the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 20 Nov 2025
Heat transport to the Central Arctic is Reduced by the Barents Sea Cooling Machine
Abstract. The Barents Sea is a primary gateway for Atlantic Water entering the Central Arctic Ocean and ubiquitous water-mass transformation on the Barents shelf is key for mitigating increases in heat transport to the central Arctic through the St. Anna Trough. Using a mesoscale-permitting reanalysis spanning 40 years, we derive the first long-term estimate of heat transport through the St. Anna trough, finding that it has increased by 0.15 TW/year since 1980. However, this is only half of the 0.28 TW/year trend in increasing heat transport into the Barents Sea through the Barents Sea Opening. Decomposing the heat transports reveals that these trends are entirely due to warming temperatures at the sections with no discernible trend in the volume transports. We find that a northward migration of the largest heat fluxes from the ocean to the atmosphere have resulted in cooler and denser Northern Barents Shelf Water, mitigating the heat transported through the St. Anna trough. However, despite functioning properly, the "Barents Cooling Machine" has been unable to keep pace with the dramatic warming of the Atlantic Water inflow, resulting in the residual trend in heat transport to the central Arctic. Finally, we present the first observational evidence for the “ocean feedback” hypothesis, indicating that it modulates volume transport into and out of the Barents Sea on sub-decadal timescales.
-
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
(15406 KB)
-
Supplement
(4953 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(15406 KB) - Metadata XML
-
Supplement
(4953 KB) - BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-5737', Anonymous Referee #1, 19 Dec 2025
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-5737/egusphere-2025-5737-RC1-supplement.pdfCitation: https://doi.org/
10.5194/egusphere-2025-5737-RC1 -
AC1: 'Reply on RC1', Shaun Eisner, 14 Jan 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-5737/egusphere-2025-5737-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Shaun Eisner, 14 Jan 2026
-
RC2: 'Comment on egusphere-2025-5737', Anonymous Referee #2, 05 Jan 2026
Review of the manuscript “Heat transport to the Central Arctic is Reduced by the Barents Sea Cooling Machine”, by S. Eisner et al.
In this manuscript, the authors utilize a state-of-the-art ocean model reanalysis to investigate a postulated feedback mechanism in the flow of warm and saline Atlantic Water through the Barents Sea en-route to the Polar Basin. The authors report that the feedback mechanism, in which increased heat loss to the atmosphere leads to increased formation and export of dense water and, thus, inflow further upstream, is substantiated based on analysis of the model results. Moreover, they report a positive trend not only in the heat carried by the Atlantic Water entering the Barent Sea in the southwest, but also to a lesser degree in the transformed water masses exiting into the Polar Basin to the northeast. The manuscript provides novel findings regarding an important topic that contributes to improve our understanding of the climate system in this hotspot of Arctic climate change and amplification. However, I have some concerns regarding both the findings, their methodological approach and the interpretation of the results, that the authors need to address. Please see my specific points below.
Specific comments:
Title: I think the title needs to be reworked a bit. As it reads now, it is stating the obvious, that a cooling machine cools the heat transport to the Arctic. However, the main finding is that the cooling machine is not fully keeping up the pace of the warming in the Barents Sea region in the sense that there is a positive trend in the heat transport to the Polar Basin, although this trend is less than the trend in the upstream heat transport into the Barents Sea. This should be better reflected in the title, I think.
L1: “Central Arctic Ocean” – is this an official name? The deep part of the Arctic Ocean is often referred to as the Polar Basin. It may also be referred to as the central Arctic Ocean, but not with a capital ‘C’. And in L2 you refer to the central Arctic with a lowercase ‘c’. Please be consistent.
L2: Here and elsewhere: Sometimes you refer to the Barents Sea, and sometimes only to the ‘Barents’. I presume you refer to the same, namely the Barents Sea, in both cases. Please be consistent and spell out ‘Barents Sea’ everywhere you refer to it.
L4: Here and elsewhere: Please spell ‘St. Anna Trough’ with a capital ‘T’.
