the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 29 Aug 2022
The response of hemispheric differences in Earth’s albedo to CO2 forcing in coupled models and its implications for shortwave radiative feedback strength
Aiden R. Jönsson
Frida A.-M. Bender
Abstract. The Earth’s albedo is observed to be symmetric between the hemispheres on the annual mean timescale, despite the clear-sky albedo being asymmetrically higher in the northern hemisphere due to more land area and aerosol sources; this is because the mean cloud distribution currently compensates for the clear-sky asymmetry almost exactly. We investigate the evolution of the hemispheric difference in albedo in CMIP6 coupled model simulations following an abrupt quadrupling of CO2 concentrations, to which all models respond with an initial decrease of albedo in the northern hemisphere (NH) due to loss of Arctic sea ice. After this initial NH darkening, the evolution of the hemispheric albedo difference diverges among models, with some models remaining at their new hemispheric albedo difference, and others returning towards their pre-industrial difference through either a reduction in SH clouds or an increase in NH clouds, or a combination of the two. These responses have different implications on the reduction in global albedo, and thereby the strength of the shortwave cloud feedback: if a cross-hemispheric communicating mechanism is primarily responsible for maintaining hemispheric albedo symmetry, the total shortwave radiative feedback must be more strongly positive. We also show that in these models, there is a link between the extent of reductions in SH extratropical cloud cover and Antarctic albedo decline due to increased poleward heat transport in the SH.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(21826 KB)
-
Supplement
(4496 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(21826 KB) -
Supplement
(4496 KB) - BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Aiden R. Jönsson and Frida A.-M. Bender
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2022-811', Anonymous Referee #1, 04 Oct 2022
Review of “The response of hemispheric differences in Earth’s albedo to CO2 forcing in coupled models and its implications for shortwave radiative feedback strength.”
This manuscript frames itself as exploring the possibilities of two future scenarios: one where the hemispheric albedo symmetry persists, and one where it is does not. Unfortunately, there is no clear delineation between these two regimes in the model results. Looking at the model spread in hemispheric albedo difference changes, it appears to be a normal distribution centered around -3 W/m2 (Figure 1). There are a few cases overlapping 0 W/m2, but these may simply be due to chance. All models show an initial negative perturbation to hemispheric albedo difference owing to a reduction in clear-sky albedo in the NH high latitudes. For some models, this is partially offset by increases in NH cloud albedo (termed a local compensation). For other models, the NH albedo reduction is matched by a reduction in SH albedo – mostly by clouds in the latitude range 30-60S – termed a remote compensation. For the remaining models, there is little compensation by clouds and the hemispheric albedo difference simply persists. It does not appear that there is any strong correlation between the magnitude of hemispheric symmetry change and SW cloud feedback (as shown in Figure 8), making it difficult to conclude anything about the role hemispheric albedo difference plays for climate projection uncertainty.
It is hard to gauge what we, as readers, are to learn from this study. It seems the conclusions are limited beyond acknowledging there is a large spread in model behavior surrounding clouds, which is already well known. If anything, this work would seem to suggest that models have no inclination for maintaining a hemispheric albedo symmetry, which makes sense given that they do not generally reproduce the observed symmetry (as shown in Supplementary Figure S1). The authors acknowledge that no physical mechanism has been proposed for why the hemispheric albedos should remain balanced, so it is perhaps not surprising that the models are unconstrained for their own hemispheric albedo differences. Some of the relationships examined in this manuscript between the cloud changes and other physical processes in the models may be useful to the scientific community, but I found the interpretation questionable at times (I have added details to these issues below). In its present form, I must recommend this paper be rejected and returned to the authors.
Major issues
Eyeballing the ‘end’ period in Fig 1a, it seems like the models suggest a normal distribution of asymmetry changes centered around -3 W/m2. Those models that come in at 0 W/m2 change seem to do so by chance. I attempted to go through the various tables and figures to determine if the models that start with a symmetric albedo (Figure S1) are the same ones that have a small ‘End’ – ‘PI’ hemispheric albedo difference, but I couldn’t find such a relationship. Do the models that have a small perturbation change to warming at the end of 150 years have any consistent relation to their initial hemispheric albedo difference? Is there any change in the distribution across models of hemispheric differences with warming?
