Publisher Correction: Disproportionate exposure to urban heat island intensity across major US cities

This article not only presents a quality mapping of the intermixing between racial and economic inequalities related to UHI exposure but comes at a perfect timing when racial (and associated economic) inequalities come at the forefront of societal discussions in the US and abroad. The paper is nicely presented, with compelling data and figures. The well-balanced discussion between economic and racial factors provide a great contribution to the literature on social impacts of climate change and adaptation strategies in cities (which by the way will exacerbate UHI and its incidence inequalities I suggest the authors briefly mention and discuss this). My major criticism is related to the non consideration of age as a critical factor for the impacts of extreme heat. In principle higher income neighborhoods or localities have higher life-longevity statistics and larger aged-population. Pregnant women and toddlers are other risk groups after aged people. In that sense I would think it is important to explicitly differentiate exposure from sensitivity and adaptive capacity; the combination of these three defines vulnerability (sensu IPCC). Then the combination of vulnerability with hazard provides a risk assessment. As such the title could read "Urban Heat Island EXPOSURE/SENSITIVITY/VULNERABILITY inequalities...", or at least the proper word should be in the abstract. That said, I do not think sensitivity and adaptive capacity are completely addressed with the employed dataset, since it considers race and income but not other variables such as age that can influence the impacts of heat extremes. One other comment is regarding the possibility of validation of the presented results. What does official statistics of excess deaths or hospitalizations in extreme heat episodes say? Is there a racial and income prevalence? I suggest the authors to revise the manuscript in respect and judge wether it would be possible to include other variables in their analysis that influence sensitivity to extreme heat, and whether available data corroborates the presented results.

The authors do not analyze climate change so there is no reason to use limited space to devote the introductory paragraph to information not relevant to the analyses. Further, it would be better to not include speculations on vicious cycles that also are not relevant to the results. It would be better to use that space to provide more information on the methods.
While it is understandable for authors to want to frame their work as unique, it is quite a stretch to say that multiple studies of UHI and inequities in different cities do not provide pervasive evidence of the magnitude of these inequities.
"Major urban areas" should be shown in a map in SM. The authors fail to provide basic information on and justification for analytic choices. One of several examples is why compare individuals below the poverty line with those 2x above the poverty line? This creates the impression the authors are only reporting statistically significant results. Support is needed for assumptions such as UHI intensities are larger in boreal and tropical areas.
The analyses need to be conducted by climate zones to limit confounding from other weather variables, such as humidity.
There also needs to be comparison of the UHI across summer months: are the UHI the same in May and August? There are enough city-level publications on UHI that it should be possible to validate the results for some cities.
At some point, text implied the analyses were only for summer months, but that is not clear.
Lines 111-115 are confused. It has been known for decades that temperature-mortality relationships are J-shaped but how is that relevant to a study of summer temperatures? There is significant controversy about how cold-related mortality could change with climate change --and that is not relevant either. This information does not support the claim that the distribution not just the mean of UHI is important --although it is true the distribution is important. A better explanation is needed of exactly what was meant.
Lines 117-120 are obtuse. Who prefers higher mean temperatures with little dispersion vs lower temperatures with more dispersion? If these are individual preferences, then please cite literature to support this contention.
Line 121: how was "desirability" determined? As this is central to the analyses, much more explanation is required.
The discussion is very weak; it does not compare the results with results from similar studies and fails to fully explore limitations, such as using one-year of cross-sectional data. What does it mean for the results that satellites over-estimate land surface temperatures? There is no evidence to support the implicit assumption that people live, work, and play only within their census tract.
It was frustrating to attempt to look up references for further information, only to find that key references are under submission. Further, many references are missing. The authors are strongly suggested to update their literature review. Further, references is rather random in the sense that multiple citations do not support the statements where they are listed.
Reviewer Comments, second round -Reviewer #1 (Remarks to the Author): I applaud the authors for the effort to include the age analysis in their study and considerations about validation of the developed index. The manuscript is ready for publication from my point of view.
Reviewer #2 (Remarks to the Author): Thank you for the extensive responses to the reviewer comments. However, it would have been helpful for readers if more of the explanations were included in the publication and not just in the responses.
I remain concerned about the results that you label as being for boreal areas. How did you define boreal areas? Boreal in North America is defined as extending south to 55N. Detroit is at 42.3N. Youngstown is 41.1. According to <<https://en.wikipedia.org/wiki/K%C3%B6ppen_climate_classification>> these cities and others are not in a boreal climate. I did not check whether other cities are actually within the claimed climate zones, but clearly they need to be.
