GreenNav: Spatiotemporal Prediction of CO₂ Emissions in Paris Road Traffic Using a Hybrid CNN-LSTM Model

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The area of research is very interesting, and the model for prediction is promising. However, from a scientific point of view, there are many mistakes and doubts about the correctness of the research.

 

My main suggestions:

1. The template used for the paper is incorrect.

2. The literature review is one of the strongest aspects of the conducted research, but recent studies from the last few years are missing. The reference formatting style is also incorrect.

3. I suggest deleting the list style from the paragraph under Table 1 and Figure 1.

4. The ":" symbol is unnecessary at the end of headings.

5. I suggest adding sample data in Chapter 3.1.

6. In Figure 2, there is a lack of a start point. I understand it is showing integration, but there should be a clear starting point.

7. In Chapter 3.3, "Phase 2: Model Training" is mentioned, but no model was presented in Phase 1. The authors only presented Data Preprocessing. I recommend adding a phase between Phase 1 and Phase 2 for preparing the model.

8. Figures 7 and 9 should be divided into two separate figures, and the text within the figures should be larger, as it is currently hard to read. There is also a lack of axis descriptions and units.

9. In Figure 13, the x-axis description is cut off, and the quality of the image is poor. I suggest providing it at 300 DPI.

10. On page 20, the heading "Analysis and Results" should use the heading style. The authors likely joined the heading and paragraph by mistake. A similar situation is seen on page 22, where "Experimental Scenario for a Future Date" merges with the paragraph.

11. Figure 15 is also of poor quality, and it is hard to read the values for model predictions and actual emissions. Moreover, the image is in French, not English. What is the prediction error? From the figure, it seems the prediction is exact. Is that possible?

12. Figure 16 is not a figure, but a table. Please present it as a table instead of a JPG. Additionally, the column headings are in French, not English. Please also include the error value. What is the source of the real emissions data? The high accuracy of the prediction model seems unusual.

13. In the abstract, the authors mention achieving an R² value of 0.91 for the model, but this is not presented in the research itself—it only appears in the state of the art. The RMSE of 0.086 is also only mentioned in the abstract.

 

 

Let me know if this addresses your concerns effectively.

Author Response

Response to Reviewer Comments: Article Revisions:

Title: GreenNav: Spatiotemporal Prediction of CO2 Emissions in Paris Road Traffic Using a Hybrid CNN-LSTM Model

Youssef Mekouar^{1,2,3, *} , Imad Saleh^1,3* and Mohammed Karim ^2,3

 

We sincerely thank the reviewer for their relevant and constructive comments. These suggestions have helped improve the quality of our paper and make it more robust. We have carefully considered each of the comments, which has contributed to enriching our work and making it more complete.

 

 

1. The template used for the paper is incorrect.

We updated the paper to align with the correct template as specified by the journal’s submission guidelines. The entire document has been reformatted accordingly, ensuring compliance with the required structure and style.

 

2. The literature review is one of the strongest aspects of the conducted research, but recent studies from the last few years are missing. The reference formatting style is also incorrect.

To address Comment 2, we ensured that the literature review includes several recent works related to hybrid prediction models for CO₂ emission estimation. Specifically, we referenced models predicting CO₂ emissions [23] and hybrid models focusing on road traffic prediction, which is closely linked to CO₂ emissions. Recent articles have been cited in the state-of-the-art section, including studies published in 2024 and 2022 [24], [25], as well as article [23] published in 2024. Additionally, we corrected the reference formatting to fully comply with the journal’s style guidelines.

 

3. I suggest deleting the list style from the paragraph under Table 1 and Figure 1.

We have removed the list style from the explanatory paragraph under Table 1 as requested. However, we chose to keep the list style under Figure 1, as it allows for a clearer visualization of the different components of our dataset being described. This format helps structure the information and enhances the reader’s understanding of the dataset’s various elements.

 

4. The ":" symbol is unnecessary at the end of headings.

We reviewed the manuscript and removed the ":" symbol at the end of the headings, ensuring full compliance with the journal’s template and formatting guidelines.

 

5. I suggest adding sample data in Chapter 3.1.

 

To address these comments, we made the following adjustments: We added a figure illustrating a sample of our integrated dataset after merging data from OpendataParis, GetAmbee, Google API, and Airparif. Change made: Added Figure 2: Example of the final integrated dataset after merging data from OpendataParis, GetAmbee, Google API, and Airparif (Page 6, Line 219).

 

6. In Figure 2, there is a lack of a start point. I understand it is showing integration, but there should be a clear starting point.

We enhanced Figure 2 by adding details indicating the steps of the first iteration in the integration process. We clearly marked the starting and ending points of the data integration process while highlighting its iterative nature. Change made: Updated Figure 2 to include labeled steps (Page 8, Line 291)

 

7. In Chapter 3.3, "Phase 2: Model Training" is mentioned, but no model was presented in Phase 1. The authors only presented Data Preprocessing. I recommend adding a phase between Phase 1 and Phase 2 for preparing the model.

To address this comment, we clarify that the construction of our model architecture is implicitly included in Phase 2: Model Training (Section 3.3), which summarizes the roles of the CNN and LSTM models in learning spatial and temporal features. The detailed architecture of our model is explicitly presented in Section 3.4: Description of Our Hybrid CNN-LSTM Model Architecture, illustrated by Figure 5: Hybrid Architecture of Our CNN-LSTM Model (Page 10, Line 346), accompanied by a detailed explanatory paragraph (Pages 10 to 11, Lines 347 to 381). Additionally, Section 3.5.2: Comparison of Results from Different Model Configurations describes the steps leading to the final architecture after several iterative improvements.

 

8. Figures 7 and 9 should be divided into two separate figures, and the text within the figures should be larger, as it is currently hard to read. There is also a lack of axis descriptions and units.

We separated Figures 7 and 9 into two individual figures according to the journal’s template format. Additionally, we improved the image quality by increasing the resolution and enlarging the text within the figures. Axis descriptions and units have also been added where needed. See figure 8 and 10 (Page 13 & 14)

 

9. In Figure 13, the x-axis description is cut off, and the quality of the image is poor. I suggest providing it at 300 DPI.

·      We corrected the issue by adjusting the x-axis description to ensure full visibility. ·      The image quality was enhanced to 300 DPI as requested.                   See figure 14 ( Page 17)

 

 

 

 

10. On page 20, the heading "Analysis and Results" should use the heading style. The authors likely joined the heading and paragraph by mistake. A similar situation is seen on page 22, where "Experimental Scenario for a Future Date" merges with the paragraph.

To address this comment, we reorganized the section “4. Experimental Scenario for CO₂ Emission Prediction Through the GreenNav Application “(Pages 18-20, Lines 591-648) into two clearly defined scenarios: ·      Scenario 1: Comparison of Predicted vs. Actual CO₂ Emissions (Section 4.1) In this scenario, we used our GreenNav application to visualize CO₂ emissions for a specific road segment on a past date. This allowed us to compare the predicted results from our deployed model with actual historical data, enabling performance evaluation and validation. ·      Scenario 2: CO₂ Emission Prediction for a Future Date (Section 4.2) Here, we used GreenNav to predict future CO₂ emissions for a given road segment. Since reference data for future dates is unavailable, this scenario focuses solely on forecasting, highlighting the application’s predictive capabilities.

 

11. Figure 15 is also of poor quality, and it is hard to read the values for model predictions and actual emissions. Moreover, the image is in French, not English. What is the prediction error? From the figure, it seems the prediction is exact. Is that possible?

The poor image quality in Figure 15  resulted from taking a screenshot of the prediction results displayed on our web application. We have since improved the image quality for better readability. Regarding the prediction accuracy, the prediction error is detailed in Table 3: Comparative Table Between Predicted and Actual Values for the Given Scenario (Page 19, Line 619). In the specific road segment of Saint-Germain, our model achieved approximately 99% accuracy. However, the model's accuracy may decrease for other road segments, depending on data variability and traffic complexity. On the entire test set, our model achieved an overall prediction accuracy of 91%, reflecting its performance across diverse traffic scenarios.

 

 

12. Figure 16 is not a figure, but a table. Please present it as a table instead of a JPG. Additionally, the column headings are in French, not English. Please also include the error value. What is the source of the real emissions data? The high accuracy of the prediction model seems unusual.

To address this comment, we reformatted the generated table from our GreenNav application and presented it as a proper table with column headings in English: Table 3: Comparative Table Between Predicted and Actual Values for the Given Scenario. We provided examples in this table (Pages 19-20, Lines 621 to 626), comparing predicted and actual CO₂ emission values for the specific road segment in question. The high prediction accuracy observed for this road segment reflects favorable conditions and data coverage. However, tests conducted on other road segments, such as Boulevard Réaumur, showed a larger prediction error, with an accuracy of 89%. This highlights that while the model performs well in many scenarios, its accuracy may vary depending on the complexity and characteristics of the road segments.

