Trends in United States Human Footprint Revealed by New Spatial Metrics of Urbanization and Per Capita Land Change
Round 1
Reviewer 1 Report
The introduction clearly describes the research problem, presents the current state of knowledge, research conducted so far in the United States, and the purpose and scope of the authors' analyses. Perhaps the authors could also refer to works carried out in Europe or on other continents. The formulated conclusions are correct. The discussion presented is a valuable element of the article.
Congratulations to the authors of their comprehensive and interesting research. The only comments I have are about the layout of the article and the description of the methodology. I am slightly more concerned about the method of determining the final ranking of regions. The authors added up the positions in the ranking - which raises my doubts. Authors should adapt their paper to the editorial requirements. Detailed comments the authors can find below.
Thinking about recipients reading the paper, the authors should place the drawings directly below the text refers to them. Maps showing the results should be more visible. Instead of shades of gray, I suggest using colors. Authors should increase the contrast of the map drawing and possibly increase the scale as well.
The authors used data from censuses from 2000 and 2010. Was there a census in the USA in 2020, or did the pandemic make it impossible? I look forward to the publication that also includes newer data.
In lines 156-163, the authors wrote:
To reduce data aggregation bias introduced by shifting block group boundaries and to allow for direct comparisons of Census counts over time, we used 2000 and 2010 data from the National Historical Geographic Information System (NHGIS) that has been standardized to 2010 Census units (Manson et al. 2019). We assumed that residential housing is prohibited or greatly restricted in public open spaces (Tanner 2002), and therefore removed overlapping parks and protected areas, including water bodies and all areas managed by federal, state, district, or local agencies, from calculations of Census block group areas. This step refined the available private land area in affected block groups prior to
housing density calculations.
Thinking about readers from outside the US, I would like the authors to add a piece of short information about the reference units adopted for analysis, i.e., block group (census block?). What exactly are block groups?
In lines 174-175, the authors wrote:
We classified the density (units/hectare) of each block group into one of four density gradient levels: urban, suburban, exurban, and rural, following Theobald (2001, 2005)
The authors classified the density values into four classes: urban, suburban, exurban, and rural. The density change is gradual, fuzzy, just like the urban, suburban, exurban, and rural areas change smoothly. It is not sure whether the density values that define the boundaries of such classes are constant over time. Should urban, suburban, exurban, and rural be classified only based on the density indicator? I am asking the authors to justify the approach adopted in the article.
In lines 208-212, the authors wrote:
We then tested for similarity between the resulting 21 unique pairs of metric rankings (e.g. density and % developed, density and consumption, consumption and % efficient lands, etc.) using the Spearman's rank-order correlation on the assumption that no two rankings are strongly correlated (positively or negatively) and that each region has a unique trajectory of urbanization.
The authors should describe the purpose of determining the correlation indicators.
In lines 212-215, the authors wrote:
We summed each region's rankings across the seven metrics to derive a composite score where lower scores indicate more sustainable development. Lastly, based on the composite scores we ranked the megaregion growth trajectories from most (1) to least (11) sustainable.
Would you please justify the method of determining ranked megaregion growth trajectories? It is a classic case of multi-criteria evaluation.
Can rank values be added together? Are all indicators equally important for the sustainable assessment? Shouldn't the range of change in the value of each indicator be taken into account? For example, indicator X changes significantly. The Y indicator does not change much. Changes in the value of the X indicator and the Y indicator are used to determine the position in the ranking 1,2,…, 11. But for the final assessment, the ranking change of indicator X seems more critical than that of indicator Y due to the range of values. Maybe the authors should consider using multi-criteria analysis of outranking methods such as PROMETHEE or ELECTREE or method utility function.
Author Response
Response to Reviewer 1 Comments
We thank the reviewer for the kind words and constructive comments.
Point 1: Thinking about recipients reading the paper, the authors should place the drawings directly below the text refers to them. Maps showing the results should be more visible. Instead of shades of gray, I suggest using colors. Authors should increase the contrast of the map drawing and possibly increase the scale as well.
Response: The placement of figures in the original submission was based on the order of sections outlined in the “Free format” submission guidelines. The maps are already at full-page extent, so they cannot be made larger. This is a national scale study, so we chose to display map results at the continental scale and included zoom inset maps in some figures to provide additional detailed information for select urbanized areas. Readers will be able to download full resolution versions of the maps at time of publication. In addition, the available online mapping application at go.ncsu.edu/us-human-footprint allows users to drill down to individual counties, cities and block groups to map 23 variables (raw data and spatial metrics) and customize the color scheme.
Action: The revised manuscript has been formatted so that figures and tables are now embedded in-text.
Point 2: The authors used data from censuses from 2000 and 2010. Was there a census in the USA in 2020, or did the pandemic make it impossible? I look forward to the publication that also includes newer data.