The discussion section is improved but still includes new results and still repeats the Results section. The discussion also contains multiple unsupported normative statements. One example is that older adults may choose to live in greener areas. The following sentence states there also is evidence and then goes on to another topic. No evidence is provided. Referring to urban and rural areas does not provide support.
The fact that globally consistent data were used does not mean the data are useful for decisionmaking at the local level. Co-production studies typically conclude that such datasets are not necessarily relevant.

REVIEWER COMMENTS
Reviewer #1 (Remarks to the Author): I applaud the authors for the effort to include the age analysis in their study and considerations about validation of the developed index. The manuscript is ready for publication from my point of view.
Thank you.
Reviewer #2 (Remarks to the Author): Thank you for the extensive responses to the reviewer comments. However, it would have been helpful for readers if more of the explanations were included in the publication and not just in the responses.
We have gone back through our responses to reviewer comments from the first round of review to ensure that each point we addressed in the response is reflected in the manuscript. Where specific line numbers or sections were not mentioned in our initial response, we went back and ensured that explanations were adequately reflected in the manuscript itself.
For instance, with respect to the reviewer's comment with respect to why we highlighted those 2x above the poverty line, we responded, that for the ease of exposition, we sought to investigate the tails of the distribution, the poor (those below poverty) and the relatively rich (above 2 times), but we went back and added results for the additional ACS category of individuals 1-2 times above the poverty line in Tables 1 and 2. We have added this information in the Methods (Lines 557-558) to make this point clear.
Lines 557-558: "While results for each of these income categories is provided in our Tables, for the ease of exposition, we focus our discussion on the tails of the income distribution: the poor (those below poverty) and the relatively rich (above 2 times)." The only other response comment that was not originally reflected in the manuscript was in response to your comment regarding temperature preferences. We also now include a brief discussion of Jensen's inequality in the methods section when introducing the inequality index approach: Lines 583-597: "A simple individual-based metric such as mean exposure is potentially misleading due to non-linear adverse health impacts of summer heat. Evidence suggests that above a moderate threshold damage is an increasing convex function of temperature, i.e., a 1 degree temperature increase causes more damage at higher temperatures (50-53). In such cases, Jensen's inequality implies that, all else equal, the average health damage for a population in which everyone faces an identical summer heat exposure will be lower than that of a population with the same mean exposure but an unequal temperature distribution. It follows that for any unequal temperature distribution there exists a more desirable (from a health perspective) distribution characterized by a higher mean and no inequality. That is, a perfectly equal summer temperature distribution is generally preferable to an unequal distribution with the same mean.
I remain concerned about the results that you label as being for boreal areas. How did you define boreal areas? Boreal in North America is defined as extending south to 55N. Detroit is at 42.3N. Youngstown is 41.1. According to <https://en.wikipedia.org/wiki/K%C3%B6ppen_climate_classification>> these cities and others are not in a boreal climate. I did not check whether other cities are actually within the claimed climate zones, but clearly they need to be.
We apologize for this confusion. The link the reviewer has provided is for the Koppen climate classification, while we have used the Koppen-Geiger climate classification, which identifies the following climate zones for the U.S.: equatorial, arid, warm temperate, and snow (see: http://koeppen-geiger.vu-wien.ac.at/usa.htm). We had originally switched the 'snow' climate zone with the term 'boreal' based on a more recent paper by the same authors of the classification scheme (http://koeppen-geiger.vu-wien.ac.at/pdf/Paper_2017.pdf): 'The boreal climate is a synonym for the historically introduced term snow climate and, in a global context, the alpine climates are called polar climates.' To avoid this confusion, we have reverted back to the original naming scheme for the climate zones as defined in the Koppen-Geiger classification. All references to the 'boreal' climate zone have been replaced with 'snow,' 'tropical' has been replaced with 'equatorial.' In tables and figures we abbreviate 'warm temperate' to 'temperate' due to space limitations and note this slight abbreviation in the text: Lines 128-130: "We group urbanized areas by Köppen-Geiger (54)climate zones: arid, snow, warm temperate (henceforth referred to as temperate), and equatorial" The discussion section is improved but still includes new results and still repeats the Results section. The discussion also contains multiple unsupported normative statements. One example is that older adults may choose to live in greener areas. The following sentence states there also is evidence and then goes on to another topic. No evidence is provided. Referring to urban and rural areas does not provide support.
We were unsure exactly which new results the reviewer was referring to, but we moved the illustrative examples of Greenville and Baltimore counties to the Results section so that we are not referencing new figures in the Discussion section. Specifically, we added 'Section D: Illustrative Examples' to the Results section and moved the paragraph discussing these two cities and references to Figure 3 there.
