Towards a Walkable City: Principal Component Analysis for Defining Sub-Centralities in the Santiago Metropolitan Area

Round 1

Reviewer 1 Report

1. It could be better if the authors further clarity why apply principal component analysis and distance rather than other currently mainstream approaches. The current argument is not very convincing.
2. The literature review in the introduction can be more clear to bolster your research design, especially the part relevant to the methodology.
3. As for the 28 variables the authors used related to urban functions, the author count the number of each function. But could it be better if the authors consider other factors like the area of Green Areas or Hospital, etc.? If not necessary, I suggest to make some explanation.  
4. The authors resort to only distance and population to measure the walkability. Other indicators, like land use mix and street connectivity, may be also relevant or even critical. Additionally, the approach of distance calculating need to be clarified further so the reader will have ides of its accuracy and precision.

Author Response

1. It would be better if the authors further clarify why they have applied a principal component analysis and distance rather than other current mainstream approaches. The current argument is not very convincing.

RESPONSE 1.c: Thank you for this suggestion. However, the reasons for using a PCA have already been explained in the Materials and Method section. Further, the decision on distance was amended based on your observation and new paragraphs were written as follows:

To complement Berry and Garrison’s methodological approach, we borrowed the explanations warrants proposed by other authors. Learning from Southworh’s contribution, a high value is assigned to the grain of the city, an aspect also considered valuable by Gehl. In this regard, we aimed to incorporate all urban functions in the first stage based on the availability of data for formulating a first approach to this walkable city in the case of Santiago. As a result, 28 urban functions were mapped to assess the walkability for each neighbourhood.

To detail the discernment process, we define the primary reference distance based on our criteria that were mentioned in the considerations elaborated by Muxi et al. for the feminist urban planning approach. Empirically, we measured the time taken by a woman to walk with a trolley at a slow ralent pace for 15 minutes. The result indicated that, for a slow pace, a woman with a trolley covers about 600 meters in 15 minutes. In the primary urban functions set, we included those activities that will be carried out every day, such as those that either kids or elderly need to attend. In this case, we included places of worship because, in Santiago, churches are usually used as community centres where elderly receive everyday assistance or as spaces for social meetings. The secondary reference based on distance was measured as a bicycling distance from one point to another, defined as the distance covered in 15 minutes while riding a bike at 6 km/h or walking at 3 km/h for 30 minutes. The secondary urban functions are less demanding on an everyday basis, where the frequency of attendance by elderly and children is not daily demanded. The tertiary level is composed of those urban functions that are required quite less frequently and, instead, are required to be nearby in case of an emergency (fire fighters, police or hospitals) or for special visiting days, such as museums and theatres.

2. The literature review in the introduction can be more clear to bolster your research design, especially in the part relevant to the methodology.

RESPONSE 2.c: Thank you for this observation, which is aligned with the previous observation. Thus, to the paragraph previously mentioned, we also added the following sentences of clarification:

To map these distances, we employed a GIS-based service, generating a route layer from blocks to the different services assessed for the distance to be assessed. The analysis locates the nearby urban functions included in the dataset that are requested for each spatial assessment and determines how many are accessible within the distance under study.

A final clarification is required regarding the use of the census and its spatial representation. In our study, census tracts without population were excluded from the analysis, for instance, especially those that are referred to as natural areas and highways. In summary, 95% of the polygons analysed have inhabitants. This obliges us to inform the reader that the results for areas with low density have a two-fold interpretation; it means that, since there is a small amount of people living there, the market and state are not interested in investing resources to develop these areas, and it also means that these are areas where a significant investment may trigger new development regions for increasing density. In Appendix C, it is possible to see the density of the Metropolitan Area of Santiago; nevertheless, the correlation measured by the Pearson coefficient between the population and the resultant scores of the 15-minute city assessments is 0.041, which is considered low.

3. As for the 28 variables the authors used that are related to urban functions, the authors count the number of each function. But would it be better if the authors consider other factors such as the area of Green Areas or Hospitals, etc.? If not necessary, I suggest to providing some explanation for the same.

RESPONSE 3.c: Thanks for this suggestion, which is similar to the observation made by the Reviewer 2. This has been partially addressed as follows:

The analysis of each census tract is not compensatory. This means that if a block is surrounded by a dozen pharmacies, its score still just counts as 1, because that essential urban function is covered even if there is only one facility that provides the service. Therefore, the census tracts with higher score are those areas where all essential urban functions are available regardless of the number of facilities that provide a particular service.

