped with Ka- and Ku-band radar, and a radiometer operating at 50–183 GHz. The atmospheric analyses used to compare with the model predictions were from the Rapid Update Cycle (RUC; Benjamin et al. 2004) at 13-km grid spacing.
b. Model setup
Version 3.5.1 of the Advanced Research core in the Weather Research and Forecasting Model (WRF-ARW; Skamarock et al. 2008; hereafter WRF) was used for a 30-h run with a triple-nested grid configuration at 9-, 3-, and 1-km horizontal grid spacings (Fig. 3b). Forecasts were initialized at 1800 UTC 17 February 2012 with initial and lateral boundary conditions from 6-hourly 13-km isobaric RUC analyses. Numerous tests were performed using various configuration options, such as different initial and lateral boundary condition data [i.e., RUC-hybrid, Global Forecast System (GFS), and North American Mesoscale Forecast System (NAM) data], different vertical grid spacings (i.e., 34, 50, and 72 levels), and different planetary boundary layer, land surface physics, and microphysical options. After comparing the simulated reflectivity to the observed C-band dual polarization reflectivity at King City, Ontario (Fig. 2), and S-band Weather Surveillance Radar-1988 Doppler (WSR-88D) reflectivity at Buffalo, New York; Cleveland, Ohio; and Detroit, Michigan, we found that the model predicted the most realistic band development and structure when choosing the configuration options in Table 1. The Eta similarity (Janjić 1990), five-layer thermal diffusion (Dudhia 1996), and Mellor–Yamada–Janjić (Janjić 1994) schemes represent the preferred surface layer, land surface, and planetary boundary layer parameterizations, respectively. The Rapid Radiative Transfer Model for global models (RRTMG; Iacono et al. 2008) was used for simulating shortwave and longwave radiation, and the Grell–Freitas ensemble cumulus parameterization scheme (Grell and Freitas 2014) was activated on the outermost 9-km grid. We utilized the Predicted Particle Properties (P3) version two-cloud microphysical parameterization (Morrison and Milbrandt 2015) for simulating this warm frontal precipitation band. All ice-phase particles in the P3 scheme are represented by four mixing ratio variables (total mass, rime mass, rime volume, and number) that freely evolve in time and space.
3. Large-scale setup
At 1200 UTC 18 February 2012 there was a short-wave trough at 700 hPa over the western Great Lakes moving southeastward toward the GCPEx field study area (Fig. 3a). The short-term (18 h) 9-km WRF forecast accurately predicts the trough location and amplitude, as well as the temperature, moisture distribution, and winds at this level (Fig. 3b). The WRF slightly underpredicts the spatial extent of regions with relative humidity >80% to the east of the midlevel trough, but the dry intrusion is well forecast, which is just west of the study area at this time.
Meanwhile, there was a relatively weak surface cyclone (~1012 hPa) located near the study area of southwest Ontario (not shown), with a weak surface warm front stretching to the east of the cyclone. Nocturnal radiational cooling and surface fluxes resulted in only a 1°–2°C surface temperature difference across the warm front, so 950 hPa is shown instead to better highlight the frontal temperature gradients (Fig. 4). At 0900 UTC 18 February, which is near the time of precipitation band genesis highlighted in section 4a, the 950-hPa cyclonic circulation was located over southwestern Lake Huron in the RUC analysis (Fig. 4a), while the 9-km WRF circulation is 75–100 km to the southwest of the observed position (Fig. 4b). There was a relatively broad temperature gradient to the east of the circulation, with the warm front extending southeastward over central Lake Erie. The warm sector to the south of this front is 1°–2°C warmer in the RUC than the WRF at this time, and the RUC has a stronger temperature gradient across northwest Ohio and central Indiana.
At 1200 UTC 18 February, the 950-hPa circulation in the RUC approached the GCPEx area (Fig. 4c), but the 9-km WRF is 100–200 km too far to the southwest (Fig. 4d). In both the analysis and WRF, the meridional temperature gradient and wind shift had increased near the warm front during the past few hours. This is the period of warm frontal band maturity discussed in section 4b. By 1500 UTC 18 February, which is an hour or two after band decay (section 4c), the 950-hPa circulation was over the GCPEx area in the RUC (Fig. 4e), and just south of this area in the WRF (Fig. 4f). As a result, the warm frontal temperature gradient and associated deformation area was shifting to the east of the GCPEx region.
