Temporal Evolution of PM_2.5 Levels and COVID-19 Mortality in Europe for the 2020–2022 Period

Round 1

Reviewer 1 Report

The paper, titled "Temporal evolution of PM2.5 levels related to Covid-19 mortality in Europe for the early 2020 to end 2022 period" addresses the important topic of the relationship between air pollution and respiratory morbidity and mortality - here Covid-19. 

Unfortunately, I identified serious methodological shortcomings in the study.

1. The entire inference is based on Pearson correlation coefficient analysis between PM2.5 concentrations and mortality. The assumption that the analyzed variables follow a normal distribution, which is a necessary assumption for further correct inference, was not previously verified.

2. Linear regression analysis was used, which is fraught with five major assumptions, failure to meet which results in far-reaching incorrect conclusions.

Therefore, I am not in a position to comment on the correctness of the inference, since at the stage of statistical analysis the assumptions crucial to the applicability of the selected methods were not verified. 

Author Response

Dear reviewer

Our answer are in the file below.

Author Response File: Author Response.pdf

Reviewer 2 Report

Reviewer

The manuscript regards a Temporal Evolution of PM2.5 Levels Related to Covid-19 Mortality in Europe for the Early-2020 to End-2022 Period. It reports an interesting analysis, nevertheless, some observations are reported below:

1.      Make the abstract more discursive.

2.      Use µg/m³ or µg m^-3 please uniform the units in the manuscript

3.      The last line of the abstract is cut

4.      In the introduction mention the WHO limit for PM. Read this paper in this regard: https://www.mdpi.com/2076-3298/10/3/42

5.      Please avoid the use of “We” in the paper. Use indirect sentences like “The authors”. Lines 83 and 91 for instance.  

6.      Line 117-118, please introduce the reference properly in the sentence “For the purpose of this work, Covid-19 mortality data are retrieved from the John 117 Hopkins University datasets (https://coronavirus.jhu.edu/).”.

7.      The legend in Figure 1, 2 and 3 is not clear.

8.      Pay attention at the units in the horizontal axis in figure 4. The 2.5 should be subscript

9.      Better explain and summarize the objective of the manuscript at the beginning of the conclusion section.

Despite some minor mistakes, english is fine. 

Author Response

The manuscript regards a Temporal Evolution of PM2.5 Levels Related to Covid-19 Mortality in Europe for the Early-2020 to End-2022 Period. It reports an interesting analysis, nevertheless, some observations are reported below.

Authors answer: We want to thank reviewer for his comments that contributed to improving the paper.

1. Make the abstract more discursive. (reprendre l’abstract revu… en début de revue :))

Authors answer: We have totally rewritten the abstract:

“ Air pollution has a strong impact on human health, from respiratory and pulmonary severe diseases to heart attacks and cancers. During the 3 years of the Covid-19 pandemic, several peaks of mortality occurred, which can be related to particulate matter (PM) pollution events. The possible effect of PM (PM₁₀ and PM_2.5 of diameter lesser than 10 and 2.5 µm respectively) on Covid-19 mortality is now established. To better understand this relationship at the European level for the period 2020-2022, data on 16 representative locations in Europe (81 million people) with PM_2.5 levels (µg.m^-3) ranging from low to high values were analyzed by statistic methods. The analysis confirms a temporal relation between peaks of PM_2.5 exposure and Covid-19 mortality. The best correlation was obtained considering the history of exposition to PM2.5 pollution during a 2-month integration-time coupled with a one-week delay for the Covid-19 mortality. Although the trend of Covid-19 mortality vs. PM_2.5 levels vary among locations, the global trend was similar, giving an estimated mean value of a 40±20% mortality increase per µg.m^-3 PM_2.5 increase. The stronger the positive (negative) gradient of the PM peak, the stronger the positive (negative) gradient of the Covid-19 mortality. These results indicate that a succession of PM pollution peaks could be more dangerous than a permanent exposure to moderate pollution levels. Finally, number-concentrations of PM should be used in the future rather than the PM_2.5 mass-concentrations (µg.m-³) with the consideration of PM composition to better evaluate the .effect of submicron particles on human health, in particular for other respiratory diseases. Such results must be considered for the management of future pandemics.”

2. Use µg/m³ or µg m^-3 please uniform the units in the manuscript

Authors answer: We have corrected the mistake.

3. The last line of the abstract is cut

      Authors answer: The abstract is rewritten.

4. In the introduction mention the WHO limit for PM. Read this paper in this regard: https://www.mdpi.com/2076-3298/10/3/42

Authors answer: Indeed, this paper is interesting, but it seems that it is not in the scope of our paper.

