Analysis of Built-Up Areas of Small Polish Cities with the Use of Deep Learning and Geographically Weighted Regression

Round 1

Reviewer 1 Report

The manuscript has been substantially improved. Thanks for the effort in this round of revision. I have two additional comments on this version:

Lines 181-183: The authors mentioned that “The 2015 dataset has been split into three subsets. First subset intended for training accounted for 90% of the dataset. Validation subset was formed using 5% of the dataset”. The amount of data used in validation is only 5%? Is this sufficient? The description of the datasets used for training, validation, and testing is also unclear. If my understanding is correct, the authors used the entire dataset for 2015 in training, validation, and testing, but only 5% of 2019 in testing. What about the rest of 2019 dataset?

Table 4: The authors misunderstood my previous comment. I meant the authors should use p-value < 0.01, 0.05, or 0.1 instead of p-value → 0.

Author Response

Dear Reviewer,

We would like to thank for careful and thorough reading of this manuscript and for the thoughtful comments and constructive suggestions, which help to improve the quality of this paper. We have incorporated all the changes as suggested.

Here is a point-by-point response to your comments and concerns:

Lines 181-183: The authors mentioned that “The 2015 dataset has been split into three subsets. First subset intended for training accounted for 90% of the dataset. Validation subset was formed using 5% of the dataset”. The amount of data used in validation is only 5%? Is this sufficient? The description of the datasets used for training, validation, and testing is also unclear. If my understanding is correct, the authors used the entire dataset for 2015 in training, validation, and testing, but only 5% of 2019 in testing. What about the rest of 2019 dataset?

We confirm that the following proportion of train, validation and test split is correct. We did not explicitly describe the purpose of using each subset because of mentioning Keras as a deep learning framework where the above-mentioned subset names are the part of the naming convention. Thank you for noticing the ambiguity. We added relevant clarifications to the paper. Validation subset was also used during neural network training for measuring performance in the intermediate training steps and also for early stopping. In addition, test subsets contained data that the machine learning model had never seen during training. We used them for final performance evaluation. Both of them were large enough (more than 6.2k items) to calculate relevant metrics. We made an attempt to use a larger proportion of the 2019 dataset as test but it did not produce significantly different results. At the end we knew that perceptual evaluation is needed for such a task and didn’t want to artificially decrease the velocity of the research process.

Table 4: The authors misunderstood my previous comment. I meant the authors should use p-value < 0.01, 0.05, or 0.1 instead of p-value → 0.

Sorry, we misunderstood the previous suggestion. We made the change to the correct p-value. Thank you for noticing this issue. (Page 17, Table 4)

 

Once again, thank you for the in-depth review.

On behalf of the research team,

Marta Nalej, PhD

 

Author Response File: Author Response.docx

Reviewer 2 Report

This paper measures the statistical associations between several socioeconomic variables and model-generated measurements of the extent of built-up areas in small cities in Poland. Machine learning algorithms are applied to remote sensing data to measure built up areas, and the socioeconomic variables are drawn from administrative sources. Very extensive references to prior literature are included, and the paper benefits from useful illustrations. On the whole, the data are suitable and the methods seem to be correctly employed (though I am better placed to assess the regression methods than the pattern recognition parts of the paper). The paper is clearly written, though there are a few typos which I detail below.

The only significant issue I have with the paper is that it makes causal claims about the relationships between the socioeconomic variables and the extent of urban fabric that are not justified in the paper and almost certainly cannot be proven given the data available to the authors. This can be corrected with some rewriting of the text, after which I think the paper will be suitable for publication.

This issue arises throughout the paper, from the abstract onwards. Claims are made implying one-way causality, e.g. “population density had the greatest influence on the share of built up areas”, GWR “was used to explain the influence of demographic, social and economic conditions on the spatial variation”, “the share of built up areas…was positively influenced by population density”, etc.

In fact, no evidence of given that causation runs from current socioeconomic conditions to the extent of urban fabric and not in the opposite direction or from some other unobserved set of factors to both socioeconomic and urban landscape. Both of these other directions of causation are possible and even likely, so the authors should be more realistic about what their analysis shows. For example, population density might be increased if local development policies favour construction of built-up areas, because these areas could provide dwellings for people to move into. Then it would be more accurate to say that built-up areas cause population density, but actually both are influenced by development policies. For variables such as unemployment or flows of new enterprises that can vary dramatically over relatively short time scales, it is even more implausible to claim that they “cause” a greater or lesser extent of built-up areas. Built-up areas represent a stock of land use developed over a long period of time, so it is very unlikely that a flow variable measured over a given year could have any meaningful effect on it. Of course small cities with more built-up areas might have higher unemployment at a given time or even over a long period, but that is evidence of correlation rather than causation.

A more appropriate claim would be that the authors found significant associations between the socioeconomic variables and the extent of built-up areas, and it would also be reasonable to make claim that the extent of such areas is bigger or smaller in areas with higher or lower values of some socioeconomic variable. I suggest the authors moderate all the causal terminology in the paper in this way. It is also customary to mention in the conclusions (or a section on limitations of the work) that the paper is not able to make any causal claims. In general, a bit more discussion of the limitations of the work would be useful, perhaps in the concluding section.

On one specific point, I was not convinced about the argument on p19 that a negative coefficient on the enterprises variable implies “buildings related to economic activity do not play a significant role in the share of built-up areas” in some places. Since your variable counts the number of establishments, the more or less negative coefficient on this variable might be something to do with the scale of businesses being set up in different areas, e.g. maybe many small businesses are being set up in less built-up cities, while the larger enterprises (of which there are fewer) are concentrated in the bigger, more densely developed cities.

Specific comments

• In line 10 of the abstract, “covered entire Poland” seems to be missing a word or two.
• In line 42 on p2, “research directions” should probably be “research direction”
• In line 171 on p8, “The” is missing in front of “2015 dataset”.
• In line 177 on p8, there seems to be a word or two missing in “…a pixel will be treated as when its neural network output…”
• In line 196 on p9, the punctuation could be improved by replacing a comma with a full stop or semicolon. “…indicates a hot spot[.] the larger the…”
• In the discussion of how the average share of built-up areas varies on p11, I was not clear how the denominator if this share was determined. Is the total area encompassed by each small city an administrative measurement or was it derived from the remote sensing data in some way (as the numerator was)?
• In line 290 on p13, one doesn’t usually refer to “a larger unemployment” but rather “a higher unemployment rate”.
• In line 419 on p21, the wording of “may be an impulse to look for” is a bit odd. You might consider something like “suggests a need to examine”.
• In line 485 on p22, I suggest something like “found” in place of “diagnosed”.

Author Response

Dear Reviewer,

Thank you for providing us with the opportunity to submit a revised draft of our manuscript. We appreciate the time and effort that you have dedicated to the review process. We have strived to incorporate changes reflecting the bulk of the suggestions provided by your review.

Here is a point-by-point response to your comments and concerns:

The only significant issue I have with the paper is that it makes causal claims about the relationships between the socioeconomic variables and the extent of urban fabric that are not justified in the paper and almost certainly cannot be proven given the data available to the authors. This can be corrected with some rewriting of the text, after which I think the paper will be suitable for publication.

Thank you for pointing out this issue. We agree that the text needs to be adjusted. Causal claims about the relationships between the socioeconomic variables and the share of built-up areas in the area of small cities were removed. Therefore, major changes have been introduced to the paper (lines 361, .383, 387, 396, 409, 427, 433, 437–438, 449, 46, 479, 481, 509, 517–51, 524)

Thank you very much for your relevant and helpful insights on the impact of newly registered domestic economic entities (Page 26, lines 469–471).

In line 10 of the abstract, “covered entire Poland” seems to be missing a word or two.

Relevant line has been changed to: “The main goal of the paper was to verify the impact of selected socio-economic factors on the share of built-up areas in small 665 Polish cities in 2019.” Thank you for noticing the issue (line 10).

In line 42 on p2, “research directions” should probably be “research direction”

Relevant line has been changed to: “The study of the physical structure of the landscapes and the spatial organisation of small cities was mentioned as another much needed research direction.” (Page 2, line 45).

In line 171 on p8, “The” is missing in front of “2015 dataset”.

It has been corrected (Page 10, line 205).

In line 177 on p8, there seems to be a word or two missing in “…a pixel will be treated as when its neural network output…”

Fixed. Thank you for noticing this issue (Page 10–11, lines 205–222).

In line 196 on p9, the punctuation could be improved by replacing a comma with a full stop or semicolon. “…indicates a hot spot[.] the larger the…”

Relevant line has been changed to: “The occurrence of statistically significant positive z-scores indicates a hot spot. The larger the …” (Page 11, line 240).

In the discussion of how the average share of built-up areas varies on p11, I was not clear how the denominator if this share was determined. Is the total area encompassed by each small city an administrative measurement or was it derived from the remote sensing data in some way (as the numerator was)?

The built-up area for each city has been determined in relation to its administrative area. Clarifications have been added in the text (Page 13 lines 298–301).

In line 290 on p13, one doesn’t usually refer to “a larger unemployment” but rather “a higher unemployment rate”.

Relevant line has been changed to: “On the other hand, a higher unemployment rate (UNEMPLOYMENT) results in a smaller share of built-up areas in the area of the surveyed units.”(Page 21, lines 359–360)

In line 419 on p21, the wording of “may be an impulse to look for” is a bit odd. You might consider something like “suggests a need to examine”.

Relevant line has been changed to: “The fact that cities are built up is not debatable, but the regional variation of the built-up areas suggests a need to examine the reasons and factors influencing this fact.” Thank you for this suggestion (Page 28, line 508).

In line 485 on p22, I suggest something like “found” in place of “diagnosed”.

Relevant line has been changed to: “It was found that population density and the construction of new residential buildings had the greatest influence on the share of built-up areas in the area of the surveyed cities.” (Page 28, line 500–501)

 

Once again, thank you for the in-depth review.

On behalf of the research team,

Marta Nalej, PhD

Author Response File: Author Response.docx

Reviewer 3 Report

General remarks:

English edition needed.

I am against translating only part of regions names. Please leave original names for all.

The article is very synthetic. The beginning arouses interest, but the following chapters leave us unsatisfied. The tool is described and successfully implemented. Interesting analyzes were made for the case of a specific country, showing the applicability. A build-up analysis was performed in small cities using satellite imagery and then the results were compared with descriptive variables for these areas. There is no general conclusion. Please consider whether the problem could be solved by individual variables results description and then the (expanded) description of the global model.

 

Abstract:

Please rewrite according to the structure: Background, Methods, Results, Conclusion (see Instructions for Authors).

1. Introduction

Comment:

Adequate, interesting. Similar aspects are found in different places of the Introduction. Too much information about small cities and too briefly described the use of remote sensing in this type of research.

Remarks:

Please sort out.

Line 36: Do not start the sentence with a number.

Lines 55-58: The list is impossible to read in that form.

Line 71: Dot missing.

Lines 81-83: Not clearly written, please reedit.

Lines 81-85: The objective described too briefly, lack of highlighted the main conclusions required in the Instructions for Authors.

Lines 88-94: I do not see the need for such description of the content. Please remove.

2. Materials and Methods

Comment:

This chapter describes the software and method used. Detailed comments below.

Remarks:

Lines 97-103: Unconnected with research and unnecessary – information included in figure 1.

Lines 171, 173: Do not start the sentence with a number.

Lines 176-178: Reedit the sentence, please.

Lines 200-222 : It is a review of the approach applications, not a methodology.

3. Results

Comment:

3.1-3.2. The results of the share of buid-up areas analyzes within the small cities area are presented.

3.3. Each of the selected variables was sequentially analyzed within the designated build-up areas. The maps and brief descriptions of their content were presented. Sometimes with abbreviated interpretation. The descriptions mainly focus on whether the impact was significant or not, which is relevant to the analysis.

Remarks:

The Results chapter breaks off without summing up the entirety and the experimental conclusions.

4. Discussion

Comment:

A lot of repetitions of what we already know from the previous chapter. Please follow the Instructions for Authors where this section should include results interpretation in perspective of the working hypotheses and the findings implications in the broadest context.

Remarks:

Chapter number is 4.4., should be 4.

Lines 418-427: Good start of the Discussion.

Lines 458-470: Very interesting.

Lines 471-479: This should be placed before the previous paragraph.

5. Conclusion

A paragraph unnecessarily separated from the previous chapter. “This section is not mandatory, but can be added to the manuscript if the discussion is unusually long or complex.”

Author Response

Dear Reviewer,

Thank you for providing us with the opportunity to submit a revised draft of our manuscript. We appreciate the time and effort that you have dedicated to the review process. We have strived to incorporate changes reflecting the bulk of the suggestions provided by your review.

Here is a point-by-point response to your comments and concerns:

I am against translating only part of regions names. Please leave original names for all.

All region names were translated to Polish. Please see Figure 5 page 17 and further in the text.

Abstract: Please rewrite according to the structure: Background, Methods, Results, Conclusion (see Instructions for Authors).

We believe that the abstract was prepared according to the instructions for authors.

Background (the question addressed in a broad context and highlight the purpose of the study): The main goal of the paper was to verify the impact of selected socio-economic factors on the share of built-up areas in small 665 Polish cities in 2019.

Data (can be considered inconsistent with the instructions, but as they form the basis of the research they can be also included in the "Methods"): Data from the Database of Topographic Objects (BDOT), Sentinel-2 satellite imagery from 2015 and 2019, and Local Data Bank by Statistics Poland form 2019 were used in the research.

Methods (Methods: briefly describe the main methods or treatments applied): A machine learning segmentation procedure was used to obtain the data on the occurrence of built-up areas. Hot Spot (Getis-Ord Gi*) analysis and geographically weighted regression (GWR) was applied to explain spatially varying impact of factors related to population, spatial and economic development and living standards on the share of built-up areas in the area of small cities.

Results (summarize the article's main findings): It was confirmed that population density had the greatest influence on the share of built-up areas in the area of the cities studied. The influence of the other socio-economic factors examined, related to the spatial and economic development of the cities and the quality of life of the inhabitants, showed great regional variation.

Conclusion (indicate the main conclusions or interpretations): The results also indicated that the share of built-up areas in the area of the cities under study is a result of the conditions under which they were established and developed throughout their existence, and not only of the socio-economic factors affecting them at present.

Too much information about small cities and too briefly described the use of remote sensing in this type of research.

Including a decent number of literature references in the field of small towns was crucial from the point of view of the conclusion and discussion. In the study, we used remote sensing data (Sentinel-2 scenes) as input data. Examples of research of urbanized areas with the use of various remote sensing data, satellite scenes, but also LIDAR data are quoted in lines 62-66. The main goal of the paper was to verify the impact of selected socio-economic factors on the share of built-up areas in small Polish cities.  We believe that an in-depth analysis of the utilized remote sensing techniques, although interesting, would only distract the reader from the main purpose of this study.

Line 36: Do not start the sentence with a number.

Relevant line was changed to: “The smallest Polish city (Wyśmierzyce) had 894 inhabitants in 2019.” (Page 1, lines 39–40­)

Lines 55-58: The list is impossible to read in that form.

We decided to remove this information as ISO standards were not used further in the paper.

Removed content: “These are developed in the form of ISO ISO37120:2018 and SO/DIS37122:2018 standards and cover topics such as: economy, education, energy, environment and climate change, finance, governance, health, housing, population and social conditions, recreation, security, solid waste, sport and culture, telecommunications, transport, urban/local economy and food security, urban planning, wastewater, water [19,20].”

Lines 81-83: Not clearly written, please reedit.

and

Lines 81-85: The objective described too briefly, lack of highlighted the main conclusions required in the Instructions for Authors

Thank you for noticing this issue. We decided to change relevant parts of the text to: “The main objective of the study was to undertake research on the impact of selected socio-economic factors on the share of built-up areas in small Polish cities and their spatial differentiation across the country. Based on literature review, various variables describing the determinants of the share of built-up land in the urban area were identified and used in the study of Polish small cities (Table 1).”(Page 3, lines 108–110).

Lines 88-94: I do not see the need for such description of the content. Please remove.

The paragraph (lines 88-94) was removed from the paper.

Lines 97-103: Unconnected with research and unnecessary – information included in figure 1

As suggested, part of the description of the study area has been deleted.

The study area included small Polish cities. Poland is located in the central-eastern part of Europe. Poland is bordered to the west by Germany, to the south by the Czech Republic and Slovakia, to the east by Ukraine and Belarus, and to the north-east by Lithuania and Russia (Kaliningrad Oblast). Poland's northern border is formed by the waters of the Baltic Sea. The area of Poland is 322 thousand km². The capital and largest city is Warsaw. The population of Poland in 2019 was 38,253,955 [10], of which 60% lived in cities, and the number of administrative units defined as cities was 940.

Lines 171, 173: Do not start the sentence with a number.

Relevant lines have been changed to: “The 2015 dataset has been split into three subsets. First subset intended for training accounted for 90% of the dataset. Validation subset was formed using 5% of the dataset.” (Page 10, line 205)

Lines 176-178: Reedit the sentence, please.

Fixed. Thank you for noticing this issue (Page 10–11 , line 210–222).

Lines 200-222: It is a review of the approach applications, not a methodology.

Paragraph placement has been corrected. It was moved to the Introduction section (Page 2  lines 75–98).

The Results chapter breaks off without summing up the entirety and the experimental conclusions.

A summary of the "Results" section has been added in the lines 499–505 (Page 28).

Chapter number is 4.4., should be 4.

Chapter number has been corrected (Page 28, line 506).

Lines 471-479: This should be placed before the previous paragraph.

The indicated paragraph (lines 471–479) has been moved to the line 580–588 (Page 30).

A paragraph unnecessarily separated from the previous chapter. “This section is not mandatory, but can be added to the manuscript if the discussion is unusually long or complex.”

The " Conclusions" section has been removed.

English edition needed.

Thank you for this suggestion. We decided that the text needs to undergo a language check.

 

Once again, thank you for the in-depth review.

On behalf of the research team,

Marta Nalej, PhD

Author Response File: Author Response.docx

Round 2

Reviewer 1 Report

The manuscript entitled “Analysis of built-up areas of small Polish cities with the use of deep learning and geographically weighted regression” submitted to journal Geosciences by Adamiak et al. studied the impacts of different socioeconomic factors on the share of built-up areas in small cities of Poland. The authors identified the share of built-up areas using a machine learning model, and then determined the impacts using OLS and GWR models. In general, this study is interesting, but one of my major concerns is the accuracy of the applied machine learning model. More details on this should be provided in the revision.

Line 36: “894 people were living in Poland's smallest city” – 894 people?

Table 1: Details such as the time of datasets and their spatial/temporal resolutions should be provided.

Figure 1: It is unclear how the boundaries of cities are determined.

Lines 200-223: This part should be moved to Introduction.

Line 288 and Table 4: Please correct “p-value → 0” to the more commonly used form: “p-value < ****”.

Lines 383-393: It is rather unclear why the number of newly registered domestic economic entities can negatively impact the share of built-up areas.

Lines 418-457: Although under the section title “Discussion and future directions”, this part is more like a summary of findings. An in-depth discussion is needed here.

Lines 468-470: I believe the original natural ecosystems and the presence of national parks and preserves may also influence the share of built-up area, which should be included if possible.

Line 471-472: “The inclusion of a temporal variable can not only improve inference but also reveal new relationships.” – please elaborate.

Author Response

Dear Reviewer,

Thank you for providing us with the opportunity to submit a revised draft of our manuscript. We appreciate the time and effort that you have dedicated to the review process. We have strived to incorporate changes reflecting the bulk of the suggestions provided by your review.

Here is a point-by-point response to your comments and concerns:

In general, this study is interesting, but one of my major concerns is the accuracy of the applied machine learning model. More details on this should be provided in the revision.

Concerns regarding the accuracy of the machine learning model are fully justified. We really appreciate this comment and will be more than happy to discuss it. As stated in the paper the model reached 0.548 intersection over union (IoU) score for test data acquired in the year 2015 and 0.588 for year 2019. Contemporary state-of-the-art built-up semantic segmentation models for aerial imagery are capable of reaching around 0.6 for Sentinel-2 images [1] and around 0.8 IoU for high-resolution RGB datasets [2]. An important difference between state-of-the art architectures and our approach lies is the segmentation masks that accompany the satellite imagery dataset. This influences the objective that the segmentation neural network has to fulfil.

In our study we utilized the Database of Topographic Objects (BDOT) to prepare segmentation masks. BDOT is the largest and most precise source of information on built-up areas available in Poland. Unfortunately, it is not free from defects. During careful analysis one can discover that in multiple areas the delineated regions are visibly under or oversegmented. This is clearly visible in the layers related to built-up areas. In consequence the input segmentation masks for malformed regions are not precise enough to fully reflect the area land cover which makes the task really demanding. Manually fixing the whole BDOT dataset was not possible due to costs and time needed to complete such a task. Therefore, we decided to introduce a fine-tailored machine learning model to mitigate above mentioned issues. We have considerable experience with the DeepLabV3+ model which we applied in various scenarios related to satellite imagery interpretation. During previous research we discovered that feeding the model with a substantially large dataset and tuning it can result with an output that slightly surpasses the ground truth segmentation and corrects minor defects. This was the main reason for training the network with the dataset covering the whole country not only with areas containing small cities. Although such an approach yields reasonable results it has its limits and  affects the output metrics significantly because they are calculated in relation to the ground truth based on BDOT not for manually prepared segmentation masks. That is why we also preformed perceptual evaluation by checking the neural network output in correspondence to the Sentinel-2 images.

Please see the attachment: Figure 1. Ground-truth built-up areas segmentation (left), model inference result (right).

Thank you for pointing out the issue. We are now sure that we missed this crucial information when describing the results yielded by the machine learning model. Relevant explanation has been added to the paper (page 10-11, lines 205-222).

Reference:

Guillaume Chhor and Cristian Bartolome Aramburu, Satellite Image Segmentation for Building Detection using U-net, http://cs229.stanford.edu/proj2017/final-reports/5243715.pdf, 2017

Emmanuel Maggiori, Yuliya Tarabalka, Guillaume Charpiat and Pierre Alliez. “Can Semantic Labeling Methods Generalize to Any City? The Inria Aerial Image Labeling Benchmark”. IEEE International Geoscience and Remote Sensing Symposium (IGARSS), https://project.inria.fr/aerialimagelabeling, 2017.

Line 36: 894 people were living in Poland's smallest city – 894 people?

We confirm that Wyśmierzyce is the smallest town in Poland and has only 894 inhabitants. We have added the name of the city and its location to the relevant part of the text (Page 1 line 39).

Table 1: Details such as the time of datasets and their spatial/temporal resolutions should be provided

We consider current description as sufficient to identify the spatiotemporal resolution of the dataset. In the table description we state that the values were calculated for year 2019 and limited to the area of small cities in Poland (Page 3).

Figure 1: It is unclear how the boundaries of cities are determined

The Figure 1 caption has been fixed by adding information regarding the source of cities boundaries delineation which is the Head Office of Geodesy and Cartography (GUGiK) borders register (Page 6 line 144).

Lines 200-223: This part should be moved to Introduction.

Paragraph placement has been corrected (Page 2, lines75-98).

Line 288 and Table 4: Please correct p-value → 0 to the more commonly used form p-value < ****

Fixed. Thank you for noticing this issue (Page 21) .

Lines 383-393: It is rather unclear why the number of newly registered domestic economic entities can negatively impact the share of built-up areas.

Possible explanation of this phenomenon is that owners of newly registered in the National Business Registry Number – REGON domestic economic entities choose their place of residence as the company's headquarters. Therefore, the number of newly registered enterprises does not affect the share of built-up areas in the area of small cities. Clarifications have been added in the text (Page 26, lines 469–471).

Lines 418-457: Although under the section title “Discussion and future directions”, this part is more like a summary of findings. An in-depth discussion is needed here.

“Discussion and future directions” (section 4) was revised and expanded. The limitations of the method and references to the variables used in the study were discussed. Paragraph “The inclusion of a temporal variable (...)”was moved to lines 580–588.

Lines 468-470: I believe the original natural ecosystems and the presence of national parks and preserves may also influence the share of built-up area, which should be included if possible.

Thank you for an interesting idea. We decided to check the impact of strict protection areas i.e. national parks and preserves on the share of built-up areas of small Polish cities in 2019. Data on the area and location of strict protection areas were taken from resources of The General Directorate for Environmental Protection in Poland. The share of these areas in the area of the examined cities was then encoded in the model as the ”Protected Areas” variable. The result of performing the analysis on the extended dataset is that introducing the additional variable to the study does not significantly affect quality of the global regression model (OLS). The coefficient of determination for the global model (R²) reached 0.791, adjusted R² 0.797, but AICc (Akaike's Information Criterion) was lower and equal to 860.331.

 

Variable	Coefficient	Standard Error	t-Value	p-Value
POP_DENSITY	0.9198	0.0206	31.6659	p-value < ****
BUILDINGS	0.2030	0.0261	6.0867	p-value < ****
UNEMPLOYMENT	-0.0689	0.0183	-3.7530	p-value < ****
WORK_POP	-0.0298	0.0191	-1.3481	0.1780
ENTERPRISES	-0.2100	0.0296	-3.9534	p-value < ****
LIVING_STANDARD	-0.0148	0.0202	-0.6942	0.4877
Protected Areas	0,0187	0.0457	1.9030	0.0574

 

For geographically weighted regression (GWR), R² reached 0.875, adjusted R² 0.848 and AICc 701.831. Strict protection areas are present only in 11.7% of the examined small towns in Poland. Their occurrence is related not only to the natural conditions, but also to top-down legal decisions. Although for the “Protected Areas” variable two opposite tendencies of influence on the share of built-up area in small cities were observed. The influence is statistically insignificant, except for a few cases in Małopolski and Mazowiecki regions (Figure 2).  Therefore, we decided not to include this variable in the study.

Please see the attachment: Figure 2. Value (a) and significance (b) of the coefficient of the impact of the share of Protected Areas.

 on the share of built-up land in the area of small Polish cities in 2019.

Line 471-472: “The inclusion of a temporal variable can not only improve inference but also reveal new relationships.” – please elaborate.

Thank you for pointing out the issue. Taking into account changes in variables values over time, e.g. decades, may allow to indicate which of them affected the changes in the share of built-up areas in the area of small cities in the period under study. It can be expected that the population growth will increase the share of built-up areas in cities. On the other hand, the increase in the number of newly registered domestic economic entities may not necessarily cause an increase in share of built-up areas in the area of small cities. This creates an opportunity for further research. Relevant explanation has been added to the paper (Page 30, lines 580–588).

 

Once again, thank you for the in-depth review.

On behalf of the research team,

Marta Nalej, PhD

 

Author Response File: Author Response.docx

Reviewer 3 Report

The paper has improved. Thank you for referring to the comments. I am satisfied with the quality of the corrections. However, I am concerned about the lack of significant changes in the abstract. This is an important element affecting the readership of the article, and in this case it is not fluent to read and does not reflect the interesting content of the text. The first sentence is not an introduction to the topic, but a goal that should arise later. Please re-edit the entire abstract with particular emphasis on linguistic correctness and interesting introduction of the reader to the subject.

Author Response

Dear Reviewer,

We would like to thank for careful and thorough reading of this manuscript and for the thoughtful comments and constructive suggestions, which help to improve the quality of this paper. We have incorporated  all the changes as suggested.

Here is a point-by-point response to your comments and concerns:

The paper has improved. Thank you for referring to the comments. I am satisfied with the quality of the corrections. However, I am concerned about the lack of significant changes in the abstract. This is an important element affecting the readership of the article, and in this case it is not fluent to read and does not reflect the interesting content of the text. The first sentence is not an introduction to the topic, but a goal that should arise later. Please re-edit the entire abstract with particular emphasis on linguistic correctness and interesting introduction of the reader to the subject.

Sorry, we didn't quite understand the previous suggestion. An introduction has been added to the abstract (Page 1, lines 9–12)

Once again, thank you for the in-depth review.

On behalf of the research team,

Marta Nalej, PhD

 

Author Response File: Author Response.docx

Round 3

Reviewer 1 Report

Thanks for the explanation. I believe the paper is ready for publication.