A sounding comparison between the 1-km WRF and the aircraft vertical profile near the CARE site at 1130 UTC 18 February illustrates some of the important thermodynamic features in the vertical (Fig. 5). The locations of the aircraft and WRF soundings at 1130 UTC are just north of a developing warm frontal band (red oval in Fig. 2c), with the veering wind profile up to 850 hPa indicative of warm advection as the warm front approaches from the southwest. Both the observations and WRF are slightly more stable than moist neutral between 900 and 650 hPa, with the observed profile closer to moist neutral around 600 hPa. The WRF is near moist neutral in the lowest 1 km above the surface, where unfortunately there are no observed data. The relative humidity values with respect to water and ice in WRF are 5%–10% greater than the observed profiles below 700 hPa.
4. Warm frontal band evolution
a. Warm frontal band genesis
The warm frontal precipitation band developed between 0600 and 1100 UTC 18 February, with the most rapid intensification occurring during the final hour. At 0700 UTC 18 February (Figs. 6a,b), the WRF realistically simulated the 2-m temperature distribution as compared to the observations, except over the northeast part of the domain where the WRF results were 2°–4°C too warm. There is a widespread precipitation area (15–35 dBZ) primarily just north of the surface front in both WRF and the observations. Note that model output from the 1-km WRF domain is displayed within the domain boundaries (thick gray) while the 3-km WRF output is displayed everywhere outside these boundaries. There are embedded convective cells within this precipitation area, but no linear or bandlike features at this time. The WRF precipitation coverage is slightly greater than observed over the southeast portion of this genesis region. This precipitation area is near the 1-km WRF boundary, but there is no sign of any hypergradients or spurious solutions, as has been seen in some other studies (e.g., Andrys et al. 2016).
Figure 7a shows the 925–850-hPa-layer average temperatures, 2D frontogenetical forcing defined in Petterssen (1936), and the axis of dilatation vectors at 0700 UTC. There is relatively weak frontogenesis [1–3 K (100 km)−1 h−1 in this layer], with the precipitation band developing in a region nearly parallel to the dilatation axes. The 1-km WRF cross section AA′ illustrates the embedded convective cores from the surface to near 600 hPa within the sloping baroclinic zone (Fig. 6c), while there is a general lack of convective cores in the northern portion of this section. There is weak frontogenetical forcing from the surface to 850 hPa within the frontal zone, while some of the weak frontogenesis above this layer is caused by some locally enhanced temperature gradients around the convective cells. Moist potential vorticity (MPV*) was calculated to assess the conditional and slantwise instability aloft using the total wind and the saturation equivalent potential temperature . A layer of weakly positive to slightly negative MPV* exists from 900 to 650 hPa to the south of the front and it slopes upward to 750–550 hPa just north of the surface front. There is also a near-neutral or slightly negative MPV* layer to the north of the surface front below 1 km AGL.
At this same time, Fig. 8 compares the stability and moisture profile for a small 0.2° × 0.2° region just south of the developing warm-frontal precipitation band against a region within the developing band. To the south of the precipitation area (Fig. 8a), the equivalent potential temperature decreases with height between 900 and 725 hPa. This is the result of the dry-air intrusion between 850 and 700 hPa, which was shown above entering from the west (Fig. 3). There is potential instability (PI) between 900 and 800 hPa, while there is conditional instability (CI) between 775 and 675 hPa, where the saturation decreases with height, and this corresponds well with the negative MPV* layer. In contrast, the small region within the precipitation area is nearly saturated to 600 hPa and the profile is near moist neutral (Fig. 8b). However, cross section AA′ still illustrates some negative MPV* areas and small regions with decreasing with height that are suggestive of CI and consistent with the embedded convective cores. There is a layer of greater stability from 900 to 850 hPa within the warm frontal zone, which is 50–100 hPa higher than the region to the south of the precipitation area.