On the other hand, we must disregard the suggestion regarding areas because it implies a different type of analysis, with different methods and data.

4. The authors resort to only distance and population to measure walkability. Other indicators, such as land-use mix and street connectivity, may also be relevant or even critical. Additionally, the approach to distance calculation needs to be clarified further in order to give the reader an idea of its accuracy and precision.

RESPONSE 4.c: Thank you for this suggestion. We are not totally clear regarding exactly what has been suggested and request you to please clarify the same. On the other hand, in the new manuscript submitted there is additional information in relation to the decision made to this research that may help to clarify some points here stated. To answer this question, we provided further data about the accuracy of data and criteria for defining the different levels of analysis. The section named Materials and methods has these changes registered. We hope this addressed your request.

Author Response File: Author Response.docx

Reviewer 2 Report

The paper focuses on a topic that is particularly relevant today, linking the preparation of cities to future epidemic calamities with urban strategies useful to improve the daily life of their citizens, like the “15 minutes city” concept of Paris.

The proposal of a methodology for evaluating the potential for an urban area to become a 15-minute city, based on its existing essential urban functions, represents relevant support to the current discussion about walkable cities.

Also, the will to propose a replicable methodology is appreciable.

I found some criticalities, especially related to the lack of clarity in the theoretical definitions.

Below I report (in summary) the main notes.

1. The Introduction section is lacking literature support to scientifically ground the discussion. For instance, in line 43 “A city that is based on everyday sub-centralities has been theorized as being a walkable city” who theorized it? In which context this statement could be considered true? Etc.

Or in line 45 “A walkable city improves connectivity by promoting intermodality and increasing the fine grain of land use” Cannot be the other way around? In my view, a fine grain of land use promotes a walkable city, being the land use a more stable parameter than pedestrian movement.

Or in line 75 “the discussion about implementing the concept of a 15-minute city to reduce the risk of future pandemics has begun to attract the interest of experts and decision-makers”, this statement should be supported by literature or references.

In general, the Introduction section should be grounded over more solid literature.

2. The work of Brian Berry and William Garrison, proposed in the Method section, is quite dated. It is suggested to increase in the literature section the references to more contemporary works about central place hierarchy.

3. A more extensive explanation of the rationality behind your choice of the primary/secondary/tertiary level services is needed. In line 107 is stated: "To discern variables by level, we used a discrete discussion between the authors to decide which urban functions are needed within a 600 metre ...". Please explain how did you evaluate this selection and why for instance supermarket are not intended as primary service while, for instance, book stores yes. If the goal is to link the 15 min city with a post-pandemic preparation strategy, this point represents a crucial factor.

4. The conclusion contains reflections which have not been discussed before. It is suggested to move some of these reflections (i.e. about Zoning/sub centralities) in the discussion section.

5. In general, the article needs editing of the English language, especially in the literature section.

Author Response

1.The Introduction section is lacking literature support to scientifically ground the discussion. For instance, in line 43 “A city that is based on everyday sub-centralities has been theorized as being a walkable city”, who theorized it? In which context can this statement be considered true? Etc.

Or in line 45 “A walkable city improves connectivity by promoting intermodality and increasing the fine grain of land use” Can it not be the other way around? In my opinion, a fine grain of land use promotes a walkable city, since the land use is a more stable parameter than pedestrian movement.

Or in line 75 “The discussion about implementing the concept of a 15-minute city to reduce the risk of future pandemics has begun to attract the interest of experts and decision-makers”, this statement should be supported by literature or references.

In general, the Introduction section should be grounded using more solid literature.

RESPONSE 1.a: Thank you for this feedback. However, we would like to know what the reviewer means by “more solid literature”. We have already quoted works on walkability by Pucher and Dijkstra, Southworth, Gehl, Muxi et al., Weng et al. and Ewing, who are the most quoted authors from 2005 onwards with regard to the matter of walkable cities as per Web of Science. All these critiques to lines 43, 45 and 75 are based on the references provided in the references list. Thus, we have to disagree with this comment, and we request you to please provide further explanation for your feedback so we can address the issue.

2. The work of Brian Berry and William Garrison, proposed in the Method section, is quite dated. It is suggested that the references in the literature section be increased to include more contemporary works about central place hierarchy.

