A Review of Surrogate Safety Measures Uses in Historical Crash Investigations

Round 1

Reviewer 1 Report

This is an interesting review of surrogate safety measures used in historical crash investigations. This is a very thorough synthesis of surrogate safety measures; however, it does not introduce any new research. It also needs to be reviewed for English language content and for consistency throughout the document.

Specific comments on the paper include:

Line 8 - What is meant by "...crash data consists the main indicator..."? Should this be "...crash data is the main indicator..."? or "remains"? As it is written it does not make any sense.

Line 9 - What is meant by "exploiting Surrogate Safety Measures"? I am unable to determine what this means.

Line 29 - This sentence needs to be rewritten, it does not make sense. I would suggest something like "...which aims to reduce road fatalities and injuries by half and provide sustainable..."

Line 37 - Add "for" before "...reliable estimates of..."

Line 45 - Not all agencies have "Road Traffic Police." I think this could be "law enforcement agencies" or "police departments"

Line 48 - What is meant by "...crash records consist a reactive approach..."? This does not make sense.

Line 110 - Define the acronym for PRISMA on first use rather than later in Line 118.

Line 138 - Consider including introductory paragraphs for all main sections that then introduce subsections.

Line 147 - Is an appendix appropriate for a journal paper? I have not seen this before.

Line 174 - What is meant by "a quite similar alternative to smartphone data"?

Line 185 - Rewrite as "...through field observations are another option..."

Line 196 - Rewrite as "...(MTTC) proposed by Ozbay et al. [71] that takes into account..."

Line 201 - Remove "kind of" at the beginning of this line.

Line 204 - Replace "to" with "of" to say "...to the evolution of next generation..."

Line 280 - Capitalize "Approximation" in the acronym definition.

Line 311 - You have already defined Deep Learning as "DL" - define acronyms on first use and then use the acronym consistently throughout the remainder of the document. You have several other acronyms that are redefined.

Line 330 - Rewrite this as "...in this section, the authors attempt to shed light on this issue..."

Line 344 - For Figure 2 and Figure 3 it seems to be more applicable to list the references by their number in the References section. That would be easier to track down by the reader.

Line 351 - In the discussion for Figures 2 and 3 it is unclear to me what the temporal dimensions really mean. What/how is time period measured and interpreted? This section needs more discussion.

Line 416 - Machine learning is already defined as ML.

Line 430 - Surrogate safety measures are already defined as SSMs.

Line 435 - Rewrite as "However, there are a small number of studies that focus on serious..."

Line 463 - ML is already defined. Also, on this line, rewrite as "Over the last few years, ML models have proven to be very efficient prediction tools..."

Overall the paper needs more detail on the SSMS (the information shows up in the table in the appendix but it is important information and should not just be in an appendix. You also need to provide more information on the time component and what that means and how it can be interpreted. The biggest shortcoming of the paper; however, is the lack of new research. It is a good synthesis but not a research paper.

The English language quality needs some work. Specific suggestions were provided in the previous section. Be sure to have it reviewed for grammatical issues to clean up the remaining questions.

Author Response

Response to reviewers’ comments

Reviewers’ comments appear in bold; authors’ comments follow in regular type. In the submitted manuscript, all changes/additions to the text have been marked using track changes.

 

Reviewer #1:

This is an interesting review of surrogate safety measures used in historical crash investigations. This is a very thorough synthesis of surrogate safety measures; however, it does not introduce any new research. It also needs to be reviewed for English language content and for consistency throughout the document.

 

Response:

Thank you very much for your review constructive input for our research work. We strongly believe that based on your suggestions and our respective changes in the manuscript, its quality has been improved significantly.

 

Specific comments on the paper include:

 

Line 8 - What is meant by "...crash data consists the main indicator..."? Should this be "...crash data is the main indicator..."? or "remains"? As it is written it does not make any sense.

 

Response:

Thank you for this remark. We have replaced “Historical road crash data consist the main…” with “Historical road crash data are the main…” (page 1, line 8).

 

Line 9 - What is meant by "exploiting Surrogate Safety Measures"? I am unable to determine what this means.

 

Response:

Exploiting refers to the practice of using and taking advantage of these indicators.

 

Line 29 - This sentence needs to be rewritten, it does not make sense. I would suggest something like "...which aims to reduce road fatalities and injuries by half and provide sustainable..."

 

Response:

Thank you very much for this comment. We have rephrased it: “Improving road safety is also included as a key component of the United Nations’ Agenda, as manifested by Sustainable Development Goals (SGDs) 3.6 and 11.2, which aim to reduce road fatalities and injuries by half and provide sustainable…” (page 1, lines 27-31).

