Astronomical sites are chosen with reference to the local climate and seeing. The latter can be monitored and it can provide a key for understanding the general quality of the site: but the intrinsic excellence of a site is very often deteriorated by the astronomical facilities themselves, which introduce an additive level of seeing (Coulman 1985). The natural seeing, which is an intrisic site characteristic, has been normally considered the major limiting factor for reaching the telescope virtual top performance. There is now a general awareness about the crucial rôle which is played by the man induced seeing, and the need for minimizing it in order to get a quasi diffraction limited condition.
Several image quality series of data demonstrate that mirror seeing is a primary source of the point spread function broadening. The seeing contributed by a 3.5 m primary mirror warmer than the air outside the dome can range from 0.4 to 0.6 arcsec per Celsius degree (Zago 1986; Racine et al. 1991), a value which is comparable or higher than the natural seeing in good astronomical sites. Several laboratory studies of image deterioration by convection (Lowne 1979; Iye et al. 1991) confirm that microthermal activity grows significantly as soon as the temperature gradient between dome air and primary mirror is positive. The dome itself is a major source of seeing degradation (Bely 1987), and the seeing contribution is highly variable, depending on the telescope enclosure design. The Canadian-France Hawaii telescope and the Nordic Optical Telescope enclosures (Racine et al. 1991; Vernin & Muñoz-Tuñon 1992, 1994) show a low impact on the overall man induced seeing, while the ESO 3.6 m telescope is more critical (Faucherre 1995). Newest telescopes are designed and built taking care of this specific problem, and thermal analysis is modeled as specific item of the global dome design. Therefore, the telescope thermal environment needs a specific approach both in the dome design phase and in the telescope operational methodology. Heat sources inside and close to the observing floor are thermally controlled and air conditioning in the telescope enclosure is a normal procedure for telescopes committed to top performances.
This paper presents a feasibility study of temperature forecast in support to active air conditioning of a telescope's dome as part of a research activity which is done within the meteorological support to the italian national telescope Galileo (TNG), located at the Observatorio Roque de Los Muchachos (ORM), in the Canary Islands. A neural network modelling is presented as compared to classic linear filter algorithm.