By 0900 UTC 18 February, the warm frontal precipitation area is more elongated and linear (Fig. 9a), but this structure was not yet present in the 1-km WRF run (Fig. 9b). The 925–850-hPa frontogenesis had increased to 2–7 K (100 km)−1 h−1 along the south side of the developing precipitation band in WRF as the temperature gradient in this layer had increased 20%–30% [to ~0.8°C (25 km)−1] nearly parallel to the dilatation axis during the past 2 h (Fig. 7b). The frontogenesis in this layer and in section AA′ has a clear maximum at the leading (southern) edge of the precipitation band and slopes upward between 900 and 800 hPa. The location of this developing band is ~75 km to the north of the low-level warm front, as is evident by the 950-hPa temperature gradient in WRF and the observed and simulated surface wind shifts. There are fewer convective cells within the precipitation area at this time (Fig. 9c) as compared to 1–2 h ago because of an increase in MPV*. There is still a layer of negative MPV* approaching the band from the south at 825–675 hPa (Fig. 9d), which is associated with a CI layer (not shown), and there is still weak PI from 850 to 800 hPa with the dry intrusion around 800 hPa (not shown). The frontogenetical circulation combined with these instability layers helped trigger a few cells across the southern portion of the band. Meanwhile, at 0900 UTC the warm frontal slope is steeper than 1–2 h earlier, as indicated by the zone of sloping lines.
b. Warm frontal band maturity
The warm frontal band intensifies between 0900 and 1100 UTC. At 1030 UTC 18 February, a well-defined precipitation band stretches a few hundred kilometers from northwest to southeast in the observed radar reflectivity data (Fig. 10a). The 925–850-hPa frontogenesis is similar to 0900 UTC along the south side of the precipitation band (Fig. 7c), thus further increasing the temperature gradient in this layer [to 1.0 K (25 km)−1]. Although the WRF band at 1030 UTC has intensified and became more linear, it is still weaker than observed and too far southwest (Fig. 10b). Observed and simulated soundings from an hour earlier at 0930 UTC help explain the weaker band in the WRF. For example, the observed sounding at the CARE site (Fig. 11a) located on the northern side of the developing band shows a CI layer from 950 to 850 hPa, while the WRF (Fig. 11b) shows shallower and weaker CI in this layer. Meanwhile, both the observed and simulated soundings are weakly stable above this layer to around 500 hPa. The air is subsaturated below 900 hPa in both the observed and WRF profiles. A more complex wind profile exists in the observations, with a slight backing wind from 950 to 850 hPa, which implies weak cold advection that may have helped destabilize the temperatures in the CI layer more than WRF.
The warm frontal band reached maturity around 1130 UTC in both the observations and the 1-km WRF (Fig. 12). The observed band extends for several hundred kilometers from west-northwest across the GCPEx region and then to the east-southeast. The observed radar reflectivities within the band exceed 32 dBZ. The WRF band is also fairly well defined, but there are more convective cores on the south side of the band than observed.
A 1-km WRF cross section at 1130 UTC illustrates a narrow core of reflectivity on the south side of the band from the surface to 800 hPa (Fig. 12c). There is a well-defined sloping frontogenetical zone from the surface on the south side of the band upward to 800 hPa within the band. This is similar to Novak et al. (2008, 2010), but the frontogenesis farther north in the comma head is typically centered from 750 to 650 hPa. This forcing helps generate an ageostrophic circulation and the linear organization of the band. The warm frontal slope continued to steepen during the frontogenesis process, and the single band became collocated near the steepest part of the front (~900–800 hPa), but the front did not break down into a series of steplike features as noted in some other studies (e.g., Neiman et al. 1993). Also, no wavelike perturbations were evident in the cross section to help generate or enhance the precipitation band. There is a layer of negative MPV* between 800 and 650 hPa just to the south of the band, which favors increased convective motions near the southern part of the band. The MPV* profile is near zero within the band, as is also illustrated above with the sounding comparison between the aircraft and WRF within the band near the CARE site (Fig. 5), which is slightly more stable than moist neutral from 900 to 600 hPa. During maturity the stability across the banded region was similar to that in Schultz and Schumacher (1999; their Fig. 4), with a CI layer to the south of the band and more neutral stratification within the band.