Line 128: “While models agree on clear-sky albedo reductions in the NH in response to warming, the spread in magnitude of total albedo reductions points to differences among the models in whether clouds serve to either amplify or reduce the total albedo reduction in the hemispheric mean.” There is very little agreement in the magnitude of clear-sky albedo change in response to warming (Fig 2b). Are the authors arguing that clouds determine how much sea ice is lost? How do we know that is the case? Comparing Figs 2a and 2b, it appears that the spread in total albedo change at 90N is smaller than the clear-sky change. Wouldn’t such a result suggest that the clouds are generally offsetting the clear-sky response to minimize the change (like the local compensation discussed later)?
Lines 156-157: 17 models amplify and 16 reduce. There are 34 models… so 1 has no significant response? Looking at Fig 3a, it appears several of these bars are almost unreadably small. Is it really only one model where SW CRE change is not statistically distinguishable from zero?
Line 209: “Planetary albedo is reduced in the Antarctic sea ice zone (Figure 6a); this is most likely the result of increasing liquid-phase precipitation reducing the sea ice surface albedo, and decreasing snowfall that otherwise would stabilize sea ice albedo.” Why not simply a result of changing temperature or ocean circulation? I struggle to understand from the results shown how we can conclude the phase of precipitation falling on sea ice is the “most likely” cause of the albedo changes there. I see that SSTs are brought up in section 4, but I think it would be valuable to bring these changes into the discussion in section 3.3.
Line 211: “This allows the sea ice albedo feedback to affect the SH polar climate in models where SH extratropical SW CRE increases more strongly; the result can be seen in increased SW radiative heating at the surface (Figure 6b, e).” How do we know causality here? I don’t follow how Figure 6 demonstrates the SH polar changes’ impact on the extratropical response.
Section 3.3: I struggle to follow the argument of the poor correlation between sea ice extent and changes in extratropical SW CRE changes. Why are the authors only looking at changes in maximum sea ice extent? Why not some time-integrated sea ice extent measure? Wouldn’t the sea ice minimum be more interesting because a larger retreat during summer would have impacts on surface fluxes that could change the clouds and circulation patterns nearby? All the changes in clouds have been annual averages, so why compare them with a seasonally dependent measure of sea-ice?
Line 244: “…meaning that the perturbation in asymmetry due to strong forcing in all models 150 years after the onset of abrupt CO2 forcing is close to the interannual variability seen in the past 20 years of observations.” If all models are close to the interannual variability, what does that tell us? How do we reconcile that result with the discussion around Figure 1?
Line 255: “When the difference between NH and SH Δ(αclear −α) is larger, asymmetry is more effectively maintained.” Is this true? Eyeball estimates in Figure 8c don’t show a clear signal. Is this plotted somewhere else or has a correlation been computed?
Given that many of the comparisons involve differences between models, differences across hemispheres, differences between all-sky and clear-sky fluxes, and differences in time, it may be helpful to readers to define a consistent use of language. For example, increase & decrease could mean value changes (keeping track of the sign), while amplify and reduce could mean magnitude changes (absolute value). It would also be helpful to define early on what latitude bands the authors mean when using terms like subtropics, extratropics, mid-latitudes, polar, Arctic, and Antarctic. Finally, having grid lines (or at least a zero-line) for Figures 1 and 2 would make examining the sign of the changes easier.
Minor issues
Line 163 “on both the the degree” -> “on both the degree”
Figure 4a – is the colorscale reversed here? How do the lines peaking over +10 W/m2 have an average of -1 W/m2? The caption text suggests they are the same variable differencing the same time periods. It doesn’t match Figure 5 either.
Figure 4b-f are the bounds too narrow on these plots? Where are models 9, 7, and 1 in panels c-f?
“We henceforth use the difference in 30-60° S area mean SW CRE between the ‘End’ and ‘Mid’ periods as an indicator of the impact of cloud albedo contribution changes on TOA albedo in the SH extratropics among models.” Is 30-60S SW CRE well-correlated with the total SH SW CRE change? In other words, is it fair to focus on this region because variability here corresponds to the total variability we are concerned with (the remote/SH albedo changes)?