RESPONSE 2.a: Thank you for this suggestion, which we believe will improve the quality of the methodological presentation. We referred to the other quoted references in the previous section to better explain how we adopted the method used by Berry and Garrison method for our analysis. The following complementary paragraph was added:

To complement Berry and Garrison’s methodological approach, we borrowed the explanations warrants proposed by other authors. Learning from Southworh’s contribution, a high value is assigned to the grain of the city, an aspect also considered valuable by Gehl. In this regard, we aimed to incorporate all urban functions in the first stage based on the availability of data for formulating a first approach to this walkable city in the case of Santiago. As a result, 28 urban functions were mapped to assess the walkability for each neighbourhood.

To detail the discernment process, we define the primary reference distance based on our criteria that were mentioned in the considerations elaborated by Muxi et al. for the feminist urban planning approach. Empirically, we measured the time taken by a woman to walk with a trolley at a slow ralent pace for 15 minutes. The result indicated that, for a slow pace, a woman with a trolley covers about 600 meters in 15 minutes. In the primary urban functions set, we included those activities that will be carried out every day, such as those that either kids or elderly need to attend. In this case, we included places of worship because, in Santiago, churches are usually used as community centres where elderly receive everyday assistance or as spaces for social meetings. The secondary reference based on distance was measured as a bicycling distance from one point to another, defined as the distance covered in 15 minutes while riding a bike at 6 km/h or walking at 3 km/h for 30 minutes. The secondary urban functions are less demanding on an everyday basis, where the frequency of attendance by elderly and children is not daily demanded. The tertiary level is composed of those urban functions that are required quite less frequently and, instead, are required to be nearby in case of an emergency (fire fighters, police or hospitals) or for special visiting days, such as museums and theatres.

3. A more extensive explanation of the rationality behind your choice of the primary/secondary/tertiary level services is needed. In line 107, it has been stated, “To discern variables by level, we used a discrete discussion between the authors to decide which urban functions are needed within a 600-metre ...”. Please explain how you evaluated this selection and why, for instance, supermarkets are not intended as primary service while, for instance, book stores are. If the goal is to link the 15-minute city with a post-pandemic preparation strategy, this point represents a crucial factor.

RESPONSE 3.a: Thank you for this feedback. We have also addressed this aspect in the previous point.

4. The conclusion contains reflections that have not been discussed before. It is suggested some of these reflections (i.e. about Zoning/sub centralities) are moved to the discussion section.

RESPONSE 4.a: Thank you for this suggestion. We have reviewed the conclusions and modified certain sections accordingly.

5. In general, the article needs editing for the English language, especially in the literature section.

RESPONSE 5.a: Thank you for this suggestion. We have hired a professional proofreading service to edit the issues with the language of the text.

Author Response File: Author Response.docx

Reviewer 3 Report

I read the paper entitled "Towards a walkable city: Principal component analysis for defining sub-centralities in the Santiago Metropolitan Area" which describes the application of principal component analysis (PCA) to a dataset of 28 variables quantified for each of the census-level tracks of the Santiago Metropolitan area, in Chile.

The paper has many merits as it presents an in-depth analysis of selected urban functions that constitute the most frequent destinations for both commuting and recreational travel within any metropolitan area. The production of six new components that constitutes the core of the PCA analysis is valid and the experience went well as the six identified components could be explained easily, with meaningful and non-overlapping services/space descriptors, as demonstrated in Table 3.

While the overall results seem valid for a land management perspective, and the use of census-level tracts is adequate for the analysis, the problems of using a straight-line distance (which is not only applied in this study, but also in well-known indicators like the WalkScore, cited in this study) are not discussed in the document. In such a large and inhabited metropolitan area, with different land use patterns, terrain characteristics and road/street spatialization patterns, GIS-based service areas using streets/roads as structural data would allow for a more realistic analysis. As an example, a neighborhood (or census track) can be (its centerpoint can be...) at a straight line distance from a hospital (or any other essential urban function in Table 1) of 1000 meters but at a real street-based distance of much more, due to a barrier (motorway, river, etc.) - affecting the analysis with unrealistic distances. Distance in consolitated urban centers are not the same as in periurban areas, where the density of the pedestrian or vehicle networks are much different, with impacts of the distance choice model. The smaller the distances considered (such as the first "buffer" of 600 m), the more significative is the impact of the distance method choice. I think the authors should discuss the limitations of their analysis arising from the option to use simplified distance calculations.