5. Please avoid the use of “We” in the paper. Use indirect sentences like “The authors”. Lines 83 and 91 for instance.

Authors answer: The reviewer is right and sorry for that.  The authors have corrected the text accordingly.

6. Line 117-118, please introduce the reference properly in the sentence “For the purpose of this work, Covid-19 mortality data are retrieved from the John 117 Hopkins University datasets (https://coronavirus.jhu.edu/).”.

      Authors answer: The link is now included in the reference list.

7. The legend in Figure 1, 2 and 3 is not clear.

Authors answer: We have changed the legend to : “b: weekly integrated data after applying the smoothing, integration and shift procedure.” We have changed the size of the lines in the legends.

8. Pay attention at the units in the horizontal axis in figure 4. The 2.5 should be subscript

      Authors answer: The figure is corrected.

9. Better explain and summarize the objective of the manuscript at the beginning of the conclusion section.

      Authors answer: We have added: “This was conducted by applying an integration procedure of ~2 months on the PM_2.5 data and a ~1-week positive shift of the Covid-19 mortality data to better correlate the peaks and their gradients.” 

Reviewer 3 Report

Review Comments

This paper study the temporal evolution of PM_2.5 levels related to Covid-19 mortality in Europe (32 cities and districts of 6 countries) for the early-2020 to end-2022 period by statistical methods. Overall, the structure of the paper is complete and the language is fluent, but the whole paper is mainly qualitative analysis, quantitative analysis is less, and the main problems are as follows:

1. The abstract of this paper generally includes: objective, methods, results, conclusion. It is suggested that the structure of this abstract should be adjusted, not to appear separately as Aim, Methods, Results and Conclusion, but to describe together.

2. Please pay attention to the format of the paper, such as the upper and lower corners(Line 16), units(Line 19), etc. Please check the full text.

3. PM_2.5 mass concentration data were collected from the national air quality monitoring networks and the Pollutrack networks with higher resolution. Can PM_2.5 concentration data from different sources and different resolutions ensure uniformity ? Is the data source reliable ? And how to ensure the QA/QC of data, it is recommended to supplement.

4. Table 1: The abbreviation which first occurred in the paper should be give the full name.

5. Line 169-171:“This could be due to......” Where the basis of this sentence is?

6. The conclusion part of this paper is not a good summary of the important results of the paper, it is recommended to re-adjust.

Author Response

This paper study the temporal evolution of PM_2.5 levels related to Covid-19 mortality in Europe (32 cities and districts of 6 countries) for the early-2020 to end-2022 period by statistical methods. Overall, the structure of the paper is complete and the language is fluent, but the whole paper is mainly qualitative analysis, quantitative analysis is less, and the main problems are as follows

Authors answer: We want to thank reviewer for his comments that  helped us improving the paper.

1. The abstract of this paper generally includes: objective, methods, results, conclusion. It is suggested that the structure of this abstract should be adjusted, not to appear separately as Aim, Methods, Results and Conclusion, but to describe together.

            Authors answers: We have changed the abstract to:

“ Air pollution has a strong impact on human health, from respiratory and pulmonary severe diseases to heart attacks and cancers. During the 3 years of the Covid-19 pandemic, several peaks of mortality occurred, which can be related to particulate matter (PM) pollution events. The possible effect of PM (PM₁₀ and PM_2.5 of diameter lesser than 10 and 2.5 µm respectively) on Covid-19 mortality is now established. To better understand this relationship at the European level for the period 2020-2022, data on 16 representative locations in Europe (81 million people) with PM_2.5 levels (µg.m^-3) ranging from low to high values were analyzed by statistic methods. The analysis confirms a temporal relation between peaks of PM_2.5 exposure and Covid-19 mortality. The best correlation was obtained considering the history of exposition to PM2.5 pollution during a 2-month integration-time coupled with a one-week delay for the Covid-19 mortality. Although the trend of Covid-19 mortality vs. PM_2.5 levels vary among locations, the global trend was similar, giving an estimated mean value of a 40±20% mortality increase per µg.m^-3 PM_2.5 increase. The stronger the positive (negative) gradient of the PM peak, the stronger the positive (negative) gradient of the Covid-19 mortality. These results indicate that a succession of PM pollution peaks could be more dangerous than a permanent exposure to moderate pollution levels. Finally, number-concentrations of PM should be used in the future rather than the PM_2.5 mass-concentrations (µg.m-³) with the consideration of PM composition to better evaluate the .effect of submicron particles on human health, in particular for other respiratory diseases. Such results must be considered for the management of future pandemics.”

2. Please pay attention to the format of the paper, such as the upper and lower corners(Line 16), units(Line 19), etc. Please check the full text.

Authors answer: This problem was not present in our previous version; perhaps it occurred when the file was converted by the on-line submission system.

3. PM₅ mass concentration data were collected from the national air quality monitoring networks and the Pollutrack networks with higher resolution. Can PM_2.5 concentration data from different sources and different resolutions ensure uniformity ? Is the data source reliable ? And how to ensure the QA/QC of data, it is recommended to supplement.

            Authors answer: The reviewer rises a very interesting point. Only few measurement stations are available from air quality networks at a given location. When available, the Pollutrack dense network of PM sensors, duly evaluated in previous studies, can provide a more accurate mean value. It is why we have considered separately the various locations. Depending on the measurement methods, the absolute values of PM2.5 can differ  by a few µm^-3 for similar level of pollution. Then, working with the gradients of  PM2.5 peaks amounts to considering relative variations.

We have added “Such measurements heterogeneity can complexify the analysis when conducting a direct comparison of the PM_2.5 pollution levels from different locations  in Europe. Then, in the following, the relation between PM_2.5 levels and Covid-19 mortality will be considered separately for the different locations. Also, working on the relative variations of PM2.5 levels instead of absolute values could minimize the effect of such heterogeneity.” And later : “...considering relative variations rather than  absolute values.”.

4. Table 1: The abbreviation which first occurred in the paper should be give the full name.

Authors answer: We suppose that the reviewer speaks on the official abbreviation for the countries. We have changed them to the full name of the countries.

5. Line 169-171:“This could be due to......” Where the basis of this sentence is?

Authors answer: We have replaced it by : “It can be suggested that...”

6. The conclusion part of this paper is not a good summary of the important results of the paper, it is recommended to re-adjust.

Authors answer: As suggested by the editor, we have added in the discussion the table 3 that summarize the main results.

Table 3: Synthesis of the results

We have added in the conclusion: “This was conducted by applying an integration procedure of ~2 months on the PM_2.5 data and a ~1-week positive shift of the Covid-19 mortality data to better correlate the peaks and their gradients..” and “Such results on the effect of PM on respiratory pandemic mortality should be considered by political authorities, especially in regions where high pollution peaks occur frequently due to specific weather conditions and geographical constraints [42]. To better manage future pandemics, the authorities should reduce all polluting industrial, transportation and agricultural activities during the most critical days of (winter) anticyclonic conditions, to limit the population exposure to the fine particulate matter.”.

Reviewer 4 Report

I propose some modificatin of the manuscript. I have some comments:

line 16 "heterogeneous levels of PM2.5" what does it mean? This word we can use for evaluation heterogenity of materials.

line 17 "PM2.5 levels were assesd by various methods". Are they comparable? Where they compared to reference methods according to EU regulatons?

line 77 "0.29 correlation" Is this strong (important) correlation?

lines 99-105 Were optical counters calibrated according to EU regulations?

Figure 4 and 7 How many points were taken for preparing one fit (curve)?

Abstract should be written as one paragraph like for requriments for "Atmosphere"

How many sets of data do you have for one location? it was not mention in the text.

lines189-196 I a not sure if smoothing correlation parameters is the correct one?

In table 2 "before" you dont have strong correlation.

Author Response

I propose some modifications of the manuscript. I have some comments

Authors answer: We want to thank reviewer for his comments that contributed to improving the paper.

line 16 "heterogeneous levels of PM2.5" what does it mean? This word we can use for evaluation heterogeneity of materials.

Authors answer: We have changed the sentence to: “...with levels of PM_2.5 (µg.m^-3) ranging from...”

line 17 "PM2.5 levels were assessed by various methods". Are they comparable? Where they compared to reference methods according to EU regulations?

Authors answer: We have replaced  “assessed” by “analyzed”. There is no EU reference method for conducting weekly analysis of the measurements.

line 77 "0.29 correlation" Is this strong (important) correlation?

Authors answer: Indeed, it is  not a strong correlation, it just indicates that something could be present. We have changed the sentence to :’ A low but positive correlation of 0.29...”

lines 99-105 Where optical counters calibrated according to EU regulations?

Authors answer: The reviewer has raised an important point. In fact, there are no EU regulation for optical counters when considering the number concentrations. On the other hand, the measurements converted to mass-concentrations could follow some EU regulations. We have duly verified that the Pollutrack measurement rightly falls  within the “indicative” category for low-cost sensors.

Figure 4 and 7 How many points were taken for preparing one fit (curve)?

Authors answer: Since the data are weekly integrated, 52 points per year are available. Thus up to 150 points were used to produce the fit. We have added in the text : ”..obtained used between 100 and 150 points per location.”

Abstract should be written as one paragraph like for requirements for "Atmosphere"

Authors answer: The abstract is totally rewritten:

“ Air pollution has a strong impact on human health, from respiratory and pulmonary severe diseases to heart attacks and cancers. During the 3 years of the Covid-19 pandemic, several peaks of mortality occurred, which can be related to particulate matter (PM) pollution events. The possible effect of PM (PM₁₀ and PM_2.5 of diameter lesser than 10 and 2.5 µm respectively) on Covid-19 mortality is now established. To better understand this relationship at the European level for the period 2020-2022, data on 16 representative locations in Europe (81 million people) with PM_2.5 levels (µg.m^-3) ranging from low to high values were analyzed by statistic methods. The analysis confirms a temporal relation between peaks of PM_2.5 exposure and Covid-19 mortality. The best correlation was obtained considering the history of exposition to PM2.5 pollution during a 2-month integration-time coupled with a one-week delay for the Covid-19 mortality. Although the trend of Covid-19 mortality vs. PM_2.5 levels vary among locations, the global trend was similar, giving an estimated mean value of a 40±20% mortality increase per µg.m^-3 PM_2.5 increase. The stronger the positive (negative) gradient of the PM peak, the stronger the positive (negative) gradient of the Covid-19 mortality. These results indicate that a succession of PM pollution peaks could be more dangerous than a permanent exposure to moderate pollution levels. Finally, number-concentrations of PM should be used in the future rather than the PM_2.5 mass-concentrations (µg.m-³) with the consideration of PM composition to better evaluate the .effect of submicron particles on human health, in particular for other respiratory diseases. Such results must be considered for the management of future pandemics.”

How many sets of data do you have for one location? it was not mention in the text.

Authors answer: At each location, we have one data set for mortality per day. We have changed the text to : “For each location, Covid-19 mortality daily data are retrieved from the John Hopkins University website [37].” The situation is more complex for the PM2.5 data, as already explained in the text. For air quality network, there is few stations per location. For Pollutrack network, there are one hundred or more of measuring instruments per location. The PM data are then averaged to produce daily values.

lines189-196 I am not sure if smoothing correlation parameters is the correct one?

Authors answer: We don’t understand the reviewer comment. We never speak of this parameter.

In table 2 "before" you don’t’ have strong correlation.

Authors answer: We have changed the title of the column to Correlation without applying the adjustments“ and “Correlation when applying the adjustments”.  We agree that the correlations before applying the new procedure, although in agreement with those obtained previously by other authors, are not very strong. On the opposite, the correlations obtained with the new methods are higher.

Round 2

Reviewer 1 Report

Dear Authors,

you can find information about linear regression assumption in any statistics textbook. These are:

linear relationship,

multivariate normality,

no or little multicollinearity,

no auto-correlation.

The fact that other studies have omitted to check the assumptions of the model does not entitle subsequent researchers to make the same omission.

Homoscedasticity.

Author Response

Authors’ answer: We understand and we agree with the reviewer comments (there are so many papers on which such questions are not considered). Nevertheless, our text was not clear,  hence the difficulty to fully catch our  analytic scheme. The linear fit is not calculated from the individual (scattered) points, but from the mean data (5 or 6 points). Thus, all these statistical considerations do not apply to such low number of points, although they follow a linear trend. We have changed the text to: “The relationship between Covid-19 mortality and PM_2.5 exposure seems to follow a linear relationship, although individual measurements were scattered from one location to another. This could be due to the possible heterogeneity of the population density (for example variations between large towns and rural zones), to the health status of the populations that can influence the mortality, and to the local management of the pandemic [26]. Then, at each location, Covid-19 mortality per million inhabitants is averaged to produce 5 or 6 integrated values for PM2.5 mass-concentrations intervals of 5 to 10 µg.m^-3. The error bars are calculated by considering the standard error of the mean. Since the evolution of the mortality values per µg.m^-3 follows a linear trend at each location, at least at first order, a linear fit is applied to these data.”

Reviewer 3 Report

I suggest this paper accept in present form.

Author Response

The reviewer has no asked for new revision.

Reviewer 4 Report

I suggest to accept  the manuscript to publication

Author Response

The reviewer has not asked for a new revision.