PM2.5 Pollution Decrease in Paris, France, for the 2013–2024 Period: An Evaluation of the Local Source Contributions by Subtracting the Effect of Wind Speed
Kasni Sumeru
Round 1
Reviewer 1 Report
Comments and Suggestions for Authors1. Introduction
In lines 43-44, using references [6-13], mention the important results reported by two or three of these references.
The last sentence in lines 50-51 does not mention any references. This sentence should require references as references.
Lines 52 to 59 are not needed in the Introduction. The information is used in the discussion when comparing the measurement results with the current applicable standards.
The two paragraphs (lines 60 to 63 and 64 to 77) should appear at the beginning of the Introduction to connect with the title.
2. Materials and Methods
The explanation of Fig. 1 is not comprehensive. For ease of comparison, the X-axis should be the same for Fig. (a) and (b), i.e., from 2018 to 2024. There is no explanation of the difference between the data in (a) Airparif and (b) Pollutrack. Explain the similarities and differences. Lines 144-145 mention that pollutant levels peak during winter, but the cause is not explained comprehensively.
3. Results
It would be better to combine the Results and Discussion sections to make it easier for readers to understand. Moreover, the discussion section is not very extensive or in-depth. In addition, the results section also includes some discussion. The format and quality of the images should be improved. The image format seems outdated and lacks modernity.
4. Discussion
It should be combined with Results. The discussion should be more comprehensive. For example, by mentioning the potential negative impact of PM2.5 in Paris if it is not controlled.
5. Conclusions
The first and second paragraphs (lines 379 to 401) should be revised to be consistent with the title and purpose of this article.
Comments on the Quality of English Language--
Author Response
Authors’ comment: We thank the reviewer for these very useful comments, which have helped us to improve the paper.
Reviewer: In lines 43-44, using references [6-13], mention the important results reported by two or three of these references.
Authors’ answer: We have changed the sentence to : “They can trigger asthma attacks, respiratory diseases (including Covid-19), cancers, heart attacks, strokes, neurological disorders, alteration of placenta, worsening diseases morbidity and thus increasing the mortality rate among the population living in polluted areas [6-13].”
Reviewer: The last sentence in lines 50-51 does not mention any references. This sentence should require references as references.
Authors’ answer: The referenced number was misplaced. We have moved it to the end of the sentence.
Reviewer: Lines 52 to 59 are not needed in the Introduction. The information is used in the discussion when comparing the measurement results with the current applicable standards.
The two paragraphs (lines 60 to 63 and 64 to 77) should appear at the beginning of the Introduction to connect with the title.
Authors’ answer: We have followed a conventional approach of presenting the subject of this paper in the introduction, moving from general consideration to more specific ones. The new WHO recommendations partly justify the aim of this paper, as already stated in the sentence; “Studies must focus on urban zones, which are often the most polluted and where most of the population lives”, and should be therefore presented in the introduction. To follow this approach, we have reorganized the last 3 paragraphs to be better in line with this structure, and to present, at the end of the introduction, the effect of winds on pollution.
Reviewer: The explanation of Fig. 1 is not comprehensive. For ease of comparison, the X-axis should be the same for Fig. (a) and (b), i.e., from 2018 to 2024. There is no explanation of the difference between the data in (a) Airparif and (b) Pollutrack. Explain the similarities and differences.
Authors’ answer: The X-axis are now the same for the two figures. We have changed the text to :” Although some differences can be observed in the maximum pollution levels, due to the different number of sensors and the resulting mean values, the two datasets display similar behavior”.
Reviewer: Lines 144-145 mention that pollutant levels peak during winter, but the cause is not explained comprehensively.
Authors’ answer: We have changed the sentence to: “First, a clear annual cycle is evident, with higher pollution peaks during winter due to the additional contribution of heating to traffic emissions, but also to the lower altitude of the upper limit of the boundary layer [49]. Since the boundary layer height is lower in winter than in summer, the vertical dispersion of particulate matter is reduced, leading to higher concentrations near the surface”
Reviewer: It would be better to combine the Results and Discussion sections to make it easier for readers to understand. Moreover, the discussion section is not very extensive or in-depth. In addition, the results section also includes some discussion.
