The CFHT is a 3.6-m telescope (see fig. 
)
housed in a large dome with a diameter of
32 m, located on Mauna Kea, Hawaii island, at the altitude of 4200 m.
This site is deemed among astronomers to be the best in the world
for natural seeing.
The observations with the telescope, which started operating in 1979,
experienced during
the first years of operation an average seeing in excess of 2 arcsec
[Racine 84].
As a result of many technical actions taken to eliminate heat transfer
to the telescope air volume, the average seeing has been improved to
0.6 arcsec [Racine 92].
A main factor in this achievement was the
installation of a chilled floor for the telescope volume.
The floor creates a stable temperature vertical gradient,
thus effectively eliminating sustaining free convection flows
inside the dome. It also acts as a powerful cold sink for any
heat generated or entering in the dome volume, contributing in
particular to keeping down also the temperature of telescope structure
and primary mirror.
We have analyzed the data log files concerning all observations done during the years 1991, 1992 and 1993. They record the FWHM image quality of all observations taken with the two main imaging instruments used on the telescope prime focus, which are respectively named FOCam and HRCam. The latter one has a fast tip-tilt guider aimed at correcting automatically telescope guiding errors (wind induced fluctuations in particular). Thus image quality from HRCam is typically 20% better than from FOCam. The records of the log files include for each observation the zenithal angle of the telescope, the integration time, the number of FWHM measurements averaged in that time, meteorological data of the external environment (wind, air temperature) as well as the temperature measured at several locations around the telescope and in the dome:
Table: Location of main temperature sensors in the CFHT dome
Only the records in which the FWHM was actually recorded (> 0)
were retained for the analysis. Since the vast
majority of observations are taken at the prime focus, the relatively few
Cassegrain observations have been discarded. Also all observations
with integration periods of less than 10 seconds are discarded, since
they may not average correctly image motion due to seeing effects,
as well as all those in which
the FWHM measure was fluctuating excessively
(i.e. with 
).
A few observation sequences in which the FWHM remained at
very high values
(> 2 arcsec) were also eliminated as it was
considered that this could only be attributed to occasional
very bad natural
seeing or to a non-optimal optical setup, both effects being out of
the scope of the present analysis while they could introduce bias
in the evaluation of local seeing effects.