Figure 13 shows a more direct comparison of the band between the King City radar and the 1-km WRF along section BB′ at 1122 and 1130 UTC, respectively. The precipitation band was not associated with a midlevel trough or trowal (trough of warm air aloft) structure (Martin 1998), but rather with low-level deformation and frontogenesis along the sloping warm frontal zone, similar to Heymsfield (1979), Locatelli and Hobbs (1987), and Hudak et al. (1996). The layer of frontogenesis in the WRF slopes upward from 0.8 to 1.8 km, and the warm frontal precipitation band is 50–100 km to the north of the surface front and beneath the ascent region associated with the frontogenesis maximum around 1.2 km. The observed reflectivity maximum with the band slopes downward to the northwest, likely a result of the advection of snow hydrometeors within low-level southeasterly flow, while the WRF dry intrusion limited the precipitation above 2.5 km. The ragged cloud top in the observed reflectivity around 2.5 km is suggestive of weak convective cores (“generating cells”), which is near a moist neutral layer (Fig. 5a). The Dual-Frequency Dual-Polarized Doppler Radar (D3R in Fig. 1) better captures these cells aloft with its finer-scale resolution, but there were no apparent fall streaks connecting these cells to the heavier precipitation rates below, so it is difficult to determine their role in enhancing the ice at lower levels. The 1-km WRF could not produce these cloud-top cells, likely because WRF is more stable than observed near cloud top (Fig. 4b) and these features are of too small a scale to be resolved by the model. Overall, between 1130 and 1230 UTC the observed precipitation rate was around 2 mm h−1 at the CARE site while only 1 mm h−1 in the 1-km WRF, since the band was somewhat weaker than observed.
c. Warm frontal band decay
During the next 1–2 h, warm frontal precipitation began to decay. At 1400 UTC 18 February in the observations and in the 1-km WRF, the band is less defined and situated 50–100 km farther to the north of the surface warm front than at previous times (Fig. 14). The 925–850-hPa frontogenesis is 40%–50% weaker than during the mature phase (Fig. 7d), and there is no well-defined dilatation axis and no further increase in the temperature gradient in this layer. The frontogenesis is also weaker within section AA′, while a layer of negative MPV* still exists between 800 and 700 hPa to the south of the band, with the low-level ascent with the front triggering weak convective cells just north of the surface warm front seen in both the observed radar and WRF. These cells are occurring near the boundary of the dry intrusion and deeper frontal cloud area to the north. However, the areal extent of the negative MPV* area is less than during the mature phase (Fig. 12d). The increase in stability as the band decays is similar to the primary band work in the comma head (Novak et al. 2008, 2010).
d. Microphysical evolution
There was a rapid evolution in the snow microphysics as the band matured. Figure 15 shows the differential reflectivity (ZDR) results from the King City radar at 1124 and 1234 UTC 18 February, which correspond to the period analyzed in section 4b when the band reached maturity and began to weaken. At 1124 UTC (Figs. 15a,b), the ZDRs are near zero or slightly positive within the band, and the correlation coefficients are relatively high (>0.99) and uniform (not shown). This suggests horizontally falling snow crystals and aggregates (Kumjian 2013). Figure 16a shows an increase in ice water content within the northern part of the precipitation band from 4.0 to 1.5 km AGL (from 1122 to 1143 UTC). The most rapid ice growth is between 3.1 and 2.4 km, which is a favored layer for depositional growth (−15° to −10°C). Figures 16b and 16c show the aircraft ice habits from the two-dimensional optical array cloud probe (2DC) sampled within the band near 1.2 km in height at 1138 UTC and those collected by the precipitation video imager (PVI) a few minutes later (1140 UTC) at the surface. There are unrimed snow aggregates at this height, with minimal cloud water observed by the aircraft at this time (Fig. 16a).This is consistent with the observed ZDRs that are near zero to slightly positive.
