Lead: Srdan Dobricic, Joint Research Center


Local atmospheric pollutant forecast service

Photo by MonikaP at Pixabay


By using an AI-based approach with local observations the “Local Atmospheric Pollutant Forecast” Service will improve forecasts by global and regional chemical models of anthropogenic and natural causes of air pollution providing benefits to local communities, and supporting decisions regarding pollution reduction and prevention.


The fast environmental change in the Arctic due to global warming provides favourable conditions for new economic activities in the region and population growth. These processes lead to increased air pollution from traffic, heating and industrial production, while the higher population density increases the risk for accidental forest fires and exposure to air pollution. Air pollution forecasts are essential for mitigation of pollution and for emergency preparedness in Arctic villages and cites. We will use an AI-based approach linking in-situ observations of Arctic air pollution to large-scale predictions of air pollution by global and regional chemical models to inform on airborne pollution impacts. At stations in the Arctic, selected in the initial stage of the study, according to data availability and user requirements the Pilot Service will apply an Artificial Neural Network (ANN) methodology to combine model forecasts by the Copernicus Atmosphere Monitoring Service and other large-scale dynamical models with local observations of atmospheric pollution. We will use the ANN to improve local short-term forecasts of atmospheric pollution which are otherwise based solely on large-scale dynamical model predictions. We will include information on short-term high particulate levels, these cannot be always be accurately predicted by large-scale dynamical models due to large uncertainties in specifying local sources of air pollution. We will provide the forecasts to users in the two following forms:

  • Visual diagnostics showing temporal evolution and forecasts of air pollution at selected Arctic stations. The graphs intended for non-specialists will compare observed and forecasted air pollution concentrations to concentrations observed at stations outside of theArctic
  • Visual diagnostics and tables showing differences between local forecasts from large-scale dynamical models and forecasts produced by ANN. We will provide these to model developers for evaluating model performance at the local scale and improving air pollutio nsource specification. Theair pollution forecasts willprovide benefits to local communities, and support decisions regarding pollution reduction and prevention. The Pilot Service will further improve the use of local air pollution observations by providing more precise forecasts. It will bring feedback information to Copernicus models on their errors at the local scale and in this way improve the understanding of deficiencies in defining sources of air pollution in the Arctic.