Correlation between Land Use Pattern and Urban Rail Ridership Based on Bicycle-Sharing Trajectory
Round 1
Reviewer 1 Report
Dear Author(s), I would like to thank you for the opportunity to read your manuscript entitled “Correlation between land use pattern and urban rail ridership based on bicycle-sharing trajectory”.
The overall manuscript is well presented with minor spelling or grammar mistakes.
The overall work is very interesting, as the land use management is very relevant and necessary especially in urban transport systems.
Here are some issues concerning your paper:
1. The overall purpose of the article is stated clearly in the introduction and underlined in the abstract.
2. The Literature Review part is logical and well organized. However, I think it is preferable to mention the first author in the text. It would be especially preferred in those places where the sentence starts with the number of the publication you want to cite, e.g. line 133 [24], line 134 [26] or line 138 [27]. Please emphasize what is the knowledge gap.
3. The Methodology part is clear. I am not sure if the maps shown at Figures are visible as they are so small.
4. The date data in Table 1 is unclear.
5. Figure 12 needs further explanations in its title. Perhaps adding a,b,c and d elements to it would allow to sort out the differences between the presented graphs.
6. According to the graphs in Figure 13 and Figure 14: the units should be added on each axis.
7. The conclusions are stated clearly and are supported by research results.
8. Future research directions and the significance of the results of the research achieved should be underlined and explained in conclusion part.
Reviewer
Author Response
Response to Reviewer 1 Comments
Comment 1: The overall purpose of the article is stated clearly in the introduction and underlined in the abstract.
Response 1: Thank you for your suggestion. I modified parts of the content in the Abstract.
Comment 2: The Literature Review part is logical and well organized. However, I think it is preferable to mention the first author in the text. It would be especially preferred in those places where the sentence starts with the number of the publication you want to cite, e.g. line 133 [24], line 134 [26] or line 138 [27]. Please emphasize what is the knowledge gap.
Response 2: Thank you for your suggestion. I modified the reference of whole paper. And highlight the knowledge gap in second paragraph “Indeed, many people devote to recognize the correlation between the built environment and urban rail system/ bicycle sharing system ridership or the performance of integrated urban rail and bicycle sharing system [1, 12-15]. Nevertheless, few identify this relationship from the dock-less bicycle sharing trajectory, in which bicycle sharing is an intermediate media to connect urban rail stations and land use patterns. We would accurately understand travel behaviour and demand regarding the trip chain between urban rail and bicycle sharing.”
Comment 3: The Methodology part is clear. I am not sure if the maps shown at Figures are visible as they are so small.
Response 3: I enlarged the whole paper's figure and modified some figures, such as Figures 4 – 11.
Comment 4: The date data in Table 1 is unclear
Response 4: Thank you for your recommendation. I modified the date data in Table 1 to “21/12/2020
”.
Comment 5: Figure 12 needs further explanations in its title. Perhaps adding a,b,c and d elements to it would allow to sort out the differences between the presented graphs.
Response 5: Thank you for your recommendation. I modified the title to “Variable Importance plot for Ingress and Egress ridership”, and add the minor number as “a, b, c, d”.
Comment 6: According to the graphs in Figure 13 and Figure 14: the units should be added on each axis.
Response 6: Thank you for your recommendation. I added the unites to each axis.
Comment 7: The conclusions are stated clearly and are supported by research results
Response 7: Thank you for your recommendation. I added more explanation and discussion in “5.1 Result of trip chain” to support the conclusion.
Comment 8: Future research directions and the significance of the results of the research achieved should be underlined and explained in conclusion part.
Response 8: Thank you for your recommendation. I modified the “future research” that explains why need do that, and revised some descriptions.
Author Response File: 
Author Response.docx
Reviewer 2 Report
Major Comment
1. To provide a better understanding of methods used in the study, you need to explain the method with the data. Just describing the method (in the current form) does not help the readers to understand your methods. A clear explanation of the method could boost the reproducibility and replicability of your study. Not only that, you also could get some citations (if you kindly write your methods to help the readers understand). So, please more gently write the method parts.
2. Is there any reason to select only four urban rail station for your analysis? Compared to other station, it seems that there are only few POIs nearby Tangbian stations. If you would like to say "you selected four stations having many nearby POIs, you may need to re-select the stations.
