Analysis of Environmental Factors on Intersection Accidents

Round 1

Reviewer 1 Report

1. Examples (e.g. Aerial Images) for Intersections within your sample, where one or more of the variables (Gas Station/Side Walk etc) are true might be of interest. Intersection design varies from country to country. Interested people might want to know, if your findings could be transferred to other locations.

2. Please provide more information about the data that you have been using (data source). Is it available to the public? Can other researchers reinvestige on the data?

3. The link between formulas of the general ZIP (1),(2) and (3),(4) or ZINB and the variable values in Tables 3 an 4 could be clearer. ZIP and ZINB can be found in litereture. However, your very approach is what the reader is interested. How do you apply ZIP to your problem. What is the meaning of your variables. Consider including the formula of your model on the paper.

4. There might be hidden influencing factors behind your data. Expecially those intersections with >30 Accidents/year might be high accident risk spots that are already experience treatment by accident commissions. A closer look on those is desirable. It might be interesting to see what happens if those extreme ones are included/excluded from the sample

 

Author Response

I deeply appreciate the editor’s concern for revision preparation. In addition, I also thank for the valuable comments of the referees, which helped us to clarify several aspects of the presentation. In the revision, I addressed all issues raised in the comments as following in the revision paper.

 

Comments and Suggestions for Authors

Reviewer1

1. Examples (e.g. Aerial Images) for Intersections within your sample, where one or more of the variables (Gas Station/Side Walk etc) are true might be of interest. Intersection design varies from country to country. Interested people might want to know, if your findings could be transferred to other locations.

Response: Thank you for the reviewer’s valuable comment. These variables may be location specific or directly transferrable in this study. Other locations can define their own variables of interest and conduct similar analysis.

2. Please provide more information about the data that you have been using (data source). Is it available to the public? Can other researchers reinvestige on the data?

Response: Thank you for the reviewer’s valuable comment. The data are acquired from police authority in a government-funded research in order to lower the risk of accidents on roadways in Taiwan and these data are not available to the public and a new permission must be obtained for new researches.

3. The link between formulas of the general ZIP (1),(2) and (3),(4) or ZINB and the variable values in Tables 3 an 4 could be clearer. ZIP and ZINB can be found in litereture. However, your very approach is what the reader is interested. How do you apply ZIP to your problem. What is the meaning of your variables. Consider including the formula of your model on the paper.

Response: Thank you for the reviewer’s valuable comment. We have revised according to your suggestions if appropriate.

4. There might be hidden influencing factors behind your data. Expecially those intersections with >30 Accidents/year might be high accident risk spots that are already experience treatment by accident commissions. A closer look on those is desirable. It might be interesting to see what happens if those extreme ones are included/excluded from the sample

Response: Thank you for the reviewer’s valuable comment. An intersection with >30 accidents per year is high risk, no matter there were previous treatments by accident commissions or not. We still take all relevant factors of these intersections into considerations. The percentage of intersections with high risk is less than 1% so probably no big change will be seen if excluding these intersections.

Reviewer 2 Report

The main aim of this manuscript is to analyze the causes of traffic accidents at 3-way (T- or Y-type) intersections in Taiwan. Using a GIS system, authors analyzed intersections and collected different types of data from accidents (only fatal and accidents with injuries were included), traffic volume, presence of objects nearby, parking, road characteristics (width of the road, width of the lanes) etc. Also, the authors used the zero-inflated model to analyze and incorporate the characteristics of the probability distribution between Poisson and the negative binomial, and then construct zero-inflated Poisson and zero-inflated negative binomial models. Although I respect the effort made by the authors, the overall quality of the paper in its current form is not sufficient. There is a number of issues that, in my opinion, must be improved.

