| 11 Aug 2026
Thirteen years of aerosol chemical composition measurements in the Paris region: reprocessing and applications
Abstract. The recent development of very long-term Aerosol Chemical Speciation Monitor (ACSM) time series presents new insights into the spatiotemporal variability of atmospheric submicron aerosol (PM1) composition. However, challenges in analyzing long-term datasets remain due to their discontinuities. This study proposes a new methodology to homogenize long-term features monitored at the semi-urban SIRTA observatory, located in the Paris region, and then analyzes the variability of the reprocessed dataset, which can be found in EBAS database at https://doi.org/10.48597/4RKH-6RDU (Petit, J.-E. and Brochet, J. and ACTRIS and GAW-WDCA, 2025). Response factor (RF) and relative ionization efficiencies (RIEs) have been identified as major parameters to reevaluate, as the concentrations of aerosol are most sensitive to these parameters. Secondly, corrections of ion transmission and mass calibration present a smaller, yet significant, interest in the harmonization protocol. This approach shows a good consistency with the latest standard operation protocol and encouraging results when compared to previous studies. Long-term datasets allow the evaluation of parameterizations like the composition-dependent collection efficiency (CDCE) determined by Middlebrook et al. (2012). In this case, the SIRTA dataset shows good correlation with the CDCE equations, as they stay in their data inter-quartile range, despite the lack of extreme concentration values recorded at the SIRTA observatory. However, this study shows better data consistency with a default collection efficiency (CE) of 0.5 rather than 0.45.
From 2012 to 2024, PM1 have decreased by 3 μg/m3/yr. Within PM1 components, NH4 and SO4 are significantly decreasing at the annual and seasonal scale, in Autumn (NH4 -0.050 and -0.053 μg/m3/yr, SO4 -0.044 and -0.046 μg/m3/yr respectively). Spring, with the highest PM1 concentrations, demonstrates significant decreasing trends for the majority of PM1 components (OM -0.21 μg/m3/yr, NO3 -0.23 μg/m3/yr, NH4 -0.12 μg/m3/yr and SO4 -0.093 μg/m3/yr). Compared with the original dataset, these results highlight the critical importance of harmonized reprocessing for the analysis of very long-term features. Persistent Pollution Episodes (PPE) also have a decreasing trend of occurrence (5 % per year) and might correlate between their duration and maximum concentrations. These trends are placed in the context of synoptic weather regimes, in which blocking regimes dominate the formation of PPE. Finally, the long-term measurements of SO4 at SIRTA reveal the need for S-related source apportionment analysis, through the identification of the contributions of volcanic and marine biogenic aerosols. Overall, with the general reduction in atmospheric aerosol concentrations, this study shows encouraging results for air quality, emphasizing the need to maintain mitigation efforts in the Paris region.
The manuscript presents a detailed reprocessing strategy for the long-term ACSM measurements at SIRTA and provides a valuable homogeneous dataset spanning more than 13 years. I found the overall methodology and presentation generally clear. My comments focus on Section 4 (the long-term analyses) and on the overall significance of the dataset.
Major comments:
(a) apply the identical trend analysis (same period, same temporal aggregation, same Mann–Kendall/Sen settings) to both the original and the reprocessed datasets;
(b) show the two annual-mean time series and their difference;
(c) identify which processing steps are primarily responsible for the differences.
(a) The Abstract (L15) reports a PM1 decrease of 3 µg m⁻³ yr⁻¹, whereas Sect. 4.1 and Table A2 give 0.34 µg m⁻³ yr⁻¹.
(b) The NH4 and SO4 slopes appear swapped between the text (L249–250: NH4 0.05, SO4 0.044) and Table 2 (SO4 −0.05, NH4 −0.044). The global SO4 slope in Table A2 (−0.44) is presumably −0.044.
(c) The spring NRPM1 slope (−0.83 µg m⁻³ yr⁻¹) is considerably larger than the sum of the component slopes (~−0.65). Sen's slopes are not additive, but a brief comment would help readers.
Minor comments: