Integration between Dockless Bike-Sharing and Buses: The Effect of Urban Road Network Characteristics
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsIn a world where the reduction of greenhouse gas emissions is increasingly urgent, the study of ways to promote bikeability assumes particular relevance. This paper studies the influence of urban road characteristics on DBS and bus integration. It covers several perspectives, considering an exhaustive list of indicators. There are only some small issues that can be improved:
- Given the wide range of the analysis, sometimes is difficult to follow the results. Therefore, I suggest that the results section be rearranged so that clear answers are given to the questions asked on page 2, lines 54-57.
- I think that including a correlation matrix would be useful. Indeed, section 4.3 mentions negative correlations.
- I think that Tables A.1 and A.2 should be incorporated into the text, to make reading easier.
Author Response
Comments 1: Given the wide range of the analysis, sometimes is difficult to follow the results. Therefore, I suggest that the results section be rearranged so that clear answers are given to the questions asked on page 2, lines 54-57.
Response 1: Thanks very much for your insightful comments. After we have clarified the research contents and results of this paper in detail, we have adjusted the expression of the research question in the Introduction so as to make it correspond to the results section more clearly. Please kindly see the revisions with the red font on page 2, line 54-58:
"It endeavors to address the following questions: 1) What are the temporal and spatial dynamic features of DBS-bus integration across different time periods? 2) What are the characteristics of urban road network and street patterns for cycling integrated with bus service in Tianjin? 3) Is there any correlation between this integration usage and road network characteristics?"
Comments 2: I think that including a correlation matrix would be useful. Indeed, section 4.3 mentions negative correlations.
Response 2: Thank for your helpful suggestions. We agree that the calculation of the correlation matrix before ZINB modelling will provide more information about the relationship between variables as well as improve the robustness of the results by mutual verification of different regression models. So we operate it for nine dependent variables and the results are shown in the attached PDF files, since it has a few pages with nine tables of correlation tables. It can be seen that the sign of the independent variables coefficients is mostly consistent with that of ZINB model, and collinearity between independent variables can ensure the accuracy of regression results. Please kindly refer to the PDF file named matrix.
Comments 3: I think that Tables A.1 and A.2 should be incorporated into the text, to make reading easier.
Response 3: Thank for your thoughtful comments and reminders. We have incorporated the tables of ZINB results into the text. Please kindly refer to it in Chapter 4.3.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsI am satisfied with the content of the paper.
Author Response
Comment 1: I am satisfied with the content of the paper.
Response 1: Thanks so much for your comment and kind approval.
Reviewer 3 Report
Comments and Suggestions for AuthorsThe paper deals with the analysis of deckles bike-sharing and bus transportation in Tianjin. The literature review is extensie and the paper covers the gap in studying bus and bike-sharing integration. Data are properly collected from bike-sharing companies. The zero-inflated negative binomial regression is used to evaluate the influence of independent date to frequency of acces and egress bike-sharing-bus integration. To improve the quality of the paper I address the following suggestions:
1. Lines 121-122. Conversing road structures to graph with nodes and edges is a very common assertion that is useless to be cited. Reference [25] in line 122 should be removed.
2. Paragraph 3.3.1 Steps 1-3: Some simple trigonometrical formula are used to establish the parking area - orientation, size, overlapping. A sugestive image containing all these words (applied on a complex bus stop with multiple overlapping) would be more eloquent and facilitates the reading.
3. Paragraph 3.3.1 Step 4: Acces integrated and egress integrated are defined, but total integrated is not defined.
4. Lines 281-283. As mentioned in the first suggestion, covering road network conversion to graphs, reference [46] is useless.
5. Line 298, after eq. 3, is missing nij(k) in explaining the equation variables. Only nij is mentioned.
6. Lines 310-311: Interpreting the values of betweenness centrality, traffic arterials is invoked. Betweenness centrality is a topological measure independent of traffic. Traffic in a transport model is allocated on routes in a specific way, considering O-D generation, modal spilt, trip costs. I suggest to reconsider the assertion.
7. Table 2. Variables Roads features in size: Is very confusing road density and the authors express it in km/km2 or %. See major road density, branch road density to secondary road density, residential road density. Please check table 2.
8. Figure 8: In text the authors mention fig 8(a), ... 8(e). Please add corresponding letters to figure 8 for correct identification.
9. Lines 577-578: The dense gridiron coefficient (Type 3) is 0.458, p=0.108 during weekday peak hour (Table A1). So it is not significant. The authors state it is significant and small. Please recheck.
10. Paragraph between lines 565-571: The interpretation of coefficients regarding intersections should be revised. For instance 3-leg intersection number and proportion are not significant in many models (Tables A1-A3). It would be very odd that the type of intersection could provide different coefficients or signs during the study periods of time/week-day. The assertion that three-leg intersections increases the time delay of cycling trips with traffic lights is available for any intersection controlled by traffic lights. The influence of intersection type to dependent variable has to be reshaped. It is not feasible how the configuration of the intersection could influence. There is a possibility for hidden/confounded variable to play a role associated too the intersection type (accidents rate, socio-economic data of persons traveling through, type of land use area etc). You also mentioned the possible effect of other variables in conclusion.
11. Paragraph between lines 628-634: Also the results are subject of miss-interpretation. Mathematically is correct, but why results concerning centrality are opposite for acces and egress? Also some other aspects (variables) could drive the users choice. Correlation does not means causality. A more refined interpretation has to be done.
12. În appendix A, Table A2 appears twice, instead of Table A3.
Author Response
Comments 1: Lines 121-122. Conversing road structures to graph with nodes and edges is a very common assertion that is useless to be cited. Reference [25] in line 122 should be removed.
Response 1: Thanks for your kind suggestions. We have removed the reference from the text.
Comments 2: Paragraph 3.3.1 Steps 1-3: Some simple trigonometrical formula are used to establish the parking area - orientation, size, overlapping. A suggestive image containing all these words (applied on a complex bus stop with multiple overlapping) would be more eloquent and facilitates the reading.
Response 2: Thanks for your insightful comments. In order to express the calculation process of the dependent variable more intuitively, we have drawn Figure 3 and put it in the latest manuscript, which corresponds the three steps with the graphic language. Please refer to it on page 7.
Comments 3: Paragraph 3.3.1 Step 4: Access integrated and egress integrated are defined, but total integrated is not defined.
Response 3: Thanks for your comments. The definition of total integrated usage has been added as requested in 3.3.1 Step 4. Please refer to it on page 7, line 266.
Comments 4: Lines 281-283. As mentioned in the first suggestion, covering road network conversion to graphs, reference [46] is useless.
Response 4: Thanks for your helpful suggestions. We agree that there is no need to cite the reference here so have removed it as requested.
Comments 5: Line 298, after eq. 3, is missing nij(k) in explaining the equation variables. Only nij is mentioned.
Response 5: Thanks for your kind reminder. The explanation of nij(k) in the equation has been added in the latest manuscript.
Comments 6: Lines 310-311: Interpreting the values of betweenness centrality, traffic arterials is invoked. Betweenness centrality is a topological measure independent of traffic. Traffic in a transport model is allocated on routes in a specific way, considering O-D generation, modal spilt, trip costs. I suggest to reconsider the assertion.
Response 6: Thanks very much for your insightful comments. We agree that it is not proper to interpret the topological network betweenness centrality with the concept of traffic since it has more to do with dynamic transport model. So we adjusted the interpretation here to assure it's more rigorous. Please kindly refer to the red font on page 8, line 313-316:
"A higher value of network betweenness centrality indicates that there is a greater difference in point betweenness centrality among all individual roads. Typically, the Lollipops and cul-de-sac network tend to have the higher network betweenness centrality than other street patterns, showing lower inter-connectivity and accessibility."
Comments 7: Table 2. Variables Roads features in size: Is very confusing road density and the authors express it in km/km2 or %. See major road density, branch road density to secondary road density, residential road density. Please check table 2.