 

 

13. In the abstract, the authors mention achieving an R² value of 0.91 for the model, but this is not presented in the research itself it only appears in the state of the art. The RMSE of 0.086 is also only mentioned in the abstract.

We appreciate the reviewer’s valuable feedback. While we correctly mentioned the R² score in our results (Page 15, Line 541), we acknowledge that the RMSE value was unintentionally omitted from this section. We have now updated the results to include the RMSE value for completeness and clarity.

 

 

 

 

 

 

Reviewer 2 Report

Comments and Suggestions for Authors

Summary:

In this article, the authors presented and developed a method for predicting CO2 emissions in Paris using an artificial neural network model. The developed artificial neural network model uses two separate branches to process temporal data and spatial data. The data from the two branches are then combined and the final prediction is obtained from them. The authors conducted the process of analyzing and testing the model, and additionally presented its practical application in the GreenNav application.

Comments:

I have not been able to find in the text what exact function of the loss was taken, please clarify and include such information in the text.

The authors use the Accuracy metric in the text, please clarify what exact metric was adopted. Additionally, I would suggest adding Accuracy charts.

The first graphs in Fig. 7, 9, 12 are not very readable, I would suggest changing them to, for example, presenting the value as an absolute difference. This would make it easier to read.

The authors touch on the computational complexity of the models. I suggest extending the discussion of computational complexity by presenting theoretical computing power requirements expressed in FLOPs - floating point operations per second. In order to better compare computational complexity.

I would suggest presenting a summary of the selected hypermatrameters and the obtained values of the loss and accuracy functions for the tested solutions in tabular form.

I would also suggest shortening the abstract due to the fact that it is simply too long and contains too much detail.

Author Response

Response to Reviewer Comments: Article Revisions:

Title: GreenNav: Spatiotemporal Prediction of CO2 Emissions in Paris Road Traffic Using a Hybrid CNN-LSTM Model

Youssef Mekouar^{1,2,3, *} , Imad Saleh^1,3* and Mohammed Karim ^2,3

We sincerely thank the reviewer for their relevant and constructive comments. These suggestions have helped improve the quality of our paper and make it more robust. We have carefully considered each of the comments, which has contributed to enriching our work and making it more complete.

1. I have not been able to find in the exact function of the loss taken in the text. Please clarify and include such information in the text.

To address your comment, we clarified the loss function used in our model on Page 8, Lines 279-285, where we specify that the Root Mean Squared Error (RMSE) was employed during model optimization.

 

2. The authors use the Accuracy metric in the text, please clarify what exact metric was adopted. Additionally, I would suggest adding Accuracy charts.

To address your comment, we clarify that the exact metric used in our evaluation is the Coefficient of Determination (R²). We described this metric in detail on Page 9, Lines 325-336, explaining its significance and why it was chosen to evaluate the predictive performance of our model. Specifically, R² measures the proportion of variance in the actual CO₂ emissions that is explained by the model's predictions, providing a clear indication of prediction accuracy. Regarding the suggestion to include Accuracy Charts, generating such charts would require retraining our model while saving the accuracy for each epoch during training. As our model has already been trained and deployed in the GreenNav application, we acknowledge this limitation and will consider incorporating accuracy tracking in future versions of the model.

 

 

3. The first graphs in Fig. 7, 9, 12 are not very readable, I would suggest changing them to, for example, presenting the value as an absolute difference. This would make it easier to read.

The low quality of the first graphs in Figures 7, 9, and 12 was due to screenshots taken from our notebook during the experimentation process. To address this issue, we found a way to export the graphs at 300 DPI, significantly improving their clarity and readability. (see figure 8 on page 13, 10 on page 14 , 13 on page 16)

 

4. The authors touch on the computational complexity of the models. I suggest extending the discussion of computational complexity by presenting theoretical computing power requirements expressed in FLOPs - floating point operations per second. In order to better compare computational complexity.

We acknowledge that presenting the computational complexity in terms of FLOPs (Floating Point Operations per Second) is an excellent suggestion. However, we did not consider saving these parameters during the training process of our model. Implementing this measure would require retraining the model while logging FLOPs calculations at each step. We recognize the value of this approach and will consider incorporating it in future experiments.

 

5. I would suggest presenting a summary of the selected hypermatrameters and the obtained values of the loss and accuracy functions for the tested solutions in tabular form.

