Correlating Traffic Data, Spectral Noise and Air Pollution Measurements: Retrospective Analysis of Simultaneous Measurements near a Highway in The Netherlands

Round 1

Reviewer 1 Report

In the simultaneous measurements, the influencing factors that are not evaluated should be specified. For Black Carbon or NOx, Diesel Emissions measurement, the tires life and road tire interaction, road conditions, climatic conditions, size of vehicles, nature of land use, and built up characteristics do influence and therefore generic model development must specify the limitations. These models are definitely not to be used as a predictive model. 

Figure 2 should be checked 

Author Response

Reviewer 1:

In the simultaneous measurements, the influencing factors that are not evaluated should be specified. For Black Carbon or NOx, Diesel Emissions measurement, the tires life and road tire interaction, road conditions, climatic conditions, size of vehicles, nature of land use, and built up characteristics do influence and therefore generic model development must specify the limitations.

Response: The list of potential influencing factors is long but this is out of the scope of this manuscript. The focus is on the short-term variability of traffic volumes and traffic dynamics. Changes in fleets, impact of regulations, tires, road types are long-term temporal trends and/or spatial variables and thus out of the scope of the collected data. This response is include in section 4.1.

These models are definitely not to be used as a predictive model. 

Response: Applying these type of models for prediction at fixed locations is out of the scope of this publication. This requires measurement data on multiple locations. The predictive power and the potential to quantify traffic related interventions using this approach is already established in “Dekoninck, Luc, and Jelle Hofman. "Multi-disciplinary sensing for personal exposure assessments: Quantifying the impact of traffic interventions and meteorological variability." In e-Forum Acusticum 2020, pp. 3085-3090. 2020.”

Figure 2 should be checked 

Response: This figure is a visual representation of the spatio-temporal personal exposure  to black carbon  while biking, used in most of my communication to the public in many occasion. We acknowledge that this representation requires some additional information. We adjusted the figure to make it more self-explanatory.

 

We thank you for this positive review.

Reviewer 2 Report

The authors present an interesting study on the correlation of air and noise pollution from measurements carried out next to a highway. For this purpose, GAM regression models were applied.

The manuscript is well structured in sections and shows figures and tables that are helpful to the reader's understanding.

A series of comments are suggested that could help the authors to improve the quality of the submitted document. I recommend a major revision.

Introduction:

• I respectfully think that the introduction should include a more extensive literature review related to the measurement procedure for air pollution, noise and meteorological conditions. Perhaps it should be indicated whether any national or international standards were followed for the measurement procedure. Extensive scientific literature is available on these aspects.
• Lines 41-44: I miss some citations and brief discussion to support these sentences.

 

Materials and Methods

• The A2 highway seems to be undergrounded in the studied section and the study site is not visible as of 2011. Please, it would be interesting if the authors could include some pictures of the site or a diagram including the position of the different sensors for air pollution, noise and atmospheric conditions.
• I think it would be necessary to detail the specific situation of the sensors for measuring air pollution, noise, microphone and meteorological conditions. Although some data on distance to the source and to buildings are included, nothing is indicated regarding height. This is a factor that may be influential on the results according to the scientific literature. In the case of noise level measured in frequency bands, this could be sensitive due to sound reflections from nearby surfaces such as ground and walls. It would be interesting to know if any standard (ISO or similar) or literature reference was followed in this respect.
• Scientific literature shows that urban obstacles between the sound source and the microphone (such as bushes, parked vehicles,...) can generate acoustic shielding effects in some frequency bands related to the indicators used in this study (OLF, OHF and HFmLF). Please include a schematic diagram of the environment under study in 2011 and comment on the possible implications of these aspects on the measurements made.

Results

• Figure 4. In order to facilitate the understanding of potential readers, I believe that some aspects of figure 4 should be explained in a little more detail, such as: explanatory legend of solid and dashed lines, units of the magnitudes shown on the axes.
• Lines 216-219: Lines 216-219: The speed limit for vehicles on the studied section is 50 km/h, so the results obtained in this paragraph for noise OLF make sense. It would be interesting if the authors could comment on whether these results could be different at a higher speed limit.
• Table 1 shows several results. I miss a broader and more detailed discussion of the results shown.
• Line 227: I miss some citations to support this statement
• Table 2: To facilitate the understanding of the results shown, I believe it is necessary to clarify in more detail in the text the meaning of the information shown in the "Traffic" column.
• Table 3: To facilitate the understanding of the results shown, I believe it is necessary to clarify in more detail in the text the meaning of the information shown in the "Noise" column.

Discussion

• In this section, I miss a discussion with results found in other publications.

References

• Please include the DOI or web link of the references to facilitate the tracking of the manuscript.

Author Response

Reviewer 2:

The authors present an interesting study on the correlation of air and noise pollution from measurements carried out next to a highway. For this purpose, GAM regression models were applied.

The manuscript is well structured in sections and shows figures and tables that are helpful to the reader's understanding.

A series of comments are suggested that could help the authors to improve the quality of the submitted document. I recommend a major revision.

Introduction:

• I respectfully think that the introduction should include a more extensive literature review related to the measurement procedure for air pollution, noise and meteorological conditions. Perhaps it should be indicated whether any national or international standards were followed for the measurement procedure. Extensive scientific literature is available on these aspects.

Response: The measurements are performed by a regional institute for environmental monitoring with a focus on comparing various instruments. The measurements were performed within an Interreg project PMLAB project with in theregion “Euregio Meuse-Rhine” (see acknowledgements in the original publication). More information can be found in de “PM-LAB - Towards a particulate matter information system for the Euregion Meuse-Rhine Technical report 2013”,  link: https://www.researchgate.net/project/PMLAB-Towards-a-particulate-matter-information-system-for-the-Euregion-Meuse-Rhine. We can rely on the quality of the existing dataset without further comments in this manuscript.

 Lines 41-44: I miss some citations and brief discussion to support these sentences.

Response: Added references for noise (EU-cnossos methods) -and for air (Copert method)

 Materials and Methods

• The A2 highway seems to be undergrounded in the studied section and the study site is not visible as of 2011. Please, it would be interesting if the authors could include some pictures of the site or a diagram including the position of the different sensors for air pollution, noise and atmospheric conditions.

Response: The local spatial features of the original paper are not relevant for the aims of this publications: showing the potential of extracting knowledge from noise measurements to enhance the understanding of the temporal behavior of the air pollutants. Any dataset with similar (spectral) noise assessments can be evaluated in this way. The readers are pointed to the matching publication instead of inserting the figures of the original publication.  

Response: The A2 highway is redesigned shortly after this measurement campaign. The current situation is therefore not relevant any more. This is added as a comment.

• I think it would be necessary to detail the specific situation of the sensors for measuring air pollution, noise, microphone and meteorological conditions. Although some data on distance to the source and to buildings are included, nothing is indicated regarding height. This is a factor that may be influential on the results according to the scientific literature. In the case of noise level measured in frequency bands, this could be sensitive due to sound reflections from nearby surfaces such as ground and walls. It would be interesting to know if any standard (ISO or similar) or literature reference was followed in this respect.

Response: The visual situation is extensively described in the original paper on the data collected in 2011. The authors try to avoid repeating this information but added some additional information to accommodate this comment in the manuscript (no buildings nearby, height of the microphone).

Response: The heigt of the microphone was added to the information.

• Scientific literature shows that urban obstacles between the sound source and the microphone (such as bushes, parked vehicles,...) can generate acoustic shielding effects in some frequency bands related to the indicators used in this study (OLF, OHF and HFmLF). Please include a schematic diagram of the environment under study in 2011 and comment on the possible implications of these aspects on the measurements made.

Response: There are no buildings, parking spaces etc. in the close vicinity of the measurement locations. The measurement location is a well-chosen fixed measurement location of the regional government which by design excluded these potential disturbing features. This suggestion is highly relevant in a next phase of this research trajectory when multiple simultaneous measurement locations would be included in multi-site pooled models. This publication should trigger these extended data collections.  