Authors’ answer: Sometime reviewers prefer results and discussion to be separated, while others prefer to be merged. We understand the reviewer’s concern, as there are two different results to present: the wind cutoff at 6 m.s-1, and the uncorrected and corrected PM2.5 trends. It seems difficult to provide a meaningful discussion before presenting all results. Nevertheless, the reviewer is correct that a short conclusion was missing at the end of the results section. We have therefore added: “All these results could argue for a revision of the PM2.5 pollution trends and of the contribution of PM2.5 sources in Paris. This is discussed in the ensuing section.” Finally, we do not see what additional points could be addressed in the discussion (the reviewer’s comment does not suggest new angles of in-depth discussion).
Reviewer: The format and quality of the images should be improved. The image format seems outdated and lacks modernity.
Authors’ answer: We do not understand which format is considered outdated, or what kind of modernity is expected for the presentation of scientific results. We are aware that authors often use Python of Excel software to produce figures, which is not our case. We therefore propose that the editor decides whether our figures require another contemporary look.
Reviewer: It should be combined with Results. The discussion should be more comprehensive. For example, by mentioning the potential negative impact of PM2.5 in Paris if it is not controlled.
Authors’ answer: As mentioned earlier, we prefer not to combine the two sections. We have added at the beginning of the discussion: “Due to their effect on human health, the PM2.5 mass-concentrations levels must be accurately measured in major cities such as Paris. To comply with the new European Ambient Air Quality Directive [18], the sources of PM2.5 emissions must be accurately identified and controlled. The ongoing effort to reduce the PM2.5 emissions focused on ameliorating thermic motor exhausts, reducing traffic and the number of vehicles within Paris, and better controlling wood-heating.”
Reviewer: The first and second paragraphs (lines 379 to 401) should be revised to be consistent with the title and purpose of this article.
Authors’ answer: We have changed the first sentence to: “A statistical method based on wind speed, using a cutoff at 6 m.s-1, made it possible to distinguish, in Paris, between permanent background pollution during high wind speeds and pollution peaks occurring under anticyclonic conditions with low wind speeds.” All the other parts of the discussion are consistent with the title of the paper.
Reviewer 2 Report
Comments and Suggestions for AuthorsThe manuscript could benefit from several improvements that could increase its scientific value and reader interest.
1. The research proves the link between wind and pollution levels, but other strong influences are also known, such as the effect of precipitation on atmospheric pollution levels. It is not possible to choose just one element of the atmospheric system because it is a whole system of links and interdependencies between climatic factors.
2. The mathematical apparatus used is not described in great detail in the paper. It would have been useful to explore more powerful statistical tools such as correlation functions between the analyzed time series.
3. Even if the purpose and level of specificity of the research are clearly specified, the authors should still provide clearer solutions for future generalization of the results. It is the authors' duty to provide a direction for the general use of their own results.
Author Response
Authors’ comment: We thank the reviewer for these very useful comments, which have helped us to improve the paper.
Reviewer: The manuscript could benefit from several improvements that could increase its scientific value and reader interest.
1. The research proves the link between wind and pollution levels, but other strong influences are also known, such as the effect of precipitation on atmospheric pollution levels. It is not possible to choose just one element of the atmospheric system because it is a whole system of links and interdependencies between climatic factors.
