HTML
-
The case study period consisted of small-scale roll clouds embedded in a much larger-scale environment that supported their formation.
-
This study focused on a case of Arctic roll clouds or cloud streets that occurred on 2 May 2013. The atmospheric conditions for this case were analyzed with ERA-Interim data with a spatial grid of 0.75° latitude by 0.75° longitude. At 0000 UTC on 2 May 2013 there was a dominant high pressure system over the Chukchi Peninsula and Chukchi Sea, along with a low pressure system at 500 hPa over the Canadian Arctic Archipelago with a trough that extended toward the southwest, just southeast of Utqiaġvik (Fig. 1). The prevailing wind direction at Utqiaġvik was from the northeast, advecting 850-hPa moisture into the Utqiaġvik region (Fig. 1b). As Fig. 1c shows, the thickness between 500 hPa and 1000 hPa decreased from the northwest to southeast of the domain. A warm air mass was advected into the region from the north near the surface, associated with a surface warm front in the vicinity of Utqiaġvik at this time.
Figure 1. ERA-Interim data at 0000 UTC 2 May 2013: (a) 500-hPa geopotential height (units: m, black contours), (b) 850-hPa geopotential height (units: m, black contours) and relative humidity (units: %, blue shading, with darker blues indicating higher relative humidity), and (c) 1000–500-hPa thickness (units: m, red dashed contours) and sea level pressure (units: hPa, black contours). The yellow stars indicate the location of Utqiaġvik and the red Ls and blue H indicate the locations of the low and high pressure centers.
The Moderate Resolution Imaging Spectroradiometer (MODIS) onboard the NASA Terra and Aqua satellites frequently passed over Utqiaġvik and its vicinity during the case study period. MODIS band 1 (band-center wavelength of 0.65 μm) visible imagery collected at 0015 UTC 2 May and 0445 UTC 2 May is presented in Figs. 2a and b. At this time, the Beaufort and Chukchi seas were covered by sea ice but with leads evident in the sea ice throughout the region, and the North Slope of the Alaskan land surface was covered by snow.
Figure 2. Observed radiances from Terra MODIS band 1 in the vicinity of Utqiaġvik at (a) 0015 UTC 2 May and (b) 0445 UTC 2 May. (c, d) Radiances for the red boxes in (a, b), respectively. The red dots denote the location of Utqiaġvik.
Figures 2a and b reveal that many of the cloud layers to the west and south of Utqiaġvik either contained evidence of waves or were cloud streets in their entirety from 0015 UTC 2 May to 0445 UTC 2 May 2013. At 0015 UTC 2 May, Utqiaġvik was at the eastern edge of the frontal cloud band with cloud streets occurring to the west of the frontal cloud band (Fig. 2a, red box). At 0445 UTC the frontal cloud band had broadened, filled in, and moved to the east, covering Utqiaġvik in a deck of clouds. At this time, lower altitude roll clouds were present to the west of Utqiaġvik under a gap in the cloud deck (Fig. 2b, red box). To estimate the wavelengths of the cloud streets to the west of Utqiaġvik, we used Figs. 2c and d, which show detailed features of the cloud streets within the red boxes in Figs. 2a and b. We obtained a wavelength of approximately 2.5–2.8 km for the cloud streets to the west of Utqiaġvik.
-
We used a numerical weather model to investigate the formation mechanisms of the roll clouds identified in Figs. 2c and d along with the influence of different model spatial resolutions on the model results. The numerical weather model employed in this study was the Weather Research and Forecasting (WRF) model (Skamarock et al., 2008), version 3.6.1. The model spatial resolutions in our experiments ranged from the mesoscale to LEP scale in order to capture both the large-scale environment and cloud-resolving scales of the roll clouds evident on 2 May 2013.