“Note also that SW CRE at higher latitudes (> 60° S) also becomes more negative consistently in models with SW CRE increases in the SH extratropics.” Is poleward of 60S considered extratropics here?
“net poleward transport of moisture away from the SH extratropics (∼30-50° S) to the polar region (> 60° S)” Now extratropics is 30-50S?
“Atmospheric moisture content increases in the SH (Figure 5a) as clouds are lost and the atmosphere is warmed.” -> This reads as if the cloud loss helps cause the increase in atmospheric moisture, which I am guessing the authors did not mean to imply.
Figure 7 – colorbar is flipped again?
Line 268: “These two possibilities, local or remote compensation, would also mean that SW radiative feedback strengths are either strongly positive or somewhat negative, respectively.” Isn’t this flipped? Remote compensation has the strong positive SW radiative feedback.
Line 290: “role in determining the the observed” -> “role in determining the observed”
“Although tropical clouds and albedo seem to play a secondary role in determining the observed hemispheric albedo symmetry on time scales longer than a year, this should also be taken into account in understanding hemispheric albedo symmetry-maintaining mechanisms that involve the extratropics, as it can mean that some of the compensation offered by extratropical albedo reductions in one hemisphere can be buffered by tropical albedo increases, which may require more substantial high latitude albedo reductions to maintain hemispheric albedo symmetry.” -> should be separated into multiple sentences for clarity
Appendix B and Figures B1-B3 are not referenced anywhere in the text.
Citation: https://doi.org/10.5194/egusphere-2022-811-RC1 - AC1: 'Reply on RC1', Aiden Jönsson, 20 Dec 2022
-
RC2: 'Comment on egusphere-2022-811', Anonymous Referee #2, 20 Oct 2022
Jönsson and Bender explore changes in albedo, radiative fluxes and cloudiness in order to improve the understanding of the hemispheric symmetry of the planetary albedo and its possible changes in a warming climate. This is performed by investigating output from the CMIP6 in combination at some points with satellite retrievals. The topic of hemispheric symmetry of the planetary albedo is an exciting and highly debated one, in particular in light of possible changes in a warming climate. The study is of interest to the readership of EGUSphere. It is written in excellent English and the figures are in good quality.
The analysis is thoroughly conducted and broad in scope. In fact, my most important remark is that there is so much material that at multiple times I was a bit lost in understanding as to how a particular result allows to conclude about the causes for changes in hemispheric difference in planetary albedo.
The Discussion section is excellent, but does not really discuss the results in light of the literature. It would rather be better as part of the Introduction, and then the discussion of the results could refer to it. I do not provide a specific suggestion for shortening the results sections, but I propose the authors consider moving some of the material to an annex to streamline the discussion.
Besides this, I only have a number of specific remarks.
l72 It is regressing the global mean temperature against the top-of-atmosphere radiation imbalance (the effect forcing is the y-axis intercept only)
l88 I find this definition throughout the manuscript puzzling, since now all signs for CRE are the opposite ones compared to the all-sky and clear-sky differences. I think this definition requires that in Fig. 2, Fig 3 etc the reader is reminded about this difference in definition.
l154 (Fig. 1 and subsequent similar figures) – it would be useful to colour the numbers in the scatterplots by the colour used for the corresponding lines in the line plot to allow to make the association at least vaguely.
l159 I propose it might be better to use the same y-axis scaling in all panels
l173 I do not understand Fig. 3b. The three bars for each model should add up. Why is that not the case? also: Clarify in Caption that this is the difference between mid and PI
l202 Would it maybe be interesting to express precipitation and e-p in energetic units for comparison to the SW fluxes? Are the authors sure about no mistake for the models that substantially cool the NH high latitudes between mid and end?
l209 What is cause and what is effect is not fully clear. It may also be that after sea ice melting, clouds are much warmer if connected to the warm ocean rather than cold sea ice. Maybe reformulate to “this is most likely related to”
l212 To me it is not clear enough why Fig. 6b,e are not largely redundant with Fig. 6a,d
l215 What exactly are the “conditions” if not extent of sea ice?
l222 I would formulate the other way around, y-axis plotted against x-axis. Clarify that cloud fraction is from MODIS, not CERES.
l245 “within” rather than “close to”, I guess, since many models have lower values.
l315 This “model dependence” I do not understand. Of course the models show different results, so the results are model-dependent. What exactly is meant, a specific influence of the dynamical core of CESM?
