\
We have already noted in section 
the special
random nature of the telescope error components due the effect of
atmospheric turbulence, in opposition to the essentially deterministic
or stable nature of the figuring and alignment errors .
There are two aspects contributing to this random nature. The first one is the great variability of atmospheric external conditions on the site of the observatory. The second aspect is due to the operation of the telescope and its enclosure: the performance will be affected by the (statistically random) pointing direction of the telescope and by the operation of the windscreen and louvers of the enclosure.
So far there has been hardly any attempt to take the random aspect of
these effects in a concurrent engineering approach to the
design of a telescope enclosure.
For instance, engineering
computations relative to wind loading on telescopes (computations
concerning local seeing have been practically inexistent) generally
consider either "worst cases" or "average cases" and fix somehow
the input conditions .
The results of these studies are then indicative of the general
behavior and performance of the systems analyzed but can hardly claim
to be a realistic prediction of a performance which can be
later verified. For instance, suppose that conditions of strong winds
are statistically associated to high values of natural
seeing
. Then errors caused by wind buffeting
will be to some extent masked by the high seeing and therefore less
critical to the telescope quality than if wind and natural seeing were
statistically unrelated.
There are numerous possible relationships of this kind
which may have an impact on the design choices.
Also the operational aspects of the enclosure with the use of louvers, venting openings and windscreens have never been the object of engineering studies. Even in the best observatories, the criteria for operating the venting devices come in the best case from empirical experience if not simply from a longtime unquestioned routine.
As a contribution to these issues we will here introduce a system engineering approach based on the evaluation of statistical distributions of error components due to local seeing and guiding.