the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 06 Dec 2022
Tropospheric NO2 vertical profiles over South Korea and their relation to oxidant chemistry: Implications for geostationary satellite retrievals and the observation of NO2 diurnal variation from space
Abstract. Tropospheric nitrogen dioxide (NO2) is of central importance for air quality, climate forcing, and nitrogen deposition to ecosystems. The Geostationary Environment Monitoring Spectrometer (GEMS) is now providing high-density NO2 satellite data including diurnal variation over East Asia. The NO2 retrieval requires independent vertical profile information from a chemical transport model (CTM) to compute the air mass factor (AMF) that relates the NO2 column along the line of sight to the NO2 vertical column. Here, we use aircraft observations from the Korea-United States Air Quality (KORUS-AQ) campaign over the Seoul Metropolitan Area (SMA) and around the Korean peninsula to better understand the factors controlling the NO2 vertical profile, its diurnal variation, the implications for the AMF, and the ability of the GEOS-Chem CTM to compute the AMF and its variability. Proper representation of oxidant chemistry is critical for the CTM simulation of NO2 vertical profiles and is achieved in GEOS-Chem through new model developments including aerosol nitrate photolysis, reduced uptake of hydroperoxy (HO2) radicals by aerosols, and accounting for atmospheric oxidation of volatile chemical products (VCPs). We find that the tropospheric NO2 columns measured from space are mainly contributed by the planetary boundary layer (PBL) below 2 km altitude, reflecting the highly polluted conditions. Repeated measurements of NO2 vertical profiles over SMA at different times of day show that diurnal change in mixing depth affecting the NO2 vertical profile induces a diurnal variation in AMF of comparable magnitude to the diurnal variation in the NO2 column. GEOS-Chem captures this diurnal variation in AMF and more generally the variability in the AMF for the KORUS-AQ NO2 vertical profiles (2.7 % mean bias, 7.6 % precision), with some outliers in the morning due to non-systematic errors in the timing of mixed layer growth.
-
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
(2198 KB)
-
Supplement
(617 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(2198 KB) - Metadata XML
-
Supplement
(617 KB) - BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2022-1309', Anonymous Referee #1, 20 Dec 2022
Yang et al. used aircraft observations from the Korea-United States Air Quality (KORUS-AQ) campaign to get detailed insights into the diurnal variability and vertical distribution of tropospheric NO2. With this knowledge, Chemical Transport Model (CTM) simulations were optimized regarding mainly the oxidant chemistry. Furthermore, the authors analyzed the diurnal variability of the air mass factor (AMF), which is crucial when retrieving NO2 vertical column densities from the recently launched Geostationary Environment Monitoring Spectrometer (GEMS). This work is relevant since for the GEMS measurement campaign, a detailed knowledge about the AMF in this region is required and until know not known. The authors show, that their optimization of the CTM simulation within the GEOS-Chem model, significantly reduces the bias between observation and model. Nevertheless, especially in the planetary boundary layer (PBL), the model still does not reflect completely the observations. However, the AMF variability can be mostly captured by the GEOS-Chem model. There are only significant deviations in the morning due to incorrect timing of mixed layer growth. I recommend this paper to be published in Atmospheric Measurement Techniques, after the following minor points of criticism will have been addressed.
General remarks:
It is hard to follow the description of the figures. Especially for Fig. 4 and 5, I would suggest a denotation with a/b/c/d and a clear reference when discussing trends and/or comparing data.
The calculation of errors and their sources are in many cases not clear. Please clarify for the whole manuscript.
Specific remarks:
Line 103: …which leads us to evaluation… --> …which leads us to the evaluation…
Line 123: …, A high-Resolution… --> a High-Resolution
Line 132: Introduce here the abbreviation of the version v13.3.4 and not in Line 145
Line 167: The sentence “The NO3- photolysis…x jHNO3.” is not clear.
- EF relative to jHNO3 mostly means a ratio and not a multiplication
- please rephrase
Line 178: what does it mean “very high”? In comparison to what?
Line 205: Is this 50 % increase just an empirical estimation? Why this exact value?
Line 219: Is zt constant or date dependent?
