A non-stationary spatial model for annual precipitation in southern Norway: an SPDE approach using R-INLA
Speaker: Rikke Ingebrigtsen (NTNU Trondheim, Norway)
Physical knowledge about spatial phenomena often require models with non-stationary dependence structures. Geostatistical models have traditionally been stationary, thus there has been an interest in the literature to provide flexible and computationally efficient models and methods for non-stationary phenomena. In this work, we demonstrate that the stochastic partial differential equation (SPDE) approach to spatial modelling provides a flexible class of non-stationary models where explanatory variables easily can be included in the dependence structure. We explore the proposed modelling framework using a simulation study, with focus on distinguishing between a stationary and non-stationary model. Further, we model annual precipitation in southern Norway using a non-stationary model with dependence structure governed by elevation, and compare this model with a stationary model. In addition, the SPDE approach enables computationally efficient Bayesian inference with integrated nested Laplace approximations available through the R-library INLA.