Response: A 2020 census was conducted, but validated census data at the block group level will not be available until mid-2022. We echo the reviewers enthusiasm about a follow-up publication and plan to undertake such a study once both the 2020 census and concurrent 2020 NLCD land cover data become available.
Point 3: In lines 156-163, the authors wrote:
“To reduce data aggregation bias introduced by shifting block group boundaries and to allow for direct comparisons of Census counts over time, we used 2000 and 2010 data from the National Historical Geographic Information System (NHGIS) that has been standardized to 2010 Census units (Manson et al. 2019). We assumed that residential housing is prohibited or greatly restricted in public open spaces (Tanner 2002), and therefore removed overlapping parks and protected areas, including water bodies and all areas managed by federal, state, district, or local agencies, from calculations of Census block group areas. This step refined the available private land area in affected block groups prior to housing density calculations.”
Thinking about readers from outside the US, I would like the authors to add a piece of short information about the reference units adopted for analysis, i.e., block group (census block?). What exactly are block groups?
Action: In the revised manuscript we have clarified what Census block groups are and why we used block group units. [see new text in section 2.1. of the revised manuscript].
Point 4: In lines 174-175, the authors wrote:
“We classified the density (units/hectare) of each block group into one of four density gradient levels: urban, suburban, exurban, and rural, following Theobald (2001, 2005).”
The authors classified the density values into four classes: urban, suburban, exurban, and rural. The density change is gradual, fuzzy, just like the urban, suburban, exurban, and rural areas change smoothly. It is not sure whether the density values that define the boundaries of such classes are constant over time. Should urban, suburban, exurban, and rural be classified only based on the density indicator? I am asking the authors to justify the approach adopted in the article.
Response: The classification of density gradient values into the four classes is commonly used and also allowed for direct comparisons to (and building upon) previously published national-scale work in this area of research, including Theobald 2001, Theobald 2005, and Brown et al. 2005 (see paragraph 2 of Discussion). The density classification boundaries we used (shown in Figure 2b) are the same thresholds used in prior research noted above and are fixed over time (2000 and 2010) to allow consistent evaluation of change in density and the other metrics analyzed by density class over time.
Action: In the revised manuscript, we have added text clarifying the use of these classes [see changes in section 2.2. of the revised manuscript]. We have also removed the word “gradient” from the manuscript where appropriate to avoid confusion since we use fixed categories.
Point 5: In lines 208-212, the authors wrote:
“We then tested for similarity between the resulting 21 unique pairs of metric rankings (e.g. density and % developed, density and consumption, consumption and % efficient lands, etc.) using the Spearman's rank-order correlation on the assumption that no two rankings are strongly correlated (positively or negatively) and that each region has a unique trajectory of urbanization.”
The authors should describe the purpose of determining the correlation indicators.
Response: We are using a nonparametric approach to test whether any pairs of metric rankings among the regions are similar or associated. If any of the pairs of ranks are similar this suggests the metrics are providing redundant information and perhaps we should simplify our assessment of overall footprint using fewer metrics. We found no strong positive (or negative) associations among the pairs of ranked values and thus kept all seven metric rankings.
Action: In the revised manuscript, we have added text clarifying the purpose of applying the Spearman’s rank-order correlation. [see expanded explanatory text in section 2.3. of the revised manuscript].
Point 6: In lines 212-215, the authors wrote:
“We summed each region's rankings across the seven metrics to derive a composite score where lower scores indicate more sustainable development. Lastly, based on the composite scores we ranked the megaregion growth trajectories from most (1) to least (11) sustainable.”
Would you please justify the method of determining ranked megaregion growth trajectories? It is a classic case of multi-criteria evaluation. Can rank values be added together? Are all indicators equally important for the sustainable assessment? Shouldn't the range of change in the value of each indicator be taken into account? For example, indicator X changes significantly. The Y indicator does not change much. Changes in the value of the X indicator and the Y indicator are used to determine the position in the ranking 1,2,…, 11. But for the final assessment, the ranking change of indicator X seems more critical than that of indicator Y due to the range of values. Maybe the authors should consider using multi-criteria analysis of outranking methods such as PROMETHEE or ELECTREE or method utility function.
Response: The purpose of this method (and the resulting Figure 7) is not a multi-criteria decision analysis, but rather to provide a simple approach and visualization that summarizes and displays 1) the variability of each metric (values in each column), 2) the relative performance of megaregions as measured by each metric (the ranks in each column), and 3) the relative performance of megaregions across all metrics (summing the ranks into composite scores on each row). Based on prior studies of sustainable urbanization, these metrics concisely summarize different, but interlinked, dimensions of urbanization, including form, extent, rate, and efficiency, that are fundamental to understanding and managing sustainable growth. We don’t have information on (and make no claims about) the relative importance of each metric/dimension, thus they are equally important here. In paragraph’s 9 and 10 of the Discussion in the revised manuscript, we note how trends in prior studies support our assessment of the most and least sustainable megaregions, but more/future work is needed to validate.