The index of zt must not to be in italics.Line 259 and 261: How are the relative differences of the profiles calculated? At one given altitude? Or overall the PBL?
Line 277: What is the accuracy of the measurement? Is it possible to show error bars or at least give an accuracy?
Line 332ff: Is the described behavior only taken from literature (Chong/Crawford) or is it taken from the new data? Where is this time resolution shown? Figure 5 only shows the given 3 timeslots
Line 353: mention, that the model values are given in parentheses in Table 1.
Line 373f: please rephrase the sentence.
Line 364f: what are the values given in parentheses? Mean+StdDev?
Line 372: Which errors do you mean? The RRMSE from above?
Citation: https://doi.org/10.5194/egusphere-2022-1309-RC1 -
RC2: 'Comment on egusphere-2022-1309', Anonymous Referee #2, 04 Jan 2023
This manuscript investigates the effect of the diurnal variability of NO2 vertical structure on the air mass factor. It is a nice albeit brief investigation into this topic. I recommend publication after minor revisions.
The lone major suggestion is to add-on to Figure 5 to better show the diurnal variability of NO2 at various altitudes. It’s hard to tell whether GEOS-Chem is getting the diurnal patterns correctly. I do think this is a critical missing piece of the manuscript because it would more clearly demonstrate whether NO2 vertical mixing in GEOS-Chem needs to be further improved (in future work). I expand on this further in my comments below.
The Introduction also needs several minor adjustments, but otherwise the manuscript is in good shape.
Line 22. Modify “high-density” to different phrase. Maybe “hourly high-spatial resolution”
Line 22. I recommend merging sentence #2 of this paragraph into the current sentence #4 of this paragraph. The current sentence #2 is a bit out of place, and should belong in a sentence mentioning the study motivation. Similar to the sentence in Line 50.
Line 32. Mention “SMA” somewhere in this sentence.
Line 41. Never heard of NOx referred to as radicals, but I suppose an argument could be made. Instead I recommend modifying “Nitrogen oxide radicals” to “Nitrogen oxides”
Line 48. Modify “starting to provide” to “providing”
Line 61. Modify “GEMS is the first geostationary instrument” to “GEMS is the first geostationary instrument measuring trace gases, such as NO2.”
Line 62. Modify “from which the stratospheric portion is removed” to “from which the stratospheric portion is removed to estimate the tropospheric SCD”
Line 69. Modify “observed” to “observe”
Line 78 - 80. Remove the sentence “These diurnal variations…”. It is confusing and out of context in my opinion. Penn and Holloway investigated a less polluted area in the US, and should not be compared to Seoul in this context. Subsequently modify “this issue” in Line 81 “the diurnal patterns of NO2”
Line 205. It would be appropriate to add one more sentence here describing what the previous literature found. Are global CO emissions too low? Is the lifetime of CO too short? Or is the reason for the underestimate still not known?
Figure 5 (or a new figure). It would be very interesting to create a pseudo diurnal plot from this information. For example on the x axis would be time of day, and y-axis would be number density. Both the observations and GEOS-Chem would be plotted on the same panel. And then have 4 or 6 panels with the diurnal patterns at 4 or 6 different altitudes (surface+three different heights or surface+five different heights). Currently I cannot tell how well GEOS-Chem is reproducing the diurnal pattern at each altitude (only the absolute values).
Line 348. To complete this paragraph, it’d be best to mention how this will affect the VCD in a final sentence. Since VCD and AMF are inversely correlated, a 14% AMF increase will decrease the VCD 14% more in the afternoon than at mid-day, assuming an identical slant column at both hours.
Line 384. Modify “it would be rare to find a 7×8 km2 GEMS pixel without clouds” to “it would be likely that a 7×8 km2 GEMS pixel would have some amount of clouds”
Citation: https://doi.org/10.5194/egusphere-2022-1309-RC2 -
RC3: 'Comment on egusphere-2022-1309', Anonymous Referee #3, 12 Jan 2023
In this manuscript, the authors report on a study comparing air mass factors for GEMS tropospheric NO2 satellite retrievals over Korea based on measured and modelled NO2 profiles. The measurements are 63 vertical NO2 profiles from the KORUS-AQ campaign, while the modelled profiles are from a dedicated version of the GEOS-Chem model. A particular emphasis is on the evaluation of diurnal changes in the vertical NO2 profiles and thus the air mass factors. The main conclusions from the paper are that the NO2 AMFs over Seoul significantly increase over the course of the day and that the new version of the GEOS-Chem model predicts NO2 profiles, which lead to AMFs, which are in good agreement with those based on observed profiles.
