A Bayesian hierarchical model for assessing the impact of human activity on nitrogen dioxide concentrations in Europe

Abstract

Ambient concentrations of many pollutants are associated with emissions due to human activity, such as road transport and other combustion sources. In this paper we consider air pollution as a multilevel phenomenon within a Bayesian hierarchical model. We examine different scales of variation in pollution concentrations ranging from large scale transboundary effects to more localised effects which are directly related to human activity. Specifically, in the first stage of the model, we isolate underlying patterns in pollution concentrations due to global factors such as underlying climate and topography, which are modelled together with spatial structure. At this stage measurements from monitoring sites located within rural areas are used which, as far as possible, are chosen to reflect background concentrations. Having isolated these global effects, in the second stage we assess the effects of human activity on pollution in urban areas. The proposed model was applied to concentrations of nitrogen dioxide measured throughout the EU for which significant increases are found to be associated with human activity in urban areas. The approach proposed here provides valuable information that could be used in performing health impact assessments and to inform policy.

A Bayesian Spatiotemporal Model to Estimate Long-Term Exposure to Outdoor Air Pollution at Coarser Administrative Geographies in England and Wales

“…A similar comparison for our best model for NO 2 shows the RMSPE value of 17.65 which is relative to the overall data standard deviation of 37.19 as reported in Table 2 here. Thus our model has much lower relative root-mean-square error, compared with the standard deviation of the full data, than what was reported in Shaddick et al (2013). For PM 10 , modelled annual mean data from 52 monitoring sites in London and their best land use regression model, PMLUR, reported an R 2 -value of 0.58.…”

“…Thus our model has much lower relative root‐mean‐square error, compared with the standard deviation of the full data, than what was reported in Shaddick et al . (). For PM 10 , Gulliver et al .…”

“…However, many researchers model annual average data, and error rates from such modelling are available. For example, Shaddick et al (2013) modelled the 2001 annual average NO 2 data from 934 sites in the then 15 European Union countries and their best model has a root-mean-squared error of 8. This is not directly comparable with our model RMSPE for the daily data as provided in Table 3 since daily data and annual average data have different variabilities.…”

“…the continental USA, Europe and other parts of the world such as Asia and South America where air pollution is a very prominent health hazard. Methodological developments include downscaler models (Berrocal et al, 2010;Alkuwari et al, 2013), data fusion models (Sahu et al, 2010), land use regression (Jerrett et al, 2005;Hoek et al, 2008;Shaddick et al, 2013;Venegas et al, 2014;Morrison et al, 2016), hierarchical space-time auto-regressive models (Sahu et al, 2007). Model-based methods for realtime forecasting of air pollution have also been developed (Huerta et al, 2004;McMillan et al, 2005;Sahu et al, 2009).…”

Maximum likelihood estimation of the multivariate hidden dynamic geostatistical model with application to air quality in Apulia, Italy

“…() compare six models for PM 10 in Piedmont (Italy), featuring different levels of complexity either in the hierarchical structure or in the spatio‐temporal covariance function. Moreover, univariate spatio‐temporal hierarchical models were proposed to account for rural/background and urban/suburban random effects on PM 10 concentrations (Sahu et al ., ), to combine monitoring data and the output from a local‐scale air pollution model for health risk assessment (Pirani et al ., ) and to assess the effects of human activity on nitrogen dioxide (NO 2 ) pollution in European urban areas (Shaddick et al ., ), to mention a few.…”