Abstract

Detailed observations from the Global Precipitation Measurement (GPM) mission Cold Season Precipitation Experiment (GCPEx) of an intense warm frontal band on 18 February 2012 were used to evaluate several bulk microphysical parameterizations within the NASA-Unified Weather Research and Forecasting (NU-WRF) Model. These included the Predicted Particle Properties (P3), Morrison (MORR), Stony Brook University (SBU), and Goddard four-class ice (4ICE) microphysics schemes. All schemes were able to predict the snowband, but the simulated intensities varied because of various assumptions in these schemes. The saturation adjustment scheme within MORR promoted excessive amounts of cloud water evaporational cooling in the warm sector, which contributed to a decrease in midlevel instability approaching the frontal band and thus a weaker band. In contrast, the explicit calculation of cloud water condensation/evaporation in the P3 scheme produced limited amounts of evaporational cooling, which allowed for greater midlevel instability to support band development. The P3 and SBU schemes produced moderate rime/graupel mass within the band that was confirmed by observations, while the MORR and 4ICE schemes drastically underpredicted the graupel mass. The high-density, fast-falling rimed particles in P3 underwent weak sublimation and melting, which helped promote a stronger horizontal temperature gradient and greater low-level instability along the frontal band compared to the other schemes. Overall, the schemes that use specified thresholds for converting between the predefined ice-phase categories of cloud ice, snow, and graupel had the most unrepresentative hydrometeor types. These results highlight the advantage of predicting ice particle properties and explicitly calculating cloud water condensation/evaporation in the P3 scheme.

1. Introduction

Bulk microphysical parameterizations (BMPs) are a critical component of a numerical model since high-resolution forecasts can vary significantly depending on the choice of the microphysics scheme and the associated assumptions (Thompson et al. 2008; Lin and Colle 2009; Morrison et al. 2009; Milbrandt et al. 2010; Lin and Colle 2011; Molthan and Colle 2012; Lang et al. 2014). The Weather Research and Forecasting Model (WRF; Skamarock et al. 2008) has several single- and multimoment BMPs, and a detailed discussion of their critical components can be found in Molthan and Colle (2012). Extensive validation efforts have helped identify weaknesses and biases in BMPs during the last several years (Thompson et al. 2008; Morrison et al. 2009; Hong et al. 2010; Milbrandt et al. 2010; Shi et al. 2010; Lang et al. 2011; Molthan and Colle 2012), which led to the development of more advanced schemes in mesoscale models (Lang et al. 2014; Morrison and Milbrandt 2015). Lang et al. (2011) refined the single-moment Goddard three-class ice (3ICE) BMP (Lang et al. 2007) by using the temperature and mixing ratio to determine the snow (N0s) and graupel (N0g) intercepts rather than fixed values. Rime splintering, immersion freezing, and contact nucleation were also introduced into the scheme to improve the simulation of the cloud water to cloud ice transition. Finally, Goddard 3ICE was renamed as four-class ice (4ICE) after introducing a hail category and associated processes, which led to higher reflectivities aloft that better compared to observations during intense squall lines (Lang et al. 2014).

The double-moment prediction of particle mass and number concentration can lead to more realistic model simulations compared to single-moment schemes only predicting mass concentration (Morrison et al. 2009; Molthan et al. 2010; Igel et al. 2015). For double-moment schemes, the slope λ and intercept N0 parameters are derived from the predicted mass and number concentration, while single-moment schemes simply use fixed values for N0s or rely on additional parameterizations, such as the temperature and snow mixing ratio parameterization used to diagnose N0s in Goddard 4ICE (Lang et al. 2014). Morrison et al. (2009) showed that the prediction of number concentration in the double-moment Morrison (MORR) scheme promoted more realistic mean particle sizes, evaporation rates, and precipitation compared to single-moment schemes. Molthan and Colle (2012) also noted the ability of the MORR scheme to simulate the ice crystal aggregation process that led an improved prediction of snowfall sizes for a case during the Canadian CloudSat/CALIPSO Validation Program (C3VP). However, rather than predicting both graupel and hail as in 4ICE, the MORR scheme has a single rimed-ice category that can be used to represent either graupel or hail (graupel in our study), which can lead to significant differences in model simulations as a result of the differing hydrometeor densities and fall speeds (Morrison and Milbrandt 2011).

Most BMP schemes, including 4ICE and MORR, use specified thresholds to transfer ice-phase hydrometeors between multiple predefined classes (e.g., cloud ice, snow, and graupel) with fixed shapes, which can cause abrupt changes in the ice particle properties and unrealistic patterns of behavior in the model simulations (Colle et al. 2005; Morrison and Grabowski 2008). This motivated the development of the Stony Brook University (SBU) scheme (Lin and Colle 2011) that combines graupel and snow into a single precipitating ice category. Instead of using constant values for the coefficients am and bm, which characterize the particle effective density and the particle fractal dimension, respectively, in the mass–diameter relationship as done in most BMPs, including 4ICE and MORR, the SBU scheme allows these coefficients to vary based on the riming intensity Ri and temperature. The same holds true for the coefficients aυ and bυ that characterize the ice crystal habit and degree of riming, respectively, in the terminal velocity–diameter relationship. The Ri formulation in SBU is dependent on the predicted liquid and ice water content where values near one represent graupel-like particles while values near zero represent dry snow. Consequently, the SBU scheme permits nonspherical precipitating ice particles with varying densities and fall speeds diagnosed at each model grid point.

Recently, Morrison and and Milbrandt (2015) developed the double-moment predicted particle properties (P3) scheme, which assigns conservation equations accounting for advection, subgrid-scale mixing, and microphysics for the prognostic mixing ratio variables of total ice mass qi, rime ice mass qrim, rime volume, and total number concentration M0i. From these equations, the P3 scheme derives predicted particle properties (e.g., rime mass fraction, bulk density, and mean particle size), which allows the independent evolution of these properties in time and space rather than diagnosing them at each model grid point as in SBU. P3 also employs different mD relationships for small ice spheres, larger unrimed particles, and rimed particles. In particular, the mD relationship for rimed ice particles is dependent on the predicted rime mass fraction (Fr = qrim/qi) and mean particle size D along with αva and βva, which have specific values of 0.01855 and 1.9 based on measurements in midlatitude cirrus clouds (Brown and Francis 1995). For the VD relationship, aυ and bυ are derived following Mitchell and Heymsfield (2005) based on the ReX relationship, where Re is the particle Reynolds number and X is the Best (Davies) number (related to the ratio of the particle mass to its projected area). This approach is explicitly dependent on the particle density, which is advantageous over most other BMPs that ignore varying particle densities in VD relationships. By predicting particle properties, an additional advantage of the P3 scheme is that it allows smooth transitions between ice-phase hydrometeor types. However, a disadvantage of the single ice category version of P3 used here is that it may lead to unfavorable forecasts in weather conditions where higher- and lower-density rimed ice particles coexist in space and time, such as narrow hail shafts (Milbrandt and Morrison 2013).

In this study, we will investigate the ability of four BMP schemes in WRF (Table 1) in simulating a narrow, well-defined snowband during the Global Precipitation Measurement (GPM) Cold Season Precipitation Experiment (GCPEx; Skofronick-Jackson et al. 2015). The defining relationship and parameterizations in the BMP schemes are summarized in Table 2. Colle et al. (2017) used field measurements and model simulations to show the evolution of this warm frontal snowband during its passage over the GCPEx campaign region in southern Ontario, Canada, on 18 February 2012. Surface measurements highlighted the rapid transition from large snow aggregates to small graupel particles as enhanced upward motions promoted a layer of supercooled cloud water within the band. High-resolution WRF Model simulations using the P3 scheme realistically predicted the band evolution, but it was slightly weaker than that observed. This study expands on the results in Colle et al. (2017) by conducting a detailed comparison and evaluation of the more advanced BMP schemes (i.e., P3, MORR, 4ICE, and SBU) in simulating this rapidly evolving snowband. A summary of each microphysical scheme is provided in Tables 1 and 2. To summarize, high-resolution WRF simulations using these schemes are compared and evaluated against GCPEx field campaign measurements to address the following questions:

  1. How does the choice of cloud microphysics scheme impact the development of a snowband associated with a mesoscale warm front?

  2. How well do the microphysics schemes predict the ice/snow distribution and riming within the warm frontal band?

  3. How do cloud microphysical processes modify the environmental conditions, which in turn can impact the warm frontal band development?

Table 1.

Characteristics of microphysical schemes used for the WRF simulations. The subscript s refers to the predefined snow category in MORR and 4ICE. The subscript i refers to the single lumped ice category in P3 and the precipitating ice category in SBU.

Characteristics of microphysical schemes used for the WRF simulations. The subscript s refers to the predefined snow category in MORR and 4ICE. The subscript i refers to the single lumped ice category in P3 and the precipitating ice category in SBU.
Characteristics of microphysical schemes used for the WRF simulations. The subscript s refers to the predefined snow category in MORR and 4ICE. The subscript i refers to the single lumped ice category in P3 and the precipitating ice category in SBU.
Table 2.

Parameters defining relationships within each of the microphysics schemes; M0s and M0i refer to the snow and total ice (P3) number concentrations, respectively. For the MORR and 4ICE schemes with separate snow and graupel categories, the graupel parameters are shown if different from snow. The numbers to the left of the solidus represent snow and numbers to the right represent graupel. For example, MORR assumes a density of 100 kg m−3 for snow and 400 kg m−3 for graupel. Two values are shown for the graupel category in the 4ICE scheme, since the second value is valid when qg > 2 g m−3. See text for further parameter definitions.

