DYCOMS-II Field Campaign
Ninth GCSS Boundary Layer Cloud Working Group Workshop
9th GCSS BL Cloud Workshop Home
Overview
Output
Revisions
Results
Setup of Simulations
Model Grid
The model domain should be 6.4 × 6.4 × 1.5 km with Δx = Δy = 50 m (nx = ny = 128) and Δz = 5m near the surface and the initial inversion at 795 m. Those able to use a stretched grid should use nz = 96 with Δz following a sin2 stretching within the boundary layer, a 125-m deep region with Δz = 5 m, and Δz increasing above, as seen below and computed in this Fortran source code.
As guidance for Boussinesq models, the air density at the surface and at the initial altitude of the inversion are 1.21 and 1.12 kg/m3, respectively.
For our bin microphysics model, we use 20 bins (with geometrically progressing particle mass) to resolve each of the aerosol and cloud droplet size distributions over particle radii ranges of 0.01—5 and 1—500 μm respectively.
Initialization
The initial profiles of winds and thermodynamics, in which the liquid-water potential temperature &thetal = (p0/p)Rd/cp(T—L ql/cp), with p0 = 1000 hPa, Rd = 83143/28.966 = 287.04 J/kg/K, cp = 1004 J/kg/K, and L = 2.5 MJ/kg, and the mixing ratio of total moisture qt = qv + ql are given below (thanks to Margreet van Zanten and Bjorn Stevens for providing these measurements), in which z is in units of meters, inversion height zi = 795 m, and surface pressure psfc = 1017.8 hPa.
| u = | 3 + 4.3 z / 1000 m/s | ||||
| v = | –9 + 5.6 z / 1000 m/s | ||||
| &thetal = |
|
||||
| qt = |
|
We suggest translating the model domain to minimize numerical errors
associated with advection; we translate the DHARMA domain at
5 and -5.5 m/s in the x and y directions, respectively.
We pseudo-randomly perturb initial temperatures within the boundary layer
about their mean values with an amplitude of 0.1 K. We suggest
initializing TKE in prognostic subgrid-scale models at
1 m2/s2.
Models with parameterized precipitation should use a uniform cloud droplet concentration N = 55 cm-3, if possible. If sedimentation of cloud droplets is not already treated in a model, it should be treated by assuming a log-normal size distribution (geometric standard deviation &sigmag = 1.5) of droplets falling in a Stokes regime, in which the sedimentation flux is given by F = c (3/(4 π&rholN))2/3 (ρql)5/3 exp( 5 ln2&sigmag ) where c = 1.19 × 108 m-1 s-1 (Rogers and Yau, 1989), and &rhol and &rho are the density of liquid water and air. The divergence of the sedimentation flux should appear as sink and source terms in the qt and &thetal equations.
Models that predict the number concentration of droplets and thus require CCN as input should assume the bimodal distribution described below. The aerosol number concentrations in the two modes should be adjusted, through trial and error, to result in an average droplet concentration of N = 55 cm-3 in cloudy grid cells. It would be preferable to keep the total aerosol number fixed at 190 cm-3 when adjusting the numbers in the two modes; in practice it is unlikely that the particles in the nucleation mode (the smaller mode) will ever be activated, so it is likely that the total aerosol number in the accumulation mode is all that matters.
The aerosol size distribution within each grid cell should be diagnosed
at the beginning of
each microphysical (sub)step by subtracting the number of droplets
in the grid cell from the large end of the initial aerosol size
distribution. The aerosol are assumed to be
composed of ammonium bisulfate, which gives a
good fit to the CCN spectra below the cloud, as seen below
(thanks to Markus Petters and Jefferson Snider for providing
the UWyo aerosol and CCN data).
Its dry density is 1.78 g/cm3, molecular weight is
115 g/mol,
and two ions are dissolved per molecule.
 |
Sub-cloud measured aerosol size distribution (solid line) and bimodal log-normal fit (dotted), transformed to CCN activation spectra (right) by assuming ammonium bisulfate composition. Measured CCN median plotted in right panel with symbols +/- standard deviation. Dark solid line in left panel denotes range of CCN measurements. |
Parameters for log-normal aerosol size distributions, fit to measurements
below cloud base, are given below. The vertical profiles of all parameters
should be independent of altitude in the simulations.
| Number (cm-3) | 125 | 65 |
| Mode radius (μm) | 0.011 | 0.06 |
| Geom. Std. Deviation | 1.2 | 1.7 |
Forcings
Subsidence
A uniform value of the large-scale divergence of the horizontal winds
is assumed to be D = 3.75·10-6 s-1,
giving a large-scale vertical wind of w = -Dz, which should
appear as a source term of each variable &phi only
as w∂φ/∂z.
Coriolis
The geostrophic winds should be the same as the initial winds given above,
and the latitude is 31.5°N.
Radiation
Radiative heating rates should be computed every time step from the
divergence of the radiative flux profile in each model column
using the same parameterization as in the
8th BL Cloud Workshop:
where Q(z1,z2) is the vertical integral of (&kappa &rho ql) from z1 to z2, &alpha = 1 K/m1/3, ρi = 1.12 g/m3 (air density at the initial height of the inversion), H is the Heaviside step function, zi is the altitude of the lowermost crossing of the qt = 8 g/kg isosurface in each model column, and the parameters &kappa, F0 and F1 have the same values used in the 8th BL Cloud Workshop, as given below.
| &kappa | 85 m2/kg |
| F0 | 70 W/m2 |
| F1 | 22 W/m2 |
Note that the third term in the radiative flux parameterization was chosen
to preserve
the observed (z-zi)1/3 structure in the
&thetal profile for z > zi by
balancing the large-scale subsidence under the
assumption that the radiative heating rate is given by
–1/(ρicp) ∂F/∂z.
This assumption implies that for models computing radiative heating
as –&theta/(ρcpT) ∂F/∂z
the balance between subsidence and radiative cooling will not quite
preserve the observed structure, a minor inconsistency we can live with.
Surface Fluxes
The upward surface momentum fluxes should be
< w'ui' > = –ui u*2 /|U|
where the wind component ui and the magnitude of the horizontal wind |U| are defined locally, and the friction velocity is fixed at u* = 0.25 m/s.
The upward sensible and latent heat fluxes
(separate from any precipitation flux)
should be fixed at the measured values of 16 and 93 W/m2,
respectively. A surface air density of 1.21 kg/m3 should be
used to convert the surface heat fluxes into kinematic units.
Sponge
We dampen the horizontal wind components and &thetal toward
their initial values with a nudging coefficient that increases
(with a sin2 altitude dependence) from
0 at z = 1250 m to (100 s)-1 at
the top of the domain (z = 1500 m).
Experiments
We are requesting two six-hour simulations from each group: one with no sedimentation of condensed water and one with precipitation (sedimentation of cloud water and drizzle drops). We additionally request optional simulations in which (i) only the sedimentation of cloud water is included in a simulation without drizzle, and (ii) the sedimentation of cloud water is omitted in a simulation with drizzle.
