Modeling Multivariate Space-Time Anisotropic Covariance Function

In multivariate context, it is common to adopt the linear coregionalization model (LCM) based on isotropic independent hidden components underlying the phenomenon of interest. In this paper, a spatio-temporal LCM which takes into account the presence of possible spatial anisotropies is proposed and practical aspects in fitting and modeling are faced. A case study concerning daily averages of climatic variables, such as minimum and maximum temperature, and 10-centimeters soil temperature, recorded at some stations of the Irish Meteorological Service for 20-year span, is discussed. Thus, after establishing the possible presence of spatial anisotropy, the independent latent components which jointly describe the direct and cross-correlation among the variables under study, are identified; then the spatio-temporal anisotropic LCM is fitted.