Authors’ comment: The reviewer raises an important point. Regarding rain, we have demonstrated that its effect is often short-term, as described in the text and in a reference (McMullen, N.; Annesi-Maesano ,I.; Renard, J.-B. Impact of rain precipitation on urban atmospheric particle matter measured at three locations in France between 2013 and 2019, Atmosphere 2021, 12, 769). We also wrote: “By contrast, no significant correlation between PM2.5 mass-concentrations is observed with the other meteorological parameters (wind direction, humidity, temperature) beyond the usual seasonal ones”, and “This balance can, however, be disturbed by stormy conditions with rains and strong winds, which can sharply reduce PM2.5 concentrations”. Our aim was not to establish correlations with all the weather parameters, but rather to focus on the correlation between wind speed and PM pollution, which is the main parameter allowing us to distinguish between anticyclonic and cyclonic conditions, as explained in the text. Of course we have analyzed the correlations between PM2.5 levels and other weather parameters, and none of them exhibit the same impact as wind speed. Considering the entire atmospheric system could introduce errors in the analysis, since PM production and transport as more sensitive to some parameters than to others. We have therefore added at the end of the introduction: “although other weather parameters could also influence PM2.5 levels and contribute to the dispersion of their values.”
Reviewer: 2. The mathematical apparatus used is not described in great detail in the paper. It would have been useful to explore more powerful statistical tools such as correlation functions between the analyzed time series.
Authors’ comment: The mean, median and most probable calculations are widely used, and we think it is not necessary to describe them again. In addition, two paragraphs already present the method of analysis. The correlation function between the time series is not suitable for this analysis since, as written in the text, identical wind conditions can produce different PM2.5 levels, depending on the duration of the anticyclonic conditions. This effect is illustrated in Figure 6c, which shows a decreasing standard deviation with increasing wind speed. Almost no studies take into account the effect of the duration of anticyclonic conditions when analyzing time-series data. We have therefore added in the text (section 3.2): “These results show that a direct correlation between PM2.5 and wind time series cannot be established, since the consecutive number of anticyclonic days could affect the absolute value of the PM2.5 mass-concentrations and thus the correlation.”
Reviewer: 3. Even if the purpose and level of specificity of the research are clearly specified, the authors should still provide clearer solutions for future generalization of the results. It is the authors' duty to provide a direction for the general use of their own results.
Authors’ comment: Although the last paragraph heads up in this direction, we have added to the conclusion: “To ensure the constituency of the results, it is recommended to use at least two independent sets of measurements, in order to verify that the results obtained from the limited number of air quality agency monitoring stations are truly representative of the real PM2.5 pollution levels.”, and “Considering such approach, two different values should be produced for each city to better characterize the PM2.5 pollution levels and therefore its evolution over time: the annual mean value calculated under all weather conditions, and the annual value derived only from local sources during high-wind conditions, more characteristic of its background pollution footprint.”
Reviewer 3 Report
Comments and Suggestions for AuthorsThis paper presents an analysis of PM2.5 pollution trends in Paris over an 11-year period, proposing a methodology to isolate local source contributions by using wind speed as a discriminating factor. The authors identify a 6 m/s wind speed threshold to distinguish between background pollution and meteorologically-influenced pollution episodes, ultimately estimating that local emissions have decreased by approximately 4% per year.
The study addresses a relevant environmental health issue and proposes an interesting approach to disentangle meteorological influences from emission source trends. The identification of a wind speed threshold to isolate background pollution represents a potentially useful methodological contribution. The authors utilize two complementary datasets - the official Airparif network and the mobile Pollutrack sensors - which provides valuable cross-validation of their findings. The statistical approach of comparing three different annual aggregation methods (mean, median, and most probable value) demonstrates methodological rigor in handling non-normal PM2.5 distributions.
However, several methodological concerns limit the robustness of the conclusions. The fundamental assumption that PM2.5 concentrations during high wind conditions (>6 m/s) represent pure background pollution from local sources is questionable. During high wind events, air masses can transport pollutants from distant sources, potentially confounding the interpretation of these measurements as local background levels. The 6 m/s threshold appears to be empirically derived from the data inflection point rather than being grounded in atmospheric dispersion theory or validated against independent measurements of source contributions. The paper would benefit from discussing potential biases introduced by this assumption and comparing results with established source apportionment methods.
