2015/10/27 by Marc Bourotte, Denis Allard, Bourotte, Marc +3
Computer Science · Environmental Science · Mathematics · #FOS: Computer and information sciences #Geochemistry and Geologic Mapping #Methodology (stat.ME) #Soil Geostatistics and Mapping #Statistical and numerical algorithms
paper · pdf · doi:10.48550/arxiv.1510.07840
openalex publication_date 2015/10/27 · openalex created_date 2022/10/02 · openalex updated_date 2026/07/28
Multivariate space-time data are increasingly available in various scientific\ndisciplines. When analyzing these data, one of the key issues is to describe\nthe multivariate space-time dependencies. Under the Gaussian framework, one\nneeds to propose relevant models for multivariate space-time covariance\nfunctions, i.e. matrix-valued mappings with the additional requirement of\nnon-negative definiteness. We propose a flexible parametric class of\ncross-covariance functions for multivariate space-time Gaussian random fields.\nSpace-time components belong to the (univariate) Gneiting class of space-time\ncovariance functions, with Mat 'ern or Cauchy covariance functions in the\nspatial margins. The smoothness and scale parameters can be different for each\nvariable. We provide sufficient conditions for positive definiteness. A\nsimulation study shows that the parameters of this model can be efficiently\nestimated using weighted pairwise likelihood, which belongs to the class of\ncomposite likelihood methods. We then illustrate the model on a French dataset\nof weather variables.\n