Objective analysis of air pollution monitoring network data; spatial interpolation and network density

Abstract

To obtain meaningful interpretations of the results of fixed air pollution monitoring stations, quantitative information on the spatial representativity of these results is of crucial importance. On the basis of data of the Dutch Air Pollution Monitoring Network a study was undertaken to estimate the absolute and relative errors which are involved in extrapolating the results beyond the actual spatial argument of measurement. In the first place, three interpolation techniques, optimum interpolation, eigenvector interpolation and so called distance-density interpolation, were compared for SO 2, NO, NO 2 and O 3. The differences between the results of these techniques proved to be small. It was further concluded that the interpolation errors and the associated persistence in space and time, as given by mutual correlations, should be specified with respect to pure space- or space-time variability. In the second place, the interpolation errors were generalized to other network densities on the basis of the optimum interpolation scheme and empirical observations. From the resulting relation between interpolation error and network density it was concluded that the SO 2-network of 108 stations over the measurement area of 150 × 220 km 2 results in relative errors of 20 %, small enough to detect mesoscale transports downwind of major source areas. The effect of sampling error, i.e., the small scale influences of local sources combined with the effect of measurement error, appeared to be of overriding importance in the efficiency of reconstruction of pollutant concentration fields at a given confidence level from monitoring network data.