The topic of the manuscript is relevant for the atmospheric community, as the diurnal variability of the AMF needs to be understood and modelled well in order to make use of the exciting new observations from geostationary orbit, of which the Korean GEMS instrument is the first to be in operation. The manuscript is overall well written and contains clear figures illustrating the main findings. There is, however, a number of points, which the authors need to address before the manuscript can be accepted for publication.
General Comments
Throughout the manuscript, the wording is in my opinion not as accurate as it should be. There are many occasions where it is said that the “GEOS-Chem calculated AMFs are …” while I think the formulation should be “AMFs based on GEOS-Chem profiles …”. Some more examples from the abstract and summary:
“the ability of the GEOS-Chem CTM to compute the AMF” – GEOS-Chem computes the NO2 profiles, not the AMFs.
“The KORUS-AQ vertical profiles indicate that 95% of the tropospheric NO2 VCD detected from space over the Seoul Metropolitan Area (SMA) originates from the PBL“ – The VCD (if fit is correct) does not depend on how it is detected. What is meant here is probably what the authors earlier called „cumulative SCD“.
„This leads to a large diurnal variability of NO2 detected from space“ – first of all, I’m not sure if a variation of 20% is large for NO2, and second, it is not clear how this is the result of most of the NO2 being located in the PBL.
I think all authors should read the text carefully again and make sure that the wording is precise.
One of the main points of the paper is the improved GEOS-Chem version, which supposedly provides better NO2 profiles for AMF calculations of Asia. However, while the agreement of the new model version with some observed trace gases is improved, this is not obvious for NO2. I therefore suggest to add the AMFs calculated by the standard version of GEOS-Chem in Figure 4 and Table 1 to demonstrate the improvement if there is any or else explicitly discuss the lack in improvement.
Throughout the manuscript it is emphasised, how the new version of the model performs better over Asia, and that GEOS-Chem profiles result in AMFs, which are very close to those derived from measured profiles. While I do not question this, I think it should be stated very clearly, that the KORUS-AQ profiles were all taken during one season and in a very limited number of places, and it is not at all clear if the nice results shown here can be reproduced for another location in another season. To some degree, this model version has been tuned for KORUS-AQ, and I expect problems when it is applied under different conditions.
Somewhere in the manuscript, it should be stated, that all AMF calculations were made for a Rayleigh atmosphere without aerosols (at least I assume that this is the case).
Detailed comments
Line 61: ”GEMS is the first geostationary instrument“ => GEMS is the first geostationary instrument to measure tropospheric NO2
Line 70: “tropospheric NO2 VCDs vary with the time of day as driven by .. mixing depth“ => I do not think that mixing depht is a driver fort he VCD
Line 187: “we find in GEOS-Chem..:” Isn’t that a matter of the emission inventory, and not the CTM?
Line 352: “GEOS-Chem reproduces closely the observed diurnal variation of the scattering correction factor” => GEOS-Chem reproduces the profile, not the scattering factor
Line 363: “surface type” => while this is formally OK, the driver for the difference is the NO2 profile shape, determined by the presence or absence of local sources.
Line 393: “they are not an obvious source of error when comparing model and observed NO2 profiles“ => this probably depends on whether the scene in GEOS-Chem is cloudy or not
Line 394: „the afternoon formation of fair-weather cumuli would decrease the sensitivity of the satellite measurement to the PBL and therefore alias the observed diurnal variation of NO2.“ => yes, but only if no cloud correction is applied in the retrieval.
Line 401: „to better understand how the vertical distribution of NO2 affects the air mass factors (AMFs) for satellite retrievals of tropospheric NO2 vertical column density“ => I think that the effects of NO2 vertical distribution on AMFs are well understood. Isn’t the point here to look at the effect of diurnal changes in the vertical NO2 distribution on the variation of AMFs over the day?