Results

• Figure 4. In order to facilitate the understanding of potential readers, I believe that some aspects of figure 4 should be explained in a little more detail, such as: explanatory legend of solid and dashed lines, units of the magnitudes shown on the axes.

Response: This information is added in the (extended) description of the GAM models. This avoids repeating this information and enlarging the footprint of the figures throughout the document (section 2.5).

• Lines 216-219: Lines 216-219: The speed limit for vehicles on the studied section is 50 km/h, so the results obtained in this paragraph for noise OLF make sense. It would be interesting if the authors could comment on whether these results could be different at a higher speed limit.

Response: The original methodology is based on mobile measurements collected along a wide variety of speeds. Using this technique on fixed monitoring locations intrinsically limits the application to the local traffic regulation and traffic conditions. The speed related noise indicator HFmLF does detect the traffic jams, despite the relatively low speed limit on the local urban highway. The comment is nevertheless very relevant and adds value for the reader and is therefore included in the text.

• Table 1 shows several results. I miss a broader and more detailed discussion of the results shown.

Response: The authors added a summary of this section with the main conclusion: the engine noise is a better predictor for the traffic volume (combining both cars and trucks in a single measure) than LAeq.

• Line 227: I miss some citations to support this statement

Response: This is very basic knowledge in the noise discipline. Elaborating on this in this context would reduce the readability of the manuscript. Nevertheless, some of these variables are mentioned and the need to verify the impact of the variables in future work is included in the text.

• Table 2: To facilitate the understanding of the results shown, I believe it is necessary to clarify in more detail in the text the meaning of the information shown in the "Traffic" column.

Response: The title ‘Traffic’, does not relate to the column but the table as a whole, presenting results using the traffic information as covariates. It is a visual enhancement of the table description below.

• Table 3: To facilitate the understanding of the results shown, I believe it is necessary to clarify in more detail in the text the meaning of the information shown in the "Noise" column.

Response: Same clarification for ‘Noise’. Indeed, as explained above. The authors think this approach increases readability of the manuscript.

Discussion

• In this section, I miss a discussion with results found in other publications.

The authors discussed the lack of literature on correlating noise and air pollution including the spectral content of noise in the introduction. The few specific third party publications on this topic refer to the original work (Dekoninck et al., 2013). The additional summary in section 3.1.2 – based on a comment above – covers this remark largely: the low frequency noise is a better predictor of the traffic volume and this feature is the underlying reason why these modeling approach is stronger than the results from other publications. 

References

• Please include the DOI or web link of the references to facilitate the tracking of the manuscript.

The authors want to thank you for this interesting review. It addresses relevant features of future developments in this highly interdisciplinary work.

Reviewer 3 Report

I only have minor edits as the paper was very well structured and designed to help other people decide on their methodology for analysis and choice of statistics and graphical re´presentation.

I keep coming back to "deviance explained" which to a native english speaker sounds very strange. I tried to chnage it to the explained deviance but because it is used as the main parameter, I dont feel qualified to make them change it as it may be something that is used in noise studies and GAM mofdels that I was unaware of. I would like the editors to comment or the other reviewer to give their opinion. It reads badly to me but it might be the method used in other papers..

 

 

 

Comments for author File: Comments.pdf

Author Response

Reviewer 3:

I only have minor edits as the paper was very well structured and designed to help other people decide on their methodology for analysis and choice of statistics and graphical representation.

I keep coming back to "deviance explained" which to a native English speaker sounds very strange. I tried to change it to the explained deviance but because it is used as the main parameter, I don’t feel qualified to make them change it as it may be something that is used in noise studies and GAM models that I was unaware of. I would like the editors to comment or the other reviewer to give their opinion. It reads badly to me but it might be the method used in other papers..