Action: In the revision, we have added text clarifying the purpose of this method. [see expanded explanatory text in section 2.3. of the revised manuscript].
Reviewer 2 Report
1) the paper is focused on US, there international review is very poor
2) in my opinion the question of population decrease need more attention, the calculated metrics need to be strongly linked to the population trend
Author Response
Response to Reviewer 2 Comments
We thank the reviewer for the comments.
Point 1. the paper is focused on US, there international review is very poor
Response: The reviewer is correct - this manuscript is entirely focused on the U.S. and therefore relies on previous work in the U.S. to frame our contribution and place the results and discussion into context. However, many of the references in the introductory paragraphs are global studies of urbanization and impacts. We then clearly identified our geographic focus (“In the United States…”) beginning with paragraph 3 of the Introduction and reiterated again (“In the United States…” in paragraph 1 of the Discussion.
Point 2: in my opinion the question of population decrease need more attention, the calculated metrics need to be strongly linked to the population trend
Response: The reviewer is correct in that population trends can dramatically impact the calculation of some metrics, in particular the calculations of per capita consumption and land-use efficiency, which rely on population counts. Population losses and implications are mentioned throughout the Results and Discussion of the original submission:
- In paragraph 1 of the Results (section 3.1. of the revised manuscript), we note the proportion of Census block groups that lost population nationally.
- In paragraph 2 of the Results (section 3.2. of the revised manuscript) and in Figure 2, we note the proportion of population decline in U.S. rural density lands.
- In paragraph 3 of the Results (section 3.3. of the revised manuscript) and in Figure 5d, we highlight New Orleans as a metropolitan area of increasing per capita consumption due to rapid out-migration (loss) of the population.
- In paragraph 7 of the Discussion, we point out that late 20th century trends of increasing development and rural population losses throughout the Great Plains, particularly losses of population in the non-metropolitan counties, could be contributing to the widespread increase in per capita consumption that we found in this region between 2000-2010.
- In paragraph 8 of the Discussion, we note that interpreting the efficiency metric can be tenuous in areas that lose population.
- In the “Comments on Interpretation” section of the Discussion, we discuss our finding that many areas that gained both housing and impervious development also lost population over the decade, and thus population counts or population density are not always a suitable proxies for the development footprint.
- In the “Comments on Interpretation” section of the Discussion, we also discuss at length the link between population losses and increasing per capita consumption despite little or no increase in impervious development, citing and discussing New Orleans as a prime example where the loss of people due to Hurricane Katrina in 2005 and resulting mass out-migration led to an increase in per capita consumption by decades end.
Reviewer 3 Report
The manuscript is very well prepared and very interesting in content. It provides insight into current issues related to increasing urbanization on a national scale. The benefit is the publication of the results online on a website that is freely accessible to the public. The manuscript can thus serve as an inspiration and model for other countries in creating similar outputs.
However, the paper should be better structured to make it clearer. The reader may be lost in the amount of unstructured text.
I think that it would be better to place the pictures to that part where they are mentioned. For example, Fig. 1, where calculations are examined should be in the Chapter 2 Material and Methods.
Placement of pictures and tables in the end of the manuscript is unclear and chaotic.
Website https://ncsu-cga.github.io/HumanFootprint/ as one of the results of the research should also be mensioned in Chapter 2 Materials and methods (way of creating the website, what type of input data are necessary, what format – it would be helpful for another scientist from another countries in creating similar website) .
Text should be better arranged – authors should use better structure of manuscript (for example structured text with buletts).
Author Response
Response to Reviewer 3 Comments
We thank the reviewer for the kind words and constructive comments.
Point 1: I think that it would be better to place the pictures to that part where they are mentioned. For example, Fig. 1, where calculations are examined should be in the Chapter 2 Material and Methods. Placement of pictures and tables in the end of the manuscript is unclear and chaotic.
Response: The placement of figures in the original submission was based on the order of sections for the “Free Format” submission.
Action: The revised manuscript has been formatted so that figures and tables are now embedded in-text.
Point 2: Website https://ncsu-cga.github.io/HumanFootprint/ as one of the results of the research should also be mensioned in Chapter 2 Materials and methods (way of creating the website, what type of input data are necessary, what format – it would be helpful for another scientist from another countries in creating similar website).
Action: In the revised manuscript, in both the last paragraph of the Introduction and in the “Data Availability Statement”, we added a link to the public GitHub repository (https://github.com/ncsu-cga/HumanFootprint) containing the web mapping application source code and associated files that others can reference to create a similar web mapping platform.
Point 3: Text should be better arranged – authors should use better structure of manuscript (for example structured text with buletts).
Response: In the original submission, we followed the journal’s recommendations for structuring the main manuscript sections (Introduction, Materials and Methods, Results, Discussion, etc.).
Action: In the revised manuscript, we have now added additional subsections and broken up longer paragraphs where appropriate to improve the overall structure.