Parameters defining relationships within each of the microphysics schemes; M0s and M0i refer to the snow and total ice (P3) number concentrations, respectively. For the MORR and 4ICE schemes with separate snow and graupel categories, the graupel parameters are shown if different from snow. The numbers to the left of the solidus represent snow and numbers to the right represent graupel. For example, MORR assumes a density of 100 kg m−3 for snow and 400 kg m−3 for graupel. Two values are shown for the graupel category in the 4ICE scheme, since the second value is valid when qg > 2 g m−3. See text for further parameter definitions.
Parameters defining relationships within each of the microphysics schemes; M0s and M0i refer to the snow and total ice (P3) number concentrations, respectively. For the MORR and 4ICE schemes with separate snow and graupel categories, the graupel parameters are shown if different from snow. The numbers to the left of the solidus represent snow and numbers to the right represent graupel. For example, MORR assumes a density of 100 kg m−3 for snow and 400 kg m−3 for graupel. Two values are shown for the graupel category in the 4ICE scheme, since the second value is valid when qg > 2 g m−3. See text for further parameter definitions.

2. Data and methods

a. GCPEx field instrumentation

A detailed description of all instrumentation deployed during the GCPEx field campaign (January–February 2012) can be found in Skofronick-Jackson et al. (2015). Measurements from the King probe on board the University of North Dakota (UND) Citation aircraft (Delene and Poellot 2016) permitted retrievals of liquid water content (LWC). Additional measurements from the two-dimensional optical array cloud probe (2D-C) and high-volume particle spectrometer (HVPS) on board the UND aircraft (Heymsfield et al. 2014) allowed retrievals of ice particle number concentrations for 38 size bins ranging from 0.05 to 30 mm. Ice water content (IWC) was then derived from the particle number concentrations using the Heymsfield et al. (2004) methodology, which is valid for snow aggregates. Particles smaller than 100 μm were ignored during processing because of difficulties with determining particle size and probe sample area (Strapp et al. 2001). The shapes, sizes, and ice habit types of precipitation particles aloft and at the surface are diagnosed via images from the aircraft 2D-C and precipitation video imager (PVI) at the Centre for Atmospheric Research Experiments (CARE) site (Blivens 2014). The combination of a gust probe and an Applanix Position and Orientation System (POS) on board the aircraft helped determine the flow of air relative to the aircraft, which permitted the calculation of the three-dimensional winds. Furthermore, we utilize the King City radar (WKR) C-band dual-polarization measurements (Hudak 2013) of reflectivity (dBZ) and differential reflectivity (ZDR) to diagnose the evolution and microphysical characteristics of the warm frontal snowband. To supplement the WKR coverage, we use dual-polarimetric S-band Weather Surveillance Radar 1988-Doppler (WSR-88D) at Buffalo, New York; Cleveland, Ohio; Detroit, Michigan; and Exeter, Ontario, Canada. Environment Canada (EC, now known as Environment and Climate Change Canada) operated the Micro Rain Radar (MRR) instrument at the CARE site (Petersen et al. 2015) where reliable values of effective dBZ and fall velocity were derived from the postprocessing methodology of Maahn and Kollias (2012). The Pluvio 400 precipitation weighting gauge at the CARE site (Petersen et al. 2013) measured accurate snow water equivalent (SWE) amounts every minute (Skofronick-Jackson et al. 2015), which helped verify the simulated precipitation. Locations of the GCPEx field instrumentation and WSR-88D sites are depicted in Fig. 1.

Fig. 1.

Map of the GCPEx field campaign region and surrounding WSR-88D sites. WKR was deployed to the southeast of the dual-frequency, dual-polarized Doppler radar (D3R) stationed at the CARE ground instrumentation site. The WSR-88D sites used in this study included Buffalo (KBUF), Detroit (KDTX), Cleveland (KCLE), and Exeter (KWSO). The KCLE radar site is located approximately 65 km to the south of the arrow’s endpoint.

Fig. 1.

Map of the GCPEx field campaign region and surrounding WSR-88D sites. WKR was deployed to the southeast of the dual-frequency, dual-polarized Doppler radar (D3R) stationed at the CARE ground instrumentation site. The WSR-88D sites used in this study included Buffalo (KBUF), Detroit (KDTX), Cleveland (KCLE), and Exeter (KWSO). The KCLE radar site is located approximately 65 km to the south of the arrow’s endpoint.

b. Model setup

The NASA-Unified WRF (NU-WRF) Model version 3.5.1 is used to perform 30-h simulations of the warm frontal snowband. NU-WRF includes the Goddard 4ICE BMP alongside the other schemes available in the public release of WRF. The P3 scheme is not available in NU-WRF, but was added into the system for this study. We used a one-way triple-nested grid at 9-, 3-, and 1-km horizontal grid spacing with 50 vertical levels centered over the GCPEx field site. Forecasts were initialized at 1800 UTC 17 February 2012 with initial and lateral boundary conditions provided by 6-hourly Rapid Update Cycle (RUC) analyses at 13-km grid spacing that includes soil temperatures and moisture, snow cover, and lake temperatures. The Grell and Freitas (2014) ensemble cumulus parameterization scheme is only used for the outermost grid. The only configuration option that differs among our simulations is the choice of cloud microphysics scheme, which varies between the 4ICE, SBU, MORR, and P3 schemes. See Table 3 for additional configuration details.

Table 3.

Configuration used in the WRF simulations.

Configuration used in the WRF simulations.
Configuration used in the WRF simulations.

3. Microphysical scheme validation

The microphysical evolution of the mature warm frontal snowband on 18 February 2012 is discussed in Colle et al. (2017). Here, we validate the 1-km WRF simulation results for each microphysical scheme as the mature warm frontal band moved over the GCPEx field campaign area from about 1100 to 1230 UTC 18 February. The observed radar reflectivity shows a well-organized frontal band with a narrow core of enhanced reflectivity exceeding 28 dBZ (Fig. 2a). A fairly well-organized band is also predicted by the P3 scheme, with some areas of locally enhanced reflectivity developing near its back edge (Fig. 2b). The MORR scheme simulates a much broader area of stronger reflectivity compared to the observations, but lacks a distinct, narrow core of reflectivity (Fig. 2c), while the 4ICE scheme simulates a considerably weaker, disorganized frontal band (Fig. 2d). A narrow band of reflectivity exceeding 28 dBZ is predicted by the SBU scheme, which reasonably compares to the observed band (Fig. 2e).

Fig. 2.

(a) Radar reflectivity from the WKR 0.3°-elevation scan (shaded in dBZ) at approximately 1230 UTC 18 Feb 2012. The 2-m temperature (°C) and 10-m wind speed (1 full barb = 10 kt, where 1 kt = 0.51 m s−1) at 1200 UTC from surface observation station measurements are also shown. Aircraft descent spiral is shown by brown circle. WRF reflectivity (shaded) at 1230 UTC from the lowest model vertical level in the (b) P3, (c) MORR, (d) 4ICE, and (e) SBU simulations, along with the predicted 2-m temperature (red, contoured every 2°C) and 10-m wind speed. GCPEx CARE site indicated by solid black diamond. Brown circled × denotes location used to produce profiles in Figs. 3, 4, and 6.

Fig. 2.

(a) Radar reflectivity from the WKR 0.3°-elevation scan (shaded in dBZ) at approximately 1230 UTC 18 Feb 2012. The 2-m temperature (°C) and 10-m wind speed (1 full barb = 10 kt, where 1 kt = 0.51 m s−1) at 1200 UTC from surface observation station measurements are also shown. Aircraft descent spiral is shown by brown circle. WRF reflectivity (shaded) at 1230 UTC from the lowest model vertical level in the (b) P3, (c) MORR, (d) 4ICE, and (e) SBU simulations, along with the predicted 2-m temperature (red, contoured every 2°C) and 10-m wind speed. GCPEx CARE site indicated by solid black diamond. Brown circled × denotes location used to produce profiles in Figs. 3, 4, and 6.

We evaluated the mean hydrometeor mass content profiles from the 1130 UTC 18 February model output [brown circled crisscrosses (×s) in Figs. 2b–e] against the LWCs and IWCs derived from the aircraft descent profile (brown circle in Fig. 2a) between 1122 and 1143 UTC (Fig. 3). For SBU, snow and graupel mass are calculated via the qi and Ri output [e.g., qs = (1 − Ri) × qi]. Note these simulated and observed profiles were located farther northeast within the frontal band than indicated in Fig. 2 because of band movement from 1130 to 1230 UTC. Thus, the aircraft descent profile was just ahead (north) of the observed maximum reflectivity band, and the aircraft 2D-C and CARE site PVI sampled unrimed snow aggregates at this time (not shown). The aircraft IWC shows an increasing trend from values less than 0.1 g m−3 above 3.0 km to values exceeding 1.0 g m−3 near 1.2 km in height, while LWC is minimal with measurements peaking near 0.1 g m−3 between 1.0 and 1.5 km. LWC measurements below 0.05 g m−3 were neglected as a result of large uncertainties at these small amounts. Nevertheless, cloud water mass in P3 peaks at 0.02 g m−3 near 3.0 km in height (Fig. 3a) because of the efficient conversion of cloud water to ice along the southern portions of the well-defined frontal band, which effectively reduces the low-level cloud water amounts to the north of this area where the aircraft was sampling. Cloud water is also limited in SBU as a result of its efficient conversion to ice precipitation, which promotes the band of enhanced reflectivity associated with graupel production (Fig. 3d). The 4ICE (Fig. 3c) and especially MORR (Fig. 3b) schemes overpredict cloud water from 2- to 2.5-km height, with mass contents exceeding 0.05 and 0.1 g m−3, respectively.

Fig. 3.

Mean vertical profiles of hydrometeor content calculated from the 100 grid points nearest the location of the brown circled × in Figs. 2b–e, and from WRF Model output at 1130 UTC. The horizontal bars represent the range of the simulated total ice mass content in 0.5-km-altitude increments. Aircraft measurements of total ice mass content (red cross) and total LWC (blue cross) are shown for the descent profile occurring between 1122 and 1143 UTC. (a) Simulated values from the P3 scheme vs aircraft measurements. (b)–(d) As in (a), but for the MORR, 4ICE, and SBU schemes, respectively.

Fig. 3.

Mean vertical profiles of hydrometeor content calculated from the 100 grid points nearest the location of the brown circled × in Figs. 2b–e, and from WRF Model output at 1130 UTC. The horizontal bars represent the range of the simulated total ice mass content in 0.5-km-altitude increments. Aircraft measurements of total ice mass content (red cross) and total LWC (blue cross) are shown for the descent profile occurring between 1122 and 1143 UTC. (a) Simulated values from the P3 scheme vs aircraft measurements. (b)–(d) As in (a), but for the MORR, 4ICE, and SBU schemes, respectively.

Total ice content is significantly underpredicted in the MORR and 4ICE schemes below 3 km in height as a result of the schemes predicting less-organized bands with weaker vertical motions peaking below 0.2 m s−1 from 2.5 to 3.0 km compared to the aircraft measurements of around 0.3 m s−1 (not shown). Additionally, cloud ice mass is a nonnegligible fraction of the total mass in 4ICE at heights as low as 1.5 km where large snow aggregates were observed by the aircraft 2D-C [not shown; see Fig. 16 in Colle et al. (2017)], which suggests that cloud ice is overestimated in the 4ICE scheme. Although SBU predicted rather weak vertical motions less than 0.2 m s−1 (not shown), the very active snow depositional growth process in SBU leads to the overprediction in snow mass above about 2.5 km, and the efficient sedimentation of these snow particles explains the adequate comparison between SBU and the aircraft at low levels. Conversely, the P3 scheme predicted a good representation of the aircraft IWC, except below 1.5 km where the simulated peak is less than the observed peak exceeding 1.0 g m−3 as a result of the underprediction in vertical motions by about 0.2 m s−1, which is common among all the schemes.

We compare the exponential size distribution parameters estimated from the aircraft measurements against those calculated from model output to help explain the biases in the hydrometeor mass content profiles from each scheme (Fig. 4). The N0s and λs parameters are derived following the methodology in Heymsfield et al. (2002), which uses the observed particle size distribution to calculate fitting coefficients based on the first, second, and sixth moments). We also determine the mass-weighted mean diameter Dm from the aircraft measurements as , with representing the number of moments depending on the assumed or calculated particle fractal dimension bm. For determining Dm, we assume a bm = 2.0 representing unrimed snow, as minimal LWC was observed by the aircraft at this time. All the microphysical schemes in this study assign an exponential size distribution to characterize the particle size distributions of snow or precipitating ice, and exponential parameters were calculated from model output using the specific assumptions in each scheme. Note we use the subscript s for the P3 size distribution parameters in the following discussion despite referring to the single lumped ice category. The considerable decrease in the observed N0s (Fig. 4a) and λs (Fig. 4b) below 2.0 km to values less than 103 mm−1 m−3 and 0.5 mm−1, respectively, along with the increase in Dm to over 7.0 mm (Fig. 4c) indicates very efficient aggregation of snow particles in the low levels, which none of the schemes are able to fully represent. The P3 and MORR schemes predict excessive aggregation above 2 km as implied by the underestimation in N0s and overestimation in Dm for both schemes when compared to the observations, which is likely limiting the peak in aggregational growth in the low levels. Nevertheless, these schemes capture some of the observed aggregational growth below 2.0 km as indicated by the continual decrease in N0s and increase in Dm to about 1.3 km where the predicted snow mass shows minimal change. Size sorting, which occurs in double-moment schemes due to the different number- and mass-weighted sedimentation velocities in double-moment schemes (Milbrandt and Yau 2005), is also likely contributing to the continual decrease in N0s and increase in Dm shown in the P3 and MORR schemes. However, size sorting is likely a larger contributor above 2 km where a steady increase in terminal fall speeds is apparent as opposed to the lower levels where fall speeds are steady (not shown). The temperature-dependent N0s parameterization in SBU leads to a steadily decreasing profile, which fails to capture the aggregation effects in the low levels. Among the four schemes, SBU predicts the lowest λs above 3.0 km as a result of the very active snow depositional growth processes. The N0s, λs, and Dm profiles from the 4ICE scheme show large discrepancies when comparing to the aircraft estimates, which suggests refinements are needed to the temperature and mixing ratio parameterization used for diagnosing the size distribution parameters.

Fig. 4.

Mean profiles of exponential size distribution parameters calculated from the simulated snow hydrometeor species using the same averaging approach as in Fig. 3. Exponential size distribution parameters calculated from aircraft measurements of particle size information for ice crystals and aggregates are also displayed for comparison purposes (black cross). The minimum and maximum of each parameter is provided in 0.5-km increments up to 3.5 km, which is the altitude at which the microphysical schemes start predicting nonnegligible cloud ice amounts, especially 4ICE. Shown are the (a) size distribution intercept parameter N0s, (b) size distribution slope parameter λs, and (c) mass-weighted mean diameter Dm.

Fig. 4.

Mean profiles of exponential size distribution parameters calculated from the simulated snow hydrometeor species using the same averaging approach as in Fig. 3. Exponential size distribution parameters calculated from aircraft measurements of particle size information for ice crystals and aggregates are also displayed for comparison purposes (black cross). The minimum and maximum of each parameter is provided in 0.5-km increments up to 3.5 km, which is the altitude at which the microphysical schemes start predicting nonnegligible cloud ice amounts, especially 4ICE. Shown are the (a) size distribution intercept parameter N0s, (b) size distribution slope parameter λs, and (c) mass-weighted mean diameter Dm.

The ZDR is the logarithmic ratio of radar reflectivities at the horizontal and vertical polarizations, where slightly negative values can result from heavily rimed conical graupel falling in a vertically oriented manner (Hall et al. 1984; Aydin and Seliga 1984). Large, irregular-shaped hail and vertically aligned ice crystals are generally the only other hydrometeor types that influence negative ZDR (Straka et al. 2000). Graupel can also be associated with near-zero or positive ZDR values when the particles are more spherical in shape or falling in a horizontally oriented manner, but other hydrometeor types such as snow can influence similar values (Straka et al. 2000). The WKR measurements at 1224 UTC 18 February reveal a distinct narrow band of negative ZDR with values as low as −0.5 dB just to the north of the CARE site and aircraft spiral location (Fig. 5a). The negative band of ZDR coincides with the stronger reflectivity of the precipitation area where values are greater than 24 dBZ, which strongly suggests that the band is composed of a large concentration of conical graupel particles. P3 shows good qualitative agreement with the observed ZDR as a narrow band of enhanced rime/graupel mass exceeding 0.2 g m−3 is apparent along the southern portion of the simulated frontal band (Fig. 5b). Conversely, the MORR and 4ICE schemes simulate unorganized areas of small graupel amounts at the surface (Figs. 5c,d). Although peak graupel amounts in SBU are similar to P3, graupel extends from the northern to southern portions of the frontal band in SBU (Fig. 5e), which is a much larger area than the narrow band of negative ZDR observed by the WKR. It is possible that graupel is occurring in other regions of the observed precipitation area outside the negative ZDR band, particularly to the south of the band where more significant cloud water amounts, as shown in the simulations (Figs. 5b–e), coincide with near-zero or slightly positive ZDR values. Nevertheless, the WKR ZDR measurements help provide some evidence that the P3 scheme predicts the most realistic graupel band with cloud water path amounts exceeding 0.5 g m−2 along its southern boundary, while the other schemes predict overall less cloud water.

Fig. 5.

(a) WKR radar observations of ZDR (shaded in dB) and dBZ (maroon, thin contour at 18 dBZ and thick contour at 24 dBZ) from a PPI scan valid at 1224 UTC 18 Feb. Aircraft spiral flight pattern is denoted in red. (b) Simulated graupel/rime mass content at the surface (shaded) and cloud liquid water path (contoured at 0.01, 0.1, and 0.3 g m−2 in black and 0.5 and 0.8 g m−2 in red) from the P3 simulation. (c)–(e) As in (b), but for the MORR, 4ICE, and SBU simulations, respectively.

Fig. 5.

(a) WKR radar observations of ZDR (shaded in dB) and dBZ (maroon, thin contour at 18 dBZ and thick contour at 24 dBZ) from a PPI scan valid at 1224 UTC 18 Feb. Aircraft spiral flight pattern is denoted in red. (b) Simulated graupel/rime mass content at the surface (shaded) and cloud liquid water path (contoured at 0.01, 0.1, and 0.3 g m−2 in black and 0.5 and 0.8 g m−2 in red) from the P3 simulation. (c)–(e) As in (b), but for the MORR, 4ICE, and SBU simulations, respectively.