We implemented six one-way nested domains over the northern part of Alaska, the Beaufort Sea, and the Chukchi Sea starting from a horizontal grid spacing of 27 km on the outer domain and working in domain ratios of 3:1 down to an inner domain with 111-m grid spacing (Fig. 3). Domains D01 through D06 had spatial grid spacing of 27 km, 9 km, 3 km, 1 km, 0.333 km, and 0.111 km. By using one-way nested domains, no information from the inner domains was passed back to the parent domains and the results from each domain illustrate what was produced at that grid spacing. We used terrain-following vertical levels with 33-m grid spacing within the boundary layer. The vertical grid spacing increased with increasing altitude up to the top of the model at 50 hPa, where the grid spacing was about 23.6 hPa. Overall, there were a total of 130 vertical layers.
Figure 3. WRF model domain for the mesoscale and large-eddy simulations. Shading shows the terrain height.
The five outer domains D01–D05 ran as mesoscale domains while the inner domain D06 ran as an LEP domain. The physical parameterizations used in the simulations were as follows. The Grell 3D scheme was chosen for the cumulus parameterization for domains D01 and D02, and the Morrison double-moment scheme was chosen for the microphysics in all domains. The Noah land surface model and NCAR Community Atmospheric Model radiation schemes were used for land surface and atmospheric radiation processes, respectively.
For a mesoscale model the covariance terms (e.g.,
$\overline {u_i'u_j'} $ ) that represent the effects of turbulence on the mean motion in the momentum equations are unknowns and must be parameterized by a planetary boundary layer (PBL) scheme. In LEP, a subgrid-scale (SGS) parameterization is still used to represent the processes of the turbulence smaller than the grid scale. To resolve most of the large energy-containing turbulence, the grid spacing must be much smaller than the large energy-containing eddies (e.g., Wyngaard, 2004).For our case, the Mellor–Yamada–Janjic scheme, which is a turbulent kinetic energy based PBL scheme, was used for domains D01 through D05. We designed domain D06 to resolve large eddies, so we turned off the PBL scheme within this inner domain. Subgrid eddy diffusion within this LEP domain was based on the three-dimensional LES turbulent kinetic energy closure of Deardorff (1980) (Table 1).
Domain Gird spacing Domain size PBL/SGS option Parent domain D01 (PBL) 27 km 65 × 65 grid cells MYJ PBL D02 (PBL) 9 km 94 × 94 grid cells MYJ PBL D01 (PBL) D03 (PBL) 3 km 148 × 149 grid cells MYJ PBL D02 (PBL) D04 (PBL) 1 km 187 × 148 grid cells MYJ PBL D03 (PBL) D05 (PBL) 0.333 km 250 × 202 grid cells MYJ PBL D04 (PBL) D05 (LEP) 0.333 km 250 × 202 grid cells 3D LES Subgrid TKE Closure of Deardorff (1980) D04 (PBL) D06 (LEP) 0.111 km 301 × 250 grid cells 3D LES Subgrid TKE Closure of Deardorff (1980) D05 (PBL) Table 1. Dimensions, grid spacing, and PBL/SGS option for the model domains.
To examine if the horizontal resolution of domain D06 was necessary for resolving the observed roll clouds, we ran another experiment with the four outer domains (D01–D04) using the same configuration as the experiment mentioned above but with the fifth domain D05 as an LEP domain, with the same settings as the LEP domain D06. Table 1 shows that domain D05 was run both as a mesoscale domain with a PBL parameterization, D05 (PBL), and also as an LEP domain, D05 (LEP).
The initial and boundary conditions used in the simulations were a combination of water vapor mixing ratios from ERA-Interim reanalysis and the remaining meteorological variables from the NCEP Global Forecast System final analysis, because sensitivity tests with this combination led to results that best matched observations of the horizontal wind fields and vertical moisture distributions. The data prescribed sea ice across the domain, except for the southern part of the Chukchi Sea, and snow covered the model land surfaces. All of the domains in the simulations started simultaneously at 1800 UTC 1 May 2013 and ran for six hours to 0000 UTC 2 May 2013.