Impact of Satellite Aerosol Assimilation on AOD Representation and Long-Term Trends in CAMS Reanalysis

Atmospheric aerosols play a crucial role in the Earth system by influencing climate, air quality, human health and environmental processes. Aerosol loading is quantified in optical terms using Aerosol Optical Depth (AOD), a wavelength dependent measure of the attenuation of solar radiation by particles in the atmosphere. The Copernicus Atmosphere Monitoring Service (CAMS) provides global aerosol reanalyses by combining numerical modelling with satellite observations (among others) via data assimilation. In the CAMS ECMWF Atmospheric Composition Reanalysis 4 (EAC4), satellite-retrieved AOD at 550 nm is assimilated to constrain aerosol fields. A parallel control simulation (CTRL), produced without AOD assimilation, enables the direct impact of assimilation to be isolated. In this study, we assess the impact of data assimilation both in the representation of the aerosol burden and its long-term variability by comparing EAC4 and CTRL configurations over the period 2003–2024. The assimilated observations in EAC4 include AOD from MODIS aboard the Terra and Aqua satellites, complemented by AATSR on-board Envisat satellite. The AOD products from both configurations are evaluated against reference ground-based AERONET measurements at 178 stations worldwide, selected after applying a set of criteria in order to ensure a robust and internally consistent analysis of long-term AOD trends. Overall, both EAC4 and CTRL show good agreement with AERONET at low AOD values (AOD < 0.2), which account for the majority of observations. However, as aerosol loading increases, both configurations tend to underestimate the actual AOD, with substantially larger biases in CTRL. Assimilation in EAC4 systematically reduces these underestimations across all AOD ranges and results in a narrower error distribution, indicating improved stability and consistency. Evaluation metrics confirm this improvement, with EAC4 exhibiting higher correlations, lower errors, and a near-zero mean bias relative to AERONET. The assimilation of satellite AOD indirectly alters the simulated aerosol composition, with the most pronounced changes occurring over dust-dominated regions, where the relative dust contribution is substantially reduced and partially redistributed toward organic matter and other aerosol components, indicating that CTRL overestimates dust loading. Additionally, satellite AOD assimilation significantly improves the representation of long-term AOD trends. Trends derived from EAC4 show strong agreement with AERONET, correctly capturing both the magnitude and sign of observed changes over the majority of the AERONET stations. Compared to CTRL, EAC4 demonstrates markedly higher correlations, reduced trend errors, and a greater fraction of statistically significant trends consistent with observations. These results highlight that satellite AOD assimilation not only enhances present-day aerosol distributions but also plays a key role in constraining long-term aerosol variability in reanalysis products.

Acknowledgements: Part of this work was supported by the COST Action Harmonia (CA21119) supported by COST (European Cooperation in Science and Technology). This work received financial support through the ACTRIS Switzerland 2025-2028 grant (Swiss contribution to the ACTRIS ERIC) funded by the Swiss State Secretariat for Education and Research and Innovation (SERI).