 

More detailed comments are below:

1. The abstract should be improved. Authors should provide an introduction and background (1-2 sentences), then the aim of the study, methodology (2-3 sentences), results, and discussion (2-3 sentences). In current form, the form of the abstract is “mixed”. First authors state what is the aim, then the background, then again what is the focus of the study, and then again the background (number of accidents on 3-way intersections).
2. I’m not sure if I fully understand this sentence: “However, for 68% of the three-way intersections surveyed, the number of traffic accidents recorded during the observation period in 0, thus using the traditional Poisson and negative binomial counting modes may easily result in estimation errors.” – do authors mean that the number of accidents was zero or?
3. The introduction section is poorly written. Authors should give readers an introduction and background, state what is the problem related to the topic they are studying, what are the previous findings and how did previous studies addressed this issue, identify the gaps in the available literature and present the aim of this study. Section 2 should be a part of the Introduction and authors should elaborate and describe the findings of previous literature. Also, most of the references are 10+ years old. Although they are for sure valuable, authors should also analyze more recent work.
4. Sentences like this “Previous research on traffic accident factors in Taiwan has focused on the impact of driver characteristics on accidents.”, should be backed with references. Also, along with my previous comment, authors should look at literature outside Taiwan and see what has been and how it has been studied.
5. I would suggest that authors first present and describe their data set and then elaborate on why they used specific methodology. In the current form, the authors first present the methodology and readers do not even know the data set and is there a lot of 0 recordings or not.
6. In a row, 140 authors state that the data set was collected on provincial highways, however, in abstract, it is written rural roads. I’m not familiar with the road classification system in Taiwan, but highways are usually not rural roads. Authors should clearly describe the road category analyzed in the study.
7. Could authors provide more data about the roads, i.e. what is the length of road network analyzed etc. The number of intersections is huge so again I’m confused what the authors calculated, which types of roads were included, what is the length of the road network etc. More data would be beneficial to the reader in understanding the scope of the dataset since not all the readers are familiar with the characteristics of road network in Taiwan.
8. In Table 1 it would be better that authors write “Number of accidents which resulted in injuries or fatalities”. Also, why did the authors divide intersections by the number of accidents? Would results be different if the classification was not used or can maybe some categories be joined? The authors should explain what is the reason for such an approach and how does this approach differs from the literature.
9. Table 2 – what is “near”? What distance from intersection was defined as a threshold after which the presence of the object would not be defined as “near”? Furthermore, when presenting the width values authors should provide the unit.
10. The result section must be improved. Several huge tables one after another extremely complicate the reading for the reader and one must go back and forward to figure out the results. This must be improved, and authors should present the results in a more fluid way.
11. Discussion and Conclusion also must be improved. Authors should discuss the results, give their opinion on why the results point in a specific direction, how do their results “connect” with previous studies etc. The authors should also present the main limitations of this study. The conclusion should provide some guidance, general conclusions such as “reduction of speed” are well known and do not bring anything new. But maybe proposing what speed limit is proper for specific situations and intersections which for example have some objects near it would be beneficial.

Author Response

Response to Reviewer 2 Comments

I deeply appreciate the editor’s concern for revision preparation. In addition, I also thank for the valuable comments of the referees, which helped us to clarify several aspects of the presentation. In the revision, I addressed all issues raised in the comments as following in the revision paper.

 

Comments and Suggestions for Authors

Reviewer2

The main aim of this manuscript is to analyze the causes of traffic accidents at 3-way (T- or Y-type) intersections in Taiwan. Using a GIS system, authors analyzed intersections and collected different types of data from accidents (only fatal and accidents with injuries were included), traffic volume, presence of objects nearby, parking, road characteristics (width of the road, width of the lanes) etc. Also, the authors used the zero-inflated model to analyze and incorporate the characteristics of the probability distribution between Poisson and the negative binomial, and then construct zero-inflated Poisson and zero-inflated negative binomial models. Although I respect the effort made by the authors, the overall quality of the paper in its current form is not sufficient. There is a number of issues that, in my opinion, must be improved.

More detailed comments are below:

1. The abstract should be improved. Authors should provide an introduction and background (1-2 sentences), then the aim of the study, methodology (2-3 sentences), results, and discussion (2-3 sentences). In current form, the form of the abstract is “mixed”. First authors state what is the aim, then the background, then again what is the focus of the study, and then again the background (number of accidents on 3-way intersections).

