Conditional simulation in dynamic linear models for spatial and temporal predictions of diffusive phenomena

Abstract.

A spatial time series framework is used for stochastic modelling of daily average Sulphur Dioxide (SO₂) levels in the Milan district. Within a spatio-temporal Kalman filter algorithm, stochastic conditional simulation is performed to obtain spatial and temporal predictions of the observed process. Unlike other recent space-time Kalman filters, the inclusion of a point source trend model also allows the development of a spatio-temporal state-space model that achieves dimension reduction in the analysis of large data set.