Alexander de Meij

Atmospheric pollutants pose significant health risks in Europe, where many cities exceed WHO and EU air quality standards. To support informed air quality policies, simulations of air pollution levels are required. They are generally provided by Chemical Transport Models, but these can be computationally expensive at high spatial resolution. In this context, deep learning methods have emerged as alternatives to full simulations. Using low and high resolution EMEP simulations, we investigate how different training strategies affect the robustness of a Residual Channel Attention Network for downscaling hourly particulate matter, nitrogen dioxide and ozone ground concentration fields from a 0.1°×0.05° to a 0.02°×0.01° resolution over Europe. The geographic extent and temporal sampling of the training data appear to be key determinants of the performance of the downscaling process: training the network on days that are evenly distributed throughout the year generally yields a better performance than training it on the same number of days but restricted to the winter season, and, similarly, the choice of a larger spatial domain during training increases the model performance in comparison with a smaller spatial domain. The optimised model achieves better performance in reconstructing fine scale spatial gradients than a bilinear interpolation of low resolution simulations. It generalises to multiple emission reduction scenarios when the link between the perturbed emissions and the target pollutant is strong and direct, while weaker and more indirect responses are less reliably reconstructed. An optimal compromise between temporal resolution and downscaling accuracy is also identified, corresponding to approximately one day of temporal averaging. These results highlight the reliability of neural network-based downscaling techniques, to support their integration into operational air quality assessment and scenario planning tools as complements to full simulations.

Atmospheric pollutants pose significant health risks in Europe, where many cities exceed WHO and EU air quality standards. To support informed air quality policies, simulations of air pollution levels are required. They are generally provided by Chemical Transport Models, but these can be computationally expensive at high spatial resolution. In this context, deep learning methods have emerged as alternatives to full simulations. Using low and high resolution EMEP simulations, we investigate how different training strategies affect the robustness of a Residual Channel Attention Network for downscaling hourly particulate matter, nitrogen dioxide and ozone ground concentration fields from a 0.1°×0.05° to a 0.02°×0.01° resolution over Europe. The geographic extent and temporal sampling of the training data appear to be key determinants of the performance of the downscaling process: training the network on days that are evenly distributed throughout the year generally yields a better performance than training it on the same number of days but restricted to the winter season, and, similarly, the choice of a larger spatial domain during training increases the model performance in comparison with a smaller spatial domain. The optimised model achieves better performance in reconstructing fine scale spatial gradients than a bilinear interpolation of low resolution simulations. It generalises to multiple emission reduction scenarios when the link between the perturbed emissions and the target pollutant is strong and direct, while weaker and more indirect responses are less reliably reconstructed. An optimal compromise between temporal resolution and downscaling accuracy is also identified, corresponding to approximately one day of temporal averaging. These results highlight the reliability of neural network-based downscaling techniques, to support their integration into operational air quality assessment and scenario planning tools as complements to full simulations.

Air pollution remains a major global challenge, affecting human health, agricultural productivity, and climate. Reduced-form models have been widely used to support rapid assessment of emission control strategies. This study presents an updated version of a global air quality screening tool, JRC-FASST, developed using the EMEP chemical transport model. JRC-FASST is a reduced-form Source-Receptor model, which translates changes in emissions of air pollutants from different source regions into changes in atmospheric concentrations. The new model version improves the representation of air pollution, by increasing spatial detail and refining the definition of emission source regions, leading to a more accurate estimation of pollutant concentrations and population and ecosystem exposure. A key enhancement of JRC-FASST is the expanded and refined set of source regions. JRC-FASST employs 76 source regions globally simulated at 0.5° × 0.5° resolution, including 41 individual countries in Europe simulated at 0.1° × 0.1° resolution. This allows a more detailed representation of national emissions and their transboundary impacts. It incorporates an expanded set of air quality indicators, including particulate matter, nitrogen dioxide, and multiple ozone metrics relevant for both health and crop impacts. JRC-FASST reproduces the results of the full atmospheric model with high accuracy, particularly for particulate matter and ozone, while maintaining computational efficiency suitable for policy screening. Overall, the updated model provides a more robust and flexible tool for exploring air quality policies at national, regional and global scales. Moreover, the current version establishes the methodological foundation for a continuously evolving platform.

Air pollution remains a major global challenge, affecting human health, agricultural productivity, and climate. Reduced-form models have been widely used to support rapid assessment of emission control strategies. This study presents an updated version of a global air quality screening tool, JRC-FASST, developed using the EMEP chemical transport model. JRC-FASST is a reduced-form Source-Receptor model, which translates changes in emissions of air pollutants from different source regions into changes in atmospheric concentrations. The new model version improves the representation of air pollution, by increasing spatial detail and refining the definition of emission source regions, leading to a more accurate estimation of pollutant concentrations and population and ecosystem exposure. A key enhancement of JRC-FASST is the expanded and refined set of source regions. JRC-FASST employs 76 source regions globally simulated at 0.5° × 0.5° resolution, including 41 individual countries in Europe simulated at 0.1° × 0.1° resolution. This allows a more detailed representation of national emissions and their transboundary impacts. It incorporates an expanded set of air quality indicators, including particulate matter, nitrogen dioxide, and multiple ozone metrics relevant for both health and crop impacts. JRC-FASST reproduces the results of the full atmospheric model with high accuracy, particularly for particulate matter and ozone, while maintaining computational efficiency suitable for policy screening. Overall, the updated model provides a more robust and flexible tool for exploring air quality policies at national, regional and global scales. Moreover, the current version establishes the methodological foundation for a continuously evolving platform.