Response: Thank you for the reviewer’s valuable comment. We have rewritten the abstract according to your suggestions.

 

2. I’m not sure if I fully understand this sentence: “However, for 68% of the three-way intersections surveyed, the number of traffic accidents recorded during the observation period in 0, thus using the traditional Poisson and negative binomial counting modes may easily result in estimation errors.” – do authors mean that the number of accidents was zero or?

Response: Yes.

 

3. The introduction section is poorly written. Authors should give readers an introduction and background, state what is the problem related to the topic they are studying, what are the previous findings and how did previous studies addressed this issue, identify the gaps in the available literature and present the aim of this study. Section 2 should be a part of the Introduction and authors should elaborate and describe the findings of previous literature. Also, most of the references are 10+ years old. Although they are for sure valuable, authors should also analyze more recent work.

Response: Thank you for the reviewer’s valuable comment. The main purpose of this study is to identify factors contributing to the happening of accidents and we have conducted extensive literature reviews and only most relevant publications can be included.

 

4. Sentences like this “Previous research on traffic accident factors in Taiwan has focused on the impact of driver characteristics on accidents.”, should be backed with references. Also, along with my previous comment, authors should look at literature outside Taiwan and see what has been and how it has been studied.

Response: Thank you for the reviewer’s valuable comment. We have included more literatures if we see fit, thanks for your suggestion.

 

5. I would suggest that authors first present and describe their data set and then elaborate on why they used specific methodology. In the current form, the authors first present the methodology and readers do not even know the data set and is there a lot of 0 recordings or not.

Response: Thank you for the reviewer’s valuable comment. We will revise as you suggested.

 

6. In a row, 140 authors state that the data set was collected on provincial highways, however, in abstract, it is written rural roads. I’m not familiar with the road classification system in Taiwan, but highways are usually not rural roads. Authors should clearly describe the road category analyzed in the study.

Response: Thank you for the reviewer’s valuable comment. Thanks for corrections. Provincial roadways are subjects of study and they constitute the backbone of road network in Taiwan. Some of them are intercity roadways, some are urban corridors and some are rural roads. There is also highway systems in Taiwan. Sorry for this confusion.

 

7. Could authors provide more data about the roads, i.e. what is the length of road network analyzed etc. The number of intersections is huge so again I’m confused what the authors calculated, which types of roads were included, what is the length of the road network etc. More data would be beneficial to the reader in understanding the scope of the dataset since not all the readers are familiar with the characteristics of road network in Taiwan.

Response: Thank you for the reviewer’s valuable comment. We have added relevant information in background.

 

8. In Table 1 it would be better that authors write “Number of accidents which resulted in injuries or fatalities”. Also, why did the authors divide intersections by the number of accidents? Would results be different if the classification was not used or can maybe some categories be joined? The authors should explain what is the reason for such an approach and how does this approach differs from the literature.

Response: Thank you for the reviewer’s valuable comment. This classification is for descriptions to readers instead for analysis. We did not conduct analysis on this classification. Sorry for this confusion.

 

9. Table 2 – what is “near”? What distance from intersection was defined as a threshold after which the presence of the object would not be defined as “near”? Furthermore, when presenting the width values authors should provide the unit.

Response: Thank you for the reviewer’s valuable comment. After carefully considering the influence of various objects on intersection safety, we define “near” as 300 meters in radius.

 

10. The result section must be improved. Several huge tables one after another extremely complicate the reading for the reader and one must go back and forward to figure out the results. This must be improved, and authors should present the results in a more fluid way.

Response: Thank you for the reviewer’s valuable comment. We have re-arrange tables in a way easier to understand.

 

11. Discussion and Conclusion also must be improved. Authors should discuss the results, give their opinion on why the results point in a specific direction, how do their results “connect” with previous studies etc. The authors should also present the main limitations of this study. The conclusion should provide some guidance, general conclusions such as “reduction of speed” are well known and do not bring anything new. But maybe proposing what speed limit is proper for specific situations and intersections which for example have some objects near it would be beneficial.