We further evaluate the WRF Model output at 1230 UTC 18 February against an aircraft spiral ascent occurring between 1223 and 1242 UTC by applying the same methodology used for the earlier aircraft descent (Fig. 6). P3 exhibits the strongest vertical motions among the simulations as the peak values greater than 0.5 m s−1 are only slightly lower than the aircraft measurements (Fig. 6a), while significant underpredictions are shown in the other schemes, especially 4ICE and SBU. As a result, these schemes also underpredict the observed low-level cloud water from the aircraft, while P3 shows a closer comparison with a peak near 0.5 g m−3 (Fig. 6b). Meanwhile, imagery from the 2D-C onboard the aircraft (Fig. 6c) and PVI at the surface (Fig. 6d) reveal precipitation dominated by heavily rimed, graupel-like particles, which provides further evidence that the P3 and SBU rime/graupel masses exceeding 0.1 g m−3 (Fig. 6b) are realistic. The heavily rimed particles were all less than about 1 mm in diameter according to the 2D-C imagery. The inability of the MORR and 4ICE schemes to produce significant graupel mass is primarily associated with their use of specified thresholds for converting between the predefined ice-phase categories of snow and graupel via the collection of cloud water. For example, both schemes require cloud water of at least 0.5 g m−3 to initiate this process of converting snow to graupel, and MORR and 4ICE struggle to attain this threshold in the simulations. Interestingly, the rime or graupel mass in SBU is only slightly lower than P3 despite the much higher vertical motions and cloud water in the P3 simulation, which is attributed to the prognostic versus diagnostic approaches. The diagnostic rime mass fraction parameterization in SBU uses the snow and cloud water mass to estimate the rime or graupel mass content at a grid point while P3 represents rime mass as a conserved prognostic variable that evolves in space and time via advection, subscale mixing, and microphysical tendencies. The fact that SBU generates moderate amounts of rime or graupel mass in a region of relatively small vertical motions suggests that the diagnostic approach is less realistic than the prognostic approach in P3.

Fig. 6.

Mean vertical profiles calculated from WRF Model output at 1230 UTC 18 Feb 2012 using the same averaging approach as in Fig. 3. Aircraft profiles are from the spiral ascent between 1223 and 1242 UTC. Aircraft measurements vs (a) simulated vertical velocity and (b) cloud liquid water. Simulated graupel/rime mass (dashed) is included in (b), but note these measurements were not available from the onboard instrumentation. (c) Aircraft ice habits from the 2DC near 1.4 km in height at 1225 UTC within the band and (d) those collected at the surface by the CARE site PVI (camera 1 with horizontal optical axis) at the same time are shown. 2DC has a buffer width of 960 μm and samples particles in size that range from 75 to 1900 μm at 30-μm resolution. Samples were acquired every 10 s. All particles observed by PVI were <5 mm in size.

Fig. 6.

Mean vertical profiles calculated from WRF Model output at 1230 UTC 18 Feb 2012 using the same averaging approach as in Fig. 3. Aircraft profiles are from the spiral ascent between 1223 and 1242 UTC. Aircraft measurements vs (a) simulated vertical velocity and (b) cloud liquid water. Simulated graupel/rime mass (dashed) is included in (b), but note these measurements were not available from the onboard instrumentation. (c) Aircraft ice habits from the 2DC near 1.4 km in height at 1225 UTC within the band and (d) those collected at the surface by the CARE site PVI (camera 1 with horizontal optical axis) at the same time are shown. 2DC has a buffer width of 960 μm and samples particles in size that range from 75 to 1900 μm at 30-μm resolution. Samples were acquired every 10 s. All particles observed by PVI were <5 mm in size.

Terminal fall speeds from the snow/ice categories in the schemes are evaluated against the MRR measurements at the CARE site (Fig. 7a). The MRR measured rather steady fall speeds between about 1.0 and 1.5 m s−1 during the period when snow aggregates were dominant prior to about 1200 UTC. Fall speeds increased to nearly 3.5 m s−1 as the observed band of negative ZDR associated with heavily rimed, conical graupel particles moved over the CARE site. The schemes show overall good agreement with MRR during the period of steady fall speeds, except for the significant underprediction in SBU where fall speeds are generally less than 0.5 m s−1. During the period of observed graupel, fall speeds in P3 increase to over 2.8 m s−1 with the arrival of the predicted band of rimed mass around 1230 UTC, while peak speeds are only 1.5 m s−1 in SBU. The snow categories (solid) from the MORR and 4ICE schemes have fall speeds consistently around 1.0 m s−1 during this 3-h time period. Graupel fall speeds are faster in these schemes (dashed lines in Fig. 7a; 2.5 m s−1 in 4ICE and 1.5 m s−1 in MORR), but have little impact because of their limited production. All schemes underpredict precipitation rates during this 3-h period from 1000 to 1300 UTC when comparing against the Pluvio 400 gauge measurements at the CARE site (Fig. 7b). Not surprisingly, P3 predicts the highest precipitation rates among the schemes with rates of 1.2 mm h−1 occurring from 1200 to 1300 UTC when the ice particles reached their peak terminal fall speed. Precipitation rates in P3 show the closest comparison to the observations from 1100 to 1200 UTC as the partially rimed snow aggregates (Fr ~ 0.05) are associated with higher terminal fall speeds than the dry snow particles simulated in the other schemes. Although precipitation rates are drastically unpredicted in SBU between 1100 and 1200 UTC, the scheme predicts a substantial increase to nearly 1.1 mm h−1 as the graupel band arrives at 1200 UTC. The lack of graupel production in MORR and 4ICE leads to the lowest precipitation rates from 1200 to 1300 UTC.

Fig. 7.

(a) Observed terminal fall speeds from the MRR at the CARE site vs the mean fall speeds from each scheme during the period of observed snow aggregates from about 1000 to 1200 UTC and graupel after 1200 UTC. Mean fall speeds from the schemes were calculated using the nearest 25 grid points to the brown circled ×s in Figs. 2b–e. The dashed lines refer to the fall speeds for graupel in MORR and 4ICE. A nonnegligible range in the simulated fall speeds at each time is indicated by a vertical bar. (b) Observed precipitation from the Pluvio 400 gauge at the CARE site vs simulated hourly precipitation for the same time period using same methodology as in (a).

Fig. 7.

(a) Observed terminal fall speeds from the MRR at the CARE site vs the mean fall speeds from each scheme during the period of observed snow aggregates from about 1000 to 1200 UTC and graupel after 1200 UTC. Mean fall speeds from the schemes were calculated using the nearest 25 grid points to the brown circled ×s in Figs. 2b–e. The dashed lines refer to the fall speeds for graupel in MORR and 4ICE. A nonnegligible range in the simulated fall speeds at each time is indicated by a vertical bar. (b) Observed precipitation from the Pluvio 400 gauge at the CARE site vs simulated hourly precipitation for the same time period using same methodology as in (a).

Overall, the P3 scheme produces the most realistic warm frontal band structure according to the WKR observations. The narrow band of heavily rimed, graupel-like precipitation simulated in P3 along with the associated vertical motions and cloud water amounts are confirmed by the GCPEx ground-based and aircraft instrumentation.

4. Detailed analysis of the microphysical schemes

a. Early genesis stage

To help understand the physical reasons for the differences between the simulated mature bands, we analyze the model output during the genesis stages. The NEXRAD sites observe the developing precipitation area advancing into the 1-km inner nest at 0700 UTC 18 February 2012 (Fig. 8a), and all schemes predict the precipitation slightly farther to the southwest (Figs. 8b–e). The P3 and MORR schemes show the closest comparison to the observations as the precipitation areas consist of numerous embedded convective cells with reflectivity >28 dBZ, but MORR appears to overestimate the areal coverage of these cells. Conversely, 4ICE and SBU predict less convective precipitation with weaker reflectivity.

Fig. 8.

As in Fig. 2, but zoomed out at 0700 UTC 18 Feb 2012 to capture the bands as they approached the GCPEx CARE site from the southwest. Radar reflectivity (shaded in dBZ) is a composite of 0.5°-elevation scans from KBUF, KCLE, KDTX, and KWSO along with the 0.3° scan from WKR. WSR-88Ds (black crosses) are also included in (a). Model output along the cross section (solid black) is displayed in Figs. 9 and 10. Brown circled × denotes the location used to produce profiles in Fig. 11.

Fig. 8.

As in Fig. 2, but zoomed out at 0700 UTC 18 Feb 2012 to capture the bands as they approached the GCPEx CARE site from the southwest. Radar reflectivity (shaded in dBZ) is a composite of 0.5°-elevation scans from KBUF, KCLE, KDTX, and KWSO along with the 0.3° scan from WKR. WSR-88Ds (black crosses) are also included in (a). Model output along the cross section (solid black) is displayed in Figs. 9 and 10. Brown circled × denotes the location used to produce profiles in Fig. 11.

Cross sections of the saturation equivalent potential temperature [, where θ is the potential temperature, L is the latent heat of vaporization, qυs is the saturation mixing ratio, cp is the specific heat of dry air at constant pressure, and T is the air temperature], saturated moist potential vorticity (, where g is the gravitational acceleration and ζa is the absolute vorticity), and horizontal wind through the simulated precipitation areas are shown in Fig. 9. Cells with reflectivity values >18 dBZ are denoted by the gray dashed lines to highlight the location of the precipitation area. Although all the simulations show some low- and midlevel instability approaching the precipitation area from the southwest, P3 predicts a steeper-sloping, more vertically inclined warm front in the lowest 1.5 km from about 40 to 100 km along the cross section, which influences a greater amount of instability in this area. Observations are limited at this time as a result of the precipitation being located about 200 km to the southwest of the GCPEx field campaign site. Nevertheless, the 2-m temperature measurements from the automated surface observation stations to the south of the warm front suggest that the MORR, 4ICE, and SBU schemes are ~1°–2°C too cool, while P3 predicts an area of 4°C temperatures protruding into the domain similar to the observations.

Fig. 9.

(a) Cross section (SW–NE) across the precipitation area in Fig. 2 showing MPV* (shaded), (red, contoured every 2 K), horizontal winds, and simulated reflectivity >18 dBZ (gray contours). MPV* has units of 10−6 K kg−1 m2 s−1. (b)–(d) As in (a), but for cross sections in the MORR, 4ICE, and SBU simulations, respectively. Topography is shown in black.

Fig. 9.

(a) Cross section (SW–NE) across the precipitation area in Fig. 2 showing MPV* (shaded), (red, contoured every 2 K), horizontal winds, and simulated reflectivity >18 dBZ (gray contours). MPV* has units of 10−6 K kg−1 m2 s−1. (b)–(d) As in (a), but for cross sections in the MORR, 4ICE, and SBU simulations, respectively. Topography is shown in black.

To help explain the differences in the frontal slope and instability between the schemes, we assessed the total accumulated latent heating–cooling due to the following microphysical processes: deposition, sublimation, freezing, melting, condensation, and evaporation. At each model time step (4 s), the thermodynamic tendencies from these six terms were added to those at the previous time step to calculate the total accumulated latent heating/cooling throughout each simulation, and these terms were output every 30 min for analysis. Figure 10 shows the differences in accumulated latent cooling rate due to both sublimation and melting between the schemes overlaid with the heating rate due to deposition from 0600 to 0630 UTC (left panels) and the temperature differences at 0700 UTC (right panels). There is a dipole structure in the difference in latent cooling rate between the P3 and MORR schemes (P3–MORR; Figs. 10a) as the strongest cooling signal exceeding 1.4 K h−1 in MORR is below 0.5 km in height at a horizontal distance of around 50 km (back or southern edge of the precipitation area) whereas the P3 is lofted between 0.5 and 1.0 km farther northeast along the cross section. The signals above the 0°C isotherms in the P3 (solid white) and MORR (dashed white) schemes are due entirely to sublimation, while melting is significant in MORR below this level. This stronger latent cooling farther to the southwest in MORR promotes cooler air along the front and in the warm sector compared to P3 as indicated by the positive P3–MORR temperature differences of greater than 1.4 K along this section (Fig. 10b). Therefore, MORR predicts a weaker low-level temperature gradient in the developing band region, which results in a weaker frontal slope and less low-level instability near 100 km along the section than in the P3 (i.e., Figs. 9a,b). The differences in vertical motion (e.g., circulation vectors in Fig. 10b) and depositional heating (Fig. 10a) between the schemes influences an opposite temperature pattern from 2.0 to 3.5 km in height, where P3 is from 0.4 to 0.8 K warmer than MORR along the leading (northern) edge of the precipitation band while MORR is warmer than P3 by similar values along the back (southern) edge.

Fig. 10.

Cross section across the precipitation area in Fig. 8 showing the difference in (a) latent cooling rate due to sublimation and melting (shaded) and the difference in latent heating rate due to deposition (contoured every 0.4 K h−1, with positive and negative values in yellow and red, respectively) between the P3 and MORR schemes for the 30-min period from 0600 to 0630 UTC 18 Feb 2012. The freezing-level isotherms (0°C) for P3 and MORR are denoted by solid and dashed magenta contours, respectively. (b) As in (a), but showing the temperature difference between the P3 and MORR schemes at 0700 UTC. (c)–(f) As in (a),(b), but showing the difference between the (c),(d) P3 and 4ICE schemes and (e),(f) P3 and SBU schemes. Freezing-level isotherms for 4ICE and SBU are denoted by dashed magenta contours in (c) and (e).

Fig. 10.

Cross section across the precipitation area in Fig. 8 showing the difference in (a) latent cooling rate due to sublimation and melting (shaded) and the difference in latent heating rate due to deposition (contoured every 0.4 K h−1, with positive and negative values in yellow and red, respectively) between the P3 and MORR schemes for the 30-min period from 0600 to 0630 UTC 18 Feb 2012. The freezing-level isotherms (0°C) for P3 and MORR are denoted by solid and dashed magenta contours, respectively. (b) As in (a), but showing the temperature difference between the P3 and MORR schemes at 0700 UTC. (c)–(f) As in (a),(b), but showing the difference between the (c),(d) P3 and 4ICE schemes and (e),(f) P3 and SBU schemes. Freezing-level isotherms for 4ICE and SBU are denoted by dashed magenta contours in (c) and (e).

Sublimation and melting results are similar in 4ICE and MORR in the lowest 1 km of the atmosphere, which leads to a frontal slope and low-level instability that compares closely to that of MORR. However, the latent cooling differences between P3 and 4ICE (P3–4ICE; Fig. 10c) show negative latent cooling signals of −0.4 to −0.8 K h−1 above 1 km in height that are absent in P3–MORR because of enhanced sublimation in 4ICE. As a result, the P3–4ICE temperature differences (Fig. 10d) depict warm perturbations of 0.4–0.8 K at these higher heights from 0 to 50 km along the cross section, and these cooler temperatures in 4ICE promote a reduction in midlevel instability approaching the precipitation band (Fig. 9c) compared to MORR (Fig. 9b). It is also important to note that P3–4ICE shows a reduction in temperature differences from about 2.0 to 3.5 km in height within the precipitation band compared to P3–MORR, as depositional heating rates are more comparable between these schemes. This suggests that the enhanced sublimation and melting in MORR is not directly related to the more active depositional growth processes, but rather tied to the physical characteristics (density, terminal fall speeds, etc.) of the ice-phase particles.

The SBU scheme predicts stronger latent heating/cooling rates than in P3 above 3.0 km in height as indicated by the P3–SBU heating and cooling rate differences exceeding −1.2 and −0.8 K h−1, respectively, at a horizontal distance of 30–80 km along the section (Fig. 10e). Although depositional heating is stronger in SBU than in P3, vertical motions are very similar between the schemes (<0.01 m s−1; not shown) as shown by the minimal differences in the circulation vectors at these heights (Fig. 10f). As a result, depositional heating outweighs sublimation cooling in SBU, which promotes mostly warmer temperatures above 3.5 km in height in SBU compared to P3, as indicated by the negative P3–SBU temperature differences of −0.2 to −0.8 K in this region (Fig. 10f). Interestingly, positive P3–SBU temperature differences exceeding 0.8 K are evident to the southwest of the band from 0 to 80 km along the cross section, despite the lack of a strong latent cooling signal in the low levels in SBU. We found that the cooler temperatures in SBU are due to low-level sublimation and melting processes that occurred just to the south of this cross section, and southerly winds in the lowest 1 km of the atmosphere propagated the cooler air into the section. Thus, SBU predicts a stable low-level environment protruding into the precipitation band from the southwest along with a weak frontal slope (Fig. 9d).

We also calculated the mean hydrometeor mass, particle density, terminal fall speed, and diameter (not shown) from convective cells near the back edge of the precipitation area [brown circle with crisscross (×) in Figs. 8b,e] from the WRF Model output at 0630 UTC 18 February (Fig. 11). To better highlight the low-level latent cooling occurring in the SBU scheme, we used grid points slightly to the southeast of the other schemes. Total ice content in P3 (Fig. 11a) undergoes minimal mass decreases throughout the profile, except in a very shallow layer at about 0.7 km in height where above freezing temperatures promote melting and sublimation of mostly unrimed mass. Below this shallow layer, the total ice mass consists entirely of rimed mass. On the other hand, MORR predicts substantial decreases in the graupel mass below about 1.0 km in height (Fig. 11b) as a result of the enhanced sublimation highlighted in Fig. 10a. In the lowest 0.5 km of the atmosphere, both the snow and graupel mass show drastic decreases as melting processes play a large role in this near-surface layer as indicated by the increase in rain mass to about 0.08 g m−3. The total ice content in 4ICE is dominated by snow (Fig. 11c) as a result of the lower cloud water amounts of 0.1 g m−3 with a decreasing trend in the snow mass below about 2 km in height from the enhanced sublimation processes discussed in Fig. 10c. Melting processes are even more active in 4ICE compared to MORR in the lowest 0.5 km of the profile as the decrease in snow mass below 0.01 g m−3 is associated with an increase in rain mass to 0.1 g m−3. For SBU, the peak snow mass (Fig. 11d) together with a snow diameter of 2 mm (not shown) at 3.5 km in height suggests snow depositional growth processes are active while the significant decrease in snow mass above this height layer is related to the sublimation shown in Fig. 10e. Minimal rain mass is predicted above 0.5 km in SBU because of the more efficient autoconversion of rain compared to the other schemes. Sublimation also leads to significant decreases in total ice in the layer from 2.8 to 3.4 km while strong melting in the near-surface layer causes reductions in the snow and graupel mass and, consequently, increases in rain mass comparable to MORR.

Fig. 11.

(a) Mean vertical profiles of simulated hydrometeor mass calculated from WRF Model output at 0630 UTC 18 Feb 2012 from 50 grid points within the precipitation area near the southwest portion of the cross section in Fig. 8. (b)–(d) As in (a), but for MORR, 4ICE, and SBU, respectively. (e) Mean vertical profiles of mass-weighted mean particle density from each scheme. (f) As in (e), but for terminal fall speed. Letters I, PI, S, and G in parentheses after the scheme names indicate the ice, precipitating ice, snow, and graupel categories, respectively.

Fig. 11.

(a) Mean vertical profiles of simulated hydrometeor mass calculated from WRF Model output at 0630 UTC 18 Feb 2012 from 50 grid points within the precipitation area near the southwest portion of the cross section in Fig. 8. (b)–(d) As in (a), but for MORR, 4ICE, and SBU, respectively. (e) Mean vertical profiles of mass-weighted mean particle density from each scheme. (f) As in (e), but for terminal fall speed. Letters I, PI, S, and G in parentheses after the scheme names indicate the ice, precipitating ice, snow, and graupel categories, respectively.