The limited spatial representation poses another significant concern. Despite the authors' acknowledgment of spatial heterogeneity in urban PM2.5 concentrations, the analysis treats Paris as a spatially homogeneous entity. The Airparif network consists of only 3-6 stations within Paris proper, which may not adequately capture the city's pollution variability. While the Pollutrack mobile sensors provide broader spatial coverage, their shorter temporal record (2018-2024) limits the reliability of trend calculations. The substantial differences in calculated trends between the two datasets (4.1-4.3% vs 4.5-6.0% per year) suggest either methodological inconsistencies or real spatial variability that undermines the citywide generalization.
The statistical analysis requires strengthening in several areas. The paper lacks confidence intervals for the trend estimates and provides no significance testing to determine whether observed trends are statistically meaningful. The choice of linear regression may not be appropriate given the acknowledged non-normal distribution of PM2.5 concentrations and potential nonlinear relationships between emissions and concentrations. The correlation analysis between PM2.5 and meteorological variables beyond wind speed is superficial and could benefit from more sophisticated multivariate approaches that account for interaction effects.
The validation of the wind speed methodology is insufficient. The paper does not compare results against established source apportionment techniques such as chemical mass balance, positive matrix factorization, or dispersion modeling that could provide independent verification of local vs. regional source contributions. The assumption that high wind conditions eliminate the accumulative effects of local emissions oversimplifies complex atmospheric mixing processes and boundary layer dynamics that can vary seasonally and diurnally.
The temporal scope and data quality considerations need better treatment. The study period includes significant policy interventions, economic fluctuations, and the COVID-19 pandemic, yet these external factors receive minimal discussion regarding their potential impact on emission trends. The paper also lacks adequate treatment of measurement uncertainties, particularly for the Pollutrack sensors, which could significantly affect trend calculations given the relatively small annual changes being detected.
The authors should incorporate relevant recent literature on urban air quality monitoring and source apportionment methodologies, including https://doi.org/10.1016/j.scitotenv.2024.174888. The discussion of implications could be strengthened by addressing the policy relevance of the findings and comparing the proposed methodology's advantages and limitations relative to established approaches for tracking emission reduction progress.
Several technical issues require attention. The paper contains inconsistencies in notation and terminology, particularly regarding the distinction between PM2.5 mass concentrations and number concentrations. The figures, while informative, could benefit from improved clarity in labeling and statistical representation of uncertainties. The conclusion that 2023-2024 background levels approach WHO guidelines (5 μg/m³) may be overly optimistic given the methodological limitations and should be presented with appropriate caveats.
The paper makes a reasonable contribution to understanding PM2.5 trends in Paris, but the methodological limitations significantly constrain the reliability of the quantitative conclusions. The wind speed threshold approach represents an interesting concept that warrants further development, but requires more rigorous validation against established source apportionment methods. The work would benefit from expansion to include chemical speciation data, comparison with other European cities, and more sophisticated statistical approaches that account for the multivariate nature of air quality determinants.
Author Response
Authors’ comment: We thank the reviewer for these very useful comments, which have helped us to improve the paper.
Reviewer: This paper presents an analysis of PM2.5 pollution trends in Paris over an 11-year period, proposing a methodology to isolate local source contributions by using wind speed as a discriminating factor. The authors identify a 6 m/s wind speed threshold to distinguish between background pollution and meteorologically-influenced pollution episodes, ultimately estimating that local emissions have decreased by approximately 4% per year.
The study addresses a relevant environmental health issue and proposes an interesting approach to disentangle meteorological influences from emission source trends. The identification of a wind speed threshold to isolate background pollution represents a potentially useful methodological contribution. The authors utilize two complementary datasets - the official Airparif network and the mobile Pollutrack sensors - which provides valuable cross-validation of their findings. The statistical approach of comparing three different annual aggregation methods (mean, median, and most probable value) demonstrates methodological rigor in handling non-normal PM2.5 distributions.
However, several methodological concerns limit the robustness of the conclusions. The fundamental assumption that PM2.5 concentrations during high wind conditions (>6 m/s) represent pure background pollution from local sources is questionable. During high wind events, air masses can transport pollutants from distant sources, potentially confounding the interpretation of these measurements as local background levels. The 6 m/s threshold appears to be empirically derived from the data inflection point rather than being grounded in atmospheric dispersion theory or validated against independent measurements of source contributions. The paper would benefit from discussing potential biases introduced by this assumption and comparing results with established source apportionment methods.