With regards to the writing of the Discussion section, we were not sure exactly which statements were considered repetitive, so we took care to ensure that any repetitive statements were removed. Because the main purpose of this section was to elaborate and contextualize our Results, we elected to keep some references to our main findings to set up the discussion points.
For example, in the first paragraph of the Discussion section (starting Line 285), we kept one sentence succinctly summarizing our Results as the core main finding of our analysis: "We find the distributions of summer daytime SUHI intensity, taking into account both the mean and dispersion, is worse for both people of color and the poor, compared to white and wealthier populations in nearly all major U.S. cities. As we illustrated in Figure 2, this pattern holds not only at the national level, but in almost all major urban areas regardless of geographical location or climate zones, with a particularly intense difference in the Northeast and upper Midwest of the continental U.S." And in lines 302-307, the first several sentences of the second paragraph, we edited to avoid repeating the Results too much but to set up discussion for why we found less average SUHI intensity exposure for elder populations than what we expected, given the heat-related mortality rates for this population subgroup: "Although age presents a vulnerability to SUHI and elderly individuals aged 65 and older comprise a substantial percentage (39 percent) of heat-related deaths in the U.S. (Vaidyanathaen et al., 2020), our finding that populations over 65 are on average slightly less exposed (1.84 C versus 2.06C for those under 65) could have several explanations." We then edited the following sentences in that same paragraph regarding elder populations to avoid the sentence that may have appeared normative (i.e., previously "older populations may choose to live in areas of cities with more greenery" has been removed). We support our observation that older populations are not more exposed to higher SUHI intensity by referencing a Harvard study that found that a substantial segment of populations over 65 live in suburban, less dense areas where research has found, and by definition, are typically greener than builtup and denser urban environments, with the exception of arid climates where rural areas are often desert areas. (Lines 307-315): "Because SUHI intensity and greenness (as measured by Normalized Difference Vegetation Index or NDVI) are negatively correlated (Chakraborty et al., 2020), cooler areas tend to be greener. There is evidence that older populations over the age of 65 tend to live in suburban areas in the U.S. Approximately half live in rural areas or in urban areas with less than 1 housing unit per acre, and 28 percent live in suburban areas (Joint Center for Housing Studies of Harvard University, 2016), which are typically greener than denser urban areas, except in arid climates (Chakraborty et al, 2019b;Nitoslawski et al., 2016;and Hansen et al., 2005)." We also removed a sentence after Line 393 that repeated a finding in the Results regarding our finding that only in 1 city (McNally, TX) white populations have a higher exposure than those living in poverty. (Sentence from previous version now removed from Discussion: "By comparison, in only one city do white populations have a higher average exposure than those living below poverty (Table S1).) The fact that globally consistent data were used does not mean the data are useful for decisionmaking at the local level. Co-production studies typically conclude that such datasets are not necessarily relevant.
Thank you for this comment, we agree that there are specific factors at the local level (e.g., urban form, building materials, stakeholder preferences, planning and design processes) that are relevant in the consideration of SUHI mitigation and management and that globally consistent datasets are unable to measure or take this into consideration. Similar critiques have been raised regarding the utility of satellite-derived globally consistent SUHI data for decisionmaking and planning for 'climate- Therefore, from Lines 362-376, we have removed the statement regarding our dataset and analysis's utility for SUHI mitigation "tailored for local conditions" and modified it as such, including a referencing to Manoli et al. (2019) who have also caveated the limits of a globally consistent dataset for UHI management in cities around the world.
Lines 365-376: "Decision-makers and urban planners can utilize this information as a starting point to identify best practices and strategies for mitigating the SUHI and the inequalities in its distribution, although there are certainly localized, contextually-specific factors that must be considered when determining SUHI management strategies. As Manoli et al. (2019), who used similar globally-consistent data to evaluate drivers of SUHI in 30,000 cities around the world, acknowledge, these data can provide a 'first-order' analysis to understand base-level SUHI exposures and differences to complement more fine-grained data on local factors that influence the SUHI (see Limitations section for more discussion on data issues).
Co-production is indeed a relevant approach used in urban planning centered in the active involvement and engagement of actors in the production of knowledge in topics such as urban forestry, urban development, waste management and climate adaptation (Campbell et. al 2016;Frantzeskaki, N., & Kabisch, N. 2016;Gutberlet 2015;Satorras et.al 2020). Furthermore, recent literature exemplifies that this type of data could be useful as a starting point or as inputs in coproduction initiatives, alongside or supporting locally-generated datasets. (Iwaniec et.at 2020, Anenberg et.al 2020. However, the scope of this research is not to address the relevance of this dataset in this type of approach, hence it is not specifically mentioned in the discussion. Reviewer Comments, third round -Reviewer #2 (Remarks to the Author): Thank you for addressing most of my comments.