Citation: https://doi.org/10.5194/egusphere-2022-811-RC2 - AC2: 'Reply on RC2', Aiden Jönsson, 20 Dec 2022
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2022-811', Anonymous Referee #1, 04 Oct 2022
Review of “The response of hemispheric differences in Earth’s albedo to CO2 forcing in coupled models and its implications for shortwave radiative feedback strength.”
This manuscript frames itself as exploring the possibilities of two future scenarios: one where the hemispheric albedo symmetry persists, and one where it is does not. Unfortunately, there is no clear delineation between these two regimes in the model results. Looking at the model spread in hemispheric albedo difference changes, it appears to be a normal distribution centered around -3 W/m2 (Figure 1). There are a few cases overlapping 0 W/m2, but these may simply be due to chance. All models show an initial negative perturbation to hemispheric albedo difference owing to a reduction in clear-sky albedo in the NH high latitudes. For some models, this is partially offset by increases in NH cloud albedo (termed a local compensation). For other models, the NH albedo reduction is matched by a reduction in SH albedo – mostly by clouds in the latitude range 30-60S – termed a remote compensation. For the remaining models, there is little compensation by clouds and the hemispheric albedo difference simply persists. It does not appear that there is any strong correlation between the magnitude of hemispheric symmetry change and SW cloud feedback (as shown in Figure 8), making it difficult to conclude anything about the role hemispheric albedo difference plays for climate projection uncertainty.
It is hard to gauge what we, as readers, are to learn from this study. It seems the conclusions are limited beyond acknowledging there is a large spread in model behavior surrounding clouds, which is already well known. If anything, this work would seem to suggest that models have no inclination for maintaining a hemispheric albedo symmetry, which makes sense given that they do not generally reproduce the observed symmetry (as shown in Supplementary Figure S1). The authors acknowledge that no physical mechanism has been proposed for why the hemispheric albedos should remain balanced, so it is perhaps not surprising that the models are unconstrained for their own hemispheric albedo differences. Some of the relationships examined in this manuscript between the cloud changes and other physical processes in the models may be useful to the scientific community, but I found the interpretation questionable at times (I have added details to these issues below). In its present form, I must recommend this paper be rejected and returned to the authors.
Major issues
Eyeballing the ‘end’ period in Fig 1a, it seems like the models suggest a normal distribution of asymmetry changes centered around -3 W/m2. Those models that come in at 0 W/m2 change seem to do so by chance. I attempted to go through the various tables and figures to determine if the models that start with a symmetric albedo (Figure S1) are the same ones that have a small ‘End’ – ‘PI’ hemispheric albedo difference, but I couldn’t find such a relationship. Do the models that have a small perturbation change to warming at the end of 150 years have any consistent relation to their initial hemispheric albedo difference? Is there any change in the distribution across models of hemispheric differences with warming?
Line 128: “While models agree on clear-sky albedo reductions in the NH in response to warming, the spread in magnitude of total albedo reductions points to differences among the models in whether clouds serve to either amplify or reduce the total albedo reduction in the hemispheric mean.” There is very little agreement in the magnitude of clear-sky albedo change in response to warming (Fig 2b). Are the authors arguing that clouds determine how much sea ice is lost? How do we know that is the case? Comparing Figs 2a and 2b, it appears that the spread in total albedo change at 90N is smaller than the clear-sky change. Wouldn’t such a result suggest that the clouds are generally offsetting the clear-sky response to minimize the change (like the local compensation discussed later)?
Lines 156-157: 17 models amplify and 16 reduce. There are 34 models… so 1 has no significant response? Looking at Fig 3a, it appears several of these bars are almost unreadably small. Is it really only one model where SW CRE change is not statistically distinguishable from zero?