In Europe, excessive nutrient pollution in freshwater and marine ecosystems poses a significant threat to human uses of water resources and disrupts aquatic ecosystems, compromising essential ecosystem services. The effectiveness of existing EU policies aimed at mitigating nutrient pollution will be influenced by the combined impact of climate variability and change. It remains uncertain whether current policies will suffice to reduce nutrient pollution to levels compatible with good environmental conditions in freshwaters and coastal waters. This study quantifies nitrogen and phosphorus fluxes in European freshwater systems and their delivery to the sea, considering the effects of existing EU policies and climate change up to 2050. Using the GREEN model, we simulated total nitrogen and phosphorus loads from 1990 to 2021 and projected future changes until 2050 using climate projections from 11 regional climate models (EURO-CORDEX) under the RCP 4.5 scenario. Our results indicate that implementing current EU policies—Common Agricultural Policy, Urban Waste Water Treatment Directive, FitFor55 package for climate neutrality—will reduce nutrient loads to European seas by 2050, with a 10% decrease in nitrogen and 8% decrease in phosphorus loads compared to 2020 values, albeit with significant uncertainty related to climate variability. Regional differences are also explored. Nutrient concentrations in the EU river network are expected to decrease slightly, as well as the potential risk for coastal eutrophication (ICEP index). However, further reductions of 10%–20% beyond existing policies may be necessary to achieve good conditions in large rivers. These reductions could come from improved agricultural practices, nature-based solutions and wetland restoration. Our study suggests that while current measures will lead to improvements by 2050, additional efforts will be required in some regions to achieve good environmental conditions, highlighting the need for integrated nutrient management to avoid imbalances between nitrogen and phosphorus in receiving waters.

In this study, we assess the relevance and utility of several performance indicators (model quality (bias) and model performance (temporal and spatial) indicators), developed within the FAIRMODE framework by evaluating eight Copernicus Atmospheric Monitoring Service (CAMS) models and their ensemble in calculating concentrations of key air pollutants, specifically NO₂, PM_2.5, PM₁₀ and O₃. The models' outputs were compared with observations that were not assimilated into the models. For NO₂, the results highlight difficulties in accurately modelling concentrations at traffic stations, with improved performance when these stations are excluded. While all models meet the established criteria for PM_2.5, indicators such as bias and winter–summer gradients reveal underlying issues in air quality modelling, questioning the stringency of the current criteria for PM_2.5. For PM₁₀, the combination of model quality indicators, bias, and spatial-temporal gradient indicators prove most effective in identifying model weaknesses, suggesting possible areas of improvement. O₃ evaluation shows that temporal correlation and seasonal gradients are useful in assessing model performance. Overall, the indicators provide valuable insights into model limitations, yet there is a need to reconsider the strictness of some indicators for certain pollutants.

Ambient fine particulate matter (PM2.5) poses a significant risk to health in Europe, where many cities are exposed to levels above World Health Organization guidelines. To support the implementation of optimal PM2.5 reduction policies, air quality models are necessary. In such context, source-receptor relationships (SRRs) are models that can be used to replace fully-fledged Chemical Transport models, and save significant computation time when simulating various emission reduction scenarios. They allow calculating the relative potential of a given source at a receptor, i.e. the share of PM2.5 concentration at a given receptor that results from the complete removal of the emissions from that source. Here, we use the SRR model SHERPA, based on the EMEP Chemical Transport Model for four different meteorological years (2015, 2017, 2019 and 2021), to evaluate relative potentials for 150 European cities. These potentials are evaluated for five different emission precursors, twelve emission sectors, and four reduction scopes (city core, commuting zone, remaining national territory and rest of Europe). Results show that relative potentials vary little between meteorological years for most cities. The industry, transport and residential sectors generally bear the highest values of relative potential for most cities through emissions of primary particulate matter. Cities near important ports where the shipping sector exhibits high values through sulfur oxides. High potentials are observed for agriculture at the national and international scales. Cities in Southern Europe have low reduction potentials due to high PM2.5 levels originating from natural sources. Smaller cities have a higher relative potential for national and international emissions, while larger cities have high relative potentials for city core emissions. Relative potentials are generally low for commuting zones. These results underline the reliability of SRRs in guiding targeted air quality interventions, thereby helping to reduce PM2.5 exposure effectively across diverse urban settings in Europe.