.. _idealized-tutorial-label: Idealized LES: beyond the drycblles ============================================= .. toctree:: :maxdepth: 2 :caption: Contents: :ref:`Running your first case` describes how to run an LES of a dry convective boundary layer. This tutorial explores the building blocks of such a simulation, such as advection and diffusion and the different options that are available (for LES). This tutorial contains examples from the Weisman Klemp case (`Weisman and Klemp (1982)`_) This is a case of an idealized convective storm, which is initiated by releasing a warm bubble close to the surface. A video of a MicroHH simulation of this bubble can be found here: https://vimeo.com/901517622. For the examples, the y-dimension was reduced to one grid cell, to create a 2D simulation. Furthermore, the fifth order advection scheme ``swadvec=2i5`` and a horizontal resolution of 600 m were used, unless mentioned otherwise. .. _Weisman and Klemp (1982): https://doi.org/10.1175/1520-0493(1982)110<0504:TDONSC>2.0.CO;2 .. note:: This tutorial shows examples of commonly used options and their impact, to help you choose suitable settings for your own simulations. However, what the most suitable settings are depends on the case. Apart from looking at the examples here, you can base your choices on the available example cases or test the sensitivity of your own case yourself. Advection ----------------- The different :ref:`Advection ``[advec]``` schemes in MicroHH differ in their order of the interpolations. In general, higher order schemes are more accurate, but also come at higher costs. The odd ordered schemes have hyperdiffusion included and the ``2i5`` and ``2i62`` schemes have the possibility to provide a list of the scalars for which monotonic advection is guaranteed. Both the hyperdiffusion and the flux limiters dampen the variance and smooth out sharp gradients. In some cases, e.g. when using a coarse resolution, this can be advantageous. However, in other cases, it can hamper the existence of small scale structures and result in overly smooth fields. More information about the second order schemes can be found in `Wicker and Skamarock (2002)`_. .. _Wicker and Skamarock (2002): https://doi.org/10.1175/1520-0493(2002)130<2088:TSMFEM>2.0.CO;2 .. admonition:: Example :class: tip The plot below shows xz cross sections after half an hour of the concentration of a passive scalar that was added to the warm bubble for different advection scheme and two resolutions. .. figure:: figures/advection.png :width: 800 | Buffer ------- The :ref:`Buffer layer ``[buffer]``` is a layer at the top of the domain that prevents the reflection of gravity waves back into the domain. The buffer layer should be thick enough and the buffer strong enough to dampen the gravity waves before they reach the top of the domain. At the same time, a too strong buffer or starting the buffer at a too low level can hamper the convection below the buffer layer. The strength of the buffer layer can be estimated with :math:`\frac{N}{2 \pi}`, where N is the Brunt-Vaisala frequency :math:`N = \sqrt{\frac{g}{\theta_v} \frac{\Delta \theta_v}{\Delta z}}`. .. admonition:: Example :class: tip For the Weisman Klemp bubble, a rough estimate the buffer strength for the layer between 20 and 25 km is :math:`\frac{\sqrt{\frac{10}{570} \frac{140}{5000}}}{2 \pi}\ \approx\ 0.0035`. The figure below shows the domain mean vertical velocity variance averaged between 1 and 1.5 hours. The line show different combinations of buffer strength and height. The buffer height is also indicated by the dashed lines. .. figure:: figures/buffer.png :width: 480 | Diffusion ------------ | The :ref:`Diffusion ``[diff]``` scheme parameterizes the subgrid scale turbulence and diffusion in LES through an eddy diffusivity. The required eddy viscosity/diffusivity coefficients can be calculated in two ways: | 1. Using Smagorinsky closure (`Lilly, 1968`_, `Mason and Thomson, 1992`_). This scheme is used in most LES example cases and also in the examples shown in this tutorial. | 2. As a function of the subgrid scale TKE (`Deardorf, 1980`_). .. _Lilly, 1968: https://doi.org/10.1002/qj.49709440106 .. _Mason and Thomson, 1992: https://doi.org/10.1017/S0022112092002271 .. _Deardorf, 1980: https://doi.org/10.1007/BF00119502 .. admonition:: Example :class: tip The figure below shows the variances in the horizontal and vertical velocity in simulations with the Smagorinsky and TKE scheme. .. figure:: figures/diffusion.png :width: 800 | .. note:: By default, MicroHH uses an adaptive time step. In some cases when a simulation crashes, it can help to enforce a shorter timestep, which you can obtain by using stricter settings for the :code:`cflmax` in the :code:`advection` group and the :code:`dnmax` in the :code:`diffusion` group. Fields: random perturbations ------------------------------ The :ref:`Fields ``[fields]``` class contains, among others, the settings that control the generation of the random perturbation in the initialization. These perturbations are required to generate turbulent structures. By rerunning a simulation with different random seeds, an ensemble can be formed. .. admonition:: Example :class: tip The figure below shows time series of water paths of hydrometeors. The individual lines are repetitions of the same simulation with rndseed = 1, 2, and 3, respectively. .. figure:: figures/fields.png :width: 400 | .. note:: In most cases, the random perturbations are applied in the model simulation, controlled by the settings in :ref:`Fields ``[fields]```. The Weisman Klemp bubble show here is an exceptional case, for which the random perturbations are applied together with the warm bubble in a separate python script. Microphysics ------------- For the :ref:`Microphysics ``[micro]```, two schemes are available. One is a single moment scheme (``swmicro=nsw6``, `Tomita (2008)`_), which in short means that the scheme predicts only the mixing ratios of the hydrometeors. The other is a double moment scheme (``swmicro=2mom_warm``, `Seifert and Beheng (2006)`_), which means that the scheme predicts the mixing ratios and number concentrations of the hydrometeors. Note that ``swmicro=2mom_warm`` does not include ice processes, only ice created by the saturation adjustment, and the cloud droplet number is fixed. .. admonition:: Example :class: tip The figure below shows time series of water paths of hydrometeors. .. figure:: figures/microphysics.png :width: 800 | .. note:: The most visible difference between the schemes is currently the lack of snow and graupel in the double moment scheme. This will change once the complete double moment scheme is available. .. note:: Both schemes currently require specifying the cloud droplet number concentration. The typical order of magnitude for :code:`Nc0` is 10-1000 cm\ :sup:`-3`, ranging from clean to polluted air, and the input value must be m\ :sup:`-3`. .. _Tomita (2008): https://doi.org/10.2151/jmsj.86A.121 .. _Seifert and Beheng (2006): https://doi.org/10.1007/s00703-005-0112-4 Thermodynamics ---------------- | In the :ref:`Thermodynamics ``[thermo]``` there are two main choices: | 1. :code:`swthermo`. The most used options here are ``swthermo=moist`` for simulations that include moisture, and ``swthermo=dry`` for dry cases. | 2. :code:`swbasestate`. Both moist and dry thermodynamics require specifying the approximation for the base state. There are two options: | - the anelastic approximation. Under this approximation, the state variables density, pressure, and temperature are described as small fluctuations from corresponding vertical reference profiles that are functions of height only. | - the Boussinesq approximation, which is the same as the anelastic approximation with the additional assumption that the density is always 1. .. note:: When using the anelastic approximation, the base state used for the moist thermodynamics can optionally be taken constant (``swupdatebasestate=false``). A time dependent base state is mainly important in cases where the domain mean state changes significantly over time. Updating the base state makes sure that the pressure is in balance with the temperature, which is of influence on the condensation. .. admonition:: Example :class: tip In the Weisman-Klemp case, the domain mean temperature increases over time at most levels. This is visible in the pressure when using a time dependent base state. The plot below shows the change in pressure relative to the initial pressure profile. In this short and simple example, the changes over time are limited and therefore the condensation is hardly affected. .. figure:: figures/thermo.png :width: 800 |