Response: ‘Deviance explained’ is the official parameter describing the outcome quality of a GAM model. This phrasing is not related to the application of GAM in the noise context. Therefore it is natural to use it in this way, despite it doesn’t feel natural English.

The authors want to thank you for this positive review.

Author Response File: Author Response.docx

Reviewer 4 Report

Abstract

 

The abstract need to be improve, a great part of abstract describes previous work or state of art. There are not clear about methodology, statistical process or main results. another aspect is the abbreviation used.

 

Lines 9-10 are not relevant information and the readers will be expected to find some references about it in the introduction section, but only they list some references.

Lines 10-11 are not clear.

Lines 11-14 What mean UFP? What is the relevant aspect about this study with the article that proposed the authors?

Lines 19-22 the authors proposed that the previous methodology explain the results, but this methodology is not described.

 

Introduction

 

In general, In the first paragraph the authors list a few references related to the topic of their proposal. The rest of introduction describes their previous works, in fact, made asseveration about the quality and advisability of the used their methodology. This is not a good practice. In another hand, the aim describe in the introduction does not congruent  with the title and the showed on the abstract.

 

Methodology

 

All this section needs to improve also. The details about study area are necessary (general map and scheme to location of measurement instrument), potential sources of emission of pollutants and noise. About samples it is necessary to describe the time resolution of each value and the methodology to obtain data introduced to the models. In the case of noise the lack of details does not allow to infer the quality and the representativeness of levels reports. Likewise, the way to describe the way to infer the OLF and HFmLF are opposite with the showed on the introduction. The few details about how the author used GAMs, the reason to select and the way to show the future results does not allow to understand and verify their findings. Other questions are: from where they obtained the meteorological data? Which year is the traffic data? The author knows about models, kind of fuels and others characteristics about vehicles that affects the emission of pollutants (combustion gas/particles) and noise.

 

Finally, the authors need to test for multicollinearity between the intrinsically related variables. E.g. temperature and relative humidity, OLF and HFmLF. This could be resulted on over fitting on the models or a false correlation.

 

They need to pay attention to details e.g. presents and describe result in the methodology section among others.

 

Results

 

This section is very difficult to understand, first all the authors describe main three moments, but when you read the first paragraph (lines 183-191), you can find five. The way to presents the results are very confused, disorganized and the figures and tables do not reflect the information described by the authors. E.g. fig 4 what values represents y axis, why sometime used the expect values (wind direction) and sometime used increased quantity (car and trucks) in the x axis? Which is/are the independent, dependent variables and covariates. It is confusing because the meteorological parameters are not used as covariates, or are they? Why use meteorology in the correlation model, this information is presumed but it is necessary that the authors justify it in the introduction and the methodology.

 

The results describe on 3.1.1. about the flow and quantity of vehicles can be expected and is known from multiple simpler correlation studies. What are the relevant findings of these models? Please, take account that the tittle proposed a correlation analysis.

 

The section 3.1.2 will be 3.1.1.1, or not, see lines 194-196. Please explain lines 232-233 about temperature and physical proprieties of the vehicle noise emission. The information of lines 231-243 sounds imprecise and can be attributed to another physical aspect as the chain on the tire when is a cold season.

 

For sections 3.2. and 3.3, the observations are similar to those already presented.

 

In all cases the authors must be careful to explain the results of the "model". They have not stated how the values entered into the model are not affected by background values or other sources.

 

Another important aspect on 3.3 section is the large range of deviance values, what is mean? This is the main part of the article that supposed to respond to the aim. But the authors do not describe in necessary detail.

 

Discussion

 

The limitations of section 4.1 present arguments that largely contradict what was stated in the introduction and reveal the shortcomings in the controls of the data used in the models, mainly due to the sampling methodology and the absence and lack of identification of the values, background levels or other potential sources. Likewise, sections 4.2 and 4.3 show that the "models" do not allow inferring the physical and/or chemical processes that may take place, every time the pollutants are emitted. Finally the unique reference to support their discussion and results is a self-citation.

