Linear transformation methods for multivariate geostatistical simulation
Date of Award
Doctor of Philosophy
School of Natural Sciences
Faculty of Computing, Health and Science
Multivariate geostatistical techniques take into account the statistical and spatial relationships between attributes but can be inferentially and computationally expensive. One way to circumvent these issues is to transform the spatially correlated attributes into a set of decorrelated factors for which the off diagonal elements of the spatial covariance matrix are zero. This requires the derivation of a transformation matrix that exactly or approximately diagonalises the spatial covariance matrix for all separation distances. The resultant factors can then analysed using the more straightforward univariate techniques. This thesis is concerned with the investigation of linear decorrclation methods whereby the resulting factors are linear combinations of the original attributes.
Bandarian, Ellen, "Linear transformation methods for multivariate geostatistical simulation" (2008). Theses: Doctorates and Masters. Paper 191.