Authors’ comment: The reviewer raises an important point, and we must be more cautious with our results. Accurately determining source contributions is challenging in a large city due to the high number and diversity of emission sources. Of course, rough estimate can be established for traffic, but other sources, such as construction sites and heating, are difficult to monitor. Furthermore, to our knowledge, there is currently no atmospheric dispersion theory accurate enough to be compared with our results, particularly for cities with specific geography features, although some works are in progress. We have added in part 3.2: “This decrease falls within the uncertainties of the instruments at low concentration levels, of a few µg.m-3, but the possibility of systematic errors cannot be excluded, such as the effect of pollutant transport from distant sources.
These results show that a direct correlation between PM2.5 and wind time series cannot be established, since the consecutive number of anticyclonic days could affect the absolute value of the PM2.5 mass-concentrations and thus the correlation. Nevertheless, this 6 m.s-1 cutoff is empirically determined and just for Paris. This value should be corroborated by future studies that take into account measurements of all possible sources despite their large number and spatial variability, as well as from atmospheric transport theory considering the city’s specific topography. Such modeling works on transport are still in progress for the city of Paris [50-52].”
Reviewer: The limited spatial representation poses another significant concern. Despite the authors' acknowledgment of spatial heterogeneity in urban PM2.5 concentrations, the analysis treats Paris as a spatially homogeneous entity. The Airparif network consists of only 3-6 stations within Paris proper, which may not adequately capture the city's pollution variability. While the Pollutrack mobile sensors provide broader spatial coverage, their shorter temporal record (2018-2024) limits the reliability of trend calculations. The substantial differences in calculated trends between the two datasets (4.1-4.3% vs 4.5-6.0% per year) suggest either methodological inconsistencies or real spatial variability that undermines the citywide generalization.
Authors’ comment: The reviewer is right concerning the low number of reference measurements stations, although all official air quality agencies provide references values with such low number of stations, certainly insufficient but in line with the current requirements from the Joint Research Centre of the European Commission . It is why we have used two different sets of data. If we consider the error bars, similar trends are obtained from the two different data sets, thus there are no methodological inconsistencies. We have added in the conclusion: “To ensure the constituency of the results, it is recommended to use at least two independent sets of measurements, in order to verify that the results obtained from the limited number of air quality agency monitoring stations are truly representative of the real PM2.5 pollution levels.”
Reviewer: The statistical analysis requires strengthening in several areas. The paper lacks confidence intervals for the trend estimates and provides no significance testing to determine whether observed trends are statistically meaningful. The choice of linear regression may not be appropriate given the acknowledged non-normal distribution of PM2.5 concentrations and potential nonlinear relationships between emissions and concentrations.
Authors’ comment: We have provided errors bars for the trends in Table 1 and in the text, confirming that the trends are statistically significant. In addition, the linear trend is commonly used both by air quality agencies and by scientific papers on this subject. It is not obvious which other type of regression could be used to describe the trend, given the data. We have added at the beginning of part 3.1: “Using a linear fit, as is commonly done for this kind of analysis regardless the origins of the trend,”
Reviewer: The correlation analysis between PM2.5 and meteorological variables beyond wind speed is superficial and could benefit from more sophisticated multivariate approaches that account for interaction effects.
The validation of the wind speed methodology is insufficient. The paper does not compare results against established source apportionment techniques such as chemical mass balance, positive matrix factorization, or dispersion modeling that could provide independent verification of local vs. regional source contributions. The assumption that high wind conditions eliminate the accumulative effects of local emissions oversimplifies complex atmospheric mixing processes and boundary layer dynamics that can vary seasonally and diurnally.
