Machine Learning-Enhanced 3D GIS Urban Noise Mapping with Multi-Modal Factors

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The paper presents an urban noise simulation method that combines the traffic noise propagation model with machine learning, optimizes the prediction accuracy by integrating multi-source data such as topography, vegetation and road network, and develops a visualization system that integrates the noise simulation and 3D scene. The study provides innovative ideas for the simulation of noise propagation and the application of GIS technology.
The topic is quite interesting and fits the scope of journal well. There are several comments and suggestions for authors to further address.
1. The keywords are too limited. Additional terms like “3D visualization” and “machine learning optimization” are suggested to append.
2. The paper discusses 'multi-modal factors' but does not explicitly define them in the introduction. Please specify.
3. The authors should incorporate recent research on machine learning optimization for noise prediction to enhance the completeness of the related work.
4. Figure 1, spatial data is suggested to change to “spatial data preparation”. Text case patterns in full-text images should be uniform.
5. Eq.(1)~(5), corresponding reference should be cited.
6. the legend bar for subfigure 5(b) and 5(d) should be added.
7. Section 3.3, why choose ridge regression model?
8. Section 3.4, the system design diagram as well as the relationship between the system and the previous models and Noise Optimization methods should be given. Otherwise, the relevance of section 3.4 to the preceding sections is really weak.
9. the second paragraph of section 4.4 is quite similar with the text of section 3.4, please modify. At the same time, the English abbreviation only needs to be given once (when used for the first time), and the LOD abbreviation does not need to be given in sections 3.4 and 4.4. The abbreviation of the word LiDAR is not given once. It is recommended to check the full text carefully.
10. Language polishing of the paper is suggested.

Comments on the Quality of English Language

Language polishing of the paper is suggested.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

The authors propose a method for noise mapping that incorporates multiple influencing factors and utilizes 3D scenery within their modelling approach. This is an intriguing and timely topic, with potential implications for urban soundscape assessment and management. The integration of 3D elements is particularly interesting and shows promise. However, several critical issues need to be addressed before the paper can be considered for publication, specially regarding the first point.

• A fundamental concern is that the study is based on data collected from only one day. Given the well-known temporal variability of traffic noise, this raises substantial doubts about the representativeness and robustness of the findings. Traffic conditions fluctuate greatly depending on the day of the week, season, weather, or unexpected events. The authors should either justify the use of single-day data with supporting literature or ideally expand the dataset to multiple time points to strengthen their conclusions.
• The noise optimisation results are not well presented. Presenting the resulting raster maps and showing how the optimization works spatially would greatly enhance clarity. Furthermore, some sections in the Results (e.g., Lines 358–362) are methodological and should be moved accordingly.
• The selection of the three predictor variables (LVD150, WAVD50, len100) lacks statistical justification. Are significance tests, correlation metrics, or feature importance measures supporting their choice? As it stands, the explanation feels arbitrary.
• The reported R² values are moderate, which raises concerns regarding the model’s predictive power. The authors should discuss this in more depth: Are these values acceptable for this type of modelling? Could this be improved with more data or different methods? Also, how does this uncertainty affect the reliability of the "noise optimization"?

• A critical omission is the lack of comparison with existing approaches. The paper should explicitly address. How does this method outperform other traditional noise modelling tools? Why should one use this method? Is it faster, more accurate, and easier to deploy?

• Table 1 presents a noticeable discrepancy between maximum and minimum values across certain variables. Are these values due to extreme outliers, data collection inconsistencies, or inherent environmental variation? 
• The authors should elaborate on the rationale for selecting five-fold cross-validation. While this is a common technique, its effectiveness depends on the dataset size and variance. Was this choice based on convention, performance considerations, or other criteria?
• Figure 5 is not sufficiently informative:
  • 5a: Presents the scenery, but does not offer any additional insight. What does it add to the narrative?
  • 5b: Lacks legend — what do the colors denote?
  • 5c: Traffic flow simulation is hard to interpret. Please clarify or annotate.
  • 5d: Again, the legend is unreadable, making interpretation impossible.
• The paper refers to the use of "Unreal Engine". The authors should concisely describe the platform, its role in the research, and its benefits in this context. Readers from environmental sciences may not be familiar with this tool.

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The paper approaches noise mapping with GIS technique and 3D rendering which is interesting. Besides the results are good I wonder if the approach can be replicated due to the large amount of measurements to refine model and to the lack of some methodologies details. In the following some issues are listed to try to specify some points of the research in order to allow replication.

minor :

some citations in introdution are old (eg n.24) please consider also https://doi.org/10.1145/3659994.3660312,https://doi.org/10.1371/journal.pone.0248939

 

line 254 check words order. 

major:

par. 3.4 please detail the format in which geoJSON are converted and the system to change lthe light according to noise.

in par. 4.3 please detail if the Ltnpm levels compared to measures are compared for each hour of measure or are average dayly values for both measured and predicted values.

In par. 4.3 you mentioned that predicted mean value is "slightly higher" than measured one.  This is not true since 14 dB means that you are predicting a traffic that is 32 times higher than reality (eg 30 cars vs more than 580 cars). Please change text accordingly

In par 4.4 most of description is already in methods (lines 410-436), please move all details about methodology  to methods section and remove from results.

About the conclusions I think it is worth to mention that an huge number of mesurements was performed to improve the model performances. 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

Thanks to the authors for having improved the paper. All suggestions and comments have been addressed and now methods are more largely described, and results are more clearly presented. Authors also include improved figures and tables which are highly appreciated.