L5: I think you should change ‘increasing’ with ‘the’: “… trend in the heat transport into the Barents Sea …”
L7-8: Has the migration of the “center of action” of the net heat loss from ocean to the atmosphere resulted in a cooler and denser Barents Shelf Water? Rather, you have shown that the temperature and density, in general, have increased. Regarding the density of the Barents Shelf Water, see also my comment to lines 169-171.
L14: The Bering Strait is not a primary gateway for Atlantic Water to the Arctic Ocean, so the reference to the Bering Strait does not belong here.
L16-17: While the Atlantic Water carries a substantial amount of heat and salt through the Barents Sea Opening, the contribution of heat to the Arctic Ocean has been questioned (e.g., Gammelrsød et al., 2009), although data sources in the exit region of the Barents Sea are scarce.
L25-26: On line 19 you state that Atlantic Water are transformed into Barents Shelf Water within the Barents Sea, while here you state that both BSW and AW leaves the Barents Sea and are carried by the ACBC as far as the Amerasian Basin. There is little evidence that water masses leaving the Barents Sea through the At. Anna Trough can still be classified as Atlantic Water.
L38 (and L164-165): This interpretation of the feedback loop proposed by Smedsrud et al. (2013) is not entirely correct. One main argument in the proposed feedback loop is that increased density in the BSX causes a density gradient between the BSX and the Polar Basin, and therefore also an accelerated outflow through the St. Anna Trough. This, in turn, creates an SSH gradient and an associated barotropic forcing through the Barents Sea (from the BSO to the BSX) that favours stronger throughflow.
L42: “Northern Barents Sea”: The region of the Barents Sea that is the focus area of this study is often referred to as the northeastern Barents Sea, whereas “northern Barents Sea” often refers to the area between the Svalbard and the Franz Josef Land archipelagos. Therefore, I recommend referring to your focus area as “the northeastern Barents Sea”.
L57: “Barents Sea cooling system” – I would rather use the term “cooling processes”.
L65: The last sentence in this paragraph clearly belongs to the paragraph below, because I assume you are not using monthly averaged atmospheric reanalysis as forcing for the ocean model!
L77: The acronym ‘STA’ is not defined yet (it is defined on line 89), although the St. Anna Trough is first mentioned on line 26 (except for in the abstract) and also on several occasions before the acronym is established.
L117: There is a reference to Fig. 2b, but Fig. 2 does not have any panel b as far as I can see.
L119: Note, that the areas with the greatest increases in heat content are also the deepest parts of the Barents Sea (the Bear Island Trough and the Hopen Trench, the Central Basin, and the St. Anna Trough), with the exception being the Northeast Basin. Normalizing the change in heat content by calculating per meter water depth may provide a somewhat different picture.
L124-126: So, a conclusion from this comparison with observations is that the heat loss in the model is smaller than the observed heat loss. I think this observation deserve a bit more discussion considering that the heat loss in the Barents Sea is the main topic of the paper.
L128-129: Change “upward” and “downward” trends with “positive” and “negative” trends, respectively. For clarity, you should also consider removing the reference to anomalies, and instead report that there are trends in the temperature and salinity (and, hence, also a trend in anomalies). Because a negative trend in anomalies could be interpreted as a trend towards more negative anomalies, but it could also be interpreted as a reduction in the magnitude of the anomalies.
L138: I suppose you mean “There is a clear positive trend in the average temperature”. An increasing trend, strictly speaking, means an accelerating trend (i.e., a positive second derivative). Please clarify.
L149: I suggest changing “heat content” with “temperature”, because that is what you have shown when decomposing the heat transport in Fig. 4.
L153: I suggest replacing “The reduced increase” with “The smaller increase”.