Authors’ comment: We adopted a “physicist’s” rather than a “mathematician’s” approach, considering one variable at a time instead of using multivariable approach. Such approach could be inappropriate because it mixes everything together without accounting for possible physical relationship between weather parameters and PM2.5 pollution. As shown in the McMullen et al. paper (reference 4), rainy conditions do not significantly change the pollution levels over several hours, whereas storm conditions completely clean the atmosphere. Fog typically occurs during anticyclonic conditions where the PM2.5 concentrations are higher. Yet these 3 cases exhibit the same humidity level. PM2.5 concentrations can be high or low independently of the temperatures and of the boundary layer dynamics. Moreover, a multivariable approach does not necessarily capture the accumulation of pollution due to several consecutive windless days. How, then, can such an approach resolve these difficulties without introducing bias? It is well known that multivariable models have yielded very promising results, but not in all situations. Our objective was not to solve the entire complex relationship between meteorological conditions and PM2.5 levels, but rather to show that wind speed is a reliable parameter for distinguishing between PM2.5 levels and their sources. We have added in the text: “The aim of this paper is not to evaluate the effect of all weather parameters on PM2.5 levels, since this could be a complex problem with multiple solutions. For example, humidity close to 100% may indicate rain, storms or fog, each producing very different PM2.5 levels. This paper will focus solely on evaluating the effect of wind speed on PM2.5 levels.”
Reviewer: The temporal scope and data quality considerations need better treatment. The study period includes significant policy interventions, economic fluctuations, and the COVID-19 pandemic, yet these external factors receive minimal discussion regarding their potential impact on emission trends.
Authors’ comment: The effect of Covid-19 was not clearly visible in the PM2.5 data for Paris, unlike for NO2 (see for example reference 26). No significant economic fluctuations were detectable in Paris pollution, which mainly originates from traffic and heating. Some policy interventions were implemented to reduce the diesel exhausts and the traffic, which explain the decreasing trend. We have added at the beginning of the discussion: ” Due to their effect on human health, the PM2.5 mass-concentrations levels must be accurately measured in major cities as Paris. To comply with the new European Ambient Air Quality Directive [18], the sources of PM2.5 emissions must be accurately identified and controlled. The ongoing effort to reduce the PM2.5 emissions focused on ameliorating thermic motor exhausts, reducing traffic within Paris, and better controlling wood-heating.”
Reviewer: The paper also lacks adequate treatment of measurement uncertainties, particularly for the Pollutrack sensors, which could significantly affect trend calculations given the relatively small annual changes being detected.
Authors’ comment: We have written: “A mean difference [between Airparif and Pollutrack] of 0.1±3.5 µg.m-3 was observed over several months of comparison.” This value indicates the systematic errors (close to zero) and the standard deviation between the two sets of data. Pollutrack sensors have uncertainties, comparable to the reference air quality monitoring sensors. The observed decreases are far above this uncertainty. We have also written: “This decrease falls within the uncertainties of the instruments at low concentration levels, of a few µg.m-3”; We do not see which adequate treatment of measurements uncertainties is missing, particularly for Pollutrack. We have changed the text to: “Thus, the two datasets are consistent when considering a large number of measurements, as done in this paper, and can be used together to analyze PM2.5 temporal trends”
Reviewer: The authors should incorporate relevant recent literature on urban air quality monitoring and source apportionment methodologies, including https://doi.org/10.1016/j.scitotenv.2024.174888.
Authors’ comment: We have provided a large number of references on this subject, focusing on previous studies in Paris and in Europe. We don’t understand why the reviewer wants to cite this specific paper that is out of the scope of our paper. Here are the title and the abstract of this paper: “Towards space-time modelling of PM2.5 inhalation volume with ST-exposure. Air quality (AQ) is directly relevant with people's health while implementing effective methods for acquiring pollution details and assessing health impact are very important for public health management. In this paper, we design an end-to-end space-time modelling framework to estimate pixelwise PM2.5 inhalation volume, called ST-Exposure which goes over the model's practicality and benefits on the following aspects: (1) Use a combination of fixed and mobile AQ sensors, we estimate PM2.5 inhalation volume based on the inference of PM2.5 exposure in Beijing (3025 , 19 Jun 16 Jul 2018) with the space-time resolution of 1 km 1 km and 1 h, with <15 % SMAPE (%). (2) Achieve pixelwise PM2.5 inhalation volume to be inferred with high-resolution (1 km 1 km, hourly) at city scale, even with sparse space-time coverage. (3) Propose a new calculation mechanism of population distribution which is better than the traditional census-based method, and can achieve more reliable estimation of the total PM2.5 inhalation volume over the whole region.”
