Abstract

The representation of the marine boundary layer (BL) clouds remains a formidable challenge for state-of-the-art simulations. A recent study by Bodas-Salcedo et al. using the Met Office Unified Model highlights that the underprediction of the low/midlevel postfrontal clouds contributes to the largest bias of the surface downwelling shortwave radiation over the Southern Ocean (SO). A-Train observations and limited in situ measurements have been used to evaluate the Weather Research and Forecasting Model, version 3.3.1 (WRFV3.3.1), in simulating the postfrontal clouds over Tasmania and the SO. The simulated cloud macro/microphysical properties are compared against the observations. Experiments are also undertaken to test the sensitivity of model resolution, microphysical (MP) schemes, planetary boundary layer (PBL) schemes, and cloud condensation nuclei (CCN) concentration. The simulations demonstrate a considerable level of skill in representing the clouds during the frontal passages and, to a lesser extent, in the postfrontal environment. The simulations, however, have great difficulties in portraying the widespread marine BL clouds that are not immediately associated with fronts. This shortcoming is persistent to the changes of model configuration and physical parameterization. The representation of large-scale conditions and their connections with the BL clouds are discussed. A lack of BL moisture is the most obvious explanation for the shortcoming, which may be a consequence of either strong entrainment or weak surface fluxes. It is speculated that the BL wind shear/turbulence may be an issue over the SO. More comprehensive observations are necessary to fully investigate the deficiency of the simulations.

1. Introduction

Clouds over the Southern Ocean (SO) exert an enormous influence on the regional radiative budget (e.g., Haynes et al. 2011; Mace et al. 2007), as well as the energy and water transport to the Antarctic (Zelinka and Hartmann 2012). Yet as detailed in Trenberth and Fasullo (2010), the radiative budget over the SO has been found to be poorly represented in both state-of-the-art reanalysis and coupled global climate models, and is directly linked to the simulations of these clouds.

Cloud thermodynamic phase has a direct impact on the precipitation efficiency (e.g., Han et al. 1998; Choi et al. 2010; Storelvmo et al. 2011) and radiation transmissions (e.g., Curry and Ebert 1992; Sedlar et al. 2012). The thermodynamic phase affects the shortwave radiation because of the difference in the scattering properties between spherical water droplets and nonspherical ice crystals. Further, liquid-phase clouds normally comprise a high concentration of small droplets, giving them a greater albedo, optical depth, and lower transmittance. High-resolution radiometer measurements over the Arctic and subarctic indicate that an incorrect determination of cloud phase can result in errors of 20%–100% in the retrieved particle effective radius and optical depth for scenes with approximately equal amounts of liquid- and solid-phase clouds. The translation of these errors into the calculation of downwelling shortwave could result in a bias of 5%–20% (Key and Intrieri 2000).

A striking feature detected from earlier cloud seeding experiments over Tasmania in Australia is the in situ observation of large quantities of supercooled liquid water (SLW; Smith et al. 1979; Ryan and King 1997; Morrison et al. 2010). The concentration of SLW was commonly found to be greater than 0.3 g m−3 between −6° and −8°C for 5-min flight legs. However, compared to other regions (e.g., the tropics), the study of SO clouds has received far less scrutiny. Field experiments such as the International Global Atmospheric Chemistry (IGAC) Project's First Aerosol Characterization Experiment (ACE 1; Bates et al. 1998) and the Southern Ocean Cloud Experiments (SOCEX-I and -II; Boers et al. 1998) are now over 15 year old.

The observation of clouds over the globe has been significantly improved in recent years as a result of the boost of passive spaceborne technologies (e.g., O'Dell et al. 2008; Greenwald 2009; Seethala and Horváth 2010). Further, the active sensors aboard the A-Train satellite constellation [CloudSat (Stephens et al. 2002) and Cloud–Aerosol Lidar and Infrared Pathfinder Satellite Observations (CALIPSO) (Winker et al. 2007)] provide unprecedented insights into the vertical distributions of microphysical and radiative properties of clouds and aerosols over the globe.

With CALIPSO observations, Hu et al. (2010) discovered the extensive presence of SLW cloud tops over the SO, consistent with findings with the Moderate Resolution Imaging Spectroradiometer (MODIS; Morrison et al. 2011). Using an A-Train merged product, the radar/lidar mask (DARDAR-MASK). Huang et al. (2012a) also highlighted the prevalence of SLW over the SO, particularly during summertime. Low-elevation clouds (below 1 km) with little seasonal cycle are commonly found and identified as problematic from a remote sensing perspective because of the inability of Cloud Profiling Radar (CPR) on CloudSat to distinguish cloud echoes from surface clutter, and the Cloud–Aerosol Lidar with Orthogonal Polarization (CALIOP) on CALIPSO commonly suffering from heavy extinction. The CPR is further limited for the low-altitude clouds (below 3 km) given their frequent occurrence between freezing and −20°C, where understanding the CPR reflectivity becomes ambiguous (Huang et al. 2012b).

Despite the satellite observations, an accurate partition of cloud thermodynamic phase, particularly for low/midlatitude maritime clouds, remains a challenging frontier for current state-of-the-art climate models (e.g., Hannay et al. 2009; Wyant et al. 2010). Although significant progress has been made in developing more physical treatments of the nucleation processes (Jacobson 2006; Storelvmo et al. 2008; Gettelman et al. 2008), only limited success has been achieved. Recently, Bodas-Salcedo et al. (2012) use the Met Office United Model (MetUM) to study the role of clouds around midlatitude cyclones in the persistent bias of surface downwelling shortwave radiation over the SO. They found that the largest bias is due to the model failing to simulate sufficient boundary layer and mid-top clouds in the postfrontal air mass. These clouds are typically associated with SLW. Franklin et al. (2013) evaluated the cloud properties simulated by Australian Community Climate and Earth System Simulator (ACCESS1.3) using the Cloud Feedback Model Intercomparison Project (CFMIP) Observation Simulator Package (COSP) and have consistent findings.

Over the past decade, mesoscale numerical weather prediction (NWP) models have become able to better predict hydrometeor species because of the improved microphysical parameterization and increased resolution (e.g., Hong and Lin 2006; Thompson et al. 2008; Morrison et al. 2009). Efforts have been made to simulate SLW in the Colorado Rocky Mountains (Reisner et al. 1998), freezing drizzle and aircraft icing events over southeastern Canada (Guan et al. 2001; Vaillancourt et al. 2003), mixed phase Arctic stratus clouds (Jiang et al. 2000; Morrison and Pinto 2005), and to evaluate the simulations against field observations. Similar studies over the SO are still lacking because of the absence of field observations. The only study in recent years (Morrison et al. 2010) identified that a major difficulty in modeling SLW off the coast of Tasmania is primarily due to the inability of the reanalysis to capture the wind shear and temperature inversion across the cloud top.

To directly address the outstanding issue discussed by Bodas-Salcedo et al. (2012) regarding the failure to simulate sufficient stratocumulus and mid-top clouds in the postfrontal air, the aim of this study is to evaluate the Weather Research and Forecasting Model (WRF) NWP in simulating postfrontal clouds over Tasmania and the SO. A-Train satellite observations and limited in situ observations are used for the evaluation. Two cases when in situ icing conditions were recorded are investigated. The first case, during the wintertime period 30 June–2 July 2008 (case A), is characterized by postfrontal midlatitude clouds. The second case, during the spring period 7–8 November 2008 (case B), is primarily featured by the widespread low-altitude marine boundary layer (BL) clouds that are not immediately associated with fronts. While large amount of SLW was recorded by in situ measurements during these periods, it was rare to have the combination of the SLW observations and the coincident A-Train overpass.

The remainder of this paper is organized as follows: section 2 describes the synoptic meteorology of the cases. Section 3 presents the model configuration and experimental design. The evaluation study is given in section 4. Section 5 discusses the results from sensitivity experiments. Discussions regarding the large-scale forcing and physical processes in relation to the prediction of clouds are given in section 6.

2. Synoptic meteorology

The synoptic meteorology of Tasmania and the SO is primarily defined by the frequent passage of midlatitude cyclones and fronts year-round (Simmonds and Keay 2000). Mid- and low-level clouds, in which SLW often occurs, are found to be frequently present over the pre- and postfrontal areas (Haynes et al. 2011; Bodas-Salcedo et al. 2012), while stratiform BL clouds commonly exist under the high pressure ridges between fronts (Boers et al. 1998).

The mean sea level pressure (MSLP) analysis (Fig. 1a) for case A at 0000 UTC 2 July 2008 shows a high pressure system stretching from central Australia to the southwest of the continent. A cold front has passed through Tasmania, leaving it exposed in the postfrontal field. After 12 h (Fig. 1b), a second front is approaching. Tasmania was situated between two fronts. In situ observations from the Hydro Tasmania cloud seeding aircraft indicated that extensive SLW was encountered over western Tasmania between 1.5 and 3 km from 1300 to 1500 UTC 2 July. During this period an A-Train track (Fig. 2a track 3) was passing over Tasmania approximately 200 km away from the aircraft. SLW was observed again the following day over the broader area. Unfortunately this flight did not coincide with any nearby satellite overpass.

Fig. 1.

MSLP analysis provided by the Bureau of Meteorology, Melbourne, Australia: (a) 0000 UTC 2 Jul, (b) 1200 UTC 2 Jul, (c) 0600 UTC 7 Nov, and (d) 0000 UTC 8 Nov 2008.

Fig. 1.

MSLP analysis provided by the Bureau of Meteorology, Melbourne, Australia: (a) 0000 UTC 2 Jul, (b) 1200 UTC 2 Jul, (c) 0600 UTC 7 Nov, and (d) 0000 UTC 8 Nov 2008.

Fig. 2.

Maps showing WRF domain settings with A-Train tracks overlaid during the simulation periods for (a) case A and (b) case B.

Fig. 2.

Maps showing WRF domain settings with A-Train tracks overlaid during the simulation periods for (a) case A and (b) case B.

The MSLP chart (Fig. 1c) for case B at 0600 UTC 7 November 2008 shows that a front is approaching Tasmania from the northwest. The front occupies a narrow latitude band with the low pressure center situated at ~40°S. A cloud seeding aircraft sampled the clouds over western and central Tasmania at 0430 UTC but only glaciated cumuli were encountered. The timing of this flight was in coincidence with an A-Train track (Fig. 2b track 1) passing over the northeast. At 0000 UTC 8 November 2008 (Fig. 1d), the front moved farther eastward, leaving Tasmania in the postfrontal air mass. Another seeding flight detected a large amount of SLW from 2230 UTC 7 November to 0100 UTC 8 November. This flight coincided with the MODIS observation on Terra.

As a result of data quality issues, the in situ observations are not used for quantitative comparison in this study. Nevertheless, they are used for a qualitative assessment of the simulated thermodynamic properties and dominant cloud features.

3. Experimental design

a. Model configuration

The Advanced Research WRF (ARW) version 3.3.1 is a nonhydrostatic Eulerian solver developed by multiple government agencies in the United States (Skamarock et al. 2008). In this study, the model is configured with an outer domain that covers a broad area of Australia and the SO (Fig. 2a). Three nested domains (one-way nesting) were applied with 64 η-level and the horizontal spacing of 30, 10, 3.3, and 1.1 km. For case A, the innermost domain is set up to incorporate Tasmania and its surrounding ocean. For case B, a second innermost domain is added to capture one of the A-Train tracks off the west coast of Tasmania (Fig. 2b track 3). The model used the Interim European Centre for Medium-Range Weather Forecasts (ECMWF) Re-Analysis (ERA-Interim) (1.5° × 1.5° grid, 38 pressure levels, and 6-hourly updates) for initialization and lateral boundary conditions. Because of the deteriorating accuracy for simulations beyond 48 h, the simulation of case A (0000 UTC 30 June–1800 UTC 2 July) was broken down into two periods with each period integrated for 36 h. The simulation of the second period is resumed at 0600 UTC 1 July allowing 6-h spinup time that overlaps with the first period. The single simulation for case B is 48 h and initialized at 1200 UTC 6 November.

For the initial evaluation, the model was run with 64 η-level with a vertical resolution of 40 m at the surface extending to 2 km for the upper levels. Simulations were configured with the Rapid Radiative Transfer Model for GCMs (RRTMG) shortwave and longwave radiation scheme (Mlawer et al. 1997; Iacono et al. 2008), the Noah land surface model [from the National Centers for Environmental Prediction (NCEP)–Oregon State University–U.S. Air Force–National Weather Service Office of Hydrologic Development; Chen and Dudhia 2001], Yonsei University (YSU; Hong et al. 2006) planetary boundary layer (PBL) scheme, the new Thompson (Thompson et al. 2008) microphysical (MP) scheme, and the Simplified Arakawa–Schubert (SAS) cumulus scheme (Pan and Wu 1995) for the outer three domains.

A suite of sensitivity experiments were also undertaken. A modified 64 η-level spacing with increased vertical levels (21 levels) within the lowest kilometer was applied to test the sensitivity of model resolution. The frequently used Lin (Lin et al. 1983) MP scheme, and other two PBL schemes, Mellor–Yamada–Janjic (MYJ; Mellor and Yamada 1982) and Mellor–Yamada–Nakanishi–Niino (MYNN; Nakanishi and Niino 2004), were implemented to test the model sensitivity to different physical treatments. An overview of the MP and PBL schemes tested in this study is given in Table 1. As the atmosphere over the SO is essentially free of terrestrial and anthropogenic aerosol emissions (e.g., Yum and Hudson 2005; Gras 1995), simulations with lower cloud condensation nuclei (CCN) concentration were also examined.

Table 1.

A list of microphysical (MP) schemes and planetary boundary layer (PBL) schemes used for the model experiments in the study.

A list of microphysical (MP) schemes and planetary boundary layer (PBL) schemes used for the model experiments in the study.
A list of microphysical (MP) schemes and planetary boundary layer (PBL) schemes used for the model experiments in the study.

b. Model evaluation

Observations employed for the model evaluation include the upper-air routine soundings at Hobart Airport (42.83°S, 147.5°E), A-Train satellite observations, and limited aircraft measurements. Our assessment allows 1-h lead time centered at the times of the soundings and the satellites to capture the maximal resemblance. The A-Train datasets employed contain the 94-GHz cloud profiling radar (CPR) reflectivity from the CloudSat “2B-GEOPROF” product, the thermodynamic variables from CloudSat “ECMWF-AUX” (from ECMWF), the vertical feature mask (VFM) profile (V3.1.1; Hu et al. 2009) from CALIOP, cloud properties (Platnick et al. 2003) from MODIS (both Aqua and Terra), and the simplified classification from the DARDAR-MASK (Delanoë and Hogan 2010), a merged product of CloudSat, CALIPSO, and MODIS observations. The DARDAR-MASK returns a range of categories: clear, ground, stratospheric features, insects, aerosol, rain, supercooled liquid water (SLW), liquid warm (LW), mixed phase (mixed) and ice (ice), and uncertain (UN) classification. The cloud-seeding aircraft provided measurements of cloud and aerosol microphysical properties.

To enable a direct comparison between satellite observations and the simulations, a radar simulator QuickBeam (Haynes et al. 2007) was used to simulate the CPR reflectivity using the simulated mixing ratios of the hydrometeor species. Further, a simple diagnostic scheme based on the relative fraction of liquid water to total water (RLW) was employed to define the simulated cloud phase as liquid (RLW > 0.7), mixed phase (0.3 ≤ RLW ≤ 0.7), or glaciated (RLW < 0.3). Cloud top defined at 0.1 cloud optical thickness (COT) from the top of the model is applied to examine cloud-top properties. This is an analog of the MODIS observation as the radiant fluxes received by the spaceborne radiometer depend strongly upon the bulk emissivity of the hydrometers through a certain optical depth. Note that thresholds of 1, 0.5, and 0.25 were also tested but they were all shown to produce cloud fractions (defined by the frequency of occurrence of clouds seen from a passive remote sensing satellite from the top of the atmosphere) too low, or cloud-top heights too low compared against the MODIS observations. Further assumptions are that the zenith angle is zero and cloud absorption coefficients are constant and dependent on specific hydrometers (i.e., cloud water, rain, ice, and snow). The absorption coefficients are 0.145, 0.000 33, 0.0735, and 0.002 34 m2 g−1, respectively (Dudhia 1989).

In the evaluation, comparisons are made for the scenes that best match the MODIS observations within the lead time allowed. It should be noted that frontal clouds at the early stages of both cases are highly glaciated and well represented in the simulations. The results presented in this paper are focused on the postfrontal cloud fields.

As the simulations with the modified vertical grid display greater skills, it is adopted for the evaluation. The sensitivity study is mainly focused on testing the MP schemes, PBL schemes, and CCN concentration (Table 2).

Table 2.

A list of configuration settings for numerical studies of the two cases.

A list of configuration settings for numerical studies of the two cases.
A list of configuration settings for numerical studies of the two cases.

4. Results of evaluation study

a. Atmospheric profiles

The simulated sounding profiles are compared with routine upper-air soundings at Hobart Airport to evaluate the simulation of thermodynamical structures.

Figure 3 compares the temperature and dewpoint temperature profiles from observations (purple) and simulations (black) for both cases. For case A (Figs. 3a,b), the simulated soundings for both frontal and postfrontal periods generally agree with the observations, although a lack of variability is noticed. During the frontal period (Fig. 3a), the modeled surface temperature is within ~3°C of observation. Both the simulation and observation suggest a deep moist layer up to ~400 hPa. However, the simulation fails to develop a wind shear layer between the surface and 780 hPa. This weakness, also highlighted in Morrison et al. (2010), was persistent to changes of vertical resolution and the PBL scheme. The simulation shows considerably less skill in simulating the lower atmospheric structure in the postfrontal flow (Fig. 3b). The model overpredicts both the temperature and moisture below 670 hPa, and the location of condensation. The simulation also fails to reproduce the strong inversion between 670 and 600 hPa and the multilayer structure.

Fig. 3.

Thermodynamic profiles obtained from upper-air routine soundings (purple lines) at Hobart Airport (42.83°S, 147.5°E) and simulated temperature and dewpoint profiles (black lines) at the same location. Wind profiles are shown on the right in each panel. (a) Upper-air sounding at 0000 UTC 2 Jul 2008, model sounding at 2330 UTC 1 Jul 2008; (b) upper-air sounding at 1200 UTC 2 Jul 2008, model sounding at 1100 UTC 2 Jul 2008; (c) upper-air sounding at 1200 UTC 7 Nov 2008, model sounding at 1200 UTC 7 Nov 2008; and (d) upper-air sounding at 0000 UTC 8 Nov 2008, model sounding at 0030 UTC 8 Nov 2008.

Fig. 3.

Thermodynamic profiles obtained from upper-air routine soundings (purple lines) at Hobart Airport (42.83°S, 147.5°E) and simulated temperature and dewpoint profiles (black lines) at the same location. Wind profiles are shown on the right in each panel. (a) Upper-air sounding at 0000 UTC 2 Jul 2008, model sounding at 2330 UTC 1 Jul 2008; (b) upper-air sounding at 1200 UTC 2 Jul 2008, model sounding at 1100 UTC 2 Jul 2008; (c) upper-air sounding at 1200 UTC 7 Nov 2008, model sounding at 1200 UTC 7 Nov 2008; and (d) upper-air sounding at 0000 UTC 8 Nov 2008, model sounding at 0030 UTC 8 Nov 2008.

For case B (Figs. 3c,d), the model again displays reasonable skill in simulating the frontal passage at the early stage (Fig. 3c). Both the simulation and observation suggest a neutrally stable moist layer deepening up to 250 hPa. However, a temperature inversion of ~5°C between 800 and 750 hPa is again missed in the simulation. The simulated dewpoint temperature near the surface is ~2°C lower than the observed while the ground temperature is overestimated by a similar magnitude. The simulated wind profiles closely resemble the observations, shearing from northeast to northwest. At the later stage (Fig. 3d), the model sounding shows general consistency with the observations, but the cloud-top height is overpredicted by ~80 hPa and the wind profiles below 750 hPa are again poorly represented.

The difficulty in simulating the soundings at Hobart, Tasmania, may, in part, be due to the challenging orography or coastal location. By comparing the ERA-Interim profiles against the Hobart routine soundings, we found that the lack of variability in the simulations is likely due to the inability of the reanalysis to represent the common multilayer structure over the SO (Hande et al. 2012).

b. Cloud structure and microphysics

1) Case A track 3

A-Train observations along track 3 for case A at 0410 UTC 2 July (Fig. 4) occurred roughly 8 h after the frontal passage. The MODIS cloud-top phase (CTP; Fig. 4a) reveals the extensive cloud cover over the marine area with CTP being primarily composed of liquid water. A complex cloud formation over the mountainous southwest Tasmania and the adjacent SO is classified as a mix of ice, mixed, and UN. The relative frequency (RF) of UN within the cloud-top temperature (CTT) range between 0° and −20°C (Fig. 4b) is found to be 52%, while the RF of SLW is 39%. The A-Train track passed through various cloud types across the domain. The CPR (Fig. 4c) detects a cloud field over Tasmania and a group of deep maritime cumulus with reflectivities between −10 and 20 dBZ. From 44.5° to 45°S, only a few isolated clouds are detected by the CPR while continuous warm clouds are recorded by MODIS. The CALIOP classification (Fig. 4d) is generally consistent with the MODIS CTP, except that the UN classified by MODIS over land is characterized as SLW by CALIOP. Over the water, the intermittently distributed marine cumuli are shown to be dominated by SLW. The DARDAR-MASK categorization (Fig. 4e) preferentially returned mixed at the cloud tops, which is likely to be a consequence of how the CPR return is resolved within the DARDAR-MASK algorithm (Huang et al. 2012a). Beneath the cloud tops, the DARDAR-MASK mainly returns ice and UN once the signals of the CPR and CALIOP both drop out.

Fig. 4.

A composite of A-Train satellite observations for case A track 3 at roughly 0420 UTC 2 Jul 2008. (a) Cloud-top phase (CTP) from MODIS, (b) Cloud-top temperature (CTT) from MODIS, (c) along-track CPR reflectivities from CloudSat, (d) along-track vertical feature mask (VFM) from CALIPSO, and (e) simplified classification from the DARDAR-MASK.

Fig. 4.

A composite of A-Train satellite observations for case A track 3 at roughly 0420 UTC 2 Jul 2008. (a) Cloud-top phase (CTP) from MODIS, (b) Cloud-top temperature (CTT) from MODIS, (c) along-track CPR reflectivities from CloudSat, (d) along-track vertical feature mask (VFM) from CALIPSO, and (e) simplified classification from the DARDAR-MASK.

The corresponding simulations are displayed in Fig. 5. Overall, the simulated cloud fields (Figs. 5a,b) are well placed compared with MODIS observations. The simulation underestimates the prevalence of the marine liquid water clouds, resulting in widespread cloud-free areas. Further, more glaciated cloud tops are produced. The simulated CPR reflectivities (Figs. 5c,d) show that the convective marine cumuli are well reproduced, but the attenuation is overestimated, suggesting that there may be uncertainties associated with the attenuation treatment by the simulator, or the prescribed empirical parameters may not well represent the microphysical properties of the simulated hydrometeor species. The simulated categorization (Fig. 5e) captures the overall cloud structure and phase partitioning. SLW is reproduced but with much lower frequency. In addition, more precipitating marine cumuli are produced and the simulated UN (associated with low mixing ratio) is widely present over the water within the lowest kilometer.

Fig. 5.

A composite of the corresponding simulated cloud properties for case A track 3 at 0415 UTC 2 Jul 2008. (a) Simulated CTP, (b) Simulated CTT, (c) simulated along-track CPR reflectivities with attenuation, (d) simulated along-track CPR reflectivities without attenuation, and (e) simulated cloud phase classification.

Fig. 5.

A composite of the corresponding simulated cloud properties for case A track 3 at 0415 UTC 2 Jul 2008. (a) Simulated CTP, (b) Simulated CTT, (c) simulated along-track CPR reflectivities with attenuation, (d) simulated along-track CPR reflectivities without attenuation, and (e) simulated cloud phase classification.

2) Case A track 4

Track 4 in case A (Fig. 6) occurs about 20 h after the front passage. Again, the pervasive liquid water clouds are detected by MODIS, except over the southwest and northeast corner. The majority of the cloud tops over the southwest quadrant are UN (Fig. 6b) given that MODIS has difficulties in determining CTP between −25° and −15°C (Morrison et al. 2011).

Fig. 6.

As in Fig. 4, but for case A track 4 at roughly 1530 UTC 2 Jul 2008.

Fig. 6.

As in Fig. 4, but for case A track 4 at roughly 1530 UTC 2 Jul 2008.

The CPR (Fig. 6c) returns moderate reflectivities (~0 dBZ) over Tasmania indicating a potential of drizzling clouds or a large quantity of liquid water droplets, which is confirmed by the CALIOP VFM (Fig. 6d). More notably, a stratified cloud layer with very weak radar returns (approximately −20 dBZ) is observed between −10° and −20°C at ~4 km over the southern portion of the domain. These hydrometers are categorized as SLW by CALIOP (Fig. 6d). The DARDAR-MASK (Fig. 6e), on the other hand, breaks this cloud field into two layers with a thin layer of ice underlying SLW, which is commonly observed over the SO (Huang et al. 2012a).

The simulation (Fig. 7) manages to replicate much of the cloud pattern seen by MODIS. Most of the UN class in MODIS is assigned to SLW (Figs. 7a,b). However, the simulated cloud fraction remains appreciably low (53%), compared to MODIS observation (96%). While the overall cloud field is captured, the layer structure from 42.6° to 45.5°S is poorly reproduced (Figs. 7c,d). The low-level clouds over Tasmania are also underrepresented. The simulated categorization (Fig. 7e) suggests a reasonable amount of SLW between freezing and −15°C to the south of Tasmania. The simulated UN again occupies the lowest kilometer.

Fig. 7.

As in Fig. 5, but at for case A track 4 at 1545 UTC 2 Jul 2008.

Fig. 7.

As in Fig. 5, but at for case A track 4 at 1545 UTC 2 Jul 2008.

3) Case B track 3

The persistent underprediction of the low-altitude clouds is more vividly demonstrated in track 3 of case B (Fig. 8), 22 h after the frontal passage. The A-Train track traversed a large field of BL warm clouds (Figs. 8a,b) to the west of Tasmania. While the LWP (Fig. 8c) is generally below 160 gm−2, the frequency of missing values is relatively high due to the failure of MODIS in retrieving the cloud effective radius (Seethala and Horváth 2010). The MODIS COT (Fig. 8c) of these clouds is typically below 20. While the CPR (Fig. 8e) could not accurately distinguish the returns due to the surface clutter, the lidar (Fig. 8f) returns a near-continuous BL cloud field below 1 km, consistent with the MODIS CTP. The DARDAR-MASK returns a continuous layer of BL clouds, consisting primarily of LW and occasionally aerosol and rain. However, the ECMWF temperatures fail to capture a clear capping inversion associated with these BL clouds.

Fig. 8.

A composite of A-Train satellite observations for case B track 3 at roughly 0505 UTC 8 Nov 2008. (a) CTP from MODIS, (b) CTT from MODIS, (c) cloud LWP from MODIS (the gray shadings represent ice clouds where estimate of LWP is not applicable; the wheat color indicates missing values), (d) cloud optical thickness (COT) from MODIS, (e) along-track CPR reflectivities from CloudSat, (f) along-track VFM from CALIPSO, and (g) simplified classification from the DARDAR-MASK.

Fig. 8.

A composite of A-Train satellite observations for case B track 3 at roughly 0505 UTC 8 Nov 2008. (a) CTP from MODIS, (b) CTT from MODIS, (c) cloud LWP from MODIS (the gray shadings represent ice clouds where estimate of LWP is not applicable; the wheat color indicates missing values), (d) cloud optical thickness (COT) from MODIS, (e) along-track CPR reflectivities from CloudSat, (f) along-track VFM from CALIPSO, and (g) simplified classification from the DARDAR-MASK.

The simulation is only able to reproduce patches of cumulus over the water (Fig. 9b). Note that lower thresholds of the COT (0.05 and 0.01) were also tested to explore potentially optically thin clouds but no substantial difference was found. Similarly, the simulated classification (Fig. 9e) shows little cloudiness near the surface, except some shallow cumuli (with drizzle) below 1.5 km. The simulated radar reflectivities are not shown as they only offer marginal insights to these low-level BL clouds. The simulated temperatures indicate a capping inversion of ~2°C at 1.5–2 km, which appeared to be deeper than that implied by CALIPSO observation. The bias of the inversion height in the lower atmosphere might be a key element for the disappearance of the very low BL clouds in the simulation. Alternatively, the absence of BL clouds may directly result from the initialization taken from the ERA-Interim, as discussed previously.

Fig. 9.

A composite of simulated cloud properties for case B track 3 at 0515 UTC 8 Nov 2008 with different planetary boundary layer (PBL) schemes. (a),(b) As in Figs. 8a,b, but simulated CTP with YSU, (c) simulated CTP with MYJ, (d) simulated CTP with MYNN, (e) simulated cloud phase classification with YSU, (f) simulated cloud phase classification with MYJ, and (g) simulated cloud phase classification with MYNN.

Fig. 9.

A composite of simulated cloud properties for case B track 3 at 0515 UTC 8 Nov 2008 with different planetary boundary layer (PBL) schemes. (a),(b) As in Figs. 8a,b, but simulated CTP with YSU, (c) simulated CTP with MYJ, (d) simulated CTP with MYNN, (e) simulated cloud phase classification with YSU, (f) simulated cloud phase classification with MYJ, and (g) simulated cloud phase classification with MYNN.

c. Cloud integrated properties

The current study case B has been extended to examine the simulated column integrated cloud water mixing ratio (simulated LWP) in comparison to MODIS observation. Cloud liquid water content is a critical link between cloud dynamics and cloud radiative effects (Stephens 1978; Wood 2012). Both longwave cooling and shortwave warming are strongly dependent on cloud LWP, particularly for stratiform BL clouds with moderately low LWP (e.g., Garrett and Zhao 2006; Petters et al. 2012). Previous study over Arctic shows that the downwelling longwave radiation is most sensitive when the LWP is below 30 g m−2 (Shupe and Intrieri 2004).

Figure 10a shows the LWP observed by MODIS at 0055 UTC 8 November. The MODIS LWP is a function of cloud water path and cloud phase optical properties (CPOP). Note that the CPOP only indicates the majority of the cloud phase to the optical level that MODIS reaches, which may lead to an underestimate of the LWP for multilayer clouds complicated by the heavy presence of ice clouds overhead. In Fig. 10a, the gray shading represents the areas where ice is assigned to CPOP hence the LWP is not available for evaluation. The most prominent feature, again, is the prevalence of the widespread marine BL clouds rolling over the ocean, with the LWP broadly below 300 g m−2. MODIS shows that the CTTs of these homogeneous stratiform clouds are between 0° and 8°C. High values of LWP are found over western Tasmania, peaking at roughly 1000 g m−2, reflecting the orographic enhancement by the western mountain range. Ice-dominating clouds spread over southwest Tasmania and the ocean off the south and southeast coasts.

Fig. 10.

The observed and simulated cloud LWP. The observed LWP is from MODIS (on Terra) at 0055 UTC 8 Nov 2008. The simulations are at 0100 UTC 8 Nov 2008. (a) Cloud LWP from MODIS (the gray shadings represent ice clouds where estimate of LWP is not applicable; the wheat color indicates missing values), (b) simulated LWP with TH, (c) simulated LWP with TC, (d) simulated LWP with TH_MYJ, (e) simulated LWP with LH, and (f) simulated LWP with TH_CCN25. The wheat color in (a) indicates missing values.

Fig. 10.

The observed and simulated cloud LWP. The observed LWP is from MODIS (on Terra) at 0055 UTC 8 Nov 2008. The simulations are at 0100 UTC 8 Nov 2008. (a) Cloud LWP from MODIS (the gray shadings represent ice clouds where estimate of LWP is not applicable; the wheat color indicates missing values), (b) simulated LWP with TH, (c) simulated LWP with TC, (d) simulated LWP with TH_MYJ, (e) simulated LWP with LH, and (f) simulated LWP with TH_CCN25. The wheat color in (a) indicates missing values.

Apart from the satellite observation, the cloud-seeding aircraft encountered extensive areas of SLW over western Tasmania persisting from 2230 UTC 7 November to 0100 UTC 8 November up to ~3.5 km. The majority of SLW clouds were found to reside between −15° and −7°C. A cloud-seeding operation was conducted during this period.

Figure 10b is the simulated LWP at 0100 UTC 8 November. Overall, the spatial distribution of the LWP has been reasonably reproduced, consistent with the satellite measurements. The heavy cloudiness over southwest Tasmania and off the southwest coast are well represented. High values of LWP are reproduced over these heavily cloudy areas. However, the LWP over central and eastern Tasmania are overpredicted, and the convective clouds to the east and southeast of Tasmania are not produced. The most striking feature, again, is that the LWP of the marine BL clouds broadly observed in the postfrontal air mass are vastly underestimated, although the cloud pattern and the roll-like structure are analog to the observed.

The evaluation with the MODIS LWP is only available for daytime observations. Unfortunately, the daytime tracks from Aqua in this study contain large amount of missing values, hence offer limited insights to the comparison.

d. Cloud-top phase properties

Haynes et al. (2011) suggest that the largest bias in the absorbed shortwave radiation over the SO is linked to the bright mid/low-top clouds that have been significantly underrepresented by the climate models. The net radiative budget is particularly sensitive to the thermodynamic phase at cloud tops.

Similar to Huang et al. (2012a), histograms of the relative frequencies (RFs) of the CTP decomposed as a function of temperature are calculated for both MODIS observations and model simulations (Fig. 11). To provide a more definitive classification, the class of UN is not considered in the simulated CTP. As the CTT is defined by the cloud temperature at the level of the specified COT while the CTP is considered as a bulk property of the hydrometeors above this level, the CTT is typically warmer than the actual mean temperature of these hydrometeors. The simulated cloud fraction (CF), defined by the ratio of cloudy model columns to the total number of model columns (innermost domain only) is shown on each panel.

Fig. 11.

Histograms showing the relative frequencies (RFs) of CTP (ice, mixed phase, and liquid water) decomposed into 5°C temperature bins. The observation is from MODIS (on Terra) at 0055 UTC 8 Nov 2008. The simulations are at 0100 UTC 8 Nov 2008. The simulated cloud top is defined at 0.1 and 0.5 COT, respectively. Cloud fraction (CF) is displayed in each panel. (a) RFs of CTP from MODIS, (b) RFs of the simulated CTP at 0.1 COT cloud top, and (c) RFs of the simulated CTP at 0.5 COT cloud top.

Fig. 11.

Histograms showing the relative frequencies (RFs) of CTP (ice, mixed phase, and liquid water) decomposed into 5°C temperature bins. The observation is from MODIS (on Terra) at 0055 UTC 8 Nov 2008. The simulations are at 0100 UTC 8 Nov 2008. The simulated cloud top is defined at 0.1 and 0.5 COT, respectively. Cloud fraction (CF) is displayed in each panel. (a) RFs of CTP from MODIS, (b) RFs of the simulated CTP at 0.1 COT cloud top, and (c) RFs of the simulated CTP at 0.5 COT cloud top.

The MODIS CTP histograms suggest a total cloud fraction of 95%, with an outstanding amount (~40%) residing between 0° and 5°C. A bimodal distribution is noticed with the second peak occurring at ~ −30°C and the minimum between −10° and −15°C. Note that Holz et al. (2008) investigated cloud detection and height evaluation using both MODIS and CALIOP and found that MODIS overestimates cloud-top height in regions with low-level temperature inversions. The observed cloud tops are dominated by SLW (28% with respect to the total cloud cover) between 0° and −20°C. In temperature ranges between −15° and −30°C, the RFs of SLW decrease with the increased proportion of UN. Mixed and ice class dominate the temperature ranges below −30°C.

Histograms for the simulated cloud top defined at 0.1 COT (Fig. 11b) display notable similarities to that observed in terms of the shape of the distribution, although the cloud fraction is significantly lower (70%). Taking CTT as a proxy of altitude, the prevalence of low-level cloud tops is captured, but they are produced at colder temperatures compared to the observation. The warm clouds (CTT > 0°C) are underpredicted by 15%, which is compensated by the increased midlevel cloud tops between freezing and −20°C.

With cloud top defined at 0.5 COT (Fig. 11c), the shape of the CTP distribution changes accordingly. While the majority of the simulated cloud tops shift to warmer temperature ranges (from 0° to −20°C), it does not lead to an increase of warm clouds, as some of the optically thin clouds are simply missed resulting in slightly lower cloud fraction (65%).

In summary, the evaluation of the two cases indicates that the simulations are able to reproduce the large-scale dynamics associated with the frontal passages over the SO. The evolution of the atmospheric state can be reasonably represented although a lack of wind shear and temperature inversions within the lower atmosphere are pronounced. The simulations have also demonstrated fair skill in predicting the macro and microphysical properties of the postfrontal stratocumuli. The CTP can be represented relatively well but glaciation is more commonly reproduced in colder cloud tops. The simulations, however, show less skill in predicting the midlatitude stratiform clouds in the postfrontal air mass. The simulations have even further difficulty in reproducing the widespread marine BL clouds that cover vast tracts of the SO.

5. Sensitivity study

A series of numerical experiments was conducted to examine the sensitivity of the simulations to model resolution, MP schemes, PBL schemes, and CCN concentration (Table 2).

a. Case B track 3

As track 3 of case B is most poorly simulated with respect to cloud cover, further investigations were made to examine the sensitivity to different PBL schemes (Fig. 9). Again, as shown by the simulated CTP (Figs. 9c,d), the three PBL schemes test are able to produce patches of marine cumuli over the ocean, followed by a convective cloud band moving eastward. The pervasive low-lying warm clouds, however, remain significantly underrepresented. Similarly, the simulated classifications (Figs. 9f,g) only show little condensate, with only some drizzling shallow cumuli forming sparsely. The capping inversion is reproduced in all the simulations, but consistently placed at 1.5–2 km. Essentially none of the PBL schemes tested is able to substantially improve the simulation. This challenge is disconcerting given the inherent differences between the three PBL schemes. While the nonlocal YSU scheme might potentially overestimate the vertical flux transport hence appearing to be least valid under nonconvective condition, the local closure models (MYJ and MYNN), together with the turbulent kinetic energy (TKE) prediction, have been shown to demonstrate more realistic behaviors for stable and neutral flow (Mellor and Yamada 1982; Janjic 2002). Having three PBL schemes tested, our analysis does not intend to imply that no PBL scheme exists is able to parameterize the marine BL processes in a realistic manner. Recent development of the higher-order PBL schemes (e.g., Larson et al. 2012) has shown better representation of mean and turbulence flux. However, the constant failures in the simulations suggest that more important deficiencies (e.g., large-scale structures of temperature and moisture field) might be preventing the formation and maintenance of these clouds, regardless of the artificial modularity of the PBL parameterizations.

b. Cloud integrated properties

As a direct comparison to Figs. 10a,b, Figs. 10c,f are the simulations of cloud LWP. In general, all the sensitivity runs have successfully reproduced the overall patterns of cloud fields. However, none of the sensitivity runs is able to predict even a comparable amount of LWP for those ubiquitous marine BL clouds that are present in TH. The contrast between TC and TH suggests that increased vertical grid spacing has notably improved the simulated LWP. This might be a result of turbulence and cloud-top entrainment being better resolved. Bretherton et al. (1999) show that larger cloud-top entrainment from the coarser vertical grid spacing experiments may contribute to the dissipation of the stratocumulus clouds.

It is interesting to note that even with the modified vertical spacing, the LWP reproduced by simulations TH_MYJ (Fig. 12d) and LH (Fig. 12e) are much lower compared against TH. This suggests that the representation of LWP is not only attributable to the refined vertical resolution, but also tied to the physical treatment in the MP and PBL schemes. Meanwhile, the plethora of parameterizations and their coupling increase the difficulty in interpreting their physical implications.

Fig. 12.

Histograms showing the relative frequencies (RFs) of CTP (ice, mixed phase, and liquid water) decomposed into 5°C temperature bins for the two cases. The calculations are made across the full duration of the simulations (after spinup) with different configuration settings. The simulated cloud top is defined at 0.1 COT. Cloud fraction (CF) is displayed on each panel. (a)–(c) CTP statistics from sensitivity tests for case A. (d)–(f) CTP statistics from sensitivity tests for case B.

Fig. 12.

Histograms showing the relative frequencies (RFs) of CTP (ice, mixed phase, and liquid water) decomposed into 5°C temperature bins for the two cases. The calculations are made across the full duration of the simulations (after spinup) with different configuration settings. The simulated cloud top is defined at 0.1 COT. Cloud fraction (CF) is displayed on each panel. (a)–(c) CTP statistics from sensitivity tests for case A. (d)–(f) CTP statistics from sensitivity tests for case B.

Experiments with CCN = 50 (not shown) and 25 cm−3 (Fig. 10f) are also tested to examine the impact of aerosol-rare conditions on the simulations (Yum and Hudson 2004; Bennartz 2007). The decreased CCN concentration results in a significantly reduced LWP for the marine BL clouds even further from the observations. A plausible explanation for the considerable loss of the liquid water over the ocean is that the condensate may have been removed quickly by precipitation rapidly formed by larger liquid water droplets. Note that a simulation with higher CCN (150 cm−3, now shown), although less realistic, was also explored, but it did not lead to a significant increase of cloudiness. Recent observational and modeling results suggest that the sensitivity of precipitation rate to CCN varies strongly with LWP (Jiang et al. 2010; L'Ecuyer et al. 2009).

c. Cloud-top phase properties

Similar to Fig. 11, the RFs of CTP for the two cases are shown in Fig. 12. These calculations are made across the full duration of the simulation (after spinup). As such, they can no longer be compared to any instantaneous observations. However, they permit a broader appreciation of the model's performance in representing the CTP properties. The top panels (Figs. 12a–c) present the results for case A while the bottom panels (Figs. 12d–f) are the results for case B.

Starting from case A, the simulated cloud fraction of simulation TC (Fig. 12a) is 44% with the majority of clouds (58%) residing between 0° and −20°C. The peak is found at temperatures between 0° and −5°C. The RFs remain fairly constant (~7%) from −20° to −40°C and start to decrease beyond −40°C. Virtually no cloud tops are simulated at temperatures warmer than 0°C. Between 0° and −20°C, SLW and mixed are found but only occur 13% and 6% of the time, respectively.

The simulation of warmer clouds is found to be sensitive to the vertical spacing of the model. Comparing simulation TH to TC (Fig. 12b), the cloud fraction reproduced by the TH increases by 17%, although the shape of the distributions is broadly similar. The increase of cloud fraction is mainly due to the increase of SLW to 17% and, to a lesser extent, mixed to 9%, in the temperature range between 0° to −20°C. The RFs of warm clouds also increase slightly. Compared to TC and TH, simulation LH (Fig. 12c) demonstrates a bimodal behavior, with a cloud fraction of 46%. The lack of clouds between −15° and −35°C is generally compensated by higher RFs of glaciated cloud tops at temperatures below −35°C and warmer cloud tops above −10°C. There is also a significant increase of SLW (to 23%).

The difference between TH and LH CTP statistics is worth noting. Compared to the single-moment Lin scheme, in which the size distributions of all the hydrometeors are represented as exponential functions, a large number of improvements to physical parameterizations are implemented in the new Thompson scheme. For example, number concentration of ice and rain are treated explicitly in addition to mixing ratios, while others are forced to behave more like two-moment schemes. Also, snow aggregates are treated as fractal-like, with a bulk density that varies inversely with diameter. Their size distribution is represented as a sum of exponential and gamma distributions. These differences are important to understanding the physical causes of the divergence in the CTP statistics.

Different from case A, the bimodal distribution is evident for all sensitivity runs for case B. The RFs of ice shift to colder temperature ranges with the peaks occurring between −45° and −55°C. Midlatitude cloud tops between −10° and −35°C are not widely produced and are exclusively dominated by ice. SLW cloud tops only form at temperatures warmer than −10°C and become dominant between 0° and −5°C. There is a substantial increase of warm clouds between 0° and 15°C, which is not found in case A. Looking at the individual tests, simulation TH again reproduces a higher cloud fraction compared to the TC (58% vs 54%), most prominent between 5° and 10°C. TH_MYJ produces a much lower cloud fraction (48%), although the pattern of distribution strongly resembles that of TH. Compared to the sensitivity tests on MP schemes, the PBL schemes tested are only shown to have marginal impacts on the CTP statistics.

In summary, the sensitivity study shows that the representation of the widespread marine BL clouds over the SO is a formidable challenge for the simulation. No single physical treatment could independently lead to a significant broad improvement of the model performance. The simulated LWP indicates that although the simulations are able to reproduce the geographical distribution of the liquid-phase clouds, the integrated amount is inherently underrepresented. The simulated LWP is sensitive to the model's vertical resolution, CCN concentration, and a great range of the physical processes represented by different parameterization schemes. The simulated CTP properties reveal that while SLW and warm clouds could be reproduced, their amounts are lower compared to the observations. Further, the statistics are sensitive to the definition of cloud top, MP schemes, vertical resolution, and to a lesser extent, PBL schemes. The limited skill of these simulations across a range of parameterizations suggests that the model initialization and forcing may be of greater concern than the model physics.

6. Discussion and conclusions

The A-Train observations have been used to evaluate the Weather Research and Forecasting Model (WRFV3.3.1) in simulating the postfrontal clouds over Tasmania and the Southern Ocean (SO). Two cases were chosen where supercooled liquid water was observed in the region with in situ instrumentation. The simulated cloud structure, radar reflectivity, cloud phase composition, column integrated, and cloud-top phase properties are compared against the observations from upper-air soundings at Hobart Airport, A-Train satellite observations, and limited aircraft measurements. Experiments are also conducted to test the sensitivity of the simulations to model resolution, microphysical (MP) schemes, planetary boundary layer (PBL) schemes, and cloud condensation nuclei (CCN) concentration.

For both case studies, the simulations demonstrate a considerable level of skill in representing the cloud fields during the frontal passages and, to a lesser extent, in the postfrontal air mass. Although the merged product, DARDAR-MASK, did not produce any SLW beyond cloud top, the aircraft encountered a substantial amount of SLW during the periods of simulation. This feature could be captured in the simulations, although the concentration is sensitive to the definition of cloud top, MP schemes, and vertical resolution.

The simulations, however, have great difficulties in simulating the common marine boundary layers (BL) clouds found hours after a front has passed through. Huang et al. (2012a) highlighted the widespread presence of BL clouds over the SO, yet Bodas-Salcedo et al. (2012) found that the largest bias of surface downwelling shortwave radiation over the SO is due to the model failing to simulate sufficient stratocumulus and mid-top clouds in the postfrontal air mass. In the winter case study (case A track 4), the simulations produced a cloud fraction of 53% while 96% coverage was observed by MODIS.

The postfrontal BL clouds observed during the case B track 3 were even more problematic as they were not even reproduced by the simulation, and this result was largely insensitive to changes in the PBL scheme, MP scheme, vertical resolution, and CCN concentration. Regardless of the parameterization, a critical factor that leads to the success of the simulation is that the large-scale environment is represented adequately. Also, one can argue that the more sophisticated a parameterization is, the more accurate the initial conditions need to be. This is particularly true for prognostic schemes that can suffer from stability problems (Price and Bush 2004). Therefore, the widespread failure shown in this study suggests that the problem resides in the representation of the physical processes that govern the boundary layer dynamics: the surface fluxes, boundary layer turbulence, wind shear, entrainments, inversion strength, and subsidence rate. However, a major impediment for further evaluation of these simulations is that in situ BL observations of these cases do not exist. Satellite observations, such as the Atmospheric Infrared Sounder (AIRS) or the Infrared Atmospheric Sounding Interferometer (IASI), are limited in the boundary layer.

To further explore these issues, Fig. 13 shows the profiles of simulated thermodynamic variables over point 44°S along track 3 of case B. As discussed before, a capping inversion is reproduced in the simulated potential temperature profile (Fig. 13a, simulation TH) between 1.5 and 2 km, which is notably higher than could be presumed from CALIPSO observation (~1000 m). However, simply elevating the height of the inversion without further mixing should have led to a deeper cloud deck, which was not simulated. An estimate of the inversion strength using the method introduced by Wood and Bretherton (2006) reveals that the simulated inversion intensity is comparable to that observed over the subtropics (3–6 K). The simulated large-scale subsidence rate (represented by the vertical velocity at 700 hPa) is roughly a factor of 2 higher than that observed over the subtropical marine boundary layer (Myers and Norris 2013). This difference is not unreasonable given that the subtropical oceans are largely dominated by Hadley circulation while our cases are of midlatitude cyclones.

Fig. 13.

Profiles of simulated thermodynamic variables at 44°S for case B track 3 with TH (solid lines) and TH_MYJ (dashed lines) configurations. (a) Simulated potential temperatures, (b) simulated dewpoint temperatures, (c) simulated relative humidity, (d) simulated wind speeds, and (e) simulated wind directions.

Fig. 13.

Profiles of simulated thermodynamic variables at 44°S for case B track 3 with TH (solid lines) and TH_MYJ (dashed lines) configurations. (a) Simulated potential temperatures, (b) simulated dewpoint temperatures, (c) simulated relative humidity, (d) simulated wind speeds, and (e) simulated wind directions.

The most obvious and superficial explanation for the absence of BL clouds in the simulations is a lack of moisture in the boundary layer, as suggested by the simulated relative humidity profile (Fig. 13c), which only has a peak of ~90% at ~800 m. This could be a result of either strong entrainment or weak surface fluxes. A key element that links to the moisture flux is the sea surface temperature (SST). During the first 24 h of the simulation the SST increases 1°C, although the change in association with warm/cold-air advection is ambiguous. It is also interesting to note that a decoupled structure near 800 m is clearly reproduced in the profiles of simulated wind speed and direction (Figs. 13d,e). This decoupling indicates that the boundary layer in the simulations is not well mixed, which might suggest a lack of turbulence mixing, hence the inadequate supply of moisture. The same examination is also made for simulations TH_MYJ (Fig. 13) and TH_MYNN (not shown). While the simulated inversion height is better placed, the boundary layer remains too dry to allow condensate.

Again, without any in situ observations of the boundary layer, our discussions above are simply speculation. Hande et al. (2012) employed the upper-air soundings at Macquarie Island to look for systematic biases in ERA-Interim reanalysis in the low-level thermodynamics over the SO and noted a large discrepancy in the boundary layer wind shear that potentially has an effect on entrainment and surface fluxes. They further noted a bias in the boundary layer decoupling, cloud placement (BL clouds are found to commonly be unassociated with the height of main inversion), and the boundary layer inversion. These biases might significantly affect the representation of boundary layer processes.

The aim of this study is to highlight that the representation of the widespread boundary layer clouds over the SO remains a formidable challenge for the state-of-the-art model simulations. Large-scale forcing and boundary layer processes are the crucial factors that significantly influence the success of simulations. To fully investigate exactly from where the model errors are likely to originate, more comprehensive observations are desired. This analysis supports the call for a dedicated long-term field campaign with the application of ground-based radar–lidar–microwave radiometer and aircraft in situ observations, which would offer tantalizing possibilities to address these challenges.

Acknowledgments

This work is supported by Australian Research Council Linkage Project LP120100115. The authors thank the three anonymous reviewers for their constructive insights and comments. The Australian National Computational Infrastructure is also thanked for providing the computational resources.

REFERENCES

REFERENCES
Bates
,
T. S.
,
B. J.
Huebert
,
J. L.
Gras
,
F. B.
Griffiths
, and
P. A.
Durkee
,
1998
:
International Global Atmospheric Chemistry (IGAC) Project's First Aerosol Characterization Experiment (ACE 1): Overview
.
J. Geophys. Res.
,
103
(
D13
),
16 297
16 318
.
Bennartz
,
R.
,
2007
:
Global assessment of marine boundary layer cloud droplet number concentration from satellite
.
J. Geophys. Res.
,
112
, D02201, doi:10.1029/2006JD007547.
Bodas-Salcedo
,
A.
,
K.
Williams
,
P.
Field
, and
A.
Lock
,
2012
:
The surface downwelling solar radiation surplus over the Southern Ocean in the Met Office model: The role of midlatitude cyclone clouds
.
J. Climate
,
25
,
7467
7486
.
Boers
,
R.
,
J. B.
Jensen
, and
P. B.
Krummel
,
1998
:
Microphysical and short-wave radiative structure of marine stratocumulus clouds over the Southern Ocean: Summer results and seasonal differences
.
Quart. J. Roy. Meteor. Soc.
,
124
,
151
168
.
Bretherton
,
C. S.
, and
Coauthors
,
1999
:
An intercomparison of radiatively driven entrainment and turbulence in a smoke cloud, as simulated by different numerical models
.
Quart. J. Roy. Meteor. Soc.
,
125
,
391
423
,
doi:10.1002/qj.49712555402
.
Chen
,
F.
, and
J.
Dudhia
,
2001
:
Coupling an advanced land-surface/hydrology model with the Penn State/NCAR MM5 modeling system. Part I: Model description and implementation
.
Mon. Wea. Rev.
,
129
,
569
585
.
Choi
,
Y.
,
C.
Ho
,
S.
Kim
, and
R. S.
Lindzen
,
2010
: Observational diagnosis of cloud phase in the winter Antarctic atmosphere for parameterizations in climate models. Adv. Atmos. Sci.,27 (6), 1233–1245.
Curry
,
J. A.
, and
E. E.
Ebert
,
1992
:
Annual cycle of radiation fluxes over the Arctic Ocean: Sensitivity to cloud optical properties
.
J. Climate
,
5
,
1267
1280
.
Delanoë
,
J.
, and
R. J.
Hogan
,
2010
:
Combined CloudSat-CALIPSO-MODIS retrievals of the properties of ice clouds
.
J. Geophys. Res.
,
115
, D00H29,
doi:10.1029/2009JD012346
.
Dudhia
,
J.
,
1989
:
Numerical study of convection observed during the winter monsoon experiment using a mesoscale two-dimensional model
.
J. Atmos. Sci.
,
46
,
3007
3107
.
Franklin
,
C. N.
,
Z.
Sun
,
B.
Bi
,
M.
Dix
,
H.
Yan
, and
A.
Bodas-Salcedo
,
2013
:
Evaluation of clouds in ACCESS using the satellite simulator package COSP: Global, seasonal, and regional cloud properties
.
J. Geophys. Res. Atmos.
,
118
,
732
748
,
doi:10.1029/2012JD018469
.
Garrett
,
T. J.
, and
C.
Zhao
,
2006
:
Increased Arctic cloud longwave emissivity associated with pollution from mid-latitudes
.
Nature
,
440
,
787
789
,
doi:10.1038/nature04636
.
Gettelman
,
A.
,
H.
Morrison
, and
S. J.
Ghan
,
2008
:
A new two-moment bulk stratiform cloud microphysics scheme in the Community Atmosphere Model, version 3 (CAM3). Part II: Single-column and global results
.
J. Climate
,
21
,
3660
3679
.
Gras
,
J. L.
,
1995
:
CN, CCN and particle size in Southern Ocean air at Cape Grim
.
Atmos. Res.
,
35
,
233
251
.
Greenwald
,
T. J.
,
2009
: A 2 year comparison of AMSR-E and MODIS cloud liquid water path observations. Geophys. Res. Lett.,36, L20805, doi:10.1029/2009GL040394.
Guan
,
H.
,
S. G.
Cober
, and
G. A.
Isaac
,
2001
:
Verification of supercooled cloud forecasts with in situ aircraft measurements
.
Wea. Forecasting
,
16
,
145
155
.
Han
,
Q. Y.
,
W. B.
Rossow
,
J.
Chou
, and
R. M.
Welch
,
1998
:
Global survey of the relationships of cloud albedo and liquid water path with droplet size using ISCCP
.
J. Climate
,
11
,
1516
1528
.
Hande
,
L. B.
,
S. T.
Siems
,
M. J.
Manton
, and
D.
Belusic
,
2012
:
Observations of wind shear over the Southern Ocean
.
J. Geophys. Res.
,
117
, D12206,
doi:10.1029/2012JD017488
.
Hannay
,
C.
,
D. L.
Williamson
,
J. J.
Hack
,
J. T.
Kiehl
,
J. G.
Olson
,
S. A.
Klein
,
C. S.
Bretherton
, and
M.
Köhler
,
2009
:
Evaluation of forecasted southeast Pacific stratocumulus in the NCAR, GFDL, and ECMWF models
.
J. Climate
,
22
,
2871
2889
.
Haynes
,
J. M.
,
Z.
Luo
,
G. L.
Stephens
,
R. T.
Marchand
,
A.
Bodas-Salcedo
,
2007
:
A multipurpose radar simulation package: QuickBeam
.
Bull. Amer. Meteor. Soc.
,
88
,
1723
1727
.
Haynes
,
J. M.
,
C.
Jakob
,
W. B.
Rossow
,
G.
Tselioudis
, and
J.
Brown
,
2011
:
Major characteristics of Southern Ocean cloud regimes and their effects on the energy budget
.
J. Climate
,
24
,
5061
5080
.
Holz
,
R. E.
,
S. A.
Ackerman
,
F. W.
Nagle
,
R.
Frey
,
S.
Dutcher
,
R. E.
Kuehn
,
M. A.
Vaughan
, and
B.
Baum
,
2008
:
Global Moderate Resolution Imaging Spectroradiometer (MODIS) cloud detection and height evaluation using CALIOP
.
J. Geophys. Res.
,
113
,
D00A19
,
doi:10.1029/2008JD009837
.
Hong
,
S. Y.
, and
J. O.
Lin
,
2006
:
The WRF single-moment 6-class microphysics scheme (WSM6)
.
J. Korean Meteor. Soc.
,
42
(
2
),
129
151
.
Hong
,
S. Y.
,
Y.
Noh
, and
J.
Dudhia
,
2006
:
A new vertical diffusion package with an explicit treatment of entrainment processes
.
Mon. Wea. Rev.
,
134
,
2318
2341
.
Hu
,
Y.
, and
Coauthors
,
2009
:
CALIPSO/CALIOP cloud phase discrimination algorithm
.
J. Atmos. Oceanic Technol.
,
26
,
2293
2309
.
Hu
,
Y.
,
S.
Rodier
,
K.
Xu
,
W.
Sun
,
J.
Huang
,
B.
Lin
,
P.
Zhai
, and
D.
Josset
,
2010
:
Occurrence, liquid water content, and fraction of supercooled water clouds from combined CALIOP/IIR/MODIS measurements
.
J. Geophys. Res.
,
115
, D00H34,
doi:10.1029/2009JD012384
.
Huang
,
Y.
,
S. T.
Siems
,
M. J.
Manton
,
A.
Protat
, and
J.
Delanoë
,
2012a
:
A study on the low-altitude clouds over the Southern Ocean using the DARDAR-MASK
.
J. Geophys. Res.
,
117
,
D18204
,
doi:10.1029/2012JD017800
.
Huang
,
Y.
,
S. T.
Siems
,
M. J.
Manton
,
L. B.
Hande
, and
J. M.
Haynes
,
2012b
:
The structure of low-altitude clouds over the Southern Ocean as seen by CloudSat
.
J. Climate
,
25
,
2535
2546
.
Iacono
,
M. J.
,
J. S.
Delamere
,
E. J.
Mlawer
,
M. W.
Shephard
,
S. A.
Clough
, and
W. D.
Collins
,
2008
:
Radiative forcing by long-lived greenhouse gases: Calculations with the AER radiative transfer models
.
J. Geophys. Res.
,
113
,
D13103
,
doi:10.1029/2008JD009944
.
Jacobson
,
M.
,
2006
:
Effects of externally-through-internally-mixed soot inclusions within clouds and precipitation on global climate
.
J. Phys. Chem.
,
110
,
6860
6873
,
doi:10.1021/jp056391r
.
Janjic
,
Z. I.
,
2002
: Nonsingular implementation of the Mellor–Yamada level 2.5 scheme in the NCEP Meso Model. NCEP Office Note 437, 61 pp.
Jiang
,
H.
,
G.
Feingold
, and
A.
Sorooshian
,
2010
:
Effect of aerosol on the susceptibility and efficiency of precipitation in warm trade cumulus clouds
.
J. Atmos. Sci.
,
67
,
3525
3540
.
Jiang
,
H.
,
W. R.
Cotton
,
J. O.
Pinto
,
J. A.
Curry
, and
M. J.
Weissbluth
,
2000
:
Cloud resolving simulations of mixed-phase Arctic stratus observed during BASE: Sensitivity to concentration of ice crystals and large-scale heat and moisture advection
.
J. Atmos. Sci.
,
57
,
2105
2117
.
Key
,
J. R.
, and
J. M.
Intrieri
,
2000
:
Cloud particle phase determination with the AVHRR
.
J. Appl. Meteor.
,
39
,
1797
1804
.
Larson
,
V. E.
,
D. P.
Schanen
,
M.
Wang
,
M.
Ovchinnikov
, and
S.
Ghan
,
2012
: PDF parameterization of boundary layer clouds in models with horizontal grid spacing from 2 to 16 km. Mon. Wea. Rev.,140, 285–306.
L'Ecuyer
,
T. S.
,
W.
Berg
,
J.
Haynes
,
M.
Lebsock
, and
T.
Takemura
,
2009
:
Global observations of aerosol impacts on precipitation occurrence in warm maritime clouds
.
J. Geophys. Res.
,
114
,
D09211
,
doi:10.1029/2008JD011273
.
Lin
,
Y. L.
,
R. D.
Farley
, and
H. D.
Orville
,
1983
:
Bulk parameterization of the snow field in a cloud model
.
J. Climate Appl. Meteor.
,
22
,
1065
1092
.
Mace
,
G. G.
,
R.
Marchand
,
Q.
Zhang
, and
G.
Stephens
,
2007
:
Global hydrometeor occurrence as observed by CloudSat: Initial observations from summer 2006
.
Geophys. Res. Lett.
,
34
,
L09808
,
doi:10.1029/2006GL029017
.
Mellor
,
G. L.
, and
T.
Yamada
,
1982
:
Development of a turbulence closure model for geophysical fluid problems
.
Rev. Geophys. Space Phys.
,
20
,
851
875
.
Mlawer
,
E. J.
,
S. J.
Taubman
,
P. D.
Brown
,
M. J.
Iacono
, and
S. A.
Clough
,
1997
:
Radiative transfer for inhomogeneous atmosphere: RRTM, a validated correlated-k model for the longwave
.
J. Geophys. Res.
,
102
(
D14
),
16 663
16 682
.
Morrison
,
A. E.
,
S. T.
Siems
, and
M. J.
Manton
,
2010
:
A modeling case study of mixed phase clouds over the Southern Ocean and Tasmania
.
Mon. Wea. Rev.
,
138
,
839
862
.
Morrison
,
A. E.
,
S. T.
Siems
, and
M. J.
Manton
,
2011
:
A three-year climatology of cloud-top phase over the Southern Ocean and North Pacific
.
J. Climate
,
24
,
2405
2418
.
Morrison
,
H.
, and
J. O.
Pinto
,
2005
:
Mesoscale modeling of springtime Arctic mixed-phase stratiform clouds using a new two-moment bulk microphysics scheme
.
J. Atmos. Sci.
,
62
,
3683
3704
.
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
.
Myers
,
T. A.
, and
J. R.
Norris
,
2013
: Observational evidence that enhanced subsidence reduces subtropical marine boundary layer cloudiness. J. Climate,26, 7507–7524
Nakanishi
,
M.
, and
H.
Niino
,
2004
:
An improved Mellor–Yamada level-3 model with condensation physics: Its design and verification
.
Bound.-Layer Meteor.
,
112
,
1
13
.
Nakanishi
,
M.
, and
H.
Niino
,
2006
: An improved Mellor–Yamada level-3 model: Its numerical stability and application to a regional prediction of advection fog. Bound.-Layer Meteor.,119, 397–407.
O'Dell
,
C. W.
,
F. J.
Wentz
, and
R.
Bennartz
,
2008
:
Cloud liquid water path from satellite-based passive microwave observations: A new climatology over the global oceans
.
J. Climate
,
21
,
1721
1739
.
Pan
,
H.
, and
W. S.
Wu
,
1995
: Implementing a mass flux convection parameterization package for the NMC medium-range forecast model. Office Note 409, NOAA/EMC, 40 pp.
Petters
,
J. L.
,
J. Y.
Harrington
, and
E. E.
Clothiaux
,
2012
:
Radiative-dynamical feedbacks in low liquid water path stratiform clouds
.
J. Atmos. Sci.
,
69
,
1498
1512
.
Platnick
,
S.
,
M. D.
King
,
S. A.
Ackerman
,
W. P.
Menzel
,
B. A.
Baum
,
J. C.
Riedi
, and
R. A.
Frey
,
2003
:
The MODIS cloud products: Algorithms and examples from Terra
.
IEEE Trans. Geosci. Remote Sens.
,
41
(
2
),
459
473
,
doi:10.1109/TGRS.2002.808301
.
Price
,
J. D.
, and
M. R.
Bush
,
2004
: Comparison between model forecast and observed boundary layer profiles and related comments on cloud prediction. Wea. Forecasting,19, 959–969.
Reisner
,
R.
,
R. M.
Rasmussen
, and
R. T.
Bruintjes
,
1998
:
Explicit forecasting of supercooled liquid water in winter storms using the MM5 mesoscale model
.
Quart. J. Roy. Meteor. Soc.
,
124
,
1071
1107
.
Ryan
,
B. F.
, and
W. D.
King
,
1997
:
A critical review of the Australian experience in cloud seeding
.
Bull. Amer. Meteor. Soc.
,
78
,
239
254
.
Sedlar
,
J.
,
M. D.
Shupe
, and
M.
Tjernström
,
2012
:
On the relationship between thermodynamic structure and cloud top, and its climate significance in the Arctic
.
J. Climate
,
25
,
2374
2393
.
Seethala
,
C.
, and
Á.
Horváth
,
2010
:
Global assessment of AMSR-E and MODIS cloud liquid water path retrievals in warm oceanic clouds
.
J. Geophys. Res.
,
115
, D13202,
doi:10.1029/2009JD012662
.
Shupe
,
M. D.
, and
J. M.
Intrieri
,
2004
:
Cloud radiative forcing of the Arctic surface: The influence of cloud properties, surface albedo, and solar zenith angle
.
J. Climate
,
17
,
616
628
.
Simmonds
,
I.
, and
K.
Keay
,
2000
:
Mean Southern Hemisphere extratropical cyclone behavior in the 40-year NCEP–NCAR reanalysis
.
J. Climate
,
13
,
873
885
.
Skamarock
,
W. C.
, and
Coauthors
,
2008
: A description of the Advanced Research WRF version 3. NCAR Tech. Note NCAR/TN-475+STR, 113 pp.
Smith
,
E. J.
,
L. G.
Veitch
,
D. E.
Shaw
, and
A. J.
Miller
,
1979
:
A cloud-seeding experiment in Tasmania
.
J. Appl. Meteor.
,
18
,
804
815
.
Stephens
,
G. L.
,
1978
:
Radiation profiles in extended water clouds: 1. Theory
.
J. Atmos. Sci.
,
35
,
2111
2122
.
Stephens
,
G. L.
, and
Coauthors
,
2002
:
The CloudSat Mission and the A-Train: A new dimension of space-based observations of clouds and precipitation
.
Bull. Amer. Meteor. Soc.
,
83
,
1771
1790
.
Storelvmo
,
T.
,
J. E.
Kristjánsson
, and
U.
Lohmann
,
2008
:
Aerosol influence on mixed-phase clouds in CAM-Oslo
.
J. Atmos. Sci.
,
65
,
3214
3230
.
Storelvmo
,
T.
,
C.
Hoose
, and
P.
Eriksson
,
2011
:
Global modeling of mixed phase clouds: The albedo and lifetime effects of aerosols
.
J. Geophys. Res.
,
116
, D05207,
doi:10.1029/2010JD014724
.
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
.
Trenberth
,
K. E.
, and
J. T.
Fasullo
,
2010
:
Simulation of present-day and twenty-first-century energy budgets of the Southern Oceans
.
J. Climate
,
23
,
440
454
.
Vaillancourt
,
P. A.
,
A.
Tremblay
,
S. G.
Cober
, and
G. A.
Isaac
,
2003
:
Comparison of aircraft observations with mixed-phase cloud simulations
.
Mon. Wea. Rev.
,
131
,
656
671
.
Winker
,
D. M.
,
B. H.
Hunt
, and
M. J.
McGill
,
2007
:
Initial performance assessment of CALIOP
.
Geophys. Res. Lett.
,
34
,
L19803
,
doi:10.1029/2007GL030135
.
Wood
,
R.
,
2012
:
Stratocumulus clouds
.
Mon. Wea. Rev.
,
140
,
2373
2423
.
Wyant
,
M. C.
, and
Coauthors
,
2010
:
The PreVOCA experiment: Modeling the lower troposphere in the southeast Pacific
.
Atmos. Chem. Phys.
,
10
,
4757
4774
,
doi:10.5194/acp-10-4757-2010
.
Yum
,
S. S.
, and
J. G.
Hudson
,
2004
:
Wintertime/summertime contrasts of cloud condensation nuclei and cloud microphysics over the Southern Ocean
.
J. Geophys. Res.
,
109
, D06204, doi:10.1029/2003JD003864.
Yum
,
S. S.
, and
J. G.
Hudson
,
2005
:
Adiabatic predictions and observations of cloud droplet spectral broadness
.
Atmos. Res.
,
73
,
203
223
,
doi:10.1016/j.atmosres.2004.10.006
.
Zelinka
,
M. D.
, and
D. L.
Hartmann
,
2012
:
Climate feedbacks and their implications for poleward energy flux changes in a warming climate
.
J. Climate
,
25
,
608
624
.