Exploring the Relationship between the Spatial Distribution of Different Age Populations and Points of Interest (POI) in China

Round 1

Reviewer 1 Report

This paper introduces a very specific study that evaluates the relations and correlations between different population groups and POIs. The approach is applied at the county scale in China.

Overall the approach is well described and comprehensive. On the methodological and applicable principles, I have several comments:

• First of all, the reference data is taken from a single time in 2010 that makes the interest of the whole approach quite limited. It would have been much better to evaluate the same patterns and correlations over time as this might lead to more useful comparisons and conclusions.
• POIs is considered as an explanatory variable but the reverse might be true and/or considered as well.
• Correlations: it appears to me that most of the correlation analysis apply some well- known indices, is there any contribution in this regard?
• The level of analysis is relatively macro both in space where additional criteria might be considered (very urban, urban, rural, peri-urban etc.).
• POIs might be also categorised and the way correlations between POI and population patterns might be presented in a more comprehensive manner as figures 7 to 12 are not completely explicit regarding the patterns that appear.
• Regarding the findings, the authors might stress the unexpected results and also discrepancies - if any - between what might be expected from the POI-population model and what might happen is some specific places.

Overall the paper might be of interest if the considered data might be not limited to a single time in order to provide more useful insights and possible valuable correlation patterns. Also the authors should stress the methodological contribution apart from the application of quite common correlation analysis. These are major issues to address to provide a valuable contribution to IJGI.

Additional comments:

The starting paragraph of section 2 is general material and recommendation for publication regulations and should be withdrawn.

 

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 2 Report

1. Detailed comments to the text have been provided in the attached PDF file.

2. I had mixed feelings about reading the reviewed article. I have an overwhelming impression that an apparent problem is being solved in it.

A. The aim of the analyzes was to show that the POI data can be used to assess the age structure of the population. The first question that arises in this connection is: where is a country where accurate data on the amount, location and nature of POIs are more readily available than demographic data?

B. The work does not mention such aspects as population movement (migrations and birth rate) and its ethnic and cultural diversity. In a country as vast and diverse as China, this must be of great importance.

C. I was a bit surprised that among so many types of POIs, sometimes quite marginal (for example catering services), there are no places of worship: churches, mosques, Buddhist temples, necropolises, etc. Do they not exist in China at all, or did not have social significance? However, this is not the case in many countries.

D. I believe that it is a mistake to combine in one category, such POIs that are easy to create and relatively mobile, such as many types of commercial services with material objects, costly technical infrastructure or institutions requiring highly qualified staff. Launching an ATM or catering service is possible practically overnight. Building an airport or establishing a university are activities that require many years of planning and implementation. Such investments are made for decades, also taking into account demographic forecasts. At the time of launch, their potential is often not fully realized yet.

E. Most of the analyzed POIs relate to various types of social services. The existence of a demand for a given service is satisfied by the construction / organization of a specific type of POI. On the other hand, being able to access a particular service attracts people and increases the population. The presence of a relationship between the number of people and the number of POIs is obvious. More interesting are the cases (administrative units) which deviate from this rule. The authors, analyzing the results presented in Figures 7-9, completely omit this aspect. Reading the abstract of the article, I expected an attempt to create some kind of prognostic model that would allow to estimate the share of individual age groups based on POI data. There is nothing like that in the text. I repeat once again - there is only a statement of certain accounts that do not explain anything.

F. I believe that the reference of the results to the existing regional divisions and the classification of cities in terms of population is insufficient. In my opinion, it would be much more important to divide the analyzed administrative units according to the type of age structure - proportions of individual groups. This would be much more important than operating on absolute counts. More broadly, I believe that the analysis of correlations, factors and interactions should also be carried out on relative values ​​- the percentage share of a given age group in the entire population. Since there is a (obvious) relationship between the number of POIs and the total population in a given administrative unit, and the number of people in a given age group increases with the size of the population, there must of course also be a relationship between the POI and the number of people in that age category. The relationship between the number of POIs and the percentage share of a given age group is not at all obvious. The finding of such a relationship would be a real discovery of the peer reviewed work.

I have the impression that this study cannot be improved because its basic assumptions are false. However, I would like to give the authors a chance to convince me that I am wrong and that the most important shortcomings can be corrected.

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 3 Report

My remarks are as followes:

1. there is no description how the POI influence the spatial distribution of different age groups. How the relationship between POI and age groups has been determined, please explain the research assumptions and discuss the results.
2. please delete author instruction in lines 116-130.
3. there is no explanation or data discussion why population data is for 2010 and POI data for 2015, please add an explanation about it.
4.  in tables 1 and 2 please provide the number of objects
5. Figures 1, 2, 3, 5 could you please explain what is shown in the extra frames with islands? What does it show?
6. Figures 2, 3, 5 there is no explanation what the selected regions show, please explain it in the map legend

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The authors made every efforts to respond appropriately to my previous comment. In order to give more value to their paper some of the responses made for example regarding comments 1 (i.e., on the temporal validity and limit of the data), 2 (i.e., POI as an explanatory variable), 3 (Geodetector tool), point 4 (scale effect), 5 (POI categorisation), 6 and 7 (contribution) should be reported either in the paper when appropriate or in a final discussion section. This will be of value for the readers (and for the authors to give more credit to their work).

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 2 Report

The authors of the article ignored most of my comments on the logical and methodological foundations of their analyzes. They did it very politely, most often agreeing with various doubts, but finally stating that "it cannot be done". Either the reason is the lack of relevant data or the statement that "this is a topic for a separate study". In some cases, I can agree with this, and most not.

1. For example, I do not quite believe that it was impossible to account for population movements (birth rate, migrations) and cultural and ethnic diversity. The effect of the census in most countries is also such data. Demographic changes are determined in relation to the previous census (in this case it would be the census from 2000), and the collected data include information about ethnicity and religious denomination, etc. And what is the differentiation of gender proportions in the analyzed spatial units? Could it make any difference?

2. Also, my remark that it was necessary to refer not only to the regional classification and the classification of cities in terms of population size, but also in terms of the age structure of the population, was ignored. My point was that having demographic data, it was possible to do your own regionalization based on different types of the population "age pyramid" and check whether there are differences between the regions distinguished in such a way in the relationship between the number of age groups and the type and number of POIs. The authors focused on showing that the correlation between different POI categories and the percentage of different age groups is weak or absent. For example, they used the argument that the planning of a network of schools, kindergartens and nurseries is related to the number of children, not to their percentage. This is, in my opinion, a flawed argument, because in one administrative unit (large city) we can have 5 POIs of schools for each of which 1,000 children attend, and in another (rural) POI there can be 50 schools, with 100 students in each.

3. I have the impression that the authors of the article did not understand my remark from point 6 of the previous review, concerning the attention to cases deviating from a relationship between the number of people and the number of POIs shown in Figures 7-9. I meant points that represent single administrative units that are very far from the regression line. Pearson's linear correlation coefficient is very sensitive to outliers. On the one hand, the elimination of such cases may significantly increase the "r" value, and on the other hand, very valuable interpretations may result from the identification of these cases and the analysis of the reasons for their occurrence. For example, in Figure 7, the second graph in the first row (POPA1 vs. POIE2 in the SEC region) shows a data point with a disproportionately small number of POIs relative to the largest population. What is an administrative unit? Why is it so different from others?

In the attached PDF file I have marked two minor typing errors in the new parts of the text.

In the previous review, I noticed the lack of information about the principle of class divisions on both population and POI maps. In the new version of the article, explanations have been added in the appropriate places that the natural breaks method was used. It doesn't quite do the trick. The fact that the classification "natural breaks" is considered the best does not mean that it is optimal. Its optimality depends on the completely arbitrary number of classes determined by the operator. To properly select the number of classes, several divisions should be made by comparing the goodness of variance fit (GVF) index obtained for them - see https://en.wikipedia.org/wiki/Jenks_natural_breaks_optimization.

Comments for author File: Comments.pdf

Author Response

Please see the attachment

Author Response File: Author Response.docx