L154: Missing ‘o’ in “accommodate”
L169-171: You state that you find an increase in the average water column density in the northeastern Barents Sea from the 1980s to the 2020s, and a collocated reduction in the SSH, which is shown in Fig. 7 (left panels). I have two main concerns with these findings and your interpretation of them. The first concern is related to the effect of the SSH difference on the circulation. A change in the density difference between two regions, here the northern Barents Sea and the Eurasian Basin, also warrants an associated change in the SSH difference between the two regions. Here, the lower SSH in the northern Barents Sea together with the larger density in the same region may cause a net zero change in the pressure difference between the two regions (a higher water column of lower density vs. a lower water column of higher density). Thus, the conclusion that these changes in SSH will cause a geostrophic circulation anomaly is not necessarily substantiated. Moreover, Figure 7, lower left panel, shows a negative “SSH curl” along the eastern flank of the St. Anna Trough (where the Barents Sea Water exits the Barents Sea for the Polar Basin). This would imply a negative contribution to the flow in the cyclonic direction (if I understand the interpretation of the “SSH curl” on lines 79-80 correct). This would be in agreement with a circulation anomaly with the higher SSH on the right-hand side, as also implied by the SSH contours in the panels on the left-hand side in Fig. 7. But this would imply weaker, not stronger, outflow through the eastern St. Anna Trough. And, as a consequence, I have a bit of a hard time following your interpretation of the results depicted in Fig. 7 as synthesized in Fig. 10. Also, the change in density as shown in Fig. 7, upper left panel, seems very large. A density decrease representing a full unit (1 kg/m3) when averaged over the full water column is substantial. And this leads to my second concern. You state, based on Fig. 7, that the density of the Barents Shelf Water and the dense water formation in the Barents Sea have increased. However, such changes will be disguised due to the fact that several water masses are included in your calculation of total water column density change. In the area where you find the largest density increase, in the Northeast Basin in the northeastern Barents Sea, the water column may consist of dense bottom water formed through brine rejection during ice formation, overlaid by Atlantic Water (Barents Sea/Shelf Water), overlaid by Arctic Water. Any change in the mass distribution between these three water masses will also affect the average density of the water column. Indeed, Lind et al. (2018) reported a decline in the amount of Arctic Water in recent decades in this region, which by itself would increase the water column density because the colder and fresher (and less dense) Arctic Water was replaced by warmer and more saline (and denser) Atlantic Water.
L176-177: Usually, the timeseries are detrended before the correlation analysis is performed. Thus, the different trends in the two timeseries should not affect the correlation between them. Also here, you state a “decreasing trend”, while a presume you mean a “negative trend”.
L195-L198: While you have found some support for an ocean feedback mechanism, the underlying hypothesis and the associated analysis is slightly different from that of Smedsrud et al. (2013). Also, your analysis is based on model results, hence, you have not presented observational evidence for an ocean feedback mechanism, but rather empirical evidence, which was also, to some degree, presented in Smedsrud et al. (2013).
L199: Again: “increasing trend” -> “positive trend”.
L210-211: The hypothesis proposed by Smedsrud et al. (2013) postulated that an increase in the dense water formation in the BSX area will cause a density-driven acceleration of the outflow through the St. Anna Trough, and a subsequent increase in the BSO inflow due to mass continuity.
L212: The reference to Fig. 10 should be a reference to Fig. 9, I believe.
L215: Again, I would rather use the term “cooling processes”.
Figures:
Figure 3: I think you got the references (left/right and top/bottom) to the variables and geographical locations mixed up. Please correct. Also, it is not stated which period the anomalies are relative to. Even if it is mentioned in the main text, please repeat it in the figure caption.
Figure 7: a), b), etc. missing on figure panels. Also, the colour legend is opposite in the top left and right panels, which is confusing when comparing the two. Moreover, the unit in the lower left panel is stated to be ‘m’, but dSSH/dx should be dimensionless.
Figure 8: The unit on the x-axis in the left panel (m/km) must be wrong. In the associated bottom left panel in Figure 7, the unit is 10-6 m. Please check.
Figure 10: Your analysis is based on model results, but here you state that this is a diagram of the observed ocean feedback loop. It would be more consistent to state that the diagram is showing the modelled ocean feedback loop.