Also within this point, authors should describe if the analysis is compensatory, for example, if a specific census tract is surrounded by dozens of "pharmacies" but has no other "local services" nearby, does that give it a good "local services" score? There was no mention to that in the text, I think. Also, does a "hospital" count "as much" as a cinema for the "Metropolitan services" total, that is, within the same components, are all the functions equally valued?

2) Another point is that there is no description of Chilean census tracts (CT): do all of the CT polygons have inhabitants? are they more of less homogeneous in terms of population, with small CT units in city centers, larger in the outskirts of the metropolitan area or within "empty" or natural areas? I suggest the authors to include a small introduction as there are no numerical outputs (besides percentages) and because having different levels of access to essential urban functions in different parts of the metropolitan area is in itself an unavoidable characteristic of such a large ad heterogeneous area. The type of map that expresses this (Fig. 7) is expected: higher scores in the central areas (downtown Santiago and downtown of the major municipalities within the metropolitan area) and lower scores in the outskirts of the area. But if the map describes "low"-level of access areas is important in terms of population, it remains to be seen.

3) The paper has some speculative sentences that are not supported by the type of analysis that was developed. Examples are lines 262-265 (concentration on a metropolitan mayor vs. local mayors) or 75-77 (the pandemic has triggered the discussion of the implementation of the 15-min. city). Please refrain from unsubstatiated speculation as the focus of the paper is the extraction of components based on the 28 selected variables and mapping of the PCA results.

4) The biggest contribution of the paper is the presentation of a methodology to reduce a bunch of variables to a tractable set of components and map them to support decision on territory, possibly on prioritizing state-driven interventions, or pointing out business opportunities for the private sector, if and where they are needed. This being said, I think the discussion around the composite indicator of figure 7 could be extended as it is a mere analysis of population within each interval. We do not know, for instance, if there are significative urban centers/municipalities that have large fractions of population with low access levels.

To conclude, as a minor suggestion, I do not think that authors must indicate the licensee/license holder of the used software (lines 140-143).

Author Response

1. While the overall results seem valid from a land-management perspective, and the use of census-level tracts is adequate for the analysis, the problems of using a straight-line distance (which is not only applied in this study but also in well-known indicators such as the WalkScore, cited in this study) are not discussed in the document. In such a large and inhabited metropolitan area, with different land-use patterns, terrain characteristics and road/street spatialisation patterns, GIS-based service areas that using streets/roads as structural data would allow for a more realistic analysis. As an example, a neighbourhood (or census track) can be (its centre point can be...) at a straight-line distance of 1000 meters from a hospital (or any of the other essential urban functions mentioned in Table 1) but is at a much greater real street-based distance due to a barrier (motorway, river, etc.), which affects the analysis that uses unrealistic distances. The distance in consolidated urban centres are not the same as those in periurban areas, where the density of the pedestrian or vehicle networks is quite different, having impacts of the distance-choice model. The smaller the distances considered (such as the first “buffer” of 600 m), the more significant the impact of the distance method choice. I think the authors should discuss the limitations of their analysis that arise from the choice to use simplified distance calculations.

RESPONSE 1.b: Thank you for this observation. We agree; we did not mention it, but we did use the distance based on roads and streets rather than a circular radius from each point. Of course, this was not clearly stated in the method or in the criteria for the analysis observed here. We have added this information to the Materials and Methods section as follows:

To map these distances, we employed a GIS-based service, generating a route layer from blocks to the different services assessed for the distance to be assessed. The analysis locates the nearby urban functions included in the dataset that are requested for each spatial assessment and determines how many are accessible within the distance under study.

2. Also within this point, authors should describe if the analysis is compensatory. For example, if a specific census tract is surrounded by dozens of “pharmacies” but has no other “local services” nearby, does that give it a good “local services” score? There was no mention of this in the text, I think. Also, does a “hospital” count “as much” as a cinema for the “Metropolitan services” total; that is, within the same components, are all the functions equally valued?

RESPONSE 2.b: Thank you for the feedback. We have explained it better in the Materials and Method section as follows:

The analysis of each census tract is not compensatory. This means that if a block is surrounded by a dozen pharmacies, its score still just counts as 1, because that essential urban function is covered even if there is only one facility that provides the service. Therefore, the census tracts with higher score are those areas where all essential urban functions are available regardless of the number of facilities that provide a particular service.