Author Response

Reviewer 4:

Abstract

 The abstract need to be improve, a great part of abstract describes previous work or state of art. There are not clear about methodology, statistical process or main results. another aspect is the abbreviation used.
 Lines 9-10 are not relevant information and the readers will be expected to find some references about it in the introduction section, but only they list some references.
Lines 10-11 are not clear.
Lines 11-14 What mean UFP? What is the relevant aspect about this study with the article that proposed the authors?
Lines 19-22 the authors proposed that the previous methodology explain the results, but this methodology is not described.

R: The remarks are very relevant. The abstract is reworked to reduce the prior knowledge on the topic to understand the problem addressed in this manuscript. The new version should accommodate the comments.

Introduction

 In general, In the first paragraph the authors list a few references related to the topic of their proposal. The rest of introduction describes their previous works, in fact, made asseveration about the quality and advisability of the used their methodology. This is not a good practice.

R: This topic is highly interdisciplinary. The intended audience are both air pollution specialists and noise specialist, but the focus is air pollution specialists. Therefore, a thorough description of the noise components, presented in simple visualizations is very relevant to achieve the goals of this manuscript. The term ‘asseveration’ doesn’t comply with the repeated validations of the presented methodology. As mentioned, this is a highly interdisciplinary method and this requires a short but insightful visualization of the approach. By using this approach the threshold for in-dept reading of the manuscript and understanding of the potential of this method is harmed significantly.

In another hand, the aim describe in the introduction does not congruent  with the title and the showed on the abstract.

R: The modifications in the abstract and the inclusion of ‘Spectral Noise’ in the title should accommodate this remark.

 Methodology

All this section needs to improve also. The details about study area are necessary (general map and scheme to location of measurement instrument), potential sources of emission of pollutants and noise. About samples it is necessary to describe the time resolution of each value and the methodology to obtain data introduced to the models.

R: A similar comment was made by another reviewer. The local spatial features of the original paper are not relevant for the aims of this publications: showing the potential of extracting knowledge from noise measurements to enhance the understanding of the temporal behavior of the air pollutants is the main aim. Any dataset with similar (spectral) noise assessments can be evaluated in this way. The readers are pointed to the matching publication instead of inserting the figures of the original publication.  

In the case of noise the lack of details does not allow to infer the quality and the representativeness of levels reports. Likewise, the way to describe the way to infer the OLF and HFmLF are opposite with the showed on the introduction.

The data collection of both noise and air pollution was performed by professional in an official institute with knowledge of both noise and air pollution (co-author Marcel Severijenen). All measurements are conform the best practice in both domains. The measurements are performed by a regional institute for environmental monitoring with a focus on comparing various instruments. The measurements were performed within an Interreg project PMLAB project with in theregion “Euregio Meuse-Rhine” (see acknowledgements in the original publication). More information can be found in de “PM-LAB - Towards a particulate matter information system for the Euregion Meuse-Rhine Technical report 2013”,  link: https://www.researchgate.net/project/PMLAB-Towards-a-particulate-matter-information-system-for-the-Euregion-Meuse-Rhine.

We can rely on the quality of the existing dataset without further comments in this manuscript.

Likewise, the way to describe the way to infer the OLF and HFmLF are opposite with the showed on the introduction.

R: This is a very relevant remark. Too much emphasis was put on the traffic dynamics in the abstract. The traffic dynamics is the strongest modifier, the engine related noise is the stronger traffic volume covariate. This should be resolved in the new version of the abstract.

The few details about how the author used GAMs, the reason to select and the way to show the future results does not allow to understand and verify their findings.

R: A similar remark was made by another reviewer. The description of the GAM technique and how to interpret both the outcomes and the splines is added to section 2.5.

Other questions are: from where they obtained the meteorological data?

R: The information on the meteorological data was missing and is now added to the text.

Which year is the traffic data?

R: The traffic data is (instantaneous) hourly data, matching the aggregation level of both the noise and air pollution measurements.