 

Line 37 - Add "for" before "...reliable estimates of..."

 

Response:

Thank you. We have added it (page 1, line 37).

 

Line 45 - Not all agencies have "Road Traffic Police." I think this could be "law enforcement agencies" or "police departments"

 

Response:

Thank you for this remark. We have replaced it with “Police Departments” (page 2, line 45).

 

Line 48 - What is meant by "...crash records consist a reactive approach..."? This does not make sense.

 

Response:

Thank you for this remark. We have replaced “consist” with “are”. As we also explain in lines 49-50, this is a reactive approach as it forces road safety analysts to wait for road crashes to occur.

 

Line 110 - Define the acronym for PRISMA on first use rather than later in Line 118.

 

Response:

Thank you very much for this comment. In the revised version of the paper, we define the PRISMA acronym on first use (page 3, line 110).

 

Line 138 - Consider including introductory paragraphs for all main sections that then introduce subsections.

 

Response:

Thank you very much for your feedback. We can understand your point of view and appreciate your suggestion. However, introductory paragraphs are not mandatory based on the journal guidelines and we also believe that the current structure of the paper is effective in conveying the information and ideas we wish to present in a logical flow.

 

Line 147 - Is an appendix appropriate for a journal paper? I have not seen this before.

 

Response:

Yes. According to the guidelines of Sustainability, we can include an Appendix.

 

Line 185 - Rewrite as "...through field observations are another option..."

 

Response:

Thank you. We have changed it based on your comment (page 5, line 186).

 

Line 196 – Rewrite as “…(MTTC) proposed by Ozbay et al. [71] that takes into account…”

 

Response:

Thank you. We have rewritten this based on your remark (page 5, lines 197-198).

 

Line 201 - Remove "kind of" at the beginning of this line.

 

Response:

Thank you. We have removed it (page 5, line 202).

 

Line 204 - Replace "to" with "of" to say "...to the evolution of next generation..."

 

Response:

Thank you. We have changed it based on your remark (page 6, line 205).

 

Line 280 - Capitalize "Approximation" in the acronym definition.

           

Response:

Thank you. We have corrected this in the revised version of the paper (page 7, line 281).

 

Line 311 - You have already defined Deep Learning as "DL" - define acronyms on first use and then use the acronym consistently throughout the remainder of the document. You have several other acronyms that are redefined.

Response:

Thank you. We have replaced it with the acronym “DL” (page 8, lines 312-313).

 

Line 330 - Rewrite this as "...in this section, the authors attempt to shed light on this issue..."

Response:

Thank you very much for this comment. We have rephrased it based on your remark (page 8, line 331).

 

Line 344 - For Figure 2 and Figure 3 it seems to be more applicable to list the references by their number in the References section. That would be easier to track down by the reader.

Response:

Thank you very much for this comment. We totally agree. In the revised version of the paper, we include the number of each reference.

 

Line 351 - In the discussion for Figures 2 and 3 it is unclear to me what the temporal dimensions really mean. What/how is time period measured and interpreted? This section needs more discussion.

Response:

Thank you very for this comment. In order to clarify this, additional information on the difference in the time periods and temporal ratio are provided in lines 339-344. Moreover, the calculation of the Temporal Ratio column of Table A1 is provided in lines 149-152 “It should be noted that the column named “Temporal Ratio” of Table A1 has been calculated due to the observed discrepancies of data collection period lengths for crashes and SSMs. The values of this column are dimensionless numbers as they have been calculated by converting the crash and SSMs data collection periods into the same time units.” For readability reasons Figure 2 and 3 were also provided.

 

Line 416 - Machine learning is already defined as ML.

 

Response:

Thank you. We have replaced it with the acronym ML (page 11, line 421; page 12, line 472; page 13, line 528).

 

Line 430 - Surrogate safety measures are already defined as SSMs.

 

Response:

Thank you very much for this remark. We have made the appropriate modifications (page 2, line 82; page 3, line 126; page 11, line 435).

 

Line 435 - Rewrite as "However, there are a small number of studies that focus on serious..."

Response:

Thank you very much for this comment. We have rephrased it based on your suggestion (page 11, lines 440-441).

 

Line 463 - ML is already defined. Also, on this line, rewrite as "Over the last few years, ML models have proven to be very efficient prediction tools..."

 

Response:

Thank you. We have replaced it with the acronym ML (page 11, line 421; page 12, line 472; page 13, line 528).

 

Overall the paper needs more detail on the SSMS (the information shows up in the table in the appendix but it is important information and should not just be in an appendix. You also need to provide more information on the time component and what that means and how it can be interpreted. The biggest shortcoming of the paper; however, is the lack of new research. It is a good synthesis but not a research paper.

Response:

Thank you once again for your constructive feedback. We strongly believe that our manuscript presents a novel contribution to the existing literature on SSMs by exclusively investigating studies that use both SSMs and historical road crash data. Moreover, our review article extends beyond measures with predefined thresholds for the identification of conflicts to measures that can be extracted from smartphone sensors or instrumented vehicles related to harsh driving behaviour events and sheds light on the temporal periods dedicated to data collection for both SSMs and road crashes. The Summary Table (A1) has been placed in the Appendix as it should be placed on a landscape format and we would not like to interrupt the natural and logical flow of the text with a such long table. With regard to the time component, additional information have added to lines 339-344.

Reviewer 2 Report

This study reviews literature related to historical crash record investigations. The study is meaningful, straightforward, and a good candidate for a literature review article.

Author Response

Thank you very much for your positive assessment.

Reviewer 3 Report

The paper presents a review of Surrogate Safety Measures studies in road safety domain, where real environmental conditions were collected and analyzed. The review is based on PRISMA flow and investigates 34 studies that resulted from PRISMA. The authors take a systematic approach that is easy to follow and verify. The review also includes their own contribution, such as determination of aspects worth of comparison among the studies, temporal ratio evaluation and individual perspectives on what can be inferred and concluded from the reviewed papers, even with some future perspectives. I really enjoyed reading it, especially because it reads much more natural than a typical review paper and because the authors did a very good job delivering a clear value to the community. Congratulation on this!

I only have few, minor comments to incorporate before publishing:

-          “Temporal ratio” column in table A1 makes a first impression, that absolute counts of crashes to SSMs are used and only later (when reading the text further), I realized that what is divided are time periods. I suggest specifying this in the table heading.

-          Page 5, line 188: “… the scope of the current research …” should be “… the scope of the presented research …” as what the authors mean is not all the research, but just their research.

-          Page 8, line 337-338: there is a motivation for why temporal ratio is calculated, but it is underspecified for the reader to understand. Only later I realized that the time discrepancy is due to new technology being able to run similar analyses with shorter time periods and that the ratio interprets as by how much more time is needed to collect equivalent sample with road crash data. In my opinion, this should be explained on page 8 with the motivation as it helps reader to better comprehend.

-          Page 12, line 493: I suggest deleting the word “even” at the end of the line as an analysis complement is less than analysis replacement. It sounds more natural/logical to the reader.

Author Response

Response to reviewers’ comments

Reviewers’ comments appear in bold; authors’ comments follow in regular type. In the submitted manuscript, all changes/additions to the text have been marked using track changes.

The paper presents a review of Surrogate Safety Measures studies in road safety domain, where real environmental conditions were collected and analyzed. The review is based on PRISMA flow and investigates 34 studies that resulted from PRISMA. The authors take a systematic approach that is easy to follow and verify. The review also includes their own contribution, such as determination of aspects worth of comparison among the studies, temporal ratio evaluation and individual perspectives on what can be inferred and concluded from the reviewed papers, even with some future perspectives. I really enjoyed reading it, especially because it reads much more natural than a typical review paper and because the authors did a very good job delivering a clear value to the community. Congratulation on this!

 

Response:

Thank you very much for your review and your constructive input for our research work.

 

I only have few, minor comments to incorporate before publishing:

“Temporal ratio” column in table A1 makes a first impression, that absolute counts of crashes to SSMs are used and only later (when reading the text further), I realized that what is divided are time periods. I suggest specifying this in the table heading.

 

Response:

Thank you very much for your remark. We can understand your point of view. Therefore, we have changed the name of the respective column of Table A1 to “Temporal Ratio (crashes period / SSMs period)”.

 

Page 5, line 188: “… the scope of the current research …” should be “… the scope of the presented research …” as what the authors mean is not all the research, but just their research.

Response:

Thank you. We have rephrased this according to your remark (page 5, line 189).

 

Page 8, line 337-338: there is a motivation for why temporal ratio is calculated, but it is underspecified for the reader to understand. Only later I realized that the time discrepancy is due to new technology being able to run similar analyses with shorter time periods and that the ratio interprets as by how much more time is needed to collect equivalent sample with road crash data. In my opinion, this should be explained on page 8 with the motivation as it helps reader to better comprehend.