3. Another point is that there is no description of Chilean census tracts (CT): do all of the CT polygons have inhabitants? Are they more or less homogeneous in terms of population, with small CT units in city centres and larger one in the outskirts of the metropolitan area or within “empty” or natural areas? I suggest that the authors include a small introduction, as there are no numerical outputs (besides percentages) and because having different levels of access to essential urban functions in different parts of the metropolitan area is in itself an unavoidable characteristic of such a large and heterogeneous area. The type of map that expresses this (Fig. 7) is expected: higher scores in the central areas (downtown Santiago and downtown of the major municipalities within the metropolitan area) and lower scores in the outskirts of the area. But if the map describes that a “low” level of access areas is important in terms of population, it remains to be seen.

RESPONSE 3.b: Thank you for this observation. We have added a clarification note as follows, along with a new map of density in the appendix:

A final clarification is required regarding the use of the census and its spatial representation. In our study, census tracts without population were excluded from the analysis, for instance, especially those that are referred to as natural areas and highways. In summary, 95% of the polygons analysed have inhabitants. This obliges us to inform the reader that the results for areas with low density have a two-fold interpretation; it means that, since there is a small amount of people living there, the market and state are not interested in investing resources to develop these areas, and it also means that these are areas where a significant investment may trigger new development regions for increasing density. In Appendix C, it is possible to see the density of the Metropolitan Area of Santiago; nevertheless, the correlation measured by the Pearson coefficient between the population and the resultant scores of the 15-minute city assessments is 0.041, which is considered low.

4. The paper has some speculative sentences that are not supported by the type of analysis that was conducted. Examples of these are lines 262–265 (concentration on a metropolitan mayor vs. local mayors) and 75–77 (the pandemic has triggered the discussion of the implementation of the 15-minute city). Please refrain from unsubstantiated speculation, as the focus of the paper is the extraction of components based on the 28 selected variables and the mapping of the PCA results.

RESPONSE 4.b: Thank you for this observation, but we have to disagree on refraining from stating these sentences because they represent relevant reflections that have emerged from the findings. Nevertheless, we have rephrased these sentences as follows:

Nowadays, each commune has its own municipal administration, and there is no need to coordinate urban transformations. Each district has its own objectives, budgets and priorities. Such governance may hinder the actual implementation of the 15-minute city on a broader scale because of the lack of metropolitan planning apparatuses. The idea of a metropolitan mayor or a metropolitan planning office may be suitable for resolving this problem. We believe that a new manner of governance is required to organise strategic decision-making to guide the common good of the 34 districts that currently lack such coordination.

Further, regarding the pandemic that triggered the discussion on the 15-minute city, we have included a footnote that indicates how the National Council of Urban Development fostered this discussion and suggested its implementation for the post-pandemic scenario.

5. The biggest contribution of the paper is the presentation of a methodology to reduce a bunch of variables to a tractable set of components and map them to support decisions on territory, possibly on prioritizing state-driven interventions, or pointing out business opportunities for the private sector, if and where they are needed. This being said, I think the discussion around the composite indicator of Figure 7 could be extended, as it is a mere analysis of the population within each interval. We do not know, for instance, if there are significant urban centres/municipalities that have large factions of population with low access levels.

RESPONSE 5.b: Thank you for highlighting this. We agree that there is more to be said about this result. We have added some reflections as follows:

On the other hand, peripheral districts, such as Quilicura, El Bosque, La Granja, Lo Espejo, Renca, La Cisterna and San Ramón, are facing a severe lack of essential urban functions in this aggregate indicator. These are low-income districts for which the possibility of managing funds to develop new centralities is actually extremely difficult to achieve because of the manner in which municipalities receive their budgets in the Chilean system, all based on a self-managed way of creating wealth and increasing their capital for making investments. Being low-income districts, the market is not quite interested in investing in these areas. This is problematic, considering that the less available functions in these areas are privately funded initiatives in a city where the central planning system cannot force the market to develop new areas nor can municipalities generate initiatives that may compete with private companies.

6. To conclude, as a minor suggestion, I do not think that authors have to indicate the licensee/license holder of the software used (lines 140–143).

RESPONSE 6.b: This has been amended. Thank you for pointing this out.

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

Authors have answered the main concerns and have improved the content, clarifying my previous doubts.