Preliminary results on tapering multivariate spatio temporal models for exposure to airborne multipollutants in Europe

The population exposure distribution is important for assessing the human risk related to air pollution. Since it requires multipollutant concentrations over a fine spatial grid at daily level, spatio-temporal modelling is required to interpolate ground level monitoring network data. The same authors recently introduced a trans-Gaussian model able to model multipollutant concentration over Europe in high resolution and, thanks to this, able to estimate the population exposure distribution, uncertainty included. Since modelling multivariate daily data at continental level may be a challenging computationl issue, we assess here the consequences of using the so-called one-taper tapering. It is concluded that for the data at hand and software considered, one-taper tapering is not a viable solution as deteriorates inference without a relevant reduction of computation time.