Meanwhile, to the south of the band at 1124 UTC the ZDRs are slightly negative and the correlations are relatively high (~0.99; not shown), suggesting a fairly uniform area of graupel, some of which may be orientated more vertically (e.g., conical graupel). By 1230 UTC 18 February (Figs. 15c,d), the area of negative ZDR increases over the central and southern band region, and the correlations are still relatively high (~0.99; not shown). This suggests a continued evolution toward more rimed snow and graupel. The aircraft 2DC observations indicate small graupel particles at this time (Fig. 17c), as well as the PVI observations at the surface (Fig. 17d). Some of the graupel are conical, confirming the negative ZDR. The aircraft profile on the south side of the band oscillates between regions of large cloud water amounts (0.6 g m−3) on the south side of the spiral and less water and more snow content (to 1.0 g m−3) on the north side of the spiral. The aircraft upward motion increases from 0.2–0.4 m s−1 at 1124 UTC to 0.4–0.7 m s−1 around 1225 UTC (Figs. 18a,b). This is consistent with the increase in supercooled cloud water and riming observed at the later time over the south side of the band.
The 1-km WRF has a similar microphysical evolution as the observations with larger vertical motion over the southern part of the band (Figs. 17a,b). There is primarily snow within much of the band area in the model around 1130 UTC (Fig. 18c), while there is a sharp transition to more cloud water along the southern edge of the band that increases by 1230 UTC (Fig. 18d). Overall, the increased vertical motion and rimed snow on the south side of the precipitation band is similar to the observations by Stark et al. (2013) for snowbands in the cyclone comma head, in which the stronger vertical motions associated with the frontogenetical circulation lead to increased riming on the warm side of the band.
5. Impact of latent heating and cooling processes
Latent heating has been shown to increase the frontogenesis and frontal movement for warm and cold fronts (Hsie et al. 1984; Szeto and Stewart 1997; Reeves and Lackmann 2004), as well as to decrease the warm frontal slope (Igel and Van den Heever 2014). To quantify the impacts of latent heating and cooling processes on the warm frontal structure and band development, separate WRF simulations were performed with these processes turned off in the model. First, a WRF run was completed by turning off the condensational and depositional heating (NOCD run) in all WRF domains starting at hour 6 (0600 UTC 18 February). This allows the initial warm frontal precipitation area to develop, but afterward it illustrates the role of heating on the warm frontal band’s evolution. By 1130 UTC 18 February, the time of band maturity in the control run (cf. Fig. 12b), there is little or no warm frontal band in the NOCD run (Figs. 19a,b), but rather a relatively broad area of precipitation with some embedded convective cells. The latent heating is concentrated in a sloping layer from 900 hPa south of the warm front to 800 hPa just north of the front (Fig. 19c). As a result, the stability increased around 800 hPa relative to the control run, while the horizontal temperature gradient is reduced around 900 hPa, such that there are no frontlike temperature gradients or frontogenesis. There is weak conditional instability (negative MPV*) below 850 hPa in the NOCD run (Fig. 19d), which was released with the weak warm advection (veering wind profile) in this layer. A separate simulation was completed by simply turning off the latent heating, which weakened the band by nearly 50% and it was less organized (not shown). Overall, latent heating from both condensation and deposition decreases the stability and increases the frontal circulations, thus promoting band development. This result is likely sensitive to the freezing level. For example, Igel and Van Den Heever (2014) showed little impact from depositional heating on the frontal structure when the freezing level was 2–3 km above the surface.
Another sensitivity run was completed by turning off all the latent cooling processes (NOLC) at hour 6. By 1130 UTC, the additional cooling to the north of the front increased the temperature gradient in the control run relative to the NOLC run (Figs. 20a,b). As a result, the frontogenesis is weaker in the NOLC run than the control (cf. Fig. 12c), and the frontal precipitation band is ~50 km farther to the south (slower) than the control. However, the precipitation band is slightly stronger in the NOLC than the control run. The layer below 800 hPa to the south of the warm front is somewhat more stable as a result of cloud water evaporation in this layer (warm layer around 900–850 hPa) and more unstable from 800 to 700 hPa (Figs. 20c,d). This was confirmed by a separate run turning off only evaporative cooling (not shown), which resulted in a similar response. As a result of the low-level stabilization, there are fewer cells triggered to the south of the warm front in the NOLC run, while some instability is removed in the control run as cells are triggered to the south of the front. The additional instability between 800 and 700 hPa in the NOLC compensates for some of the lost frontogenetical forcing and results in a well-defined warm frontal precipitation band. Other runs without melting or sublimation cooling had more limited impact (not shown), with the band similar to the control. Although latent cooling increases the band intensity slightly, only a relatively weak (<20 dBZ) and unorganized frontal band develops if both latent heating and cooling are turned off (not shown), so latent heating processes appear to dominate over the cooling.