 

Response:

Thank you for your feedback. We agree that the motivation for why the 'Temporal Ratio' is calculated could be better explained in the paper. The following text has been added (page 8, lines 339-344):

“The difference in time periods between the collection of historical road crash data and SSMs is mainly attributed to the emergence of new technologies, which allow for rapid collection of SSMs data and the conduction of analyses with shorter time periods. The ‘Temporal Ratio’ column could be interpreted as by how much more time is needed to collect an equivalent sample of SSMs with road crash data.”

 

Page 12, line 493: I suggest deleting the word “even” at the end of the line as an analysis complement is less than analysis replacement. It sounds more natural/logical to the reader.

 

Response:

Thank you. We agree. Τhe word “even” has been deleted in the revised version of the manuscript (page 12, line 505).

Reviewer 4 Report

The study presents a systematic review of surrogate safety measures (SSM) uses in historical crash investigations. A detailed review of 34 papers shortlisted using the PRISMA technique is presented. Based on the review, future research directions are discussed. Overall, the paper is interesting and timely, considering that the world is now moving towards using SSM for proactive traffic safety assessment. Following are my comments on the work:

·         The authors mention that there are a limited number of studies focusing on estimating crashes by severity using SSM. However, recently a few studies listed below have attempted to estimate crashes by severity using different SSMs. The authors are advised to refer to these studies and subsequently add them.

1.      Arun, A., Haque, M. M., Bhaskar, A., Washington, S., & Sayed, T. (2021). A bivariate extreme value model for estimating crash frequency by severity using traffic conflicts. Analytic methods in accident research, 32, 100180.

2.      Arun, A., Haque, M. M., Bhaskar, A., & Washington, S. (2022). Transferability of multivariate extreme value models for safety assessment by applying artificial intelligence-based video analytics. Accident Analysis & Prevention, 170, 106644.

3.      Goyani, J., Paul, A. B., Gore, N., Arkatkar, S., & Joshi, G. (2021). Investigation of crossing conflicts by vehicle type at unsignalized T-intersections under varying roadway and traffic conditions in India. Journal of transportation engineering, Part A: Systems, 147(2), 05020011.

·         The authors mention that SSM-based studies for crash investigations are largely focused on vehicular traffic, and very few focus on pedestrians. The authors are advised to refer to the following works for pedestrians and bicyclists where the studies have used SSM for crash investigation.

1.      Ali, Y., Haque, M. M., & Mannering, F. (2023). A Bayesian generalised extreme value model to estimate real-time pedestrian crash risks at signalised intersections using artificial intelligence-based video analytics. Analytic Methods in Accident Research, 38, 100264.

2.      Johnsson, C., Laureshyn, A., & Dágostino, C. (2021). Validation of surrogate measures of safety with a focus on bicyclist–motor vehicle interactions. Accident Analysis & Prevention, 153, 106037.

·         The authors are advised to articulate their discussion on future research direction in four to five major domains, such as (a) the Temporal period required for SSM uses in crash investigation, (b) Modelling approaches, (c) data collection and processing, (d) real-time application, (e) integration of SSM for reliable crash investigation, and (f) level of analysis (for instance, network level analysis).

·         Recently, a hybrid modeling approach integrating the statistical and machine learning approaches has been used. For instance, Hussain, F., Li, Y., Arun, A., & Haque, M. M. (2022). A hybrid modelling framework of machine learning and extreme value theory for crash risk estimation using traffic conflicts. Analytic methods in accident research, 36, 100248, integrated the machine learning and extreme value theory for estimating crashes from SSMs. The authors are advised to refer to this study and add subsequent discussions to their paper.

Author Response

Response to reviewers’ comments

Reviewers’ comments appear in bold; authors’ comments follow in regular type. In the submitted manuscript, all changes/additions to the text have been marked using track changes.

The study presents a systematic review of surrogate safety measures (SSM) uses in historical crash investigations. A detailed review of 34 papers shortlisted using the PRISMA technique is presented. Based on the review, future research directions are discussed. Overall, the paper is interesting and timely, considering that the world is now moving towards using SSM for proactive traffic safety assessment. Following are my comments on the work.

Response:

Thank you very much for your review and your constructive input for our research work.

 

The authors mention that there are a limited number of studies focusing on estimating crashes by severity using SSM. However, recently a few studies listed below have attempted to estimate crashes by severity using different SSMs. The authors are advised to refer to these studies and subsequently add them.