Line 209: “Planetary albedo is reduced in the Antarctic sea ice zone (Figure 6a); this is most likely the result of increasing liquid-phase precipitation reducing the sea ice surface albedo, and decreasing snowfall that otherwise would stabilize sea ice albedo.” Why not simply a result of changing temperature or ocean circulation? I struggle to understand from the results shown how we can conclude the phase of precipitation falling on sea ice is the “most likely” cause of the albedo changes there. I see that SSTs are brought up in section 4, but I think it would be valuable to bring these changes into the discussion in section 3.3.
Line 211: “This allows the sea ice albedo feedback to affect the SH polar climate in models where SH extratropical SW CRE increases more strongly; the result can be seen in increased SW radiative heating at the surface (Figure 6b, e).” How do we know causality here? I don’t follow how Figure 6 demonstrates the SH polar changes’ impact on the extratropical response.
Section 3.3: I struggle to follow the argument of the poor correlation between sea ice extent and changes in extratropical SW CRE changes. Why are the authors only looking at changes in maximum sea ice extent? Why not some time-integrated sea ice extent measure? Wouldn’t the sea ice minimum be more interesting because a larger retreat during summer would have impacts on surface fluxes that could change the clouds and circulation patterns nearby? All the changes in clouds have been annual averages, so why compare them with a seasonally dependent measure of sea-ice?
Line 244: “…meaning that the perturbation in asymmetry due to strong forcing in all models 150 years after the onset of abrupt CO2 forcing is close to the interannual variability seen in the past 20 years of observations.” If all models are close to the interannual variability, what does that tell us? How do we reconcile that result with the discussion around Figure 1?
Line 255: “When the difference between NH and SH Δ(αclear −α) is larger, asymmetry is more effectively maintained.” Is this true? Eyeball estimates in Figure 8c don’t show a clear signal. Is this plotted somewhere else or has a correlation been computed?
Given that many of the comparisons involve differences between models, differences across hemispheres, differences between all-sky and clear-sky fluxes, and differences in time, it may be helpful to readers to define a consistent use of language. For example, increase & decrease could mean value changes (keeping track of the sign), while amplify and reduce could mean magnitude changes (absolute value). It would also be helpful to define early on what latitude bands the authors mean when using terms like subtropics, extratropics, mid-latitudes, polar, Arctic, and Antarctic. Finally, having grid lines (or at least a zero-line) for Figures 1 and 2 would make examining the sign of the changes easier.
Minor issues
Line 163 “on both the the degree” -> “on both the degree”
Figure 4a – is the colorscale reversed here? How do the lines peaking over +10 W/m2 have an average of -1 W/m2? The caption text suggests they are the same variable differencing the same time periods. It doesn’t match Figure 5 either.
Figure 4b-f are the bounds too narrow on these plots? Where are models 9, 7, and 1 in panels c-f?
“We henceforth use the difference in 30-60° S area mean SW CRE between the ‘End’ and ‘Mid’ periods as an indicator of the impact of cloud albedo contribution changes on TOA albedo in the SH extratropics among models.” Is 30-60S SW CRE well-correlated with the total SH SW CRE change? In other words, is it fair to focus on this region because variability here corresponds to the total variability we are concerned with (the remote/SH albedo changes)?
“Note also that SW CRE at higher latitudes (> 60° S) also becomes more negative consistently in models with SW CRE increases in the SH extratropics.” Is poleward of 60S considered extratropics here?
“net poleward transport of moisture away from the SH extratropics (∼30-50° S) to the polar region (> 60° S)” Now extratropics is 30-50S?
“Atmospheric moisture content increases in the SH (Figure 5a) as clouds are lost and the atmosphere is warmed.” -> This reads as if the cloud loss helps cause the increase in atmospheric moisture, which I am guessing the authors did not mean to imply.
Figure 7 – colorbar is flipped again?
Line 268: “These two possibilities, local or remote compensation, would also mean that SW radiative feedback strengths are either strongly positive or somewhat negative, respectively.” Isn’t this flipped? Remote compensation has the strong positive SW radiative feedback.
