A statistical method to estimate PM2.5 concentrations from meteorology and its application to the effect of climate change

Abstract

A statistical algorithm was developed to estimate PM2.5 concentrations over Europe based on a weather-type representation of the meteorology. We used modeled PM2.5 concentrations as pseudoobservations, because of a lack of PM2.5 speciated measurements over Europe, and included four meteorological variables. This algorithm was evaluated on the learning period (2000–2008) to test its ability to reproduce the pseudoobserved data set and then applied for two climatological scenarios (RCP4.5 and RCP8.5) and one historical (1975–2004) and two future periods (2020–2049 and 2070–2099). In Italy, Poland, and northern, eastern, and southeastern Europe, all future scenarios lead to decreases in PM2.5, whereas in the Balkans, Benelux, the UK, and northern France, they lead to increases in PM2.5. Considering each season separately shows stronger responses, which may vary for a given region and scenario. Decomposing the changes in PM2.5 concentrations as the sum of intertype and intratype changes, and a residual term shows that (1) the residual term is negligible; (2) intertype changes affect more the regions along the Atlantic Ocean; and (3) in most other regions, intertype and intratype changes are often on the same order of magnitude. The relationship between the atmospheric circulation and weather types evolves and therefore modifies the mean of meteorological variables and PM2.5 concentrations. This algorithm offers a novel approach to investigate the effect of climate change on air quality and can be applied to other pollutants, regions, and meteorological models. Furthermore, this approach can be applied using actual speciated PM2.5 observations, if a sufficiently dense monitoring network were available.