To address this comment, we created Table 2: Comparison of the Three Configurations of the Hybrid CNN-LSTM Model (Page 17, Line 576), which summarizes the selected hyperparameters and the corresponding values of the loss and accuracy functions obtained during testing. This table provides a clear overview of the tested configurations and their respective performances.

 

 

 

6. I would also suggest shortening the abstract due to the fact that it is simply too long and contains too much detail.

To address this comment, we shortened the abstract by reducing its length to comply with the journal’s template requirement of 200 words maximum. The revised abstract now focuses on the most relevant aspects of our research while omitting excessive details.

 

 

 

 

 

Reviewer 3 Report

Comments and Suggestions for Authors

Review

In the paper „Spatiotemporal Prediction of CO2 Emissions in Paris Road Traffic Using a Hybrid CNN-LSTM Model” the authors proposed a model for predicting CO2 on the streets of Paris, in which they took into account spatial and temporal traffic data using deep learning networks LSTM and CNN. The article is interesting, but many things must be improved.

 

My comments:

1. In the abstract, the authors write that they will compare their prediction model with a linear regression model. There is no such comparison in the article for the same measurement data.

2. The literature in the references is not numbered consecutively. For example, at the beginning of page 4 there is a reference to item 36, and previously on page 3 the last reference has number 23.

3. The authors repeatedly repeat the same statement, which is obvious, that the increase in CO2 emissions increases with the increase in traffic intensity. It might be worth defining what this relationship is with a formula.

4. Due to the fact that the authors studied the impact of many factors on CO2 emissions, it seems reasonable to determine which of the factors (data) has the greatest impact on the accuracy of CO2 prediction.

5. The authors presented the next steps of optimizing the network training parameters that affect the accuracy of the prediction, but no examples of the actual model input data were provided. Please add a table containing sample spatial and temporal input data before normalization, which are fed to the input of the LSTM network and to the input of the CNN network. Provide a method of data normalization.

6. In Fig. 14, the data presented concerns 01/01/2023, and the authors write below and in several other places that the test date is January 6, 2023!

7. Similarly, in Fig. 15, the date is 7/10/2023 at the top and the date is 01/01/2023 on the right?

In the table Fig. 16 (the table should be numbered) there are missing units. In addition, it is not necessary to write numbers with an accuracy of 6 digits after the dot if the data represent integers (e.g. in column q instead of 517.000000 enter 517). This makes it difficult to analyze the table content.

8. The illegible figures (7, 9, 12 – descriptions of the coordinate axes in the drawing on the left) also require correction.

 

Author Response

Response to Reviewer Comments: Article Revisions:

Title: GreenNav: Spatiotemporal Prediction of CO2 Emissions in Paris Road Traffic Using a Hybrid CNN-LSTM Model

Youssef Mekouar^{1,2,3, *} , Imad Saleh^1,3* and Mohammed Karim ^2,3

 

We sincerely thank the reviewer for their relevant and constructive comments. These suggestions have helped improve the quality of our paper and make it more robust. We have carefully considered each of the comments, which has contributed to enriching our work and making it more complete.

 

 

1. In the abstract, the authors write that they will compare their prediction model with a linear regression model. There is no such comparison in the article for the same measurement data.

You are correct, the comparison with a linear regression model was mistakenly mentioned in the abstract. During the development of the article, we found it more relevant to compare our hybrid model with other existing hybrid models in the state-of-the-art section. To address this, we removed the reference to linear regression from the abstract. The relevant comparisons are detailed in Table 1: Review of Existing Artificial Intelligence Models for Predicting CO₂ Emissions: Approaches and Methodologies (Page 3, Line 133).

 

2. The literature in the references is not numbered consecutively. For example, at the beginning of page 4 there is a reference to item 36, and previously on page 3 the last reference has number 23.

To address this comment, we corrected the numbering of all references in the manuscript, ensuring that they are listed in consecutive order throughout the document ( Page 3, line 113 and 120). Additionally, we corrected the reference formatting to fully comply with the journal’s style guidelines.

 

3. The authors repeatedly repeat the same statement, which is obvious, that the increase in CO2 emissions increases with the increase in traffic intensity. It might be worth defining what this relationship is with a formula.

To address this comment, we introduced a new section in our article (Page 6, Lines 221 to Page 7, Line 240) where we describe the CO₂ emission modeling process we implemented in our previous work. This section includes a probabilistic equation that models CO₂ emissions based on various emission factors, including traffic intensity, vehicle characteristics, and driving conditions.

 

4. Due to the fact that the authors studied the impact of many factors on CO2 emissions, it seems reasonable to determine which of the factors (data) has the greatest impact on the accuracy of CO2 prediction.

We referenced our previous work ([16], [17], [18]), where the selection of CO₂ emission factors for each road segment in Paris was studied in detail. In the current article (Page 7, Lines 231 to Line 233), we listed the emission factors considered in our model. The key factors influencing CO₂ emissions are N, representing the number of vehicles in the segment, and the occupancy rate, reflecting how long these vehicles occupy the segment, which increases with traffic density.

 

5. The authors presented the next steps of optimizing the network training parameters that affect the accuracy of the prediction, but no examples of the actual model input data were provided. Please add a table containing sample spatial and temporal input data before normalization, which are fed to the input of the LSTM network and to the input of the CNN network. Provide a method of data normalization.

 

To address this comment, we described our temporal and spatial input datasetsused for model training on Page 8, Lines 301-310, as well as in the model architecture description on Page 10, Lines 349-359. Additionally, we provided an example of the integrated dataset after merging data from multiple APIs (OpendataParis, Get-ambee, Google API, and Airparif) in Figure 2 (Page 6, Line 219). To clarify our data preprocessing approach, we detailed the normalization strategies applied to our dataset to enhance model learning. These data preprocessing methods, including the normalization functions used, are described on Page 10, Lines 363-382, and further explained on Page 11, Lines 398-413.

 

6. In Fig. 14, the data presented concerns 01/01/2023, and the authors write below and in several other places that the test date is January 6, 2023!

We apologize for the confusion caused by the incorrect date 01/01/2023 displayed now in Figure 15 ( Page 20 at line 628) , which was due to a default setting in our application’s data display component. We appreciate the reviewer for pointing out this oversight. To resolve the issue, we corrected the application’s default date configuration to ensure accurate data display in future results.

 

7. Similarly, in Fig. 15, the date is 7/10/2023 at the top and the date is 01/01/2023 on the right? In the table Fig. 16 (the table should be numbered) there are missing units. In addition, it is not necessary to write numbers with an accuracy of 6 digits after the dot if the data represent integers (e.g. in column q instead of 517.000000 enter 517). This makes it difficult to analyze the table content.

We corrected the inconsistent dates in the former Figure 15, now updated as Figure 16: Component: CO₂ Emission Predictions for a Street at a Future Time. Additionally, the former Figure 16 has been renumbered as Table 3: Comparative Table Between Predicted and Actual Values for the Given Scenario, with missing units added and numerical precision corrected.

 

8. The illegible figures (7, 9, 12 – descriptions of the coordinate axes in the drawing on the left) also require correction.

We corrected the illegible axis descriptions in Figures 7, 9, and 12 by increasing the font size and improving the overall image quality for better readability.

 

Reviewer 4 Report

Comments and Suggestions for Authors

This paper tackles the problem of predicting CO2 emissions in urban road traffic by using a hybrid CNN-LSTM model that employs (street level) spatial and temporal traffic data. Moreover, the model is integrated into the GreenNav application, which allows precise estimations of  CO2 emissions for every street in Paris, France in real time. From a practical perspective, this is a big plus.

However, for improving the clarity of the presentation, I have a few suggestions:

-          The manuscript is rather lengthy, which may hinder the overall readability. Certain verbose paragraphs may be condensed, other (useful) details, like how the data from different sources has been preprocessed to prepare the training dataset(s), may be placed in an Appendix, or in Supplementary materials. This observation is also valid for the description of the refinement process that improved the accuracy of the model from 61% to 91%. A summary of iterations might suffice, focusing more on the final model and its validation.

-          Several figures (3,4,7,11) are not directly referenced in the text, and the figure captions are – in my opinion – rather eliptic.

-          Though the practical approach is obvious and welcome, the problem of reproducibility and scalability of the solution is not approached in sufficient detail. While the manuscript discusses adaptation to other urban environments, the proposed strategies (transfer learning and federated learning) remain hypothetical and lack the details that might help other researchers to reproduce the study in another city.

-          Certain limitations of the model (e.g. the model’s reliance on high-resolution data,  raising scalability concerns for cities with limited data access, and the influence of external factors such as wind or other weather conditions on CO2 dispersion) are acknowledged but not explicitly addressed in the modelling. Maybe a discussion of how data availability impacts model accuracy would be helpful.