Line 290: “role in determining the the observed” -> “role in determining the observed”
“Although tropical clouds and albedo seem to play a secondary role in determining the observed hemispheric albedo symmetry on time scales longer than a year, this should also be taken into account in understanding hemispheric albedo symmetry-maintaining mechanisms that involve the extratropics, as it can mean that some of the compensation offered by extratropical albedo reductions in one hemisphere can be buffered by tropical albedo increases, which may require more substantial high latitude albedo reductions to maintain hemispheric albedo symmetry.” -> should be separated into multiple sentences for clarity
Appendix B and Figures B1-B3 are not referenced anywhere in the text.
Citation: https://doi.org/10.5194/egusphere-2022-811-RC1 - AC1: 'Reply on RC1', Aiden Jönsson, 20 Dec 2022
-
RC2: 'Comment on egusphere-2022-811', Anonymous Referee #2, 20 Oct 2022
Jönsson and Bender explore changes in albedo, radiative fluxes and cloudiness in order to improve the understanding of the hemispheric symmetry of the planetary albedo and its possible changes in a warming climate. This is performed by investigating output from the CMIP6 in combination at some points with satellite retrievals. The topic of hemispheric symmetry of the planetary albedo is an exciting and highly debated one, in particular in light of possible changes in a warming climate. The study is of interest to the readership of EGUSphere. It is written in excellent English and the figures are in good quality.
The analysis is thoroughly conducted and broad in scope. In fact, my most important remark is that there is so much material that at multiple times I was a bit lost in understanding as to how a particular result allows to conclude about the causes for changes in hemispheric difference in planetary albedo.
The Discussion section is excellent, but does not really discuss the results in light of the literature. It would rather be better as part of the Introduction, and then the discussion of the results could refer to it. I do not provide a specific suggestion for shortening the results sections, but I propose the authors consider moving some of the material to an annex to streamline the discussion.
Besides this, I only have a number of specific remarks.
l72 It is regressing the global mean temperature against the top-of-atmosphere radiation imbalance (the effect forcing is the y-axis intercept only)
l88 I find this definition throughout the manuscript puzzling, since now all signs for CRE are the opposite ones compared to the all-sky and clear-sky differences. I think this definition requires that in Fig. 2, Fig 3 etc the reader is reminded about this difference in definition.
l154 (Fig. 1 and subsequent similar figures) – it would be useful to colour the numbers in the scatterplots by the colour used for the corresponding lines in the line plot to allow to make the association at least vaguely.
l159 I propose it might be better to use the same y-axis scaling in all panels
l173 I do not understand Fig. 3b. The three bars for each model should add up. Why is that not the case? also: Clarify in Caption that this is the difference between mid and PI
l202 Would it maybe be interesting to express precipitation and e-p in energetic units for comparison to the SW fluxes? Are the authors sure about no mistake for the models that substantially cool the NH high latitudes between mid and end?
l209 What is cause and what is effect is not fully clear. It may also be that after sea ice melting, clouds are much warmer if connected to the warm ocean rather than cold sea ice. Maybe reformulate to “this is most likely related to”
l212 To me it is not clear enough why Fig. 6b,e are not largely redundant with Fig. 6a,d
l215 What exactly are the “conditions” if not extent of sea ice?
l222 I would formulate the other way around, y-axis plotted against x-axis. Clarify that cloud fraction is from MODIS, not CERES.
l245 “within” rather than “close to”, I guess, since many models have lower values.
l315 This “model dependence” I do not understand. Of course the models show different results, so the results are model-dependent. What exactly is meant, a specific influence of the dynamical core of CESM?
Citation: https://doi.org/10.5194/egusphere-2022-811-RC2 - AC2: 'Reply on RC2', Aiden Jönsson, 20 Dec 2022
Peer review completion
Journal article(s) based on this preprint
Aiden R. Jönsson and Frida A.-M. Bender
Aiden R. Jönsson and Frida A.-M. Bender
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 923 | 167 | 25 | 1,115 | 39 | 18 | 8 |
- HTML: 923
- PDF: 167
- XML: 25
- Total: 1,115
- Supplement: 39
- BibTeX: 18
- EndNote: 8
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(21826 KB) - Metadata XML
-
Supplement
(4496 KB) - BibTeX
- EndNote
- Final revised paper