Figure S3 label: The figure label states “decadally averaged Turbulent (top), Radiative (middle) and total (bottom) …”, but it does not tell turbulent or total of what. I presume it is heat fluxes, but please state it explicitly.
Citation: https://doi.org/10.5194/egusphere-2025-5737-RC2 -
AC2: 'Reply on RC2', Shaun Eisner, 14 Jan 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-5737/egusphere-2025-5737-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Shaun Eisner, 14 Jan 2026
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2025-5737', Anonymous Referee #1, 19 Dec 2025
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-5737/egusphere-2025-5737-RC1-supplement.pdf
-
AC1: 'Reply on RC1', Shaun Eisner, 14 Jan 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-5737/egusphere-2025-5737-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Shaun Eisner, 14 Jan 2026
-
RC2: 'Comment on egusphere-2025-5737', Anonymous Referee #2, 05 Jan 2026
Review of the manuscript “Heat transport to the Central Arctic is Reduced by the Barents Sea Cooling Machine”, by S. Eisner et al.
In this manuscript, the authors utilize a state-of-the-art ocean model reanalysis to investigate a postulated feedback mechanism in the flow of warm and saline Atlantic Water through the Barents Sea en-route to the Polar Basin. The authors report that the feedback mechanism, in which increased heat loss to the atmosphere leads to increased formation and export of dense water and, thus, inflow further upstream, is substantiated based on analysis of the model results. Moreover, they report a positive trend not only in the heat carried by the Atlantic Water entering the Barent Sea in the southwest, but also to a lesser degree in the transformed water masses exiting into the Polar Basin to the northeast. The manuscript provides novel findings regarding an important topic that contributes to improve our understanding of the climate system in this hotspot of Arctic climate change and amplification. However, I have some concerns regarding both the findings, their methodological approach and the interpretation of the results, that the authors need to address. Please see my specific points below.
Specific comments:
Title: I think the title needs to be reworked a bit. As it reads now, it is stating the obvious, that a cooling machine cools the heat transport to the Arctic. However, the main finding is that the cooling machine is not fully keeping up the pace of the warming in the Barents Sea region in the sense that there is a positive trend in the heat transport to the Polar Basin, although this trend is less than the trend in the upstream heat transport into the Barents Sea. This should be better reflected in the title, I think.
L1: “Central Arctic Ocean” – is this an official name? The deep part of the Arctic Ocean is often referred to as the Polar Basin. It may also be referred to as the central Arctic Ocean, but not with a capital ‘C’. And in L2 you refer to the central Arctic with a lowercase ‘c’. Please be consistent.
L2: Here and elsewhere: Sometimes you refer to the Barents Sea, and sometimes only to the ‘Barents’. I presume you refer to the same, namely the Barents Sea, in both cases. Please be consistent and spell out ‘Barents Sea’ everywhere you refer to it.
L4: Here and elsewhere: Please spell ‘St. Anna Trough’ with a capital ‘T’.
L5: I think you should change ‘increasing’ with ‘the’: “… trend in the heat transport into the Barents Sea …”
L7-8: Has the migration of the “center of action” of the net heat loss from ocean to the atmosphere resulted in a cooler and denser Barents Shelf Water? Rather, you have shown that the temperature and density, in general, have increased. Regarding the density of the Barents Shelf Water, see also my comment to lines 169-171.
L14: The Bering Strait is not a primary gateway for Atlantic Water to the Arctic Ocean, so the reference to the Bering Strait does not belong here.
L16-17: While the Atlantic Water carries a substantial amount of heat and salt through the Barents Sea Opening, the contribution of heat to the Arctic Ocean has been questioned (e.g., Gammelrsød et al., 2009), although data sources in the exit region of the Barents Sea are scarce.
L25-26: On line 19 you state that Atlantic Water are transformed into Barents Shelf Water within the Barents Sea, while here you state that both BSW and AW leaves the Barents Sea and are carried by the ACBC as far as the Amerasian Basin. There is little evidence that water masses leaving the Barents Sea through the At. Anna Trough can still be classified as Atlantic Water.
