## Set up of the ASTEX large-eddy simulation

**Initial pertubations and translation velocity**

The LES model is initialised with random fluctuations of θ_{L} and q_{T} given by:

θ_{L} : [-0.1,+0.1] (K)

q_{T} : [-2.5*10^{-2}, +2.5*10^{-2}] (g kg^{-1})

In order to minimize numerical errors associated with advection we propose to translate the model domain with -2 and -7 ms^{-1} in the x and y direction, respectively.

**Radiation**

Like the CGILS experiments, we would like to calculate the radiation fields from the ground surface to the top of the atmosphere. As a suggestion, you may use the freely available radiation code used in the UCLA LES (Bjorn Stevens) which is open source. In addition, there is also a code available that is developed by Peter Blossey (University of Washington/Center for Multiscale Modeling of Atmospheric Processes) for the CGILS experiment. The value for the solar constant S_{0} = 1376 Wm^{-2}. Marat Khairoutdinov has kindly provided a piece of FORTRAN code to calculate a zenith angle dependent sea surface albedo (Briegleb, 1992). Please use the mean volume radius (r_{vol}) to compute the effective radius as follows, r_{eff}=1.03 r_{vol}. The factor 1.03 is based on a lognormal distribution for the cloud droplet sizes having a width σ_{g}= 1.2, and was kindly provided by Andy Ackerman. Note r_{vol} = ( 3ρ_{air}q_{liq} / (4πρ_{water}N_{c}) )^{1/3}
in which ρ_{air} is density of dry air, q_{liq} is mass mixing ratio of
cloud water, ρ_{water} is the density of liquid water and N_{c} is the cloud droplet
number concentration in units of #/volume of air.

**Domain Parameters and Boundary Conditions**

**Horizontal domain size**: 4.48 x 4.48 km^{2}**Number of grid points in the horizontal directions**: Nx = Ny = 128, implying a horizontal resolution: Δx = Δy = 35 m- For the
**vertical levels**we would like to use a vertical grid spacing of 5 m in order to resolve the inversion. However, as the inversion height grows from ~ 600 to about ~ 2000 m the use of a static grid requires this resolution to be used between these levels. Irina Sandu has written a short fortran code that generates a non-equidistant vertical grid which we would like to use for the ASTEX transition. Note that the last flight the observed cloud top height was below 2000 m, and if your model tends to reproduce this (for example in a test run with a coarser vertical resolution) you may start the vertical stretching from a lower height than generated by the proposed grid generator.

In addition to this control run, we would like to receive simulation results on a large horizontal domain (~ 25x25 km^{2}). Running such a case can be computionally expensive, and depending on your computational resources we ask for a simulation using at least 256^3 grid points.

**Lateral boundary condition**: Periodic**Top boundary condition**: In order to minimize spurious reflection of upward propagating gravity waves, you may want to use a sponge layer for damping perturbations. The sponge layer should not start any lower than 200 m above the mean inversion height.**Advection by subsidence**: To avoid the formation of wiggles near the inversion, it is strongly recommended to use an upwind scheme for the large-scale vertical advection term.