3. Is the bike dock-less? Or it requires the dock? Such characteristic may be responsible for different patterns.
4. How you could draw cycling trajectory (Figure 7)? How the trajectory has the same name as rail station? I couldn't understand it.
5. It is very challenging to identify the purposes of the trip through the data. Only counting the number of cyclists in each land use pattern may not be identical to the place the cyclists exactly go. Is there any thought on the validation?
6. You may need to provide more explanations of the main purposes of using random forest in your analysis. That parts look like a part of different paper. It does not seem to well connect to the previous sections.
7. You also showed the different patterns of ridership over time, however, there is no any analysis on that differences.
Minor Comment
1. The way of citing the reference.
-e.g., Cheng [27] analyzed travel mode choice ~~ (line 138)
Author Response
Response to Reviewer 2 Comments
Comment 1: To provide a better understanding of methods used in the study, you need to explain the method with the data. Just describing the method (in the current form) does not help the readers to understand your methods. A clear explanation of the method could boost the reproducibility and replicability of your study. Not only that, you also could get some citations (if you kindly write your methods to help the readers understand). So, please more gently write the method parts.
Response 1: Thank you for your suggestion. I added more explains in Methodology to explain why I select RF model, and how to apply it. I supplemented Figure 2 to describe “Flow diagram of random forest modeling methodology”, and “This study divided the dataset into test set and train set to avoid probable model overfitting because test set should not be used for model development [23]. Therefore, 30% of dataset was randomly selected as the test set, and the remaining data is train set. Firstly, this study establishes the predicting mode by bootstrap aggregating to ensure the model is suitable for predicting urban rail ridership from land use pattern. Then, this study determines the variable importance due to OOB estimates of all decision trees in the forest can be averaged to obtain the generalized error estimate of the RF model [26]. That further recognizes the critical variable of land use pattern to urban rail ridership”.
Comment 2: Is there any reason to select only four urban rail station for your analysis? Compared to other station, it seems that there are only few POIs nearby Tangbian stations. If you would like to say "you selected four stations having many nearby POIs, you may need to re-select the stations
Response 2: Thank you for your suggestion. Due to data limitations, we merely obtained four urban rail station ridership data, hence, we only analyzed these four stations. This content I supplemented in paper “Due to data limitations, we merely obtained four urban rail station ridership data, which these stations are located on the Xiamen mainland (Figure 6). Hence, this paper selects them as specific analysis stations to identify this correlation during the morning rush hour (6-10). These four stations are located in the centre of Xiamen mainland (Wushipu, Lvcuo, Tangbian, and Jiangtou station).”
Comment 3: Is the bike dock-less? Or it requires the dock? Such characteristic may be responsible for different patterns.
Response 3: Thank you for your suggestion. I supplemented this description in paper, especially in the Abstract. “Abstract: Urban rail systems as rapid mass transport has always been popular worldwide to relieve urban traffic in tight areas or reconstruct the urban structure. Land use characteristics are vital for this system and correlate with urban ridership. Dock-less bicycle-sharing extends the station service coverage range because it forms an integrated public transport system with an urban rail system that provides a convenient travel model. Hence, the land use pattern with dock-less bicycle-sharing is associated with urban rail ridership. This paper measures the correlation between land use and urban rail ridership according to the dock-less bicycle-sharing trajectory, which will precisely reflect the passenger travel behaviour of the trip chain.”
Comment 4: How you could draw cycling trajectory (Figure 7)? How the trajectory has the same name as rail station? I couldn't understand it.
Response 4: Thank you for your suggestion. I added the description of how to draw cycling trajectory in paper. “The dock-less bicycle sharing trajectory dataset has recorded the location of the cycler during the whole ride. Hence, there are more than two million bicycle sharing data per day. Meanwhile, there is a unique user ID for recognizing the riding route, which also contributes to us drawing the cycling route for everyone. We only need to connect the per point by time series to complete the cycling route.”
Comment 5: It is very challenging to identify the purposes of the trip through the data. Only counting the number of cyclists in each land use pattern may not be identical to the place the cyclists exactly go. Is there any thought on the validation?