Mass-weighted mean particle density (Fig. 11e) and terminal fall speed (Fig. 11f) profiles help explain the differences in the sublimation and melting between the schemes. After losing the remainder of the unrimed mass in the shallow layer at 0.7 km (Fig. 11a), the rimed particles in P3 show rapid increases in density and fall speed to values of 900 kg m−3 and about 5.5 m s−1, respectively, which limit further melting and sublimation. Conversely, MORR assumes constant snow (solid) and graupel (dashed) densities of 100 and 400 kg m−3, respectively, when separating between these hydrometeor types, which influences rather consistent fall speeds of 1.2 and 1.8 m s−1 for the snow and graupel profiles that permit sufficient time for enhanced melting and sublimation. Although snow and graupel densities are allowed to vary in 4ICE, graupel mass is negligible while snowfall speeds of less than 1.0 m s−1 help promote stronger sublimation and melting than in MORR. The slight reduction in overall sublimation and melting in the SBU scheme compared to MORR and 4ICE is related to the larger particle densities and terminal fall speeds of the precipitating ice category in Figs. 11e,f. Note that the primary reason for the smaller total ice mass in P3 and SBU is linked to the fact that a considerable fraction of the total ice consists of rime/graupel particles, which helps limit the mass content aloft through efficient sedimentation and fall out of particles. On the contrary, the inefficient sedimentation of the slowly falling snow particles in MORR and 4ICE helps promote more active depositional growth and mass contents aloft.

b. Late genesis stage

By 1000 UTC 18 February 2012, the well-defined band extends several hundred kilometers from the west-northwest to east-southeast across the region with reflectivity exceeding 32 dBZ (Fig. 12a), and the P3 scheme predicts the band structure reasonably well (Fig. 12b). Although more organized than in the early genesis stage, MORR predicts some convective-like cells with reflectivity >24 dBZ along the southwest portion of the precipitation area, instead of a well-defined band, as in P3 (Fig. 12c). Conversely, limited development occurs in 4ICE during this time period as the precipitation area has no banded structure with reflectivity generally below 28 dBZ (Fig. 12d). A fairly ragged band with reflectivity exceeding 28 dBZ has developed in SBU, but the overall appearance of the frontal precipitation is quite different from the observations as widespread cells are shown well to the southwest of the band.

Fig. 12.

As in Fig. 8, but at 1000 UTC 18 Feb 2012. Brown circled × denotes the location used to produce Fig. 15.

Fig. 12.

As in Fig. 8, but at 1000 UTC 18 Feb 2012. Brown circled × denotes the location used to produce Fig. 15.

We utilize the same south-southwest–north-northeast cross sections as shown for the early genesis stage to highlight the differences in band structure, frontogenetical forcing, and stability. By this time, the organized band in P3 is associated with a frontogenetical layer sloping upward from near the surface to 1.5 km with peak magnitudes reaching 10 K (100 km)−1 h−1 below 1.0 km in height (Fig. 13a). This layer of frontogenetical forcing promotes the slightly steeper warm frontal slope (Fig. 13b) than shown during the early genesis stage. The MORR (Fig. 13c), 4ICE (Fig. 13e), and SBU (Fig. 13g) schemes predict less-organized structure to the frontal precipitation with numerous cells developing to the southwest of the stronger reflectivity core along with more gentle frontal slopes (Figs. 13d,f,h) associated with frontogenetical forcing of generally less than 6 K (100 km)−1 h−1 (Figs. 13c,e,g). Furthermore, P3 predicts a more pronounced midlevel layer of conditional instability (CI; negative MPV*) with warmer, drier air extending farther northeast along the cross section than in the other schemes, as indicated by the 291- and 292-K contours. The midlevel layer of CI in SBU (Fig. 13h) is associated with a shallow warm, dry tongue with a depth of only about 0.5 km according to the 291-K contour. A well-defined layer of CI extends from the near-surface layer upward to over 2.0 km in height as it impinges upon the frontal band in MORR (Fig. 13d), while the 4ICE scheme also shows some enhanced low-level instability (Fig. 13f) in this same layer.

Fig. 13.

(a) Cross section across the frontal band in Fig. 12 showing simulated reflectivity (shaded), potential temperature (gray, contoured every 2 K), Petterssen frontogenesis [red, contoured every 4 K (100 km)−1 h−1 starting at 2 K (100 km)−1 h−1], and three-dimensional wind circulation vectors (vertical velocity multiplied by factor of 20 relative to horizontal velocity). (b) As in (a), but showing MPV* (shaded), (red, contoured every 2 K with additional contour at 281 K), and horizontal winds. (c)–(h) As in (a),(b), but for the (c),(d) MORR; (e),(f) 4ICE; and (g),(h) SBU simulations.

Fig. 13.

(a) Cross section across the frontal band in Fig. 12 showing simulated reflectivity (shaded), potential temperature (gray, contoured every 2 K), Petterssen frontogenesis [red, contoured every 4 K (100 km)−1 h−1 starting at 2 K (100 km)−1 h−1], and three-dimensional wind circulation vectors (vertical velocity multiplied by factor of 20 relative to horizontal velocity). (b) As in (a), but showing MPV* (shaded), (red, contoured every 2 K with additional contour at 281 K), and horizontal winds. (c)–(h) As in (a),(b), but for the (c),(d) MORR; (e),(f) 4ICE; and (g),(h) SBU simulations.

To understand the physical reasons for the differences between the schemes, we assess the accumulated total latent heating rates due to all microphysical processes for 30-min intervals between 0830 and 1000 UTC 18 February along this cross section (Figs. 14a,c,e). The total latent heating rate calculations include positive contributions from deposition, condensation, and freezing along with negative contributions from sublimation, evaporation, and melting. Note that the freezing contributions are very minimal (not shown). Strong evaporational cooling persists near the top of the low-level cloud layer in the warm sector in MORR, which leads to the shallow layer of positive P3–MORR latent heating rate differences exceeding 0.8 K h−1 to the southwest of the band (0–100 km along section) that slope upward between 1.5 and 2.5 km in height (Fig. 14a). As a result, the P3–MORR temperature difference cross section at 0930 UTC (Fig. 14b) shows positive perturbations of 0.4–1.4 K in this same layer. This strong cooling signal in MORR is associated with the use of saturation adjustment in the scheme, which can lead to excessive cloud water evaporation near cloud top (Grabowski and Morrison 2008). P3 does not use saturation adjustment, but rather explicitly predicts condensation and evaporation that helps to limit spurious cloud water evaporation at cloud top. Positive P3–MORR latent heating rate differences of 0.2–0.8 K due to stronger depositional heating in P3 are apparent from 1.5 to 3.5 km in height at a horizontal distance of 150–200 km along the section, and this warm signal developed after the strong cloud water evaporation in MORR. Therefore, we can conclude that the enhanced cloud water evaporation in MORR promoted cooler, more humid air associated with less midlevel instability (lower ) to impinge upon the frontal band from the southwest (Fig. 13d), which led to weaker vertical motions and depositional heating within the band compared to the P3 simulation. Furthermore, the strong cloud-top evaporation in MORR promoted the development of the low-level CI layer by causing a steeper lapse rate in this narrow, upward-sloping layer, which led to low-level vertical motions and convective cells to the southwest of the precipitation core. The frontal slope in MORR continues to be more gentle than in P3 because of the sublimation and melting during the early genesis stage as a cool signal persists below 1 km in height around 150 km into the P3–MORR cross section (Fig. 14b). Strong condensational heating in MORR near 150 km into the section (Fig. 14a) produces minimal temperature perturbations as vertical motions and adiabatic cooling offset the heating in this layer.

Fig. 14.

As in Fig. 13, but showing the difference in (a) total latent heating rate due to all microphysical processes (e.g., deposition, sublimation, condensation, evaporation, freezing, and melting) between the P3 and MORR schemes for the 30-min period from 0830 to 0900 UTC 18 Feb 2012. Solid black and white contours highlight differences in latent cooling due to evaporation for rates <−0.5 and >0.5 K h−1, respectively. (b) As in (a), but showing the temperature (shaded) and wind vector differences between the P3 and MORR schemes at 0930 UTC. (c)–(f) As in (a),(b), but showing the difference between the (c),(d) P3 and 4ICE schemes and (e),(f) P3 and SBU schemes. Note that (e) is for the 30-min period from 0900 to 0930 UTC.

Fig. 14.

As in Fig. 13, but showing the difference in (a) total latent heating rate due to all microphysical processes (e.g., deposition, sublimation, condensation, evaporation, freezing, and melting) between the P3 and MORR schemes for the 30-min period from 0830 to 0900 UTC 18 Feb 2012. Solid black and white contours highlight differences in latent cooling due to evaporation for rates <−0.5 and >0.5 K h−1, respectively. (b) As in (a), but showing the temperature (shaded) and wind vector differences between the P3 and MORR schemes at 0930 UTC. (c)–(f) As in (a),(b), but showing the difference between the (c),(d) P3 and 4ICE schemes and (e),(f) P3 and SBU schemes. Note that (e) is for the 30-min period from 0900 to 0930 UTC.

Enhanced cloud water evaporation is also occurring in 4ICE as shown in the P3–4ICE cross section where plumes of positive latent heating rate differences exceeding 0.8 K h−1 appear from 1.5 to 3.0 km along the cross section between 30 and 110 km (Fig. 14c). However, evaporation in 4ICE occurs in more sporadic plumes rather than a distinct, upward-sloping layer as in MORR. Sublimation in 4ICE is also contributing to the positive latent heating rate differences in P3–4ICE, but at slightly higher heights (2.0–3.0 km) than evaporation. As a result, the P3–4ICE temperature differences show positive perturbations extending above 2.5 km in height while those in P3–MORR are constrained to lower heights (Fig. 14d). These differences are linked to variations in the saturation adjustment techniques in each scheme as saturation adjustment in MORR only impacts evaporation while both evaporation and sublimation are impacted in 4ICE. Additionally, 4ICE applies a vertical velocity constraint to limit spurious evaporation and sublimation, which is likely the reason for the slightly reduced positive temperature perturbations in P3–4ICE. Similar to MORR, the near-surface melting and sublimation in 4ICE during the early genesis stage continues to promote the cooler temperatures around 150 km along the cross section.

