Exploring the Interoperability of Public Transport Systems for Sustainable Mobility in Developing Cities: Lessons from Johannesburg Metropolitan City, South Africa

Round 1

Reviewer 1 Report

This paper analyzes the social media posts to calculate the connectivity index between high speed rail and bus rapid transit system in the City of Johannesburg. Below please find my comments about the paper:

 

What are the limitations of using GeoWeb 2.0 data for calculating connectivity indexes? Are all age groups and socio-economic groups likely to post social media post? How does this potential bias could affect the results? Perhaps this information could be added to Section 1.1.1

Lines: 109:122. I appreciate that paragraph in line 109-122 summarizes the literature review section. Please clarify how this paper builds on work of previous researchers and what are the contributions of this paper.

Please provide a brief description and equations for conducting connectivity analysis. I appreciate that a reference is made to MATLAB toolbox in line 175, however, more description of the methods will be helpful to reader, and researchers who would like to adopt your methodology.

What are the specific lessons learned from this paper that can help improve connectivity between transit modes? Does this method help in rescheduling and redesigning BRT system? What can other regions learn from this case study and proposed methodology? The paper will be stronger if these information is added to conclusion section.  

Author Response

All the comments and concerns of the reviewer have been attended and resolved. See attached document.

Author Response File: Author Response.pdf

Reviewer 2 Report

In this paper, authors explore the interoperability of different public transport systems – two bus networks and a rail system – applied to the South African city of Johannesburg. Whilst I acknowledge the relevance of this topic, I think substantial improvements are needed in the paper before acceptance at this stage.

One of my main issues is that the paper remains very vague about the exact steps authors have undertaken, and how / why different methods are used and linked together.

• For example, I found Chapter 1 quite long and rather vague about what the authors are exactly going to do. It would be good if authors can bring across the main takeaways and study limitations more explicitly.

This is more problematic in Chapter 2, as there are hardly any formalisations and justifications of the methods used in this study, making it difficult to reproduce this work. For example:

• In step 1, which exact public transport network data is used? Is this GTFS data, or any other data source, and how is it obtained and processed?
• In Step 4+5 (based on Figure 1), I miss a clear formalisation of the connectivity closeness and clustering analysis: can you please provide mathematical formulas to show which variables you exactly used, and which method you used to perform these steps, as this remain rather vague now?
• I miss a much more extensive discussion on the choice for the Local Moran’s Index. Why did authors – for example – not use some other unsupervised learning approach (DBSCAN, hierarchical clustering) in this clustering step? What is the argument to opt for this particular index?
• For me, it remains unclear what the role of this GeoWeb analysis is in the whole study. Is this done merely as proxy to identify the busiest stops based on digital (tweeted) information about these stops? If that is the case, why not use counts, travel ticket sales data etc. for this to be more accurate?

A second issue is related to the definition of ‘high-speed train’. Can authors provide a clear definition of what they consider as ‘high-speed train’ in this study? Which design speed, stop density etc. are you referring to? The conventional definition of a HST is not to function as connection on a city or agglomeration network level, but to operate on a national or even an international network level connecting main cities. However, reading through your work, I’m not sure if you apply this same definition, as I don’t see how such higher-network level system can be relevant to discuss connectivity within Johannesburg (rather than between Johannesburg and other cities). If authors refer to a regional train (similar to the Paris RER or London Overground), please change ‘high-speed train’ to a more suitable definition throughout the paper.

The third issue relates to the contribution. In the abstract, authors mention as objective to ‘ formulate policy frameworks in integrating public transit systems in cities of the developing world, learning from the metropolitan city of Johannesburg.’ However, throughout the paper there is only focus on the Johannesburg case study, without any discussion about how to generalise this to other developing countries.

I also miss a discussion here about policy implications: what role can a transit agency play in integrating different transport networks, i.e. by means of integrated network design or integrated fare systems?