Response: Thank you for the reviewer’s valuable comment. We have elaborate our experiment results and propose our suggestions.

Reviewer 3 Report

Title of the Paper:

Analysis of environmental factors on intersection accidents

General observations: The paper analyzes the causes of traffic accidents at 3-way intersections in Taiwan to reduce the probability of the accidents.

• There are some English grammatical error and typos that need careful editing: for instance “traffic accidents has been “ should be “traffic accidents have been “
• “Previous research on traffic……….characteristics on accidents. “ The authors may provide references.
• “This study collected ….Geographic Information Systems to conduct spatial analysis”… The authors may provide the types of data collected and types of spatial analysis (for instance distances, shapes etc.) carried out.
• “Based on a review of the relevant literature, …….negative binomial” The authors may provide references.
• In section 2. Variables of Accident-prone location looks limited …the authors may provide a comprehensive list of variables. (Author may refer to Table 2 for Variables used in the analysis of accident data in “The Effect of Drivers' Demographic Characteristics on Road Accidents in Different Seasons using Data Mining” by Shokouhyar,S et al.(2017) etc…
• Geographic Information Systems (GIS) may be used in the manuscript for repetitive use in a manuscript.
• “This study uses Geographic Information Systems……….as summarized below.” Authors may replace below with Table 2 and replace Geographic Information Systems with GIS
• Authors may use the abbreviation for ZIP and ZINB for Zero-Inflated Poisson and Zero-Inflated Negative Binomial models to avoid distraction in readings.
• In Table 2 of Summary of selected variables, under descriptive statistics, a more detailed caption (like no of accidents/percentage/units etc.) is required to provide clarity of data provided. Some data like 'Road width' may need the units of meter.
• “However, an excessive …………..above two models.” Authors may replace ‘above two’ with ZIP and ZINB
• Authors may provide a reference for  “Smaller AIC and BIC index values mean that the selected model fits the original data well”
• Authors may report their findings for a variable like speed limit, fast lane width, sidewalk width on accident probability and corresponding remedies in the discussion section.
• “The results of this research ….prior findings, ……of significant variables”. Authors may provide references.
• “The present findings can ……… improve intersection design and safety”. The authors need to provide more details on their findings to emphasize the research contribution.

Author Response

Comments and Suggestions for Authors

Reviewer 3

Analysis of environmental factors on intersection accidents

General observations: The paper analyzes the causes of traffic accidents at 3-way intersections in Taiwan to reduce the probability of the accidents.

• There are some English grammatical error and typos that need careful editing: for instance “traffic accidents has been “ should be “traffic accidents have been “
• “Previous research on traffic……….characteristics on accidents. “ The authors may provide references.
• “This study collected ….Geographic Information Systems to conduct spatial analysis”… The authors may provide the types of data collected and types of spatial analysis (for instance distances, shapes etc.) carried out.
• “Based on a review of the relevant literature, …….negative binomial” The authors may provide references.
• In section 2. Variables of Accident-prone location looks limited …the authors may provide a comprehensive list of variables. (Author may refer to Table 2 for Variables used in the analysis of accident data in “The Effect of Drivers' Demographic Characteristics on Road Accidents in Different Seasons using Data Mining” by Shokouhyar,S et al.(2017) etc…
• Geographic Information Systems (GIS) may be used in the manuscript for repetitive use in a manuscript.
• “This study uses Geographic Information Systems……….as summarized below.” Authors may replace below with Table 2 and replace Geographic Information Systems with GIS
• Authors may use the abbreviation for ZIP and ZINB for Zero-Inflated Poisson and Zero-Inflated Negative Binomial models to avoid distraction in readings.
• In Table 2 of Summary of selected variables, under descriptive statistics, a more detailed caption (like no of accidents/percentage/units etc.) is required to provide clarity of data provided. Some data like 'Road width' may need the units of meter.
• “However, an excessive …………..above two models.” Authors may replace ‘above two’ with ZIP and ZINB
• Authors may provide a reference for  “Smaller AIC and BIC index values mean that the selected model fits the original data well”
• Authors may report their findings for a variable like speed limit, fast lane width, sidewalk width on accident probability and corresponding remedies in the discussion section.
• “The results of this research ….prior findings, ……of significant variables”. Authors may provide references.
• “The present findings can ……… improve intersection design and safety”. The authors need to provide more details on their findings to emphasize the research contribution.

