Using a generalized additive model to quantify the influence of local meteorology on air quality in Zagreb

This paper reports the estimated response of hourly mean concentrations of selected air pollutants, namely carbon monoxide (CO), sulphur dioxide (SO2), nitrogen dioxide (NO2) and particulate matter with an aerodynamic diameter of up to 10 μm (PM10), to local scale meteorology in Zagreb, Croatia for the period 2006–2012. A new method is applied here for the urban area of Zagreb. In a general model, a logarithm of hourly mean air pollutant concentrations is expressed as the sum of the nonlinear functions of meteorological and several time variables, with the latter included accounting for temporal variation in emissions. The model can be formulated within the framework of generalized additive models (GAMs) and is additive on the logarithmic scale, which results in multiplicative effects on the original scale. Although the model is nonlinear, it is simple and easy to interpret. It quantifies the impact of meteorological conditions and emissions on air pollution. A measure of the relative importance of each predictor, partial effects and statistical evaluation of the model are also presented. Overall, the results show that the most important predictors are those related to emissions. The aggregate impact of meteorological variables in the model explained 45% of variance in CO, 14% in SO2, 25% in NO2 and 24% in PM10. This indicates that meteorology, at least on a local scale, is a noticeable driver of air quality in Zagreb. Stable atmospheric conditions in the urban area favour the occurrence of higher concentrations of air pollutants. Convection processes dominate under unstable conditions, resulting in the dilution of pollutant concentrations within the boundary layer.