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.

Birnbaum–Saunders spatial modelling and diagnostics applied to agricultural engineering data

Bayesian statistics in spatial epidemiology]

Spatial Hyperspheric Models for Compositional Data

Spatial Functional Data analysis: Irregular spacing and Bernstein polynomials

Gaussian modeling with B-splines for spatial functional data on irregular domains

A Spatial Autoregressive Graphical Model with Applications in Intercropping

Spatial modelling of agro-ecologically significant grassland species using the INLA-SPDE approach

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

Bayesian Spectral Modeling of Microscale Spatial Distributions in a Multivariate Soil Matrix

Flexible Basis Representations for Modeling Large Non-Gaussian Spatial Data

A multivariate Gaussian random field prior against spatial confounding

Enhanced spatial modeling on linear networks using Gaussian Whittle-Matérn fields

Spatial Latent Gaussian Modelling with Change of Support

Spatial+: a novel approach to spatial confounding

Bimodal Birnbaum–Saunders generalized autoregressive score model

Crop Yield Prediction Using Bayesian Spatially Varying Coefficient Models with Functional Predictors

Bayesian spatial modelling of geostatistical data using INLA and SPDE methods: A case study predicting malaria risk in Mozambique

Bayesian Nonstationary Spatial Modeling for Very Large Datasets

Spatially Dependent Bayesian Modeling of Geostatistics Data and Its Application for Tuberculosis (TB) in China

Performance evaluation and hyperparameter tuning of statistical and machine-learning models using spatial data

Efficient Modeling of Spatial Extremes over Large Geographical Domains