Figure 1: I find it odd to mix volume mixing ratios and mass concentrations in one figure.
Figure 2: While the median profiles are nice, it would be good to add a figure also indicating the variability of these profiles
Figures 2 – 5: There are some strange steps in these curves – what is the vertical resolution / sampling?
Citation: https://doi.org/10.5194/egusphere-2022-1309-RC3 - AC1: 'Comment on egusphere-2022-1309', Laura Yang, 02 Feb 2023
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2022-1309', Anonymous Referee #1, 20 Dec 2022
Yang et al. used aircraft observations from the Korea-United States Air Quality (KORUS-AQ) campaign to get detailed insights into the diurnal variability and vertical distribution of tropospheric NO2. With this knowledge, Chemical Transport Model (CTM) simulations were optimized regarding mainly the oxidant chemistry. Furthermore, the authors analyzed the diurnal variability of the air mass factor (AMF), which is crucial when retrieving NO2 vertical column densities from the recently launched Geostationary Environment Monitoring Spectrometer (GEMS). This work is relevant since for the GEMS measurement campaign, a detailed knowledge about the AMF in this region is required and until know not known. The authors show, that their optimization of the CTM simulation within the GEOS-Chem model, significantly reduces the bias between observation and model. Nevertheless, especially in the planetary boundary layer (PBL), the model still does not reflect completely the observations. However, the AMF variability can be mostly captured by the GEOS-Chem model. There are only significant deviations in the morning due to incorrect timing of mixed layer growth. I recommend this paper to be published in Atmospheric Measurement Techniques, after the following minor points of criticism will have been addressed.
General remarks:
It is hard to follow the description of the figures. Especially for Fig. 4 and 5, I would suggest a denotation with a/b/c/d and a clear reference when discussing trends and/or comparing data.
The calculation of errors and their sources are in many cases not clear. Please clarify for the whole manuscript.
Specific remarks:
Line 103: …which leads us to evaluation… --> …which leads us to the evaluation…
Line 123: …, A high-Resolution… --> a High-Resolution
Line 132: Introduce here the abbreviation of the version v13.3.4 and not in Line 145
Line 167: The sentence “The NO3- photolysis…x jHNO3.” is not clear.
- EF relative to jHNO3 mostly means a ratio and not a multiplication
- please rephrase
Line 178: what does it mean “very high”? In comparison to what?
Line 205: Is this 50 % increase just an empirical estimation? Why this exact value?
Line 219: Is zt constant or date dependent?
The index of zt must not to be in italics.Line 259 and 261: How are the relative differences of the profiles calculated? At one given altitude? Or overall the PBL?
Line 277: What is the accuracy of the measurement? Is it possible to show error bars or at least give an accuracy?
Line 332ff: Is the described behavior only taken from literature (Chong/Crawford) or is it taken from the new data? Where is this time resolution shown? Figure 5 only shows the given 3 timeslots
Line 353: mention, that the model values are given in parentheses in Table 1.
Line 373f: please rephrase the sentence.
Line 364f: what are the values given in parentheses? Mean+StdDev?
Line 372: Which errors do you mean? The RRMSE from above?
Citation: https://doi.org/10.5194/egusphere-2022-1309-RC1 -
RC2: 'Comment on egusphere-2022-1309', Anonymous Referee #2, 04 Jan 2023
This manuscript investigates the effect of the diurnal variability of NO2 vertical structure on the air mass factor. It is a nice albeit brief investigation into this topic. I recommend publication after minor revisions.
The lone major suggestion is to add-on to Figure 5 to better show the diurnal variability of NO2 at various altitudes. It’s hard to tell whether GEOS-Chem is getting the diurnal patterns correctly. I do think this is a critical missing piece of the manuscript because it would more clearly demonstrate whether NO2 vertical mixing in GEOS-Chem needs to be further improved (in future work). I expand on this further in my comments below.
The Introduction also needs several minor adjustments, but otherwise the manuscript is in good shape.