Reviewer: The discussion of implications could be strengthened by addressing the policy relevance of the findings and comparing the proposed methodology's advantages and limitations relative to established approaches for tracking emission reduction progress.
Authors’ comment: We have added this in the discussion section, keeping in mind that it is a scientific paper and not a policy relevance-oriented paper: “Due to their effect on human health, the PM2.5 mass-concentrations levels must be accurately measured in major cities as Paris. To comply with the new European Ambient Air Quality Directive [18], the sources of PM2.5 emissions must be accurately identified and controlled. The ongoing effort to reduce the PM2.5 emissions focused on ameliorating thermic motor exhausts, reducing traffic within Paris, and better controlling wood heating.” And in the conclusion “To ensure the constituency of the results, it is recommended to use at least two independent sets of measurements, in order to verify that the results obtained from the limited number of air quality agency monitoring stations are truly representative of the real PM2.5 pollution levels” and “Considering such approach, two different values should be produced for each city to better characterize the PM2.5 pollution levels: the annual mean value calculated under all weather conditions, and the annual value derived only from the local sources during high-wind conditions.
Reviewer: Several technical issues require attention. The paper contains inconsistencies in notation and terminology, particularly regarding the distinction between PM2.5 mass concentrations and number concentrations.
Authors’ comment: We found some typo errors and we corrected them. We have also checked that the words “number concentrations” and “mass-concentrations” are properly used. We have added the definitions in the text “PM2.5 mass-concentrations (integrated mass of all particles smaller than 2.5 µm)” and “Particle number concentrations (number of particles of a given size)”.
Reviewer: The figures, while informative, could benefit from improved clarity in labeling and statistical representation of uncertainties.
Authors’ comment: We think that the reviewer is referring to Figure 3, 4, 6, 7, and 8. The main issue is that no errors bars are available for weather data and for Airparif PM2.5 data, which is indeed a limitation. All the Air quality agencies provide measurements without errors bars, so we cannot include them in the figures. For Pollutrack, error bars were estimated in our previous papers, to be of the order of 3 µg/m3 for individual measurements. When aggregating the values by day, the uncertainties, compared to the Airparif measurements, decrease to 0.1 µg/m3, as already mentioned in the text; We have added in the legend of Figures 4 and 8: “ The mean uncertainty, compared to the Airparif measurements, is 0.1 µg.m-3.”
Reviewer: The conclusion that 2023-2024 background levels approach WHO guidelines (5 μg/m³) may be overly optimistic given the methodological limitations and should be presented with appropriate caveats.
Authors’ comment: We have written a 10-line paragraph on this, starting with “Nevertheless, these optimistic results must be tempered by 3 important points”. We don’t really see what more could be added.
Reviewer: The paper makes a reasonable contribution to understanding PM2.5 trends in Paris, but the methodological limitations significantly constrain the reliability of the quantitative conclusions. The wind speed threshold approach represents an interesting concept that warrants further development, but requires more rigorous validation against established source apportionment methods. The work would benefit from expansion to include chemical speciation data, comparison with other European cities, and more sophisticated statistical approaches that account for the multivariate nature of air quality determinants.
Authors’ comment: We have done our best to address the reviewer’ comments. We believe that the interest of chemical speciation data (which are not easy to determine at a global city scale), is beyond the scope of this paper. As stated in the text, a comparison with other European cities will be carried out in a future paper. Finally, a multivariate nature of air quality determinants is also beyond the scope of this paper. Our aim is more focused: to show that interesting results can be obtained for PM2.5 trends using only wind conditions.