Response 5: Thank you for your suggestion. Indeed, this is hugely challenging to validate these movements, as there is no uniform paying measure to reflect this transfer phenomenon, just like the Integrated metro-bus system. But this study considered cycling trips whose origin or destination lay within 100 meters of the urban rail station entrance were considered bicycle-metro transferring trips, which has been adopted by Wu, X (2019); Guo, Y. and S.Y. He (2020); Guo, Y., et al. (2021). Hence, this method possible is useful currently, but your question is worth further study, but I think it should be combinate with other measurements such as Street View. I also supplement this bias in “future research” that is “Forth, recognizing the transfer activity with integrated metro-bicycle sharing that should consider more details, because this study deems the urban rail station coverage range is 100m in this transfer system, but there is bias. Hence, future research could consider more from this aspect.”.
Comment 6: You may need to provide more explanations of the main purposes of using random forest in your analysis. That parts look like a part of different paper. It does not seem to well connect to the previous sections.
Response 6: Thank you for your suggestion. I supplemented more information to explain why I selected RF model in methodology (second paragraph). “Random forest model explores more refined associations between the outcome and explanatory variables [24], nonlinear relationships, multimodal data, categorical and numerical features, missing values, and tolerance of random variables can all be handled by an RF model [25, 26], as well it has good accuracy compared with [26, 27]. Yifan Wen applied high-density traffic monitoring data and land use data to train random forest models that accurately predict dynamic, link-level vehicle emissions [23]. Long Cheng examined built environment effects on the elderly’s walking by random forests, which found that land use mix only increases older adults’ walking at certain levels [24]. They also found random forest model outperforms linear regression.”
Comment 7: You also showed the different patterns of ridership over time, however, there is no any analysis on that differences.
Response 7: Thank you for your suggestion. I added more analysis for time series changing of bicycle sharing usage in “5.1 Result of trip chain” to support the conclusion.
Comment 8: The way of citing the reference. -e.g., Cheng [27] analyzed travel mode choice ~~ (line 138)
Response 8: Thank you for your suggestion. I modified the reference in the whole paper.
Author Response File: Author Response.docx
Reviewer 3 Report
The paper introduces an interesting application on geographical data based on the correlation between land use and trajectories traveled by bicycles. The interest in the theme, in addition to the aspects properly related to sustainable mobility, concerns the analysis of geographical data for the study of social behavior. Surely some aspects of the research could be more in-depth as well as the expansion of the sample taken into analysis, but the research presents ideas of originality that could find further fields of application.
A spelling revision of some parts of the text is also required.
Author Response
Response to Reviewer 3 Comments
Comment 1: The paper introduces an interesting application on geographical data based on the correlation between land use and trajectories traveled by bicycles. The interest in the theme, in addition to the aspects properly related to sustainable mobility, concerns the analysis of geographical data for the study of social behavior. Surely some aspects of the research could be more in-depth as well as the expansion of the sample taken into analysis, but the research presents ideas of originality that could find further fields of application.
Response 1: Thank you for your suggestion. I modified this paper a lot. (1) Highlight the knowledge gap in Literature Review (second paragraph). (2) Improve the Figures. (3) Added more explanations in Methodology to explain why I selected the RF model and how to apply it. (4) Added more analysis for time series changing of bicycle sharing usage in “5.1 Result of trip chain” to support conclusion. (5) Modified the “future research” to provide more explain why need does that, and revised some describes.
Comment 2: A spelling revision of some parts of the text is also required.
Response 2: Thank you for your suggestion. I improved the English language of the whole paper.
Author Response File: Author Response.docx
Round 2
Reviewer 2 Report
Thank you for your revision.
I have the follow-up question of my first comment in the previous revision. I am still not that satisfied with your explanations about the method. You may be able to elaborate the explanations about Section 3.1, 3.2, and 3.3 with the example of the data set.
Author Response
Comment 1: I have the follow-up question of my first comment in the previous revision. I am still not that satisfied with your explanations about the method. You may be able to elaborate the explanations about Section 3.1, 3.2, and 3.3 with the example of the data set.
Response 1: Thank you for your suggestion. I re-structured the “3. Methodology”, I first described the “study area and data”, then I supplemented the correlation dataset or part of the dataset in the per Section. As well, I revised the “5.2 Result of this correlation” to delete some repeat content with “3.1 Method”.
Author Response File: 
Author Response.pdf
