A Sequence-Based Hybrid Ensemble Approach for Estimating Trail Pavement Roughness Using Smartphone and Bicycle Data

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This study fosters the investigation of realistic and accurate methods and data analysis techniques to measure the IRI of the pavement surface. The study presents an innovative method and extremely valuable comparison between different techniques. However, multiple comments needs to be addressed before accepting the manuscript for publication

·        A comprehensive explanation is required for Figure 1. For example, what do the authors mean by “bicycle model”?. It is not clear for the audience the key points of every points.

·        It is recommended that the name of the sections, the dates of their last overlay or rehabilitation, and the time of IRI survey to be reported

·        In figure 2, it is recommended to show the fixation points of the camera and iphone on the bicycle. Also, it is recommended to add some labels to show the different parts of the walking profile. The figure should be more informative.

·        What are the roles of the two cameras on the bicycle?

·        Where is the accelerometer fixed on the bicycle??. Section 2.2 mentions a measurement for the vertical acceleration, while the image of the bicycle (figure 2) does not report an accelerometer on the bicycle

·        The presentation of equations 7 and 8 may be not necessary. Meanwhile, clear and detailed explanation for the “threshold velocity” of equation 9 is very important.

·        Figure 3 and the related explanation (lines 272 to 300) are hard to understand. A different presentation method and more clear explanation are necessary. For example, it is not what does “ensemble model 1” mead?. It is not clear what does the separation “input n” and “input n-1” mean?

·        What does W in equation 15 mean?

·        The explanation for the increment of the RMSE with the length increase, reported between lines 369 to 379, is not strong enough. The authors are recommended to find more robust interpretation using references

·        As the three techniques, random forest, adaptive boosting, and gradient boosting, are used in the study, it is recommended to introduce more about their methods, concepts, equations, or algorithms in a “theoretical background section” after the introduction part

·        The conclusion section is very short to summarize the most important findings of the study. Many findings are not mentioned in the conclusion section

For the abstract, More reporting for the findings of the study is required, while the discussion of the gap in the literature can be minimized

Comments on the Quality of English Language

No English proficiency issues have been noticed

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments and Suggestions for Authors

General comments :

The paper is well written the concept and the analysis and the use case is well developed. The commercial use case for this technique is questionable given that traffic speed deflectometer is a well-established method for assesment of surface roughness for major pavement sections. In addition, with climate change ,where overheating of pavement surfaces beyond their average surface temperatures is resulting in cases of loss of texture instead of increased surface roughness being the major factor causing pavement failure.  

 

The referencing used is comprehensive and also the paper writing is upto the mark.

Line number 190 :The friction factor can also be calculated using Newtons second law. Not sure of the applied logic for the bicycle roughness index as an indicator of friction factor. The surface roughness of the road in every section can also be determined by using Image analysis techniques . The image analysis technique can capture the road profile in a much better way. There is not comparison of the various methods that could be used to estimate friction factor using a smart phone

 

Line 197 : The final correlation between BRI and IRI is derived using an empirical regression equation. This would then mean that if we have to use BRI as a indicator for road roughness then for each pavement surface the regression coefficients needs to be estimated which would then limit the usability of the method to predict surface roughness. A sampling and mathematical technique to utilize BRI for various different pavements is missing.

 

Line 253 : The SVR regression technique is a good procedure but there are about 20 input variables and one output parameter. This can result in a regression model with an inherent tendency of overfitting. Information on how the 20 input parameters were converted to a single parameter function is not provided. The sensitivity of the parameter function to the input variables is important, since that can impact the regression coefficients, since the final SVM model has only one predictor variable.

 

Line no 305  : It is a standard practice to the validate the model using 20% validation data, the validation data set used should be representative (meaning datapoint all across the range should be selected) a few lines on the validation data set used would be useful to ensure that the validation was robust

 

Line 339 :  The approach sued for assessing the impact of each variable (MSE approach) is a good robust technique. The only question being since MSE is a squared function even small changes could be amplified hence is the ratio used truly representative  or is it like case of being too cautious ?

Comments on the Quality of English Language

The paper is well written there are no specific english language comments that are required. All the references have been cited 

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The authors have made the required changes and the manuscript can be accepted for publication

Reviewer 2 Report

Comments and Suggestions for Authors

Happy to see the revised version of the manuscript, the method description are very detailed and will be useful for the community. I have gone through the revised changes highlighted in blue in the manuscript and I am okay with the changes.