1 Introduction

In the previous paper (Masciadri et al. 1998 - Paper I) we presented the general characteristics of an atmospheric numerical model (Meso-Nh) conceived to provide not only 3D maps of the classic meteorological parameters P, T, and $\mathbf{ \vec{V}}$ but also the $C_{\rm N}^2 \$ optical turbulence profiles. We described the dynamical turbulence (Sect. 3.2) and the optical turbulence (Sect. 3.3) parametrization realized in the model. We presented a complete description of the astronomical parameters coded in Meso-Nh (Sect. 3.4). Finally, we discussed the applications of this technique to site testing and flexible scheduling for optimization of the observing time of a telescope. The main questions for an efficient application of the numerical technique concern its sensitivity and statistical reliability. The latter depends mainly on the quality of the initial atmospheric parameters supplied to the model to make a simulation. We intend, in the present paper, to analyze the statistical reliability of the Meso-Nh model comparing simulations with measurements obtained during a 14-night site testing campaign PARSCA93 (Fuchs & Vernin 1994) that took place at Paranal mountain (Chile) in May 1993, in which optical measurements of the seeing were made. The model initialization data used for this study are both radiosoundings provided by Antofagasta station and ECMWF (European Center for Medium Weather Forecast) analysis computed at an appropriate grid point.