L38 (and L164-165): This interpretation of the feedback loop proposed by Smedsrud et al. (2013) is not entirely correct. One main argument in the proposed feedback loop is that increased density in the BSX causes a density gradient between the BSX and the Polar Basin, and therefore also an accelerated outflow through the St. Anna Trough. This, in turn, creates an SSH gradient and an associated barotropic forcing through the Barents Sea (from the BSO to the BSX) that favours stronger throughflow.
L42: “Northern Barents Sea”: The region of the Barents Sea that is the focus area of this study is often referred to as the northeastern Barents Sea, whereas “northern Barents Sea” often refers to the area between the Svalbard and the Franz Josef Land archipelagos. Therefore, I recommend referring to your focus area as “the northeastern Barents Sea”.
L57: “Barents Sea cooling system” – I would rather use the term “cooling processes”.
L65: The last sentence in this paragraph clearly belongs to the paragraph below, because I assume you are not using monthly averaged atmospheric reanalysis as forcing for the ocean model!
L77: The acronym ‘STA’ is not defined yet (it is defined on line 89), although the St. Anna Trough is first mentioned on line 26 (except for in the abstract) and also on several occasions before the acronym is established.
L117: There is a reference to Fig. 2b, but Fig. 2 does not have any panel b as far as I can see.
L119: Note, that the areas with the greatest increases in heat content are also the deepest parts of the Barents Sea (the Bear Island Trough and the Hopen Trench, the Central Basin, and the St. Anna Trough), with the exception being the Northeast Basin. Normalizing the change in heat content by calculating per meter water depth may provide a somewhat different picture.
L124-126: So, a conclusion from this comparison with observations is that the heat loss in the model is smaller than the observed heat loss. I think this observation deserve a bit more discussion considering that the heat loss in the Barents Sea is the main topic of the paper.
L128-129: Change “upward” and “downward” trends with “positive” and “negative” trends, respectively. For clarity, you should also consider removing the reference to anomalies, and instead report that there are trends in the temperature and salinity (and, hence, also a trend in anomalies). Because a negative trend in anomalies could be interpreted as a trend towards more negative anomalies, but it could also be interpreted as a reduction in the magnitude of the anomalies.
L138: I suppose you mean “There is a clear positive trend in the average temperature”. An increasing trend, strictly speaking, means an accelerating trend (i.e., a positive second derivative). Please clarify.
L149: I suggest changing “heat content” with “temperature”, because that is what you have shown when decomposing the heat transport in Fig. 4.
L153: I suggest replacing “The reduced increase” with “The smaller increase”.
L154: Missing ‘o’ in “accommodate”
L169-171: You state that you find an increase in the average water column density in the northeastern Barents Sea from the 1980s to the 2020s, and a collocated reduction in the SSH, which is shown in Fig. 7 (left panels). I have two main concerns with these findings and your interpretation of them. The first concern is related to the effect of the SSH difference on the circulation. A change in the density difference between two regions, here the northern Barents Sea and the Eurasian Basin, also warrants an associated change in the SSH difference between the two regions. Here, the lower SSH in the northern Barents Sea together with the larger density in the same region may cause a net zero change in the pressure difference between the two regions (a higher water column of lower density vs. a lower water column of higher density). Thus, the conclusion that these changes in SSH will cause a geostrophic circulation anomaly is not necessarily substantiated. Moreover, Figure 7, lower left panel, shows a negative “SSH curl” along the eastern flank of the St. Anna Trough (where the Barents Sea Water exits the Barents Sea for the Polar Basin). This would imply a negative contribution to the flow in the cyclonic direction (if I understand the interpretation of the “SSH curl” on lines 79-80 correct). This would be in agreement with a circulation anomaly with the higher SSH on the right-hand side, as also implied by the SSH contours in the panels on the left-hand side in Fig. 7. But this would imply weaker, not stronger, outflow through the eastern St. Anna Trough. And, as a consequence, I have a bit of a hard time following your interpretation of the results depicted in Fig. 7 as synthesized in Fig. 10. Also, the change in density as shown in Fig. 7, upper left panel, seems very large. A density decrease representing a full unit (1 kg/m3) when averaged over the full water column is substantial. And this leads to my second concern. You state, based on Fig. 7, that the density of the Barents Shelf Water and the dense water formation in the Barents Sea have increased. However, such changes will be disguised due to the fact that several water masses are included in your calculation of total water column density change. In the area where you find the largest density increase, in the Northeast Basin in the northeastern Barents Sea, the water column may consist of dense bottom water formed through brine rejection during ice formation, overlaid by Atlantic Water (Barents Sea/Shelf Water), overlaid by Arctic Water. Any change in the mass distribution between these three water masses will also affect the average density of the water column. Indeed, Lind et al. (2018) reported a decline in the amount of Arctic Water in recent decades in this region, which by itself would increase the water column density because the colder and fresher (and less dense) Arctic Water was replaced by warmer and more saline (and denser) Atlantic Water.