Author Response

Response to Reviewer Comments: Article Revisions:

Title: GreenNav: Spatiotemporal Prediction of CO2 Emissions in Paris Road Traffic Using a Hybrid CNN-LSTM Model

Youssef Mekouar^{1,2,3, *} , Imad Saleh^1,3* and Mohammed Karim ^2,3

 

We sincerely thank the reviewer for their relevant and constructive comments. These suggestions have helped improve the quality of our paper and make it more robust. We have carefully considered each of the comments, which has contributed to enriching our work and making it more complete.

 

1. The manuscript is rather lengthy, which may hinder the overall readability. Certain verbose paragraphs may be condensed, other (useful) details, like how the data from different sources has been preprocessed to prepare the training dataset(s), may be placed in an Appendix, or in Supplementary materials. This observation is also valid for the description of the refinement process that improved the accuracy of the model from 61% to 91%. A summary of iterations might suffice, focusing more on the final model and its validation.

To address this comment, we followed your suggestions and made the following improvements: ·      We expanded the Data Preprocessing section by providing more details on the normalization functions applied to each type of data, ensuring a dataset best suited for model training (Page 11, Lines 384 to Page 12, Line 435). ·      To better explain the steps and procedures that allowed us to transition from one model configuration to another, we summarized the configurations of the three implemented architectures in Table 2: Comparison of the Three Configurations of the Hybrid CNN-LSTM Model (Page 17, Line 576).

 

2. Several figures (3,4,7,11) are not directly referenced in the text, and the figure captions are – in my opinion – rather eliptic.2

We reviewed the manuscript and ensured that all figures, including Figures 3, 4, 7, and 11, are explicitly referenced in the text. Additionally, we revised the figure captions to provide clearer and more descriptive explanations.

 

3. Though the practical approach is obvious and welcome, the problem of reproducibility and scalability of the solution is not approached in sufficient detail. While the manuscript discusses adaptation to other urban environments, the proposed strategies (transfer learning and federated learning) remain hypothetical and lack the details that might help other researchers to reproduce the study in another city.

 

We referenced relevant studies closely aligned with our proposed strategies for adapting the model to other urban environments using transfer learning and federated learning (Page 21, Lines 670 to 698). These references provide a solid foundation for extending our approach in future research. Additionally, we introduced a standardized protocol that other researchers can follow to implement our CO₂ emission prediction system in different cities, considering specific constraints and data availability (Page 21, Lines 699 to 713). This protocol aims to improve the reproducibility and scalability of our approach.

 

4. Certain limitations of the model (e.g. the model’s reliance on high-resolution data, raising scalability concerns for cities with limited data access, and the influence of external factors such as wind or other weather conditions on CO2 dispersion) are acknowledged but not explicitly addressed in the modelling. Maybe a discussion of how data availability impacts model accuracy would be helpful.

In this article, we acknowledge that several external factors, such as high-resolution data availability, weather conditions (e.g., wind), and data access limitations, can impact CO₂ emissions in road traffic. A detailed study on these influencing factors was conducted in our previous works ([16], [17], [18]), where we examined their effects on CO₂ emissions modeling. To strengthen this point, we added an explanatory paragraph in the current manuscript, where we modeled CO₂ emissions for a specific road segment using relevant emission factors derived from our previous studies (Page 6, Lines 221 to 240). This addition clarifies how such factors are considered in our modeling approach and highlights their significance in predicting CO₂ emissions accurately.

 

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The paper is now better prepared for final publishing.

I suggest adding the number of formulas presented in line 227 and making the variables not bolded in the formula and definition.

Author Response

Response to Reviewer Comments: Article Revisions:

Title: GreenNav: Spatiotemporal Prediction of CO2 Emissions in Paris Road Traffic Using a Hybrid CNN-LSTM Model

Youssef Mekouar^{1,2,3, *} , Imad Saleh^1,3* and Mohammed Karim ^2,3

 

We sincerely thank the reviewer for their relevant and constructive comments. These suggestions have helped improve the quality of our paper and make it more robust. We have carefully considered each of the comments, which has contributed to enriching our work and making it more complete.

 

1. I suggest adding the number of formulas presented in line 227 and making the variables not bolded in the formula and definition.

To address this comment, we added the numbering for the formulas presented in line 227 and ensured that the variables in the formula and their definitions are not bolded.

Reviewer 2 Report

Comments and Suggestions for Authors

Well revised.

Author Response

We sincerely thank the reviewer.