The monitoring and forecast of meteorological parameters has been rarely used by astronomers: e.g., cloud cover was checked visually and water vapour content was merely used. There is now an ever growing interest and need for an extensive application of atmosphere soundings to astronomical activities; newest telescopes are demanding on this side, being devoted to do queued observations. The Gemini and VLT projects, e.g., will adopt an operational model that will include at least 50% of the allotted time as queued: they will take advantage of an active scheduling of the observing facilities. For all of these high tech telescopes and instruments, weather and site monitoring will be very important in the selection of proposals for the coming nights. Therefore prediction algorithms are under contract for ESO/VLT for clouds and water vapour, aiming at developing custom made daily weather forecasts for Paranal: although that site must be seen as a special situation, this is the expected trend for the future ground based observatories.
Astronomers can in principle get support from the local weather forecast
services. The European Center for Medium-Range Weather Forecast (ECMWF), e.g.,
predicts the behaviour of the atmosphere in the medium-range up to ten days
ahead, providing numerical weather prediction maps on a synoptic scale every 12
hours. National Meteorological institutions receive these products through a
dedicated telecommunications network and can further process the ECMWF output,
providing an higher horizontal resolution in a short range prediction. The
running atmospheric models are extremely demanding from the computing resources
side: the recent availability of more powerful computers led the ECMWF to
definitely increase the actual grid resolution to about
60 km. The latest
numeric model also improves the vertical discretization, with 31 baric levels
available instead of 19: the forecast standard error for temperature parameter
in the medium range ECMWF model is well higher than 2 degrees. In case of the
ECMWF model, the horizontal discretization grid as well as the error on the
forecast of meteorological parameters is too high for astronomical purposes.
In case of short range prediction, a local area model (LAM) may be used, which takes into account more detailed information, mostly local sampled data and an higher resolution orographic model. The output of a LAM is "limited'', both in time and space, being valid in the range from several hours to about two days and within a limited spatial area, but can offer a forecast map having grid points separated only 30 to 15 km and a prediction error on temperature as low as 2 Celsius degrees RMS on a 24-hours prediction map. This is true for an operational LAM running on a rough orography in a Mediterranean area (Picchedda and Chessa, private communication); a much better result is expected for a more stable region as the La Palma island is. The laminar flux due to the presence of the Atlantic Ocean may actually smooth the complex behaviour which is found to be the dominant characteristic on an european region. Therefore a LAM could virtually perform with a resolution approaching the Celsius degree, but the limit is still on the computing side.
As european development, the project High Resolution Limited Area Model (HIRLAM) is the most challenging operational example for a LAM. It is a project cooperation among several european countries, and the reference version is maintained at the ECMWF.