Box-Jenkins Analysis of Air Pollution Data

In a network of observation stations throughout the Federal Republic of Germany, SO₂, NO₂,O₃, and dust concentrations are continuously measured. 1)For a sample of time series from these measurements we fit univariate and multi-variate Box-Jenkins models.2)It is shown that so-called standard models exist, which have a good fit for all members of a whole class of time series. The knowledge of such models can shorten the model finding process considerably in further analyses.3)In the multivariate case we compare different types of models (transfer function, ARMAX-, and multiple regressions models).