 

Response: Thank you for the reviewer’s valuable comment. As suggested, the authors have corrected these grammatical errors and reproof the entire article. Moreover, the authors have sent our manuscript to a native English speaker for proofreading and editing. The authors are very appreciative for the reviewer’s efforts.

Round 2

Reviewer 2 Report

After reading the revised version of the manuscript I have to conclude that the authors did not improve the paper to the level where it can be accepted (at least in my opinion). There are still many issues with the paper:

1. Abstract is still confusing and now even too long. The abstract should be between 150-200 words, not +300. Also, in the abstract authors should first present the background and state what is the problem. Then, describe the aim of the study and methodology, followed by results and a short conclusion/discussion.
2. To my second comment authors just replied "Yes". If I did not understand the sentence properly probably other readers will also get confused. This means authors should rephrase the sentence to make it clearer.
3. The introduction is still in my opinion poorly written. First, although I understand the study is focused on Taiwan's database, authors should at least describe what are the findings of other studies around the world and then connect these findings with the current state in Taiwan to elaborate on why their research is important. Overall, the majority of the Introduction is related to the analysis methods and not on the topic of the study. The authors should include section 2.2. into the Introduction and elaborate on the literature.
4. My fourth comment was not taken into account and I do not see where the authors included additional literature in the Introduction - as stated in the previous comment, the authors should include section 2.2. into the Introduction and elaborate on the literature.
5. Section 2 in my opinion should not have a literature review and should be only based on the explanation of the objectives and methodology. Again, the whole Section is confusing and without any "flow". The authors should clearly explain what is the objective of the study, what is their dataset, and methodology.
6. I understand better now what types of road were included, but authors should then use the same terminology - in row 160 authors state "Taiwan’s provincial highways", however in rows 63-66 they are explaining other roads. Please, harmonize the terminology.
7. Figure 1 still has a typo (Main RADS in city)
8. The authors replied to my ninth comment and explain that they define “near” as 300 meters in radius. However, in the manuscript, this is not explained. The idea of a review is not to satisfy me, but to improve the manuscript so it can be better understood by all other readers.
9. Is really maximum road width 88 m?
10. Unit is not "1 m", unit is "m".
11. Tables 3 and 4 are extremely difficult to read and it takes a lot of time for the reader to understand them. Authors should think from the reader's perspective who is not familiar with the work and needs to understand it in an easy way. Those long tables should be reduced in some ways, and/or elaborated through the text. Longer versions of the tables can be added to the Appendix if the reader wants to have deeper look.
12. The purpose of the Discussion is to critically elaborate on the obtained results and "compare" your results with previous findings. This is not present in the current Discussion. Also, the authors must present the main limitations of this study. 
13. In the conclusion it is written "The results of this research correspond well with prior findings, such as the significance and directionality of significant variables, that comparing the results with related studies in other countries." - where exactly did authors compare their results with previous findings? This should have been done in the Discussion section.
14. There are still many typos and English language problems. Although these are minor things, authors should improve the paper in that sense also.

Author Response

Dear Reviewer:

I would like to submit the revision, “Analysis of environmental factors on intersection accidents” (Sustainability-1523243), which is resubmitted to the sustainability. Please find enclosed the revised version.

I deeply appreciate the valuable comments of the referees, which helped us to clarify several aspects of the presentation. In the revision, I addressed all issues raised in the comments as following in the revision paper.

 

More detailed comments are below:

1. Abstract is still confusing and now even too long. The abstract should be between 150-200 words, not +300. Also, in the abstract authors should first present the background and state what is the problem. Then, describe the aim of the study and methodology, followed by results and a short conclusion/discussion.

