Scalable Bayesian Inference for Large Spatial Compositional Databases

C. Guarner Giner, D. V. Conesa Guillén

The availability of large land use and land cover databases has increased substantially in recent years, posing new challenges for statistical analysis. These data are typically spatially and spatio-temporally structured, compositional in nature, and available at high spatial resolution or over heterogeneous spatial supports. As a result, standard Bayesian inference approaches may become computationally demanding when applied to such large and complex datasets. In this work, we consider the problem of evaluating large databases of spatial compositional data from the perspective of scalable Bayesian inference. Within the Integrated Nested Laplace Approximations (INLA) framework, we discuss inference strategies based on data and model partitioning, sequential updating, and consensus-type procedures, which offer practical alternatives when full joint inference is not feasible.

Keywords: Land use, Spatial compositional data, Scalable Bayesian inference, INLA

GT Bioestadística SEIO: Current Challenges and Advances in Spatial Biostatistics
September 2, 2026  11:20 AM
Aula 28

T. Kneib

A. Mateos Caballero, A. Domínguez Monterroza, A. Jiménez Martín, M. Martínez Pizarro

A. Urdangarin Iztueta