For your information, "normative" does not mean repetitive. A normative statement expresses a value judgment about whether a situation is desirable or undesirable. Whereas a descriptive statement is meant to describe the world as it is, a normative statement is meant to talk about the world as it should be. This is what I referred to when mentioning that several sentences in the Discussion were normative without references.
Also, although your dataset can not address co-production, it is still important to discuss that effective interventions are co-produced; e.g. in order to tailor the results for local levels, coproduction is needed.
We would like to thank the editor and referees for the time and effort put into helping improve the manuscript. Below we provide detailed responses to the last round of comments. Since our response refers to previous rounds of comments, for clarity we use black font for the referee's current comments, red font for the referee's previous comments and blue font for our response. Our previous responses we highlight in a green font.
Reviewer #2 (Remarks to the Author): Thank you for addressing most of my comments.
For your information, "normative" does not mean repetitive. A normative statement expresses a value judgment about whether a situation is desirable or undesirable. Whereas a descriptive statement is meant to describe the world as it is, a normative statement is meant to talk about the world as it should be. This is what I referred to when mentioning that several sentences in the Discussion were normative without references.
There were two relevant aspects to the reviewer's last round of comments that we consider here: 1. Discussion section repeats the Results section.

Unsupported normative statements
In the last set of reviews, regarding repetition the referee stated: "The discussion section is improved but still includes new results and still repeats the Results section." To which we responded: "With regards to the writing of the Discussion section, we were not sure exactly which statements were considered repetitive, so we took care to ensure that any repetitive statements were removed. Because the main purpose of this section was to elaborate and contextualize our Results, we elected to keep some references to our main findings to set up the discussion points.
For example, in the first paragraph of the Discussion section (starting Line 285), we kept one sentence succinctly summarizing our Results as the core main finding of our analysis: "We find the distributions of summer daytime SUHI intensity, taking into account both the mean and dispersion, is worse for both people of color and the poor, compared to white and wealthier populations in nearly all major U.S. cities. As we illustrated in Figure 2, this pattern holds not only at the national level, but in almost all major urban areas regardless of geographical location or climate zones, with a particularly intense difference in the Northeast and upper Midwest of the continental U.S.

…
And in lines 302-307, the first several sentences of the second paragraph, we edited to avoid repeating the Results too much but to set up discussion for why we found less average SUHI intensity exposure for elder populations than what we expected, given the heat-related mortality rates for this population subgroup: "Although age presents a vulnerability to SUHI and elderly individuals aged 65 and older comprise a substantial percentage (39 percent) of heat-related deaths in the U.S. (Vaidyanathaen et al., 2020), our finding that populations over 65 are on average slightly less exposed (1.84 C versus 2.06C for those under 65) could have several explanations." Regarding "normative" statements, the referee commented: "The discussion also contains multiple unsupported normative statements. One example is that older adults may choose to live in greener areas. The following sentence states there also is evidence and then goes on to another topic. No evidence is provided. Referring to urban and rural areas does not provide support." To which we responded: We then edited the following sentences in that same paragraph regarding elder populations to avoid the sentence that may have appeared normative (i.e., previously "older populations may choose to live in areas of cities with more greenery" has been removed). We support our observation that older populations are not more exposed to higher SUHI intensity by referencing a Harvard study that found that a substantial segment of populations over 65 live in suburban, less dense areas where research has found, and by definition, are typically greener than built-up and denser urban environments, with the exception of arid climates where rural areas are often desert areas. (Lines 307-315): "Because SUHI intensity and greenness (as measured by Normalized Difference Vegetation Index or NDVI) are negatively correlated (Chakraborty et al., 2020), cooler areas tend to be greener. There is evidence that older populations over the age of 65 tend to live in suburban areas in the U.S. Approximately half live in rural areas or in urban areas with less than 1 housing unit per acre, and 28 percent live in suburban areas (Joint Center for Housing Studies of Harvard University, 2016), which are typically greener than denser urban areas, except in arid climates (Chakraborty et al, 2019b;Nitoslawski et al., 2016;and Hansen et al., 2005)." The referee did not indicate, and we could not find, other instances of "normative" statements requiring additional support.
Also, although your dataset can not address co-production, it is still important to discuss that effective interventions are co-produced; e.g. in order to tailor the results for local levels, coproduction is needed.
Thank you for this comment -we agree and have added the following sentence: Line (370-373): "Studies have demonstrated the importance of co-production (i.e., involving citizens in the production of knowledge and planning decisions) in developing tailored urban environmental policies (Satorras et al., 2020).