Machine Learning-driven deweathering of air pollutant concentrations at Finokalia monitoring station

For the most part, air pollution is governed by emissions, but it can be affected by meteorological conditions as well. Due to the complex nature of the atmosphere, it is no easy task to decouple the effect of weather on measured concentrations of aerosols and thus uncover the true sources of pollution. In this study, the robust method of Deweathering (also referred to as Meteorological Normalization) will be used for the first time on concentrations measured at the Finokalia Monitoring Station, to examine the impact of meteorology in the area. The variables that will be considered are ground measurements from the station, elements of air mass back trajectories analysis, ERA5 reanalysis meteorological data as well as temporal variables. The pollutants of interest include the Total Particle Number concentration, the Aitken, Accumulation and Nucleation Particle Number concentrations, and finally Carbon Monoxide and Black Carbon concentrations. The deweathered values will be compared to the observations in order to draw conclusions about the sources of local pollution. Tracing the anthropogenic emission sources and investigating them separately from natural causes is crucial to evaluate environmental policies regarding the decrease of air pollution. An attempt will be made to examine the impact of the IMO2020 shipping fuel regulation. It is also discussed how the models appear to have difficulties in handling the dependencies of back trajectory clusters and signals of mixed sources.For the most part, air pollution is governed by emissions, but it can be affected by meteorological conditions as well. Due to the complex nature of the atmosphere, it is no easy task to decouple the effect of weather on measured concentrations of aerosols and thus uncover the true sources of pollution. In this study, the robust method of Deweathering (also referred to as Meteorological Normalization) will be used for the first time on concentrations measured at the Finokalia Monitoring Station, to examine the impact of meteorology in the area. The variables that will be considered are ground measurements from the station, elements of air mass back trajectories analysis, ERA5 reanalysis meteorological data as well as temporal variables. The pollutants of interest include the Total Particle Number concentration, the Aitken, Accumulation and Nucleation Particle Number concentrations, and finally Carbon Monoxide and Black Carbon concentrations. The deweathered values will be compared to the observations in order to draw conclusions about the sources of local pollution. Tracing the anthropogenic emission sources and investigating them separately from natural causes is crucial to evaluate environmental policies regarding the decrease of air pollution. An attempt will be made to examine the impact of the IMO2020 shipping fuel regulation. It is also discussed how the models appear to have difficulties in handling the dependencies of back trajectory clusters and signals of mixed sources