The author knows about models, kind of fuels and others characteristics about vehicles that affects the emission of pollutants (combustion gas/particles) and noise.

R: The approach evaluates the average fleet at that specific moment in time in general terms, without specification or other evaluations. Changes in fleet composition and other slowly varying variables can only be included in these types of analysis when longer data series are collected, thus after adapting the principles promoted in this manuscript.

Finally, the authors need to test for multi-collinearity between the intrinsically related variables. E.g. temperature and relative humidity, OLF and HFmLF. This could be resulted on over fitting on the models or a false correlation.

R: This is a very relevant remark, the authors are very aware of this risk. The real value of adding ‘noise as a proxy’ in air pollution data series is to reduce this overfitting problem within the air pollution discipline models themselves. The issue of overfitting was explicitly addressed in the ”Dekoninck, Luc, Qiang Yang, Haokai Zhao, James Ross, Darby Jack, and Steven Chillrud. "An international application of the city-wide mobile noise mapping methodology: Retro-active traffic attribution on a bicycle commuters health study in New York City." In INTER-NOISE and NOISE-CON Congress and Conference Proceedings, vol. 259, no. 6, pp. 3265-3276. Institute of Noise Control Engineering, 2019.”. In that publications, the combination of background concentrations, temperature, wind and humidity fully over fit the model.

Within this manuscript, the models including the meteorology are only added to illustrate that despite the strong influences and complex interactions between the meteorological variables and the impact of meteorology on the dispersion and chemical processes, the ‘noise as a proxy’ stays a dominant variable for the combustion related air pollutants. Technically, this remark illustrates why combining noise and air pollution measurements should be adopted at large.

They need to pay attention to details e.g. presents and describe result in the methodology section among others.

R: The authors are aware of the built-up of intermediate results throughout the manuscript. This approach is also the result of the complexity of interdisciplinary work of which several aspects are new to both disciplines. The structure is explicitly aiming on granularly building confidence in the different aspects. Another reviewer explicitly mentioned this structure as ‘helpful for the reader’s understanding’.  We don’t see how we can present this in a better way. 

Results

 This section is very difficult to understand, first all the authors describe main three moments, but when you read the first paragraph (lines 183-191), you can find five.

R: The description does math the structure but section 2 and 3 are subdivided in base and meteo-extended models. In short:

3.1: Relation traffic and noise parameters
3.2: Traffic based models (without and with temperature and humidity)  
3.3: Noise based models (without and with temperature and humidity)  
3.4: Comparison
3.5: What’s happening with UFP?

The way to presents the results are very confused, disorganized and the figures and tables do not reflect the information described by the authors. E.g. fig 4 what values represents y axis, why sometime used the expect values (wind direction) and sometime used increased quantity (car and trucks) in the x axis? Which is/are the independent, dependent variables and covariates.

Response: Each spline has its own x-axis, fitting the occurring values in the dataset. The y-axis is always a representation of the impact on the outcome variable. This y-axis is identically for all covariates. When downloading the reviewed document, the figures were found to be rotated, not presented as submitted. I think this was an action by the editors. The figures are very wide and use the margin to the left systematically to allow the best visualization. When viewed as intended, the figures should be self-explaining. To enhance reasdability, additional headers are included to the figures. The spline of each covariate are shown in a single column, nicely showing the (non)linearity matching the description. Steeper splines with smaller confidence interval identify be strongest models. 

It is confusing because the meteorological parameters are not used as covariates, or are they? Why use meteorology in the correlation model, this information is presumed but it is necessary that the authors justify it in the introduction and the methodology.

Response: The meteorological parameters are covariates, potentially influencing the outcome, as explained above in the structure summary. This is explicitly stated in the general description. For example, the wind direction covariate showed that the spectral content didn’t resolve the behavior of the UFP, which required an additional evaluation (section 3.5).This unexpected behavior shows that UFP is not explained by both the traffic count and the noise proxy. Other mechanism affect the outcome in that case. The results show that relatively clean air is required to explain the highest levels of UFP in the data collection.

