Abstract:Spatial data are often derived from multiple sources (e.g. satellites, in-situ sensors, survey samples) with different supports, but associated with the same properties of a spatial phenomenon of interest. It is common for predictors to also be measured on different spatial supports than the response variables. Although there is no standard way to work with spatial data with different supports, a prevalent approach used by practitioners has been to use downscaling or interpolation to project all the variables of analysis towards a common support, and then using standard spatial models. The main disadvantage with this approach is that simple interpolation can introduce biases and, more importantly, the uncertainty associated with the change of support is not taken into account in parameter estimation. In this article, we propose a Bayesian spatial latent Gaussian model that can handle data with different rectilinear supports in both the response variable and predictors. Our approach allows to handle changes of support more naturally according to the properties of the spatial stochastic process being used, and to take into account the uncertainty from the change of support in parameter estimation and prediction. We use spatial stochastic processes as linear combinations of basis functions where Gaussian Markov random fields define the weights. Our hierarchical modelling approach can be described by the following steps: (i) define a latent model where response variables and predictors are considered as latent stochastic processes with continuous support, (ii) link the continuous-index set stochastic processes with its projection to the support of the observed data, (iii) link the projected process with the observed data. We show the applicability of our approach by simulation studies and modelling land suitability for improved grassland in Rhondda Cynon Taf, a county borough in Wales.

Modeling massive highly-multivariate nonstationary spatial data with the basis graphical lasso

Flexible Basis Representations for Modeling Large Non-Gaussian Spatial Data

Spatial Factor Modeling: A Bayesian Matrix-Normal Approach for Misaligned Data

Bayesian Nonstationary Spatial Modeling for Very Large Datasets

Normalizing Basis Functions: Approximate Stationary Models for Large Spatial Data

A Variational Approach for Modeling High-dimensional Spatial Generalized Linear Mixed Models

Multivariate nearest‐neighbors Gaussian processes with random covariance matrices

Flexible Modeling of Nonstationary Extremal Dependence using Spatially Fused LASSO and Ridge Penalties

Highly Multivariate High-dimensionality Spatial Stochastic Processes-A Mixed Conditional Approach

Modeling Non-Stationary Processes Through Dimension Expansion

Spatial Latent Gaussian Modelling with Change of Support

Multilevel latent Gaussian process model for mixed discrete and continuous multivariate response data

Multivariate Gaussian Process Regression for Multiscale Data Assimilation and Uncertainty Reduction

Nonstationary Spatial Process Models with Spatially Varying Covariance Kernels

Estimating Basis Functions in Massive Fields under the Spatial Mixed Effects Model

Mixed-Stationary Gaussian Process for Flexible Non-Stationary Modeling of Spatial Outcomes

Nonstationary Spatial Modeling of Massive Global Satellite Data

Hierarchical Nearest-Neighbor Gaussian Process Models for Large Geostatistical Datasets

Sparse Graphical Linear Dynamical Systems

Multivariate Matérn Models -- A Spectral Approach