L176-177: Usually, the timeseries are detrended before the correlation analysis is performed. Thus, the different trends in the two timeseries should not affect the correlation between them. Also here, you state a “decreasing trend”, while a presume you mean a “negative trend”.
L195-L198: While you have found some support for an ocean feedback mechanism, the underlying hypothesis and the associated analysis is slightly different from that of Smedsrud et al. (2013). Also, your analysis is based on model results, hence, you have not presented observational evidence for an ocean feedback mechanism, but rather empirical evidence, which was also, to some degree, presented in Smedsrud et al. (2013).
L199: Again: “increasing trend” -> “positive trend”.
L210-211: The hypothesis proposed by Smedsrud et al. (2013) postulated that an increase in the dense water formation in the BSX area will cause a density-driven acceleration of the outflow through the St. Anna Trough, and a subsequent increase in the BSO inflow due to mass continuity.
L212: The reference to Fig. 10 should be a reference to Fig. 9, I believe.
L215: Again, I would rather use the term “cooling processes”.
Figures:
Figure 3: I think you got the references (left/right and top/bottom) to the variables and geographical locations mixed up. Please correct. Also, it is not stated which period the anomalies are relative to. Even if it is mentioned in the main text, please repeat it in the figure caption.
Figure 7: a), b), etc. missing on figure panels. Also, the colour legend is opposite in the top left and right panels, which is confusing when comparing the two. Moreover, the unit in the lower left panel is stated to be ‘m’, but dSSH/dx should be dimensionless.
Figure 8: The unit on the x-axis in the left panel (m/km) must be wrong. In the associated bottom left panel in Figure 7, the unit is 10-6 m. Please check.
Figure 10: Your analysis is based on model results, but here you state that this is a diagram of the observed ocean feedback loop. It would be more consistent to state that the diagram is showing the modelled ocean feedback loop.
Figure S3 label: The figure label states “decadally averaged Turbulent (top), Radiative (middle) and total (bottom) …”, but it does not tell turbulent or total of what. I presume it is heat fluxes, but please state it explicitly.
Citation: https://doi.org/10.5194/egusphere-2025-5737-RC2 -
AC2: 'Reply on RC2', Shaun Eisner, 14 Jan 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2025/egusphere-2025-5737/egusphere-2025-5737-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Shaun Eisner, 14 Jan 2026
Peer review completion
Journal article(s) based on this preprint
Data sets
SODA4 Dataset Gennady Chepurin et al. https://dsrs.atmos.umd.edu/DATA/
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 302 | 144 | 37 | 483 | 52 | 20 | 20 |
- HTML: 302
- PDF: 144
- XML: 37
- Total: 483
- Supplement: 52
- BibTeX: 20
- EndNote: 20
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Shaun A. Eisner
James A. Carton
Leon Chafik
Lars H. Smedsrud
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(15406 KB) - Metadata XML
-
Supplement
(4953 KB) - BibTeX
- EndNote
- Final revised paper