Marc Sarazin

ABSTRACT We present the concept of a new instrument dedicated to modeling turbulence inside the dome and in the surface layer. It consists of using parallel laser beams separated by non redundant baselines between 0.1 and 2-3m and measuring Angle-of-Arrival (AA) fluctuations from spots displacements on a CCD. We use weighted least-square method to fit the measured AA longitudinal and transverse covariances with theoretical forms deduced from the usual models of turbulence. Then, the whole parameters characterizing this turbulence are provided from a complete spatio-temporal analysis of AA fluctuations. Thus, the surface layer turbulence energy in terms of C2N constant is provided from the AA structure function as in the DIMM instrument.

Astronomical observatories are extremely dependent ,on sky transparency. As the expensive new ,very large telescopes enter into operation, flexible observing modes are being introduced, which allow each 'observing block' to be scheduled at the most appropriate time. In such modes, it makes sense to develop tools for forecasting ambient conditions. We,present here the operational water vapour and cirrus cloud forecast model developed for ESO observatories in Northern Chile. Because the observatories are located well above the inversion layer, nebulosity is mainly due to high altitude

The European Southern Observatory's planned Astronomical Weather Station for the Very Large Telescope which is currently under construction at Cerro Paranal in Chile includes (i) advance temperature prediction, which would permit air conditioning in the telescope enclosure to be preset as a function of the next night's expected temperature; and (ii) prediction of seeing, a few hours in advance, to allow flexible scheduling of the most appropriate instrumentation. Extensive data, collected since 1985, are being used to appraise various methodologies. A recurrent neural network is described, which uses arbitrary time-delayed connections to capture the dynamic of time series. This endows the model with a memory of its previous states. The resulting network is time- and space-recurrent, and generalizes most recurrent architectures. The performance of this network is discussed. The results are compared with the k-nearest neighbors method.