The work presented in this paper is a mixed solution between numerical and statistical models through a hybrid predictive meteorology model based on a mix of vector autoregression and hierarchical cluster analysis (which we called HP-VAR). The presented model is constituted of three steps: Hierarchical Cluster Analysis (HCA) to study the behavior of stations and remove atypical observations based on layer altitudes, Physical equations to generate virtual stations and Time Series Analysis (TSA), in particular the Vector Auto Regressive (VAR) model, to make predictions. The work proposed in this article shows the efficiency of the model for the prediction of two meteorological parameters: TMIN and TMAX.