Comparative evaluation of CHIMERE and ENSEMBLE model estimates versus in-situ measurements of particulate matter in the Iasi urban area, Romania

This research investigates the air quality in Iasi in 2024 by comparing estimations of particulate matter (PM2.5 and PM10) from the European Copernicus Atmosphere Monitoring Service (CAMS) with measurements obtained from an urban background air quality monitoring station within Romania's national network. The analysis compared the Ensemble CAMS output, representing an average of many models, to the individual CHIMERE model. Both provide a spatial resolution of 0.1° × 0.1°. The expected yearly mean for PM2.5 was 13.56 µg/m³ with CHIMERE and 13.15 µg/m³ with the Ensemble multi-model, compared to an observed value of 15.52 µg/m³. CHIMERE estimated PM10 at 21.2 µg/m³, whereas the Ensemble calculated it at 16.31 µg/m³, while the measured value was 27.57 µg/m³. The monthly average bias was negative for both models, indicating underestimation (December being an exception). The monthly root mean squared error (RMSE) varied from 3.8 to 21.7 µg/m³, depending on the model and pollutant. Optimal correlations for PM2.5 were observed in November, with both models exhibiting values of 0.9. Strong correlations were noted for PM10 in November and April for both CHIMERE and Ensemble, indicating enhanced model performance throughout seasonal transitions. Of the 25 exceedances of the daily PM10 limit value documented in 2024 at the urban background station, the CHIMERE model validated seven, while the Ensemble model validated none. The conclusions highlight that, while CAMS products are effective for delineating spatial pollution patterns, rigors local validation is necessary, especially in urban environments, where local factors and weather patterns may influence the accuracy of simulate predictions.