Environmental Data Analytics for Smart Cities: A Machine Learning and Statistical Approach
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsGeneral Comments:
Al-Salehy and Bailey evaluated the dynamics of carbon monoxide in Jubail, Saudi Arabia, with sensors and machine learning models from 2018 to 2022. Also, the authors compared the predictive accuracy of multiple machine learning models for CO concentrations with meteorological data. The manuscript is recommended for publication in Smart Cities after addressing the following comments.
- Introduction: It would be helpful to add 1-2 sentences explaining why this study only focuses on CO rather than other gaseous air pollutants.
- Section 3.2: Give more details about the sources of the gas pollutant and meteorological data. If these data were measured using your own sensors, please provide information on the calibration and validation processes.
- Section 4.3.1: For meteorological impact, only the wind speed is mentioned in the main text, but Table 1 includes only wind direction. Also, are there other meteorological factors that affect CO concentrations?
- Section 5.6: How did you assess whether these built machine learning models have an overfitting issue or not?
- Section 6: This section contains multiple repetitions with the Section 4 method and the Section 5 results. May consider combining the replicate parts to shorten this paper and make it easier to read.
Minor changes:
- Start from Line 376: Sections 4.4.1.1, 4.4.1.2, 4.4.1.3 should be renumbered as 4.3.1.1, 4.3.1.2, 4.3.1.2.
- Figure 7: Is the y-axis showing CO concentrations?
- Figures 8 and 9: Aligning the y-axis ranges in Figures 8 and 9 will clarify outlier handling. Also, consider placing the legends for all plots together, outside the figure.
- Figure 11: This figure only shows CO concentrations after outlier handling, and does not include data before outlier handling. Need to modify the caption to clarify this. Also, consider placing the legends for all plots together, outside the figure. The y-axis should be labeled with CO and unit.
- RMSE and MAE should have units (e.g., Lines 966, 973-974, 977-978, 982-983, and Table 12)
- Figures 37-42: The y-axis should include units.
Author Response
Response to Reviewer 1
Ali Suliman AlSalehy – Mike Bailey
Thursday 22nd May, 2025
We sincerely thank the Reviewer for their thorough review and valuable comments, which have helped us improve the manuscript significantly.
Comment 1:
Introduction: It would be helpful to add 1-2 sentences explaining why this study only focuses on CO rather than other gaseous air pollutants.
Response: We clarified this in the Introduction on page 2, lines 68–71. CO was selected due to its strong public health relevance and clear association with mobile and industrial sources. While data on other gases were available, CO offers the clearest signal for predictive modeling in this context. The availability of other pollutants allows for future studies, which fall beyond the scope of this paper.
Comment 2:
Section 3.2: Give more details about the sources of the gas pollutant and meteorological data. If these data were measured using your own sensors, please provide information on the calibration and validation processes.
Response: The data were sourced from the Environmental Monitoring and Control Department of the Royal Commission for Jubail. As these were external data, we could not independently verify sensor calibration. We assumed standard operational procedures were followed. This clarification appears in Section 3 on page 6, lines 213–216.
Comment 3:
Section 4.3.1: For meteorological impact, only the wind speed is mentioned in the main text, but Table 1 includes only wind direction. Also, are there other meteorological factors that affect CO concentrations?
Response: We agree this distinction may not be immediately intuitive to readers and have clarified it in the manuscript. The temperature and wind speed were categorized as "Temporal" because we applied time-based transformations like lags and rolling statistics. Wind direction, encoded using sine/cosine to preserve cyclic properties, was categorized as "Meteorological." We added a clarifying note below Table 1 on page 11.
Comment 4:
Section 5.6: How did you assess whether these built machine learning models have an overfitting issue or not?
Response: As detailed in Section 4.4.1 (page 17, lines 562–568) and Section 4.5.1 (page 19, lines 627–636), we used expanding-window TimeSeriesSplit cross-validation, preserving temporal order. We compared train and test metrics (RMSE, MAE, MAPE, R²) across all folds. Similar performance across folds and use of regularization confirmed the models were not overfitting. This has now been explicitly summarized in Section 5.6.
Comment 5:
Section 6: This section contains multiple repetitions with the Section 4 method and the Section 5 results. May consider combining the replicate parts to shorten this paper and make it easier to read.
Response: We understand the concern and have revised Section 6 to reduce repetition. While some restatement was intentional to guide the reader through a complex workflow, we’ve made it more concise while preserving interpretation and implications. Specifically, we removed overlapping text summarizing model methods and results, and instead directed readers to Sections 4 and 5 for details.
Minor changes:
Comment 1:
Start from Line 376: Sections 4.4.1.1, 4.4.1.2, 4.4.1.3 should be renumbered as 4.3.1.1, 4.3.1.2, 4.3.1.2.
Response: Corrected the numbering from 4.4.1.x to 4.3.1.x throughout the manuscript.
Comment 2:
Figure 7: Is the y-axis showing CO concentrations?
Response: The y-axis in Figure 7 (and all relevant figures) has been updated to explicitly state CO concentrations in ppm.
Comment 3:
Figures 8 and 9: Aligning the y-axis ranges in Figures 8 and 9 will clarify outlier handling. Also, consider placing the legends for all plots together, outside the figure.
Response: We adjusted the y-axis ranges in Figures 8 and 9 and repositioned all legends outside the plots to enhance clarity.