Response: Thank you for the reviewer’s valuable comment. We deleted some not-absolutely-necessary description to shorten the abstract.

 

1. To my second comment authors just replied "Yes". If I did not understand the sentence properly probably other readers will also get confused. This means authors should rephrase the sentence to make it clearer.

Response: Thank you for the reviewer’s valuable comment. The fact that the number of accidents is zero now is mentioned in abstract.

 

1. The introduction is still in my opinion poorly written. First, although I understand the study is focused on Taiwan's database, authors should at least describe what are the findings of other studies around the world and then connect these findings with the current state in Taiwan to elaborate on why their research is important. Overall, the majority of the Introduction is related to the analysis methods and not on the topic of the study. The authors should include section 2.2. into the Introduction and elaborate on the literature.

Response: Thank you for the reviewer’s valuable comment. Previous researches in Taiwan are focused on different subjects and quite different from our approach so limited review on them is given in this paper.

 

1. My fourth comment was not taken into account and I do not see where the authors included additional literature in the Introduction - as stated in the previous comment, the authors should include section 2.2. into the Introduction and elaborate on the literature.

Response: Thank you for the reviewer’s valuable comment. Previous researches in Taiwan are focused on different subjects and quite different from our approach so limited review on them is given in this paper.

 

1. Section 2 in my opinion should not have a literature review and should be only based on the explanation of the objectives and methodology. Again, the whole Section is confusing and without any "flow". The authors should clearly explain what is the objective of the study, what is their dataset, and methodology.

Response: Thank you for the reviewer’s valuable comment. This section includes background of provincial roadway system in Taiwan and important factors concluded in previous researches and explains why these factors are considered in our analysis.

 

1. I understand better now what types of road were included, but authors should then use the same terminology - in row 160 authors state "Taiwan’s provincial highways", however in rows 63-66 they are explaining other roads. Please, harmonize the terminology.

Response: Thanks for your suggestion.

 

1. Figure 1 still has a typo (Main RADS in city)

Response: Thanks for your suggestion.

 

1. The authors replied to my ninth comment and explain that they define “near” as 300 meters in radius. However, in the manuscript, this is not explained. The idea of a review is not to satisfy me, but to improve the manuscript so it can be better understood by all other readers.

Response: Thanks for your suggestion. We explain this in Table 2.

 

1. Is really maximum road width 88 m?

Response: Thank you for the reviewer’s valuable comment. There is actually one roadway in Taiwan with a width 88 meters, it is located in Chia Yi County.

 

1. Unit is not "1 m", unit is "m".

Response: Thanks for your suggestion.

 

1. Tables 3 and 4 are extremely difficult to read and it takes a lot of time for the reader to understand them. Authors should think from the reader's perspective who is not familiar with the work and needs to understand it in an easy way. Those long tables should be reduced in some ways, and/or elaborated through the text. Longer versions of the tables can be added to the Appendix if the reader wants to have deeper look.

Response: Thank you for the reviewer’s valuable comment. Since the number of factors considered in this research are so many, it is difficult to make the table more readable without losing necessary details.

 

1. The purpose of the Discussion is to critically elaborate on the obtained results and "compare" your results with previous findings. This is not present in the current Discussion. Also, the authors must present the main limitations of this study. 

Response: Thank you for the reviewer’s valuable comment. We obtained results very similar to findings of other researches and this is mentioned in the concluding remarks. The main limitation is lack of traffic volumes at the time accidents happened. AADT was used to represent the traffic volume.

 

1. In the conclusion it is written "The results of this research correspond well with prior findings, such as the significance and directionality of significant variables, that comparing the results with related studies in other countries." - where exactly did authors compare their results with previous findings? This should have been done in the Discussion section.

Response: Thank you for the reviewer’s valuable comment. The factors considered in this research are mostly from the findings of previous researches and the results indicate the significance and directionality of these factors are very similar.

 

1. There are still many typos and English language problems. Although these are minor things, authors should improve the paper in that sense also.

Response: Thanks for your suggestion.

Author Response File: Author Response.pdf