Estimation of the Measurement Uncertainty of Ambient Air Pollution Datasets Using Geostatistical Analysis

Abstract

We developed a methodology able to automatically estimate of measurement uncertainty in the air pollution data sets of AIRBase. The figures produced with this method were consistent with expectations from laboratory and field estimation of uncertainty and with the Data Quality Objectives of the European Directives. The proposed method based on geostatistical analysis is not able to estimate directly the measurement uncertainty. It estimates the nugget effect together with a micro-scale variability that must be minimized by accurate selection of the type of station. Based on the results obtained so far, it is likely that measurement uncertainty is best estimated using all background stations of whatever area type. So far the methodology has been used to estimate uncertainty in 4 different countries independently. This work should be continued for the whole Europe or for background station without national borders. The method has been shown to be also useful to compare the spatial continuity of air pollution in different countries that seems to be influenced by the topography of each country. Moreover, it may be used to quantify the trend of measurement uncertainty over long periods like decade with the possibility to evidence improvement in the data quality of AIRBase datasets. Thanks to the implemented outlier detection module that would also be of interest as the warning system when Member States report they measurement to the European Environment Agency, we have proposed an easy solution to investigate wrong classified stations in AIRBase.