The active snow depositional growth processes in SBU promote snow mass contents near 0.1 g m−3 around 3.5 km in height (not shown) that cause the negative latent heating rate differences exceeding 1.4 K h−1 at these heights along the P3–SBU cross section from 50 to 100 km (Fig. 14e). Efficient sedimentation of these snow particles into the warmer, drier midlevel layer below leads to sublimation cooling, as indicated by the positive P3–SBU latent heating rates of 0.2–0.4 K h−1 around 2.5 km in height. This layer of sublimation persists for a 2-h period from 0800 to 1000 UTC as a result of snow depositional processes remaining active during this time, which promotes the positive temperature differences of 0.4–0.8 K h−1 in this same layer along the P3–SBU cross section (Fig. 14f). The strong positive rates exceeding 2.0 K h−1 from 150 to 250 km along the P3–SBU latent heating cross section associated with stronger depositional heating within the frontal band in P3 appeared after the development of the midlevel sublimation layer in SBU. Thus, we can infer that the snow sublimation in SBU is leading to cooler, more humid (less unstable) midlevel air impinging upon the frontal band, which helps decrease the instability and depositional growth processes within the band compared to that in P3. The near-surface cooler air associated with the melting and sublimation during the early genesis stage in SBU extends much farther into the warm sector than in MORR and 4ICE, which explains why the frontal slope in SBU is the most gentle among the schemes (Fig. 13h). Overall, SBU has the most stable environment within the frontal band (Fig. 13h) among the schemes because of the lack of a low-level instability layer combined with the reduction in midlevel instability.

We compare mean vertical profiles of θe, , and winds calculated from 13-km RUC analysis data at 1000 UTC 18 February near 43.5°N, 81.6°W [brown circles with crisscrosses (×s) in Fig. 12] to those calculated from model output at the same time and location (Fig. 15). This location was chosen to help determine the scheme that predicted the most realistic instability profiles just to the southwest of the frontal band. All schemes predict a stronger midlevel dry intrusion impinging upon the frontal band than shown in the RUC profile, as indicated by the larger difference between θe and in the WRF profiles. Although the P3 scheme has the largest dry bias, the simulated profile shows that the closest comparison to the RUC profile as a stable layer from 1.5 to 2 km resides beneath a CI layer (decreasing ) from about 2 to 3 km in height (Fig. 15a). Latent cooling processes (i.e., evaporation and sublimation) are mostly inactive to the southwest of the frontal band in P3, which leads to the most representative profile, but also the largest dry bias. Conversely, the excessive evaporational cooling in MORR leads to the shallow, moist unstable layer from about 1.8 to 2.3 km in height (Fig. 15b) and, consequently, the strong stable layer above this level. The 4ICE profile also shows a shallow, moist unstable layer below a drier stable layer, but they are not as prominent as in MORR due to the weaker latent cooling processes. Latent cooling processes are mostly inactive below about 2 km in height in the SBU scheme, which leads to the low-level stability. However, SBU predicts a more lofted, drier CI layer than the MORR and 4ICE schemes due to the persistent snow sublimation occurring from about 2 to 3 km in height (Fig. 15d).

Fig. 15.

(a) Mean vertical profiles of θe (solid), θe* (dashed), and winds (1 full barb = 10 kt) from the P3 and RUC analysis at 1000 UTC 18 Feb 2012. The P3 profiles were calculated from the 100 grid points closest to the brown circled × in Fig. 12 (just to the southwest of the frontal band) while only 4 grid points were used for calculating the RUC profiles. (b)–(d) As in (a), but for MORR, 4ICE, and SBU, respectively.

Fig. 15.

(a) Mean vertical profiles of θe (solid), θe* (dashed), and winds (1 full barb = 10 kt) from the P3 and RUC analysis at 1000 UTC 18 Feb 2012. The P3 profiles were calculated from the 100 grid points closest to the brown circled × in Fig. 12 (just to the southwest of the frontal band) while only 4 grid points were used for calculating the RUC profiles. (b)–(d) As in (a), but for MORR, 4ICE, and SBU, respectively.

5. Summary and conclusions

We evaluated the P3, MORR, SBU, and 4ICE cloud microphysical schemes in simulating an intense warm frontal snowband on 18 February 2012 using aircraft and surface instrumentation deployed during the GCPEx field campaign over southern Ontario. Enhanced vertical motions and cloud water in the P3 simulation led to an organized, well-defined frontal band with a moderate band of rime mass along its southern boundary. The large rime mass fractions within this band indicated the abundance of heavily rimed, graupel-like particles, which was confirmed by aircraft and radar measurements. Although the SBU simulated an overall weaker, less-organized frontal band than P3, it was still able to produce a moderate band of graupel mass. However, the areal coverage of light-to-moderate graupel mass in SBU extended from the southern to northern portions of the frontal band, rather than the narrow, linear band of rime mass in P3. Conversely, graupel production was minimal in the MORR and 4ICE simulations as snow precipitation dominated the unorganized frontal bands.

Simulated microphysical processes and associated latent heating–cooling impacted the frontal band development and led to the differences in the band intensity and structure between the schemes. During the early genesis stage, enhanced low-level sublimation and melting occurred in the MORR and 4ICE simulations as a result of the falling precipitation being dominated by snow particles with low densities and fall speeds. Even though rime fractions were significantly higher in SBU compared to MORR and 4ICE, the graupel production led to only a small increase in particle densities and fall speeds, which allowed for enhanced sublimation and melting in the low levels. Sublimation and melting were limited in P3 because of the production of rimed particles with high densities and fall speeds in the low levels. As a result, P3 showed greater low-level instability and a stronger horizontal temperature gradient than the other schemes, which promoted frontal band development during the early genesis stage. The saturation adjustment scheme in MORR and 4ICE helped further limit the frontal band development during the late genesis stage by causing excessive cloud water evaporation at the low-level cloud-top layer in the warm sector. This cooler, more humid air associated with the evaporative layer reduced the midlevel instability approaching the frontal band in MORR and 4ICE. The saturation adjustment scheme in 4ICE accounts for both sublimation and evaporation rather than only evaporation as in MORR, which allowed for sublimation at cloud top to contribute to the cooling signal in 4ICE. The P3 scheme avoids the dependence on saturation adjustment by explicitly calculating the cloud water condensation/evaporation, which limits the cloud-top evaporational cooling. Active snow depositional growth processes in the midlevels of the warm sector in SBU promoted sedimentation and sublimation cooling below, which led to a reduction in the midlevel instability approaching the frontal band. Overall, the microphysical processes in MORR, 4ICE, and SBU inhibited band development by promoting a less unstable environment with weaker frontogenetical forcing and vertical motions, while a more unstable environment with stronger forcing was maintained in the P3 simulation.

Despite having an environment suited for frontal band development, the SBU produced a moderate band of graupel, while snow dominated the MORR and 4ICE simulations. The MORR and 4ICE schemes rely on specified thresholds for converting between the predefined ice-phase categories of cloud ice, snow, and graupel, which led to the production of mostly snow rather than a defined graupel band. On the other hand, the diagnostic approach for calculating the ice particle properties used in the SBU scheme helped produce the moderate band of graupel. However, SBU predicted graupel in regions of weak frontogenetical forcing and vertical motions, which was not confirmed by the field measurements. Similar to the observations, P3 predicted a narrow, linear band of rime mass in a region of strong frontogenetical forcing and vertical motions. This suggests that the prognostic approach for tracking the ice particle properties in the P3 scheme is more realistic than the diagnostic approach.

Acknowledgments

Model simulations were conducted on the NASA Discover Cluster. This work was supported by National Aeronautics and Space Administration Grant NNX13AF88G. We appreciate the comments and suggestions by the editor and the three anonymous reviewers, who helped improve several aspects of this paper.

