03-moist_bubble.md

Next: How to analyse output data, Previous: Setting up the environment, Up: Main page

How to prepare a simulation

In this workshop we simulate a rising moist bubble. The liquid-water buoyancy $b_l$ and the specific humidity $q$ distribution inside the moist bubble are given by

$$\begin{align} b_{l}(\vec{x}') &= b_{\circ} \left( 1 + \frac{0.3 x'y' -0.4 x'z' + 0.5 y'z'}{R^2} \right) S(h) \\\ q(\vec{x}') &= q_n + (q_{\circ}-q_n)S(h)\,, \end{align}$$

with edge-smoothing function

$$S(h) = \left\{ \begin{align} 1 &{}\qquad h \leq 0 \\\ 1 - 10 h^3 + 15 h^4 - 6 h^5 &{}\qquad 0 < h < 1 \\\ 0 &{}\qquad h \geq 1 \end{align} \right.$$

and $h\equiv(|\vec{x}'|/R-f_s)/(1-f_s)$, we choose $f_s=0.8$ and bubble radius $R = 800$. In the figure below you can see a cross section of the initial liquid buoyancy (left panel) and specific humidity (centre panel). The right panel of the figure shows the basic state stratification. This figure is taken from Dritschel et al (2018). For further information about the example, see also section 3.4 in Frey et al (2023).

We provide you with a Python script that creates the gridded input fields for this setup. After following the instructions on how to load the Python virtual environment, you can run the script input/write_moist_setup.py using the following command:

python write_moist_setup.py

which creates a file called moist_<nx>x<ny>x<nz>.nc, where <nx>, <ny> and <nz> are replaced by the number of grid cells per dimension (default: nx = ny = nz = 64). You can call the script with the --help argument to get further information.

How to run a simulation

The basic command to run a three-dimensional EPIC simulation is

epic3d --config <file.config>

where <file.config> is a placeholder for a configuration file that specifies all simulation parameters. The configuration file for the moist bubble test case is already given in input/moist.config.

The argument field_file which is currently set to 'moist_64x64x64.nc' points to a netCDF file which contains the initial gridded input data as well as domain specifications and physical quantities.

Cirrus uses the SLURM job scheduling system. To run a simulation please use the provided batch script. A job is submitted with

sbatch submit-job.sh

Important

Our resources for this workshop are limited, so we kindly ask users to only use a maximum of 1 computing node per job. In addition, jobs should not run longer than 20 minutes.

Tip

You can check the status of your submitted jobs with

squeue --me

A submitted or running job with id <jobid> is cancelled calling

scancel <jobid>

Next: How to analyse output data, Previous: Setting up the environment, Up: Main page