Parsimonious Var Models For Air Pollution Dynamic Analysis

We discuss a framework to obtain temporal predictions for an evolving spatial field regularly sampled in time at arbitrary spatial locations. Difficulties caused by large data sets and the modelling of complicated spatio-temporal interactions limit the effectiveness of traditional space-time statistical models. In this study, we propose the use of a flexible approach to deal with large and small time-scale variability of the observed data. The temporal model is applied with respect to both the observed data domain and the common component domain, to achieve a dimensionality reduction.