Comment 4:
Figure 11: This figure only shows CO concentrations after outlier handling, and does not include data before outlier handling. Need to modify the caption to clarify this. Also, consider placing the legends for all plots together, outside the figure. The y-axis should be labeled with CO and unit.
Response: Caption now specifies that the data shown are after outlier removal. We also ensured the y-axis is labeled with CO and the unit “ppm.” The renumbering (from Figure 11 to 6) was part of a broader reorganization to prevent figure order confusion.
Comment 5:
RMSE and MAE should have units (e.g., Lines 966, 973-974, 977-978, 982-983, and Table 12)
Response: We added units (ppm, ppm², and ...) to all relevant metrics in lines 966, 973–983, and Table 12 on pages 34–35.
Comment 6:
Figures 37-42: The y-axis should include units.
Response: Y-axes in Figures 37–42 have been updated to include units appropriate to each variable (e.g., ppm for CO, °C for temperature, m/s for wind speed).
Note: A detailed point-by-point response is provided in the attached PDF for the reviewer’s reference.
We have carefully considered all reviewer feedback and made revisions accordingly. The manuscript now includes expanded justification for CO selection, clarified model validation procedures, consistent figure formatting and captions, complete variable definitions, and improved visual clarity. We trust these changes address all concerns and respectfully resubmit the revised manuscript for your consideration.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThe manuscript is well written and easy to understand. The topic is relevant, and the authors have presented a very interesting approach. Below, I provide some comments to contribute to the final quality of the work.
- I suggest the authors reduce the number of keywords.
- I strongly suggest that the authors describe of all acronyms, at least the first time they appear in the manuscript (e.g., line 46: "...LSTM networks...").
- By convention, a space should be left between the numeral and the unit. I suggest the authors revise the description of the altitudes in lines 212 and 213.
- I suggest that the authors revise the figure citations. For example, Figure 2 is not cited in the text.
- Equation 4 represents three separate equations. I suggest that the authors revise this point. A similar issue occurs with equation 5
- Figure 3 has poor resolution.
- In Equations 6 through 11, it is necessary to identify the variables described explicitly
- Figures 4 and 5 are missing the variable labels on the axes. I suggest the authors revise all figures.
- Figures 6, 10 and 14 are missing.
- I found it somewhat confusing how the authors introduce figures in the text that are presented much later. For example, in line 790, Figures 10 and 11 are cited, followed shortly by Figure 7. I suggest that the text be restructured to avoid this issue.
Author Response
Response to Reviewer 2
Ali Suliman AlSalehy – Mike Bailey
Thursday 22nd May, 2025
We thank the Reviewer for their thoughtful and constructive comments. Below are our point-by-point responses.
Comment 1:
I suggest the authors reduce the number of keywords..
Response: We reduced and refined the keyword list to focus on the core concepts and removed broad terms like “Environmental Monitoring.”
Comment 2:
I strongly suggest that the authors describe of all acronyms, at least the first time they appear in the manuscript (e.g., line 46: "...LSTM networks...").
Response: All acronyms are now defined when first mentioned throughout the manuscript, including in the abstract and captions.
Comment 3:
By convention, a space should be left between the numeral and the unit. I suggest the authors revise the description of the altitudes in lines 212 and 213.
Response: All units were reviewed and corrected to include spacing, e.g., "10 m" instead of "10m." See page 6, lines 222–224.
Comment 4:
I suggest that the authors revise the figure citations. For example, Figure 2 is not cited in the text.
Response: All figure references have been reviewed and updated. See Methodology Section 4, page 7, lines 261–262 for the correction.
Comment 5:
Equation 4 represents three separate equations. I suggest that the authors revise this point. A similar issue occurs with equation 5.
Response: Equations 4 and 5 were split into individual equations for clarity. See page 12, lines 418 and 428.
Comment 6:
Figure 3 has poor resolution.
Response: Replaced with a high-resolution version to ensure legibility of labels and details.
Comment 7:
In Equations 6 through 11, it is necessary to identify the variables described explicitly.
Response: All variables used in Equations 6–11 are now defined immediately after each equation. See pages 17–18, lines 579–598.
Comment 8:
Figures 4 and 5 are missing the variable labels on the axes. I suggest the authors revise all figures.
Response: We updated Figures 4 and 5, and all other figures, to include clear axis labels with variable names and units.
Comment 9:
Figures 6, 10 and 14 are missing.
Response: These figures included multiple subfigures. We revised all captions and references to reflect this clearly (e.g., Figure 6(a), Figure 6(b) within a unified caption).
Comment 10:
I found it somewhat confusing how the authors introduce figures in the text that are presented much later. For example, in line 790, Figures 10 and 11 are cited, followed shortly by Figure 7. I suggest that the text be restructured to avoid this issue.
Response: Figure citations have been restructured to follow a sequential and logical flow that matches their appearance in the manuscript.
Note: A detailed point-by-point response is provided in the attached PDF for the reviewer’s reference.
We have carefully considered all reviewer feedback and made revisions accordingly. The manuscript now includes expanded justification for CO selection, clarified model validation procedures, consistent figure formatting and captions, complete variable definitions, and improved visual clarity. We trust these changes address all concerns and respectfully resubmit the revised manuscript for your consideration.
Author Response File: Author Response.pdf
