Brief Description of the Governing Equations in the Nonhydrostatic Regional COSMO-model

J. Alejandro Martinez
Mesoscale Meteorological Modeling: Spring Semester 2014
Assignment # 1
COSMO is the nonhydrostatic regional model of the Consortium for Small-Scale Modelling. The basic
version of the COSMO model was developed at the German Meteorological Service (DWD).
Currently, COSMO is used by several weather agencies in Europe for numerical weather prediction
(NWP) and by other institutions as a research tool of atmospheric processes at the mesoscale. In
addition, COSMO is the basis for the Climate Limited-Area Model (COSMO-CLM) and the COSMO
Aerosols and Reactive Trace gases model (COSMO-ART). For detailed information about the
COSMO model the reader is referred to the website www.cosmo-model.org. The available
documentation includes the description of the Dynamics and Numerics, Physical Parametrization, Data
Assimilation, Preprocessing and a User's Guide.
In this report we briefly describe the thermo-hydrodynamical system of equations that results from the
basic equations for conservation of momentum, mass and energy after some important modifications,
before their numerical implementation in the COSMO model. For a detailed description of all the
modifications, the reader is referred to the Dynamics and Numerics documentation. Here we describe
the resulting system after the following changes: i) the use of temperature and pressure as state
variables instead of internal energy and density; ii) the reduction of the resulting system by some
thermodynamical simplifications; iii) the transformation to a rotated spherical coordinate system; and
iv) the introduction of a base state. This material is completely based in the documentation on
Dynamics and Numerics of the COSMO model.
The resulting set of equations in a spherical coordinate system (, , z) is:

The dependent variables are u, v, w, p', T, qx, and p. Their meaning is as follows:
u: wind component along the e direction (zonal wind when (, , z) are the geographic coordinates)
v: wind component along the e direction (zonal wind when (, , z) are the geographic coordinates)
w: vertical wind component.
p: pressure
p': deviation from a base state pressure (see below)
T: temperature
: density
qx: specific content of constituent x.
x = d, dry air; x = v, water vapor; x = l, liquid water; x = f, ice.
Note that there are equations for each of these variables, except for p, that would be described below.
All the listed dependent variables have a corresponding prognostic equation, except density that is
computed from a diagnostic equation (the ideal gas law). The origin of each of the equations, in the
corresponding order, is as follows:
1. Conservation of momentum in the e direction (zonal, when (, , z) are the geographic coordinates)
2. Conservation of momentum in the e direction (meridional, if (, , z) are geographic coordinates)
3. Conservation of momentum in vertical direction.
4. Evolution of perturbation pressure: comes from ideal gas law, continuity and conservation of energy.
5. Temperature : comes from the equation for the internal energy (first law of thermodaynamics).
6. Conservation equation for water vapor
7. Conservation equations for liquid water and ice (one equation for each type (superscript))
8. Ideal gas law.
The meaning of the other terms is as follows:
t: time
f = 2 sin g =2(cos Ng cos cos + sin sin Ng ) the Coriolis parameter in the rotated system,
where Ng is the location of the geographical North Pole in the rotated system.
B: Bouyancy term (see below).
D: is the 3D wind divergence in the rotated spherical coordinate system.
Px = qx|vxT| , precipitation fluxes of water (x=l) and ice (x = f).
vxT is the corresponding terminal velocity.
The source terms are as follows:
Sx : cloud microphysical sources/sinks of constituent x, per unit mass of moist air.
Mu, Mv, Mw: sources of momentum
QT = (Lv/cpd)Sl + (Ls/cpd)Sf + MT + Qr : diabatic heating term
MT: sources of temperature (divergence of turbulent flux of sensible heat H).
Qr: temperature change due to convergence/divergence of radiation.
Mqx: sources of moisture of type x.

In addition, several constants have to be used to solve the governing equations:

a: average radius of the Earth
: angular velocity of the Earth (rotation rate)
g: acceleration due to gravity
Rv: Gas constant for water vapor
Rd: Gas constant for dry air
cpd: specific heat at constant pressure for dry air
Lv: latent heat of vaporization
Lf: latent heat of sublimation
Approximations and considerations leading to the governing equations
The system of governing equations as shown above, is obtained from the original forms of the relations
1-8 after several physical and geometric considerations. In the following we make a few comments on
some of the modifications leading to the presented governing equations (for a detailed description and
derivation, see the Dynamics and Numerics documentation).
i) The use of temperature and pressure as state variables instead of internal energy and density
The basic thermo-hydrodynamic equations include the conservation of mass (continuity) equation and
the first law of thermodynamics (conservation of energy). The latter relationship states that the
changes in internal energy are due to transfers of energy in the form of work and heat. In a gas,
changes in the internal energy are directly related to changes in temperature, while transfers of energy
due to work are related to pressure acting to change the volume of the gas. Thus, the equation for the
conservation of energy can be written in terms of temperature, pressure, density (inverse of volume)
and the sources of heat.
On the other hand, the continuity equation is formulated in terms of the density and velocity fields as
the dependent variables. Direct use of this relationship in the model imposes exact conservation of
mass. However, the numerical implementation of the resulting system does not allow the use of
efficient algorithms for the treatment of sound waves. This drawback is not present when a system of
equations can be written where pressure rather than density is the prognostic variable. The equation for
the time evolution of pressure can be obtained by combining the ideal gas law, the continuity equation
and the equation for the time evolution of temperature (as derived from the conservation of energy).
ii) Thermodynamical simplifications and grid-volume averaging
Several simplifications related to thermodynamical variables and processes lead from the basic
conservation relationships to the governing equations as listed above:
The molecular (diffusion) fluxes are neglected, because they are much smaller than the resulting
fluxes from turbulent mixing (the M terms). However, the molecular fluxes related to liquid and frozen
moisture (the Px terms above) are retained.
In principle, the latent heat of vaporization is a function of temperature. However, for the
atmospheric processes of interest this dependence is rather weak and that allows the use of a constant
value for the latent heat of vaporization. Something analogous happens for the latent heat of
sublimation.

