Constructing representative air quality indicators with measures of uncertainty

摘要

Air quality indicators summarize overall concentrations of pollution for an urban area and are calculated from routine monitoring data comprising numerous pollutants measured at many locations. The indicator is constructed by aggregating these data over space and pollutants, typically using the sample mean, median or maximum. We propose an alternative approach based on geostatistical modelling, which allows intervals of uncertainty to be calculated for the spatial aggregation stage, and hence for the final indicator. We then extend our geostatistical model by allowing for the fact that the locations that are chosen for the pollution monitors may depend on the hypothesized concentrations at these locations, a phenomenon which is known as preferential sampling. We assess the effectiveness of our methods by simulation and use them to construct an air quality indicator for Greater London, England, for the month of August 2006.