Fluid flow numerical models have been used in at least two previous occasions with the aim of predicting seeing. It appears that, while the method is applicable in principle, there are also practical difficulties linked to the size of the model, the definition of the initial and boundary conditions, and the possibility to validate the computation by means of representative tests.
We have attempted here an application to the problem of mirror seeing
in which we have parameterized the local 
profile by means of a
similarity relationship. The case of a
mirror within a closed tube and ventilated by fans was simulated with
interesting results.
The conclusions that can be drawn from this
study are that a simulation with a similarity parameterization
of the 
profile is feasible and appears to give realistic results.
The limitations of this approach are however intrinsic
to the utilization of a CFD model.
Indeed the main advantage of CFD models is in the possibility
to simulate particular cases and conditions which cannot be
represented by simple relationships or for which there are
no experimental data. On the other hand the 
parameterization
refers to a well characterized steady state condition and may not be
as accurate in the actual model conditions.