The results describe on 3.1.1. about the flow and quantity of vehicles can be expected and is known from multiple simpler correlation studies. What are the relevant findings of these models? Please, take account that the tittle proposed a correlation analysis.

Response: The traffic based models are necessary as a reference to understand the added value of the noise based models as stated at the start of the section ‘Results’. The message is: even if traffic data is available, the noise measurements add value due to (1) the OLF as the sole parameters quantifying traffic volumes, and (2)  the inclusion of the traffic dynamics through HFmLF. Reaching this conclusion is not possible without explicitly adding the models in their variant with the traffic data only. This was the scope of the original evaluation of this dataset (reference 16).
Since air pollution measurements are not always attributed with (third-party depending) traffic data, this knowledge gap can be filled by adding noise measurement to the air pollution site. It is true that the traffic data was not included in the title. We added this in the title. This is a relevant remark. This will trigger more interest from the intended public.

The section 3.1.2 will be 3.1.1.1, or not, see lines 194-196. Please explain lines 232-233 about temperature and physical proprieties of the vehicle noise emission. The information of lines 231-243 sounds imprecise and can be attributed to another physical aspect as the chain on the tire when is a cold season. 

Response: Again, this is an aspect of the complexity of interdisciplinary work. The rubber of the tires becomes softer at higher temperatures which reduces the noise emission.

Response: Chains on tires and nails on winter tires are a very specific issue, but the scope of this manuscript. For your information, this is a very specific topic in the Nordic countries because the nails and change damage the road surface and reduce the surface quality, and thus increasing noise emission also in summer. It is difficult to elaborate in this detail to all potential aspects, especially since these aspects are nog relevant for this specific dataset (normal weather in spring, no nails or chains in winter in the Netherlands.)

For sections 3.2. and 3.3, the observations are similar to those already presented.

In all cases the authors must be careful to explain the results of the "model". They have not stated how the values entered into the model are not affected by background values or other sources.

Response: All covariates are listed in the tables which match the figures. The tilting of the figures reduced the readability significantly. We hope this makes the structure and The aim of the manuscript is to test the applicability of the method for multiple air pollutants. The applicability is mainly linked to specific features of the air pollutant. Combustion related pollutants result in strong models, for both traffic and noise, but noise adds additional value on top of the traffic data. When the models fails, the local situation is affected by other (unknown) sources, or the pollutants doesn’t correlate with the magnitude of traffic volumes: the case of UFP were the well-known phenomenon of ‘scavenging’ occurs and is made visible in section 3.5.

Another important aspect on 3.3 section is the large range of deviance values, what is mean? This is the main part of the article that supposed to respond to the aim. But the authors do not describe in necessary detail.

Response: As explained above. This should also be clearer with the additional information in section 2.5 (GAMs). The deviance explained combines two features, the higher the value, the stronger the covariates predict the outcome. 

Discussion

The limitations of section 4.1 present arguments that largely contradict what was stated in the introduction and reveal the shortcomings in the controls of the data used in the models, mainly due to the sampling methodology and the absence and lack of identification of the values, background levels or other potential sources. Likewise, sections 4.2 and 4.3 show that the "models" do not allow inferring the physical and/or chemical processes that may take place, every time the pollutants are emitted. Finally the unique reference to support their discussion and results is a self-citation.

Response: The correction for background concentrations for the air pollutants - as used in the Dekoninck et al., 2013 - couldn’t be achieved for the majority of the air pollutants and is therefore omitted. This decision is supported by the short distance to the major local source for air pollution (the urban highway). The dataset is also limited to six weeks in spring. The background correction is largely implemented to address the seasonal variability (e.g. contributions of household heating systems and large-scale long distance air pollution transport). This also reduces the need and applicability of the background correction on this specific dataset. The self-citation on the methodology and the data collection are independent of each other. The lack of external references is the result of low availability of simultaneous assessments of both spectral noise and air pollution.

We want to thank the reviewer for this review. We were able to improve quality and readability based on this comments.