Line 22. Modify “high-density” to different phrase. Maybe “hourly high-spatial resolution”
Line 22. I recommend merging sentence #2 of this paragraph into the current sentence #4 of this paragraph. The current sentence #2 is a bit out of place, and should belong in a sentence mentioning the study motivation. Similar to the sentence in Line 50.
Line 32. Mention “SMA” somewhere in this sentence.
Line 41. Never heard of NOx referred to as radicals, but I suppose an argument could be made. Instead I recommend modifying “Nitrogen oxide radicals” to “Nitrogen oxides”
Line 48. Modify “starting to provide” to “providing”
Line 61. Modify “GEMS is the first geostationary instrument” to “GEMS is the first geostationary instrument measuring trace gases, such as NO2.”
Line 62. Modify “from which the stratospheric portion is removed” to “from which the stratospheric portion is removed to estimate the tropospheric SCD”
Line 69. Modify “observed” to “observe”
Line 78 - 80. Remove the sentence “These diurnal variations…”. It is confusing and out of context in my opinion. Penn and Holloway investigated a less polluted area in the US, and should not be compared to Seoul in this context. Subsequently modify “this issue” in Line 81 “the diurnal patterns of NO2”
Line 205. It would be appropriate to add one more sentence here describing what the previous literature found. Are global CO emissions too low? Is the lifetime of CO too short? Or is the reason for the underestimate still not known?
Figure 5 (or a new figure). It would be very interesting to create a pseudo diurnal plot from this information. For example on the x axis would be time of day, and y-axis would be number density. Both the observations and GEOS-Chem would be plotted on the same panel. And then have 4 or 6 panels with the diurnal patterns at 4 or 6 different altitudes (surface+three different heights or surface+five different heights). Currently I cannot tell how well GEOS-Chem is reproducing the diurnal pattern at each altitude (only the absolute values).
Line 348. To complete this paragraph, it’d be best to mention how this will affect the VCD in a final sentence. Since VCD and AMF are inversely correlated, a 14% AMF increase will decrease the VCD 14% more in the afternoon than at mid-day, assuming an identical slant column at both hours.
Line 384. Modify “it would be rare to find a 7×8 km2 GEMS pixel without clouds” to “it would be likely that a 7×8 km2 GEMS pixel would have some amount of clouds”
Citation: https://doi.org/10.5194/egusphere-2022-1309-RC2 -
RC3: 'Comment on egusphere-2022-1309', Anonymous Referee #3, 12 Jan 2023
In this manuscript, the authors report on a study comparing air mass factors for GEMS tropospheric NO2 satellite retrievals over Korea based on measured and modelled NO2 profiles. The measurements are 63 vertical NO2 profiles from the KORUS-AQ campaign, while the modelled profiles are from a dedicated version of the GEOS-Chem model. A particular emphasis is on the evaluation of diurnal changes in the vertical NO2 profiles and thus the air mass factors. The main conclusions from the paper are that the NO2 AMFs over Seoul significantly increase over the course of the day and that the new version of the GEOS-Chem model predicts NO2 profiles, which lead to AMFs, which are in good agreement with those based on observed profiles.
The topic of the manuscript is relevant for the atmospheric community, as the diurnal variability of the AMF needs to be understood and modelled well in order to make use of the exciting new observations from geostationary orbit, of which the Korean GEMS instrument is the first to be in operation. The manuscript is overall well written and contains clear figures illustrating the main findings. There is, however, a number of points, which the authors need to address before the manuscript can be accepted for publication.
General Comments
Throughout the manuscript, the wording is in my opinion not as accurate as it should be. There are many occasions where it is said that the “GEOS-Chem calculated AMFs are …” while I think the formulation should be “AMFs based on GEOS-Chem profiles …”. Some more examples from the abstract and summary:
“the ability of the GEOS-Chem CTM to compute the AMF” – GEOS-Chem computes the NO2 profiles, not the AMFs.
“The KORUS-AQ vertical profiles indicate that 95% of the tropospheric NO2 VCD detected from space over the Seoul Metropolitan Area (SMA) originates from the PBL“ – The VCD (if fit is correct) does not depend on how it is detected. What is meant here is probably what the authors earlier called „cumulative SCD“.