 The grid-cell average turbulent flux of sensible heat is parameterized as H = cpd <> <v''''>, where
<X> represents the grid-cell average of field X, and is the Exner function. In this case, the turbulent
fractional fluctuations in temperature T'' have been approximated to the turbulent fractional fluctuations
in the potential temperature , and the turbulent fractional fluctuations in pressure (i.e. the sub-grid
scale fluctuations) have been neglected (i.e. assumed to be much smaller than the corresponding
fractional fluctuations in ). Note that Exner pressure is not computed from a prognostic equation
(unlike other formulations, see e.g. Fovell (2005)), and the definition of potential temperature is used as
a diagnostic relationship to obtain from p and T.
The dissipation rate due to viscous stresses is neglected (<> = 0).
In general, the Mx source terms have several components:
a) Tendency due to small scale turbulent mixing (due to grid-cell averaging)
b) Tendency due to small scale moist convection (due to parametrization of moist convection)
c) Lateral boundary relaxation term for one-way nesting for the model (smoothing)
d) Source term representing computational mixing (resulting from numerical implementation)
e) Tendency due to a Rayleigh damping scheme applied within the upper boundary (avoid spurious
reflection of gravity waves)
iii) The transformation to a rotated spherical coordinate system
The listed governing equations are written for a spherical coordinate system. As such, the common
approximations of r = a + z a, and ()/r ()/z have been considered. Once the corresponding
simplifications have been done, the system is written in terms of z, instead of r. Because of the
transformation to a spherical coordinate system, some metric terms appear in the momentum equations
(tan terms).
In COSMO, the system of equations is formulated for a generalized spherical coordinate system. This
generalized system does not need to coincide with the geographical system, i.e. the poles of both
systems can be located in different physical locations. The advantage of using a generalized spherical
coordinate system is that the latitude-longitude grid can be as regular (nearly equal areas) over the
domain of interest as in the case of the tropics in the geographical coordinate system. Thus, problems
related to the computation of gradients over areas of convergence of parallels (i.e. longitudes) can be
avoided to a good extent.
iv) The introduction of a base state
As in the case of other models (e.g. MM5), in COSMO the state of the system is decomposed into a
base state and a perturbation. When the perturbations are not too far form the base state, the
substitution of this decomposition into the governing equations reduces the errors in the calculations of
pressure gradients. This issue is specially important in the case of using a terrain-following vertical
coordinate (see bellow). In the case of the COSMO model, the base state is assumed to be steady, at
rest and having no moisture. Therefore, the total fields can be decomposed as:
u(, , z,t) = u'(, , z,t),
v(, , z,t) = v'(, , z,t),
w(, , z,t) = w'(, , z,t),
T(, , z,t) = T0(z) + T '(, , z,t),
p(, , z,t) = p0(z) + p'(, , z,t),
(, , z,t) = 0(z) + '(, , z,t),
qx(, , z,t) = qx '(, , z,t),
x = v, l, f.

The base state is assumed to be given by (by definition, the other base state variables are zero):
T0(z) = TSL ( 1 (2gz / RdTSL2 )),
p0(z) = pSL exp [ (TSL/) (1 T0(z) /TSL) ],

TSL= 288.15K, = 42K

pSL = 1000hPa.

This base state describes an atmosphere where temperature decreases at a constant lapse rate with
logarithmic pressure. With the decomposition into a base state and a perturbation, the bouyancy term
in the governing equations has the form:
B = g (0/) [ (T '/T) (T0p'/Tp0) + ((Rv/Rd) 1)qv ql qf],
where both the fractional fluctuations in pressure ant temperature have been retained. Therefore, this
expression for the bouyancy term is not the same to that obtained from the anelastic approximation.
v) Terrain-following coordinates
Finally, because the Earth's surface is not flat but has an irregular orography that modifies the flow
patterns and imposes complicated lower boundary conditions, the introduction of a terrain-following
vertical coordinate is convenient. However this represents a transformation from an orthogonal
curvilinear system (, , z) to a non-orthogonal curvilinear system (, , ). Such a transformation
leads to the introduction of different metric terms into the governing equations:

where:

and the other terms have the same meaning as described before, and the diagnostic equation for the
density is the same as listed in the first set of governing equations. Note that the diabatic term in the
pressure tendency equation is neglected, because usually this term is much smaller than the divergence
term (D). In general, the vertical coordinate decreases with height (which is seen from z/ < 0). In
the COSMO model there are several options of vertical coordinate, including a pressure based
coordinate and a height based coordinate. For further details about the definitions of the vertical
coordinates, the reader is referred to section 3.5 of the Dynamics and Numerics documentation.
References:
G. Doms, 2011. A Description of the Nonhydrostatic Regional COSMO-model, Part I: Dynamics and
Numerics. COSMO-Model 4.20, www.cosmo-model.org.
Schttler U., G. Doms, and C. Schraff, 2013.
A Description of the Nonhydrostatic Regional
COSMO-model, Part VII: User's Guide. COSMO-Model 4.29, www.cosmo-model.org.
R. G. Fovell, 2005. Numerical Modeling notes.