6. Summary and conclusions
Using ground-based radar and aircraft data from the Global Precipitation Measurement (GPM) Mission Cold Season Precipitation Experiment (GCPEx), as well as model analyses and simulations, the structure and evolution of an intense warm frontal precipitation band is documented over southern Ontario on 18 February 2012. The precipitation band went through a life cycle of genesis, maturity, and decay over a 5–6-h period. The Weather Research and Forecasting (WRF) Model simulated this band life cycle down to 1-km grid spacing, and it was able to realistically predict the precipitation band evolution, albeit somewhat weaker and slightly farther south than observed.
Initially, there was an area of stratiform precipitation with embedded convective cells to the north of the surface warm front. Just south of this precipitation area there was potential and conditional instability between 900 and 650 hPa, where the moist potential vorticity (MPV*) was near zero. A midlevel dry intrusion approaching from the south and west reduced the stability within midlevels. Meanwhile, the stratification was more moist neutral for slantwise and vertical convective motions within the banding region. As the low-level cyclonic wave along the front approached and intensified, the deformation and frontogenesis increased around 1 km above the surface and the precipitation band rapidly developed and matured. Over the next 2 h, the low-level deformation decreased as the low-level cyclone moved east and the column stabilized, which led to weakening precipitation within the band.
The precipitation band in this GCPEx case was not associated with a midlevel trough or trowal structure, but rather with low-level deformation and frontogenesis along the sloping warm frontal zone. There was a single well-defined band generated within the sloping frontogenesis region above the surface (~800 hPa). Unlike other warm frontal multiband studies that highlighted conditional symmetric instability (CSI; e.g., Wiesmueller and Zubrick 1998), our dominant instability was CI or PI during band genesis. As a result, embedded convective cells developed earlier in our GCPEx event, but the frontogenesis was weak then and banding was limited. As the deformation increased, the stability also increased near the banding location (MPV* ~ 0), which favored the development of a single band. The increase in stability as the band matured is similar to other studies of a precipitation band in the comma head (Novak et al. 2010).
Latent heating through condensational heating and depositional ice growth has been shown to narrow the region of frontal ascent and increase the vertical velocity. We conducted simulations either turning off the condensational and depositional heating (NOCD) or the latent cooling processes (NOLC run). Similar to these other studies, we found for the 18 February 2012 case that latent heating helps increase the frontal circulations and the resulting band development. Latent cooling also helps increase the frontogenesis given the evaporative and sublimation cooling within the frontal precipitation, but the band in the NOLC run was slightly stronger than the control run because there was more available midlevel instability approaching the band. The low-level instability increased with the cooling, which prevented cells from developing ahead (south) of the band and releasing the midlevel instability as compared to the control run. Overall, this highlights the predictability issues that can arise given the interaction between the frontal circulation patterns and moist physics.
The ice microphysics on 18 February 2012 went through a rapid transition as the warm frontal band matured. Initially, there were aggregates of unrimed snow during early maturity within the band, but as the vertical motion increased over the south (warm) side of the band, more supercooled water and riming occurred in the observations and thee model. This is consistent with the frontogenetical forcing resulting in the largest upward motion on the warm side of the temperature boundary. There were generating cells near cloud top in the observations, which is consistent with many of the recent PLOWS studies of cyclones over the central United States (Rosenow et al. 2014; Plummer et al. 2014). These cells were not simulated in the 1-km WRF simulations likely because of the coarse model resolution. For example, Keeler et al. (2016) showed that WRF can realistically simulate these cells with grid spacings of a few hundred meters. Even without these cells, the WRF still developed the frontal band; thus, the seeder cells were not critical for band development. However, the absence of seeder cells may have been one of the reasons why the WRF underpredicted the intensity of the band.
This work was supported by National Aeronautics and Space Administration Grant NNX13AF88G. We appreciate the comments and suggestions made by the Editor (Dr. David Schultz) and three anonymous reviewers, who helped improve several aspects of this paper.