REFERENCES

REFERENCES
Aydin
,
K.
, and
T. A.
Seliga
,
1984
:
Radar polarimetric backscattering properties of conical graupel
.
J. Atmos. Sci.
,
41
,
1887
1892
, doi:.
Blivens
,
F.
,
2014
: GPM ground validation Precipitation Video Imager (PVI) GCPEX. NASA Global Hydrology Center DAAC, Huntsville, AL, accessed 1 August 2015, https://dx.doi.org/10.5067/GPMGV/GCPEX/PVI/DATA301.
Brown
,
P. R. A.
, and
P. N.
Francis
,
1995
:
Improved measurements of the ice water content in cirrus using a total water probe
.
J. Atmos. Oceanic Technol.
,
12
,
410
414
, doi:.
Colle
,
B. A.
,
M.
Garvert
,
J. B.
Wolfe
,
C. F.
Mass
, and
C. P.
Woods
,
2005
:
The 13–14 December 2001 IMPROVE-2 event. Part III: Simulated microphysical budgets and sensitivity studies
.
J. Atmos. Sci.
,
62
,
3535
3558
, doi:.
Colle
,
B. A.
,
A.
Naeger
, and
A.
Molthan
,
2017
:
Structure and evolution of a warm frontal precipitation band during the GPM Cold Season Precipitation Experiment (GCPEx)
.
Mon. Wea. Rev.
,
145
,
473
493
, https://doi.org/10.1175/MWR-D-16-0072.1.
Delene
,
D.
, and
M. R.
Poellot
,
2016
: GPM ground validation UND Citation cloud microphysics GCPEX V2. NASA Global Hydrology Center DAAC, Huntsville, AL, accessed 1 August 2015, https://dx.doi.org/10.5067/GPMGV/GCPEX/MULTIPLE/DATA203.
Grabowski
,
W. W.
, and
H.
Morrison
,
2008
:
Toward the mitigation of spurious cloud-edge supersaturation in cloud models
.
Mon. Wea. Rev.
,
136
,
1224
1234
, doi:.
Grell
,
G. A.
, and
S.
Freitas
,
2014
:
A scale and aerosol aware stochastic convective parameterization for weather and air quality modeling
.
Atmos. Chem. Phys.
,
14
,
5233
5250
, doi:.
Hall
,
M. P. M.
,
J. W. F.
Goddard
, and
S. M.
Cherry
,
1984
:
Identification of hydrometeors and other targets by dual-polarization radar
.
Radio Sci.
,
19
,
132
140
, doi:.
Heymsfield
,
A. J.
,
A.
Bansemer
,
P. R.
Field
,
S. L.
Durden
,
J. L.
Smith
,
J. E.
Dye
,
W.
Hall
, and
C. A.
Grainger
,
2002
:
Observations and parameterizations of particle size distributions in deep tropical cirrus and stratiform precipitating clouds: Results from in situ observations in TRMM field campaigns
.
J. Atmos. Sci.
,
59
,
3457
3491
, doi:.
Heymsfield
,
A. J.
,
A.
Bansemer
,
C.
Schmitt
,
C.
Twohy
, and
M. R.
Poellot
,
2004
:
Effective ice particle densities derived from aircraft data
.
J. Atmos. Sci.
,
61
,
982
1003
, doi:.
Heymsfield
,
A. J.
,
A.
Bansemer
, and
M. R.
Poellot
,
2014
: GPM ground validation NCAR cloud microphysics particle probes GCPEX. NASA Global Hydrology Center DAAC, Huntsville, AL, accessed 5 January 2016, https://dx.doi.org/10.5067/GPMGV/GCPEX/MULTIPLE/DATA201.
Hong
,
S.-Y.
,
K.-S. S.
Lim
,
Y.-H.
Lee
,
J.-C.
Ha
,
H.-W.
Kim
,
S.-J.
Ham
, and
J.
Dudhia
,
2010
:
Evaluation of the WRF double-moment six-class microphysics scheme for precipitating convection
.
Adv. Meteor.
,
2010
, 707253, https://dx.doi.org/10.1155/2010/707253.
Houze
,
R. A.
, Jr.
,
P. V.
Hobbs
,
P. H.
Herzegh
, and
D. B.
Parsons
,
1979
:
Size distributions of precipitation particles in frontal clouds
.
J. Atmos. Sci.
,
36
,
156
162
, doi:.
Hudak
,
D.
,
2013
: GPM ground validation dual polarized C-band Doppler radar King City GCPEX. NASA Global Hydrology Center DAAC, Huntsville, AL, accessed 1 August 2015, https://dx.doi.org/10.5067/GPMGV/GCPEX/CBAND/DATA101.
Igel
,
A. L.
,
M. R.
Igel
, and
S. C.
van den Heever
,
2015
:
Make it a double? Sobering results from simulations using single-moment microphysics schemes
.
J. Atmos. Sci.
,
72
,
910
925
, doi:.
Lang
,
S.
,
W.-K.
Tao
,
R.
Cifelli
,
W.
Olson
,
J.
Halverson
,
S.
Rutledge
, and
J.
Simpson
,
2007
:
Improving simulations of convective system from TRMM LBA: Easterly and westerly regimes
.
J. Atmos. Sci.
,
64
,
1141
1164
, doi:.
Lang
,
S.
,
W.-K.
Tao
,
X.
Zeng
, and
Y.
Li
,
2011
:
Reducing the biases in simulated radar reflectivities from a bulk microphysics scheme: Tropical convective systems
.
J. Atmos. Sci.
,
68
,
2306
2320
, doi:.
Lang
,
S.
,
W.
Tao
,
J.
Chern
,
D.
Wu
, and
X.
Li
,
2014
:
Benefits of a fourth ice class in the simulated radar reflectivities of convective systems using a bulk microphysics scheme
.
J. Atmos. Sci.
,
71
,
3583
3612
, doi:.
Lin
,
Y.
, and
B. A.
Colle
,
2009
:
The 4–5 December 2001 IMPROVE-2 event: Observed microphysics and comparisons with the Weather Research and Forecasting Model
.
Mon. Wea. Rev.
,
137
,
1372
1392
, doi:.
Lin
,
Y.
, and
B. A.
Colle
,
2011
:
A new bulk microphysical scheme that includes riming intensity and temperature-dependent ice characteristics
.
Mon. Wea. Rev.
,
139
,
1013
1035
, doi:.
Lin
,
Y.
,
L. J.
Donner
, and
B. A.
Colle
,
2011
:
Parameterization of riming intensity and its impact on ice fall speed using ARM data
.
Mon. Wea. Rev.
,
139
,
1036
1047
, https://doi.org/10.1175/2010MWR3299.1.
Maahn
,
M.
, and
P.
Kollias
,
2012
:
Improved Micro Rain Radar snow measurements using Doppler spectra post-processing
.
Atmos. Meas. Tech.
,
5
,
2661
2673
, doi:.
Milbrandt
,
J. A.
, and
M. K.
Yau
,
2005
:
A multimoment bulk microphysics parameterization. Part I: Analysis of the role of the spectral shape parameter
.
J. Atmos. Sci.
,
62
,
3051
3064
, doi:.
Milbrandt
,
J. A.
, and
H.
Morrison
,
2013
:
Predicting graupel density in a bulk microphysics scheme
.
J. Atmos. Sci.
,
70
,
410
429
, doi:.
Milbrandt
,
J. A.
,
M. K.
Yau
,
J.
Mailhot
,
S.
Bélair
, and
R.
McTaggart-Cowan
,
2010
:
Simulation of an orographic precipitation event during IMPROVE-2. Part II: Sensitivity to the number of moments in the bulk microphysics scheme
.
Mon. Wea. Rev.
,
138
,
625
642
, doi:.
Mitchell
,
D. L.
, and
A. J.
Heymsfield
,
2005
:
The treatment of ice particle terminal velocities, highlighting aggregates
.
J. Atmos. Sci.
,
62
,
1637
1644
, doi:.
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:.
Molthan
,
A. L.
,
W. A.
Petersen
,
S. W.
Nesbitt
, and
D.
Hudak
,
2010
:
Evaluating the snow crystal size distribution and density assumptions within a single-moment microphysics scheme
.
Mon. Wea. Rev.
,
138
,
4254
4267
, doi:.
Morrison
,
H.
, and
W. W.
Grabowski
,
2008
:
Modeling supersaturation and subgrid-scale mixing in a large-eddy model with two-moment bulk warm microphysics
.
J. Atmos. Sci.
,
65
,
792
812
, doi:.
Morrison
,
H.
, and
J.
Milbrandt
,
2011
:
Comparison of two-moment bulk microphysics schemes in idealized supercell thunderstorm simulations
.
Mon. Wea. Rev.
,
139
,
1103
1130
, doi:.
Morrison
,
H.
, and
J. A.
Milbrandt
,
2015
:
Parameterization of cloud microphysics based on the prediction of bulk ice particle properties. Part I: Scheme description and idealized tests
.
J. Atmos. Sci.
,
72
,
287
311
, doi:.
Morrison
,
H.
,
G.
Thompson
, and
V.
Tatarskii
,
2009
:
Impact of cloud microphysics on the development of trailing stratiform precipitation in a simulated squall line: Comparison of one- and two-moment schemes
.
Mon. Wea. Rev.
,
137
,
991
1007
, doi:.
Petersen
,
W. A.
,
A.
Tokay
, and
P. N.
Gatlin
,
2013
: GPM ground validation Pluvio precipitation gauge GCPEX. NASA Global Hydrology Center DAAC, Huntsville, AL, accessed 1 August 2015, https://dx.doi.org/10.5067/GPMGV/GCPEX/PLUVIO/DATA301.
Petersen
,
W. A.
,
C.
Pettersen
,
M.
Kulie
,
P. N.
Gatlin
, and
M. T.
Wingo
,
2015
: GPM ground validation NASA Micro Rain Radar (MRR) GCPEX V2. NASA Global Hydrology Center DAAC, Huntsville, AL, https://dx.doi.org/10.5067/GPMGV/GCPEX/MRR/DATA204.
Shi
,
J. J.
, and Coauthors
,
2010
:
WRF simulations of the 20–22 January 2007 snow events over eastern Canada: Comparison with in situ and satellite observations
.
J. Appl. Meteor. Climatol.
,
49
,
2246
2266
, doi:.
Skamarock
,
W. C.
, and Coauthors
,
2008
: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp., http://dx.doi.org/10.5065/D68S4MVH.
Skofronick-Jackson
,
G.
, and Coauthors
,
2015
:
Global Precipitation Measurement Cold Season Precipitation Experiment (GCPEX): For measurement’s sake, let it snow
.
Bull. Amer. Meteor. Soc.
,
96
,
1719
1741
, doi:.
Straka
,
J. M.
,
D. S.
Zrnić
, and
A. V.
Ryzhkov
,
2000
:
Bulk hydrometeor classification and quantification using polarimetric radar data: Synthesis of relations
.
J. Appl. Meteor.
,
39
,
1341
1372
, doi:.
Strapp
,
J. W.
,
F.
Albers
,
A.
Reuter
,
A. V.
Korolev
,
U.
Maixner
,
E.
Rashke
, and
Z.
Vukovic
,
2001
:
Laboratory measurements of the response of a PMS OAP-2DC
.
J. Atmos. Oceanic Technol.
,
18
,
1150
1170
, doi:.
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:.

Footnotes

© 2017 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses).