Decoding Urban Traffic Pollution: Insights on Trends, Patterns, and Meteorological Influences for Policy Action in Bucharest, Romania

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Focusing on traffic pollution in Bucharest, Romania, the authors used hourly NO₂ concentration data from January 2017 to September 2022 and meteorological data from the same period to analyse NO₂ pollution trends, spatial-temporal patterns, and meteorological influences in the city centre's busy traffic area, using generalised additive modelling (GAM) and generalised additive mixed modelling (GAMM).

1. Hourly data for 2017-2022 are used in the text, but the data pre-processing process, such as the method of filling missing values and the criteria for rejecting outliers, is not described in detail. In particular, a negative value of - 9.04 μg/m³ appears in the NO₂ data?
2. Model diagnostic details such as residual distribution plots and autocorrelation function plots are not provided, and the reasons why residual autocorrelation problems are not found in some models are not explained.
3. The interaction between wind speed and atmospheric pressure on NO₂ concentration lacks the explanation of the physical mechanism combined with meteorological theories; the inverted U-shape relationship between relative humidity and NO₂ is not explored, and its influence on the chemical transformation of NO₂ is not explored.
4. The differences between Bucharest as a high-density city and the rest of Romania are not taken into account, and there is a lack of differentiated policy recommendations for different regions.
5.In the introduction, it is evident that the authors do not have sufficient information on existing studies. It is suggested that the authors should explore more studies by others. For instance, refer to ‘VAR - tree model based spatio - temporal characterization and prediction of O3 concentration in China’ and ‘Multi - objective optimal dispatch strategy for power systems with Spatio - temporal distribution of air pollutants’.
6. The boundary between ‘traffic-related pollution’ and ‘urban air quality’ needs to be clarified.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report

Comments and Suggestions for Authors

This study explores traffic-related air pollution in Bucharest, Romania, which is an under-investigated European city. The authors attempted to use the Generalized Additive Modeling (GAM) framework to explore the trends, patterns, and meteorological drivers of NO2 concentrations. The topic itself is very interesting and important, and some discussions that connect research findings with local policies development are also involved. However, there are some technical shortcoming, and the clarity of some parts should be explicitly explained.

(1) The most significant technical issue lies in the handling of the GAMM results. The authors present that GAMM can be used to account for potential residual autocorrelation (Lines 483-486), which is an excellent practice. However, the results presented in Table 6 show a dramatic reduction of adjusted R-squared from 86.1% (in multifactor GAM) to 51.2% (in GAMM). The authors couldn't provide an explanations on such a significant drop. Also, the effects of several key variables change from non-linear to linear, and wind direction loses its significance. This indirectly suggests that suggests that the underlying model structures of both GAM and GAMM are quite different, thus it's a bit unfair to conclude that one model is better than the other due to the high R2 value, because one cannot compare apple with orange. Therefore, the authors must (a) provide diagnostic plots for final multifactor-GAM, so that one can visualize whether residual autocorrelation is significant in necessitating GAMM model; (b) the authors should also provide a methodological reasoning and discussion on why the results of GAM and GAMM differ that much. Also under what scenario or settings, one should adopt GAM? or one should adopt GAMM? The detailed mathematical formulation of both models should also be clearly explained too.

(2) Section 2.2 is too short, only some formula are shown, which is not good from technical perspective. The detailed steps and formulations of both GAM and GAMM must be included. It is suggested to present it in a tabular form, from the beginning of model setting (e.g., input datasets required) to methods / algorithms used (in details), then to output and statistical assessment processes.

(3) The manuscript has omitted several key details regarding the initial data processing, which are crucial for reproducibility. Also, are the datasets obtained for model building and construction undergoing QA and QC processes? Please provide justification to verify the accuracy of initial data adopted in this study, or showcase the data preprocessing / cleaning stages applied during the process.

(4) Potentially missing datasets (especially meteorological datasets) - The time series data for air pollution and meteorology are unlikely to be 100% complete over the 6-year period. The manuscript does not state how missing hourly observations were handled. Were they ignored, imputed via linear interpolation (or other numerical methods), or was an average has actually been taken? Therefore, please clarify the final date range used, and explicitly describe the protocol for handling any missing values of meteorological or air pollution attributes.

(5) The choice of the basis dimension (k) for the smooth terms is crucial for GAM model from mathematical point of view, so that no overfitting will occur. The manuscript does not mention how k was selected, or whether its adequacy was carefully and properly checked. Please add in further details.

(6) The process of selecting the most optimal model (in Line 453) is important, but the authors didn't explain such process in clear manner. Was it based on minimizing AIC, maximizing R2 of other statistical derivation? Please clarify.

(7) Introduction: The introduction is very detailed and clearer, but could be slightly condensed to highlight the aims of this study. The first two paragraphs of Introduction provide excellent global context, but could be shortened. Also, the health effects of PM2.5 and NO2 (or NOx) pollution caused by traffic emission should be referenced. Please refer to the following reference and include it into the revised manuscript: 

PM2.5 and NOx: Table 1 of https://www.mdpi.com/1660-4601/18/12/6532

Ozone: https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2019.02518/full

(8) The motivation of adopting Generalised Additive Modelling (GAM) framework to estimate nitrogen dioxide (NO2) concentrations at a major urban traffic location is not clearly explained. Some details have to be added into both Methodology section and Introduction section. In particular, what similar work has been conducted by GAM? What's the strength of GAM?

(9) The manuscript used the Variance Inflation Factor (VIF) to check for multicollinearity and based on the statistical check, two variables were removed. It would be much better if the specific threshold used for removing two variables can be added into the Methodology section.

(10) The authors could more explicitly link the meteorological findings to policy development and implementation, and provide some commendable suggestions with regard to policy formulation and laying down of strict laws in Romania etc. Some experience from other European countries should be discussed too.

(11) The core of GAM model is the smooth function, please define the smooth term clearly.

(12)  The manuscript mentions that GAM includes a penalty parameter to control the smoothness of the model. Please state the optimization problem (with constraints) explicitly.

(13) The limitations, parametric and structural uncertainty of the estimated coefficients and the shape of the smooth functions, given the dataset of traffic emission and meteorology should be discussed in details. 

(14) It would be interesting if the authors can connect current study with the use of remotely sensed instruments and satellite datasets in retrieving meteorological quantities, air quality attributes etc. Please try to cite some related articles that utilized satellite remote sensing techniques in air quality detection / prediction / forecasting etc. Please also focus more on how current study can be extended into large-scale retrieval, for example, what are the datasets needed etc.?

Minor issue: All "2" of NO_2 should be underscripted.

Minor issue: The abstract states the dataset spans "January 2017–September 2022" (line 27), while the Methods section states "1.01.2017 to 16.02.2022" (line 235). Check carefully.

Major Revision is needed at this stage, and it is suggested that the authors should address all aforementioned comments.

 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The topic presented in the paper is valid and important. In the reviewed paper, the Authors presented the problem of decoding urban traffic pollution as iInsights on trends, patterns, and meteorological influences for policy action in Bucharest in Romania. The Authors in this paper, a generalized additive modeling (GAM) framework is employed to model hourly concentrations of nitrogen dioxide (NO2), i.e., a relevant traffic-pollution proxy, at a busy urban traffic location in central Bucharest, Romania. All models are developed on a wide, fine-granularity dataset spanning January 2017–September 2022 and include meteorological covariates. Model robustness is assured by switching between the generalized additive model (GAM) framework and the generalized additive mixed model (GAMM) framework when the residual autoregressive process needs to be specifically acknowledged. Results indicate that trend GAMs explain a large amount of the hourly variation in traffic pollution. Furthermore, meteorological factors contribute to increasing the models’ explanation power, with wind direction, relative humidity, and the interaction between wind speed and the atmospheric pressure emerging as important mitigators for NO2 concentrations in Bucharest. In my opinion, the paper can be considered for publication, after taking into account the following remarks:
- before paper publishing, English should be carefully checked. As far now, some typos and grammar mistakes can be found,
- in the Keywords section, there is too much keywords. The best practise is, when the number of keywords is up to five items,

• at the end of the Introduction section, the Authors wrote about the main aim of the paper, presenting at the same time the research questions. It is very well, but also a short information what was included in each paper section should be added, at the end of the Introduction section,
• there is a lack of typical "Literature review" section usually located after Introduction section. Some literature review is located in the Introduction section, but despite the lack of separate section, a some "state-of-the-art" from the presented literature review should be added at the end of this literature review,
• in the section called "2. Materials and Methods ", the step-by-step chart should be added with presentation a step-by-step procedure/methods of analysis used in this study,
• concernig the figure called " Figure 3. Hourly, diel, weekly, and monthly pattern plots of Bucharest NO2 concentration. Source: Authors’ representation in R software. ", they are too small, and becasue of it are unreadible,

• there is a lack of explanation in the paper text for some of the used acronyms. The Authors are kindly asking to check, if all given acronyms have given their full meanings,

• the "GAM estimation results " presented in the subsection 3.2. is described in general way. Some more details should be provided in order to explain the readers the estimation results,
• concernign the figure called "Figure 7. Smooth functions of trend, month, and wind speed with the estimated 95% confidence intervals in the adjusted NO2 GAM model. Source: Authors’ representation in R software. " there is a lack of name of axis "y". It should be added. The same remark is dedicated to another similar figures, where are axises "x", and "y", and where is a lack of name/units on these axises,
• concerning the figure called "Figure 8. Bivariate smooth function of wind speed and atmospheric pressure components in the interaction effects NO2 GM model. Source: Authors’ representation in R software." there is a lack of legend, so, the readers don;t know what does it means the used colors,

• in the Conclusions section, some more detailed conclusions dedicated to obtained research results should be included.

  Thank you very much.

   

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

It is recommended to accept directly.

Author Response

Thank you for all your feedback and recommendations that have helped us improve the manuscript significantly.

Reviewer 2 Report

Comments and Suggestions for Authors

The authors have tried to address almost all my previous comments, except some more minor updates needed as follows:

(1) A new paragraph of Section 2 was added on the advancement of satellite remote sensing in air quality monitoring and prediction with regard to NO2 retrieval. Apart from TROPOMI, many other research products were established to retrieve informatics of NO2. Please acknowledge their contribution as well. See below:

https://amt.copernicus.org/articles/14/455/2021/

https://www.mdpi.com/2072-4292/10/11/1789

https://egusphere.copernicus.org/preprints/2024/egusphere-2024-617/

(2) For previous Comment (7), the authors commented that "We found the recommended reference relevant for our paper and cited it." We notice that the authors have made some edits in the manuscript, however, the two provided references have not been cited in the latest version of such manuscript actually.
PM2.5 and NOx: Table 1 of https://www.mdpi.com/1660-4601/18/12/6532
Ozone: https://www.frontiersin.org/journals/immunology/articles/10.3389/fimmu.2019.02518/full

(3) The authors repeatedly emphasize that introduction of the GAM model is necessary for NO2 retrieval, which is scientific according to the results shown, and the authors highlight some advantages of GAM model - for example, offering significant flexibility in capturing nonlinear relationships, the estimation of smooth functions is inherently subject to both statistical and structural uncertainties. However, the authors still not fully emphasize the effectiveness of the GAM model as compared with other existing machine learning based models etc. It would be better to establish a clear explanation of why GAM model performs much better than other model types in NO2 retrieval or monitoring. 

(4) In Table 4, entries in "Interaction" are presented in the form of "te(ws, wd)", please make it more layman, similar as in "Terms" in Table 5 and Table 6.

(5) Figure 10 was not shown clearly - one sub-plot was accidentally removed from the manuscript.

(6) Section 4.1 presents information from IQAir website, but are there any more official datasets / supporting documents, say from local government?

(7) For Section 4.2, it would be better if the authors can compare GAM model with the use of other models in this context. The account of meteorological variables and effects in Lines 975-890 are great!

Minor Updates:

(1) The "3" in O3 should be underscripted, for example, in Line 262 and 264.

(2) Line 358 has some problem with regard to "beta".

(3) Equations (3) and (4) have some formatting problems. For Equation (3), what is the lower limit and upper limit of integration? For Equation (4), the minimization problem should use min (short form), also the constraints of such constrained minimization problem are missing.

After the authors address aforementioned comments, I think the manuscript is good.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 3

Reviewer 2 Report

Comments and Suggestions for Authors

The quality of the manuscript is much enhanced. I recommend the publication of this manuscript.