Do rural background sites capture changes in primary PM2.5 emissions at the national scale? Recent trends in PM2.5 and its main components in metropolitan France.

Trends in daily PM_2.5 mass concentrations were assessed at 5 rural background sites in France over the period 2014-2021, together with major particulate components associated with (i) anthropogenic emissions including fossil fuels (FF) and biomass burning (BB) from primary emissions; and (ii) non-sea-salt sulphate, nitrate and ammonium, i.e., secondary particulate constituents from gaseous precursors sulphur dioxide, nitrogen oxides and ammonia, respectively. To disentangle the influence of weather, long-range transport, and the oxidative capacity of the atmosphere, boosted regression tree (BRT) models were built at each site; and normalised time series were calculated by randomising the value of the explanatory variables at a given time. Different BRT models were formulated and two types of normalised PM_2.5 time series were calculated: de-weathered time series (without the influence of the meteorological and long-range transport) and de-weathered & de-oxidised time series (randomisation of meteorology, transport and O_X (NO₂ + O₃) levels). Over the studied period, PM_2.5 concentrations decreased at approximatively -5% year^-1, almost 1.5 times faster than changes in primary emissions in France. Overall trends in de-weathered, and de-weathered and de-oxidised PM_2.5 concentrations were lower than trends in PM_2.5 observations for the same period, at -3.9% year^-1 and -3.2 % year^-1, respectively. Trends in de-weathered & de-oxidised PM_2.5 were close to those in emissions, demonstrating the importance of including variables capturing the oxidative capacity of the atmosphere in the normalising techniques to compare trends in PM_2.5 with trends in primary emissions. Trends in observations of PM_2.5 were consistent with trends in nitrate particles from reduced NO_X emissions, and to trends in ammonium particles and biomass burning.