Structure:

• I recommend making a much more explicit distinction between your general method, and the particular application to Johannesburg as case study, as these aspects are now heavily intertwined. The same applies for mentioning particular software (Geoweb, Matlab, ArcGis Pro etc.) in your method, whilst I consider this rather an application for your specific work. I recommend making Chapter 2 entirely independent from your case study, whilst adding a new Case Study chapter where you introduce all specifications of your case study, the software you used etc.
• I think the structure of several chapters is a bit odd. For example, in Chapter 1 there is first a part without subsection, followed by only sub section 1.1 (no sub section 1.2), which again is followed by a further sub section 1.1.1 only. I recommend restructuring or renumbering this in a more logical way (e.g. 1.1 ‘relevance’ and 1.2 ‘literature review’). Similar things occur for example in Chapter 2.
• I miss a section with reflection on the method and recommendations for improvements and future research.
• I’m not sure if the name of section 1.1.1 ‘GeoWeb’ does cover the content, as this section discusses literature on spatial analyses in relation to connectivity.

Style: overall, I find this paper to be well-written. However, throughout the paper there are many small typo’s and grammar errors, which need to be corrected. See for example (but not limited to) the following errors on pages 1-3:

• 1 row 44: ‘robustness of such systems’: ‘of’ missing
• 2 row 82: ‘resonate’ should be ‘resonates’
• 2 row 89: ‘increasing’ should be ‘increasingly’
• 3 row 99 ‘varies’ should be ‘various’
• 3 row 11 ‘have be used’should be ‘have been used’

Then, there are a few smaller questions:

• I find Figure 2 not very clear: can you please clarify this figure by showing the number of nodes and links for each of the three networks?
• Figure 5: as I assume the number of nodes is not equal for the three PT networks, it might be better to use a relative scale for the y-axis instead of an absolute number.
• ‘An interesting feature is that most of nodes along both the Gaubus and Rea Vaya have a level 2-degree centrality, even though the two networks are in different locations with the Gaubus located north of Johannesburg and the Rea Vaya to the south of the city.’: As Figure 5 shows node degree for each separate PT network, it is completely logical that most nodes have degree 2, as this simply reflects an inbound and outbound link of the same bus stop. I’m not quite sure how the abovementioned statement is derived?

Author Response

All the comments and concerns by the reviewer have been attended to and resolved. See attached document.

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report

Thank you for improving your paper and several of my previous comments. There are a few remaining issues.

Step 1 of the method: it remains unclear to me which data source authors use in their method to perform this step. I understand you use shapefiles and polyline data, but it is unclear which source you use: GTFS? Any other open data source or standardised format, which allows this study being replicated for other cities?

Step 1+5 of the method: how are the formulae for closeness (Step 1) and Local Moran's Index I (Step 5) related? I don't see variable C (from the first formula) appearing anywhere in the second formula?

Figure 6 (was Figure 5): I appreciate the y-axis indicates the absolute number of nodes, but I feel that using the relative number of nodes on the y-axis instead is much more informative: as the number of nodes is different for Reabaya / Gautrain / Gaubus, the distribution of degree centrality in absolute terms does not really provide a helpful comparison.

Could you add a short section discussing:
a) Policy implications: what role can a transit agency play in integrating different transport networks, i.e. by means of integrated network design or integrated fare systems?
b) Reflection: recommendations for improvements of your method
c) Generalisability: to what extent can results be generalised or extrapolated to other cities, as the focus of this work is merely South Africa.

Typo's:
- Method Step 4 (row 234): 'Euclidean distance' should be with capital E
- Method Step 5 (row 243): 'To identify the distribution of hot spots along the public transportation networks and rank these hot spots.': this is an incomplete sentence
- Results 3.1 (row 315): 'Figure 5' should be 'Figure 6'

Author Response

All the comments were considered and all the suggested corrections were done. See attached document that shows the list of all corrections done.

Author Response File: Author Response.pdf