In preceding papers we have presented comparisons between simulated and measured profiles above two of the best astronomical sites: Cerro Paranal (Chile) (Masciadri et al. [CITE]; Masciadri et al. [CITE]) and Roque de los Muchachos (Canary Islands) (Masciadri et al. [CITE]). These works aimed at validating the atmospheric model Meso-Nh which was conceived to simulate classical atmospheric parameters and was then adapted to simulate optical turbulence (3D maps of ). Good results were obtained but further improvements can be made. One of the principal limitations is the statistical reliability of the model that is, at the present time, not very high. In this paper we present some modifications introduced in the optical turbulence parameterization of the model that aim to reduce some systematic errors. Preliminary results, obtained comparing the simulations with measurements taken at the San Pedro Mártir site (Mexico) in March-April 1997, show that we are now able to obtain a better qualitative and quantitative estimate of the profiles. Finally, we show how we have been able to improve the spatiotemporal variability of the model.