Evaluation of European Air Quality Simulations in 2019 Using the CMAQ–WRF Modeling System

Air pollution levels across Europe have declined significantly over the past two decades, largely due to targeted emission controls in the energy, industry, and transport sectors. Recent monitoring data (2023–2024) show that EU air quality standards are met at 99% of stations for PM_2.5 and 98% for nitrogen dioxide (NO₂). Despite this progress, air pollution remains the leading environmental health risk in Europe. The European Environment Agency (EEA) estimates that annual exposure to PM_2.5, ozone (O₃), and NO₂ caused approximately 239,000, 70,000, and 48,000 premature deaths, respectively, in the EU–27.  Therefore, member States are required to prepare National Air Pollution Control Programmes (NAPCPs) and local action plans, with a stronger role for air quality modeling in complementing monitoring, assessing pollutant distributions, and evaluating mitigation pathways. Air quality modeling has become indispensable for understanding pollutant dynamics, quantifying the effects of emission reductions, and designing integrated air–climate policies. Large–scale initiatives such as FAIRMODE, AQMEII, EURODELT, and HTAP have advanced knowledge on model performance and uncertainties, while emphasizing the need for harmonized approaches across Europe. These efforts show that robust policy support requires continent–wide evaluations based on consistent emissions, updated meteorological drivers, and comprehensive observational datasets. In this context, a harmonized pan-European framework based on the Weather Research and Forecasting (WRF) and Community Multiscale Air Quality (CMAQ) models is developed and evaluated for 2019 using unified meteorological, chemical, and emission inputs. Model skill for O₃ and PM_2.5 is assessed against observations from over 2300 and 1700 stations, respectively, employing standard and advanced diagnostics (mean bias (MB), root mean square error (RMSE), correlation coefficient (r), Taylor, Target, and quantile-binned error (QBE) analyses). For hourly O₃, r = 0.51 and MB = +1.9 ppb; for hourly PM_2.5, r = 0.52 and MB = −5.0 µg m⁻³ (Mod–Obs). The model reproduces large-scale and seasonal patterns but underestimates PM2.5 by 4–6 µg m⁻³ and damps O₃ variability by ~40–50 %. Taylor and Target diagnostics show that random and phase errors dominate (uRMSD/σ_obs ≈ 0.85–0.9), whereas systematic bias is modest (MB/σ_obs ≤ 0.3–0.5). QBE analysis confirms amplitude compression, with underestimation of high-O₃ and high-PM_2.5 events and overprediction of low-O₃ levels. Overall, model skill is limited more by variance and episode representation than by mean offset. Improving boundary-layer dynamics, emission timing, and secondary aerosol processes will reduce seasonal and regional biases. Despite moderate underestimation, the framework provides a scientifically robust, harmonized basis for continental air-quality evaluation and scenario analysis, consistent with FAIRMODE, AQMEII, and the EU 2030 Air Quality Directive.