Response 7: Thanks for your comments. We have modified the corresponding contents in Table 2, unifying the original % to km/km2. Please check the table in the latest manuscript.
Comments 8: Figure 8: In text the authors mention fig 8(a), ... 8(e). Please add corresponding letters to figure 8 for correct identification.
Response 8: Thanks for your kind reminders. Since we also have added a new figure in the manuscript, the corresponding letters have been added in figure 9 from fig. 9(a) to fig. 9(e).
Comments 9: Lines 577-578: The dense gridiron coefficient (Type 3) is 0.458, p=0.108 during weekday peak hour (Table A1). So it is not significant. The authors state it is significant and small. Please recheck.
Response 9: Thanks for your insightful comments. We have rechecked and adjusted the conclusion here according to its regression results. Please refer to it on page 17, line 581-585 as follows:
"Among these, the coefficient is higher in the two sparse street patterns. This indicates that the larger block scale is somewhat unfriendly to walking, thereby encouraging more cycling integration behavior. In dense grid-like road networks, however, there is no significant increase in the number of people choosing to ride during periods when travel is less urgent."
Comments 10: Paragraph between lines 565-571: The interpretation of coefficients regarding intersections should be revised. For instance 3-leg intersection number and proportion are not significant in many models (Tables A1-A3). It would be very odd that the type of intersection could provide different coefficients or signs during the study periods of time/week-day. The assertion that three-leg intersections increases the time delay of cycling trips with traffic lights is available for any intersection controlled by traffic lights. The influence of intersection type to dependent variable has to be reshaped. It is not feasible how the configuration of the intersection could influence. There is a possibility for hidden/confounded variable to play a role associated too the intersection type (accidents rate, socio-economic data of persons traveling through, type of land use area etc). You also mentioned the possible effect of other variables in conclusion.
Response 10: Thanks very much for your insightful comments. We agree the interpretation of conclusions about road intersections should be optimized. Besides, we hope to clarify that the assertion here in chapter 4.3.1 is concerning on the effects on DBS-bus total-integrated usage, so the proportion of 3-leg intersections are significant during two of the three time periods. Except for the common knowledge that is available for any intersections, we also extend the conclusion to multiple geometric intersection types. Please refer to the red font on page 17, line 568-575 as follows:
"In particular, the increasing of its proportion would greatly reduce the frequency. Similarly, the proportion of 3-leg intersections is also negatively correlated on weekday off-peak hours and weekends. On one hand, this is consistent with previous research to some extent that verifies intersections typically hinder the cycling behavior by increasing the time delay of cycling trips with traffic lights equipped at street corner [17,58,59]. On another hand, it extends the conclusion to multi-model intersections. The negative effect of cul-de-sac usually tends to be greater than that of 3-leg intersections according to their coefficients’ performance."
Comments 11: Paragraph between lines 628-634: Also the results are subject of miss-interpretation. Mathematically is correct, but why results concerning centrality are opposite for access and egress? Also some other aspects (variables) could drive the users choice. Correlation does not means causality. A more refined interpretation has to be done.
Response 11: Thanks for your kind comments and reminders. It is truly improper to use the word "opposite" since we do not expect to express opposite sign of the coefficients. What it means is that there are some differences of the regression results when it comes to the egress integration situation from that of access integration. And we also added more detailed explanation and optimized the interpretation. Please refer to it on page 18, line 633-639:
"Primarily, the centrality of the road network almost does not affect people's decision whether to ride to the bus stop. But the integration preference has more to do with it in the situation departing from the bus station. This means that the demand for dockless bike-sharing integration in a bus catchment area is closely related to its surrounding road topological structure. If the road connectivity around the bus stop is poor or the hierarchical structure is simple, people are more willing to cycle to their destination after getting off at the bus stop."
Comments 12: In appendix A, Table A2 appears twice, instead of Table A3.
Response 12: Thanks for your helpful reminder. The title has been corrected in the latest manuscript as "Table 6. Modeling results of ZINB analysis for total integrated use, egress integrated use and access integrated use (weekend)", which also took into account another reviewer's suggestion to put the table into the text.