 

1. Arun, A., Haque, M. M., Bhaskar, A., Washington, S., & Sayed, T. (2021). A bivariate extreme value model for estimating crash frequency by severity using traffic conflicts. Analytic methods in accident research, 32, 100180.

 

2. Arun, A., Haque, M. M., Bhaskar, A., & Washington, S. (2022). Transferability of multivariate extreme value models for safety assessment by applying artificial intelligence-based video analytics. Accident Analysis & Prevention, 170, 106644.

 

3. Goyani, J., Paul, A. B., Gore, N., Arkatkar, S., & Joshi, G. (2021). Investigation of crossing conflicts by vehicle type at unsignalized T-intersections under varying roadway and traffic conditions in India. Journal of transportation engineering, Part A: Systems, 147(2), 05020011.

 

Response:

Thank you very much for this comment. Based on the reviewed studies that are presented in the Summary Table, we have highlighted injury severity estimation using SSMs as a critical research need. Based on your comment, we can understand that there are a few recent studies that aim to shed light on this issue. Therefore, we have added the proposed studies in the Discussion section (page 11, lines 445-446), as follows: “The inclusion of the level of injury severity in similar studies would be highly interesting for the quantification and the comparative assessment of the relationship between SSMs and different crash severity levels. Injury severity estimation using SSMs is also high-lighted as a critical research need by Arun et al. [28]. To that direction, a few recent re-search studies have attempted to estimate crashes by severity level using different SSMs [87, 88, 89].”

 

The authors mention that SSM-based studies for crash investigations are largely focused on vehicular traffic, and very few focus on pedestrians. The authors are advised to refer to the following works for pedestrians and bicyclists where the studies have used SSM for crash investigation.

 

1. Ali, Y., Haque, M. M., & Mannering, F. (2023). A Bayesian generalised extreme value model to estimate real-time pedestrian crash risks at signalised intersections using artificial intelligence-based video analytics. Analytic Methods in Accident Research, 38, 100264.

 

2. Johnsson, C., Laureshyn, A., & Dágostino, C. (2021). Validation of surrogate measures of safety with a focus on bicyclist–motor vehicle interactions. Accident Analysis & Prevention, 153, 106037.

 

Response:

Thank you very much for your comment. We had already included in the Summary Table the study of Johnsson et al. (2021) numbered as reference [66]. Based on your comment, we also added the study of Ali et al. in the Discussion section as follows (page 11, lines 456-459):

“Therefore, more research is needed on the manner in which various SSMs could be exploited to enhance the safety of VRUs. Towards this direction, Ali et al. developed a Bayesian Generalized EVT model in order to estimate real-time pedestrian crash risks at signalized intersections using Artificial Intelligence (AI)-based video analytics [92].”

 

The authors are advised to articulate their discussion on future research direction in four to five major domains, such as (a) the Temporal period required for SSM uses in crash investigation, (b) Modelling approaches, (c) data collection and processing, (d) real-time application, (e) integration of SSM for reliable crash investigation, and (f) level of analysis (for instance, network level analysis).

 

Response:

Thank you very much for your comment. The objective of Section 4.2. Future Research Directions is to outline research directions that do not appear to be sufficiently investigated from the reviewed studies of the Summary Table. To that end, we have tried to have one paragraph for each research direction ((i) severity, (ii) vulnerable road users, (iii) level of analysis, (iv) modelling approach, (v) improvements in the technological field). With regard to other issues, such as temporal period, data collection etc. we have summarized and discussed the key trends in Section 4.1. We believe that the current structure of the paper is effective in conveying the information and ideas we wish to present in a logical flow.

 

Recently, a hybrid modeling approach integrating the statistical and machine learning approaches has been used. For instance, Hussain, F., Li, Y., Arun, A., & Haque, M. M. (2022). A hybrid modelling framework of machine learning and extreme value theory for crash risk estimation using traffic conflicts. Analytic methods in accident research, 36, 100248, integrated the machine learning and extreme value theory for estimating crashes from SSMs. The authors are advised to refer to this study and add subsequent discussions to their paper.

 

Response:

Thank you very much for this suggestion. We totally agree that hybrid modelling approaches a quite promising modelling framework for estimating crashes from SSMs. To that end, we have added the proposed reference and the following in the Discussion section (page 12, lines 482-485): “Furthermore, hybrid modelling approaches integrating both statistical and ML techniques could be considered in future research studies, as this framework represents a methodological advancement in traffic conflict-based crash estimation models [94].”.

Round 2

Reviewer 1 Report

Thank you for the revisions made based on the comments.