VAR models for spatio-temporal structures: An application to environmental data

Abstract

In this work we propose the use of Vector Autoregressive Models (ST-VAR) and Generalized Vector Autoregressive Models (ST-GVAR) for spatiotemporal data in which relations among different sites of a monitoring network are not exclusively dependent on the distance among them but take also into account intensity and direction that a phenomenon measured on the whole network has on a single site of the network. We present an application of these models using data of fifteen italian meteorological locations of Mediterranean Basin.