„This leads to a large diurnal variability of NO2 detected from space“ – first of all, I’m not sure if a variation of 20% is large for NO2, and second, it is not clear how this is the result of most of the NO2 being located in the PBL.
I think all authors should read the text carefully again and make sure that the wording is precise.
One of the main points of the paper is the improved GEOS-Chem version, which supposedly provides better NO2 profiles for AMF calculations of Asia. However, while the agreement of the new model version with some observed trace gases is improved, this is not obvious for NO2. I therefore suggest to add the AMFs calculated by the standard version of GEOS-Chem in Figure 4 and Table 1 to demonstrate the improvement if there is any or else explicitly discuss the lack in improvement.
Throughout the manuscript it is emphasised, how the new version of the model performs better over Asia, and that GEOS-Chem profiles result in AMFs, which are very close to those derived from measured profiles. While I do not question this, I think it should be stated very clearly, that the KORUS-AQ profiles were all taken during one season and in a very limited number of places, and it is not at all clear if the nice results shown here can be reproduced for another location in another season. To some degree, this model version has been tuned for KORUS-AQ, and I expect problems when it is applied under different conditions.
Somewhere in the manuscript, it should be stated, that all AMF calculations were made for a Rayleigh atmosphere without aerosols (at least I assume that this is the case).
Detailed comments
Line 61: ”GEMS is the first geostationary instrument“ => GEMS is the first geostationary instrument to measure tropospheric NO2
Line 70: “tropospheric NO2 VCDs vary with the time of day as driven by .. mixing depth“ => I do not think that mixing depht is a driver fort he VCD
Line 187: “we find in GEOS-Chem..:” Isn’t that a matter of the emission inventory, and not the CTM?
Line 352: “GEOS-Chem reproduces closely the observed diurnal variation of the scattering correction factor” => GEOS-Chem reproduces the profile, not the scattering factor
Line 363: “surface type” => while this is formally OK, the driver for the difference is the NO2 profile shape, determined by the presence or absence of local sources.
Line 393: “they are not an obvious source of error when comparing model and observed NO2 profiles“ => this probably depends on whether the scene in GEOS-Chem is cloudy or not
Line 394: „the afternoon formation of fair-weather cumuli would decrease the sensitivity of the satellite measurement to the PBL and therefore alias the observed diurnal variation of NO2.“ => yes, but only if no cloud correction is applied in the retrieval.
Line 401: „to better understand how the vertical distribution of NO2 affects the air mass factors (AMFs) for satellite retrievals of tropospheric NO2 vertical column density“ => I think that the effects of NO2 vertical distribution on AMFs are well understood. Isn’t the point here to look at the effect of diurnal changes in the vertical NO2 distribution on the variation of AMFs over the day?
Figure 1: I find it odd to mix volume mixing ratios and mass concentrations in one figure.
Figure 2: While the median profiles are nice, it would be good to add a figure also indicating the variability of these profiles
Figures 2 – 5: There are some strange steps in these curves – what is the vertical resolution / sampling?
Citation: https://doi.org/10.5194/egusphere-2022-1309-RC3 - AC1: 'Comment on egusphere-2022-1309', Laura Yang, 02 Feb 2023
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 424 | 163 | 15 | 602 | 29 | 4 | 7 |
- HTML: 424
- PDF: 163
- XML: 15
- Total: 602
- Supplement: 29
- BibTeX: 4
- EndNote: 7
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| United States of America | 1 | 343 | 54 |
| China | 2 | 43 | 6 |
| Germany | 3 | 38 | 6 |
| South Korea | 4 | 35 | 5 |
| Netherlands | 5 | 21 | 3 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 343
Cited
Laura Hyesung Yang
Daniel J. Jacob
Nadia K. Colombi
Shixian Zhai
Kelvin H. Bates
Viral Shah
Ellie Beaudry
Robert M. Yantosca
Haipeng Lin
Jared F. Brewer
Heesung Chong
Katherine R. Travis
James H. Crawford
Lok Lamsal
Ja-Ho Koo
Jhoon Kim
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(2198 KB) - Metadata XML
-
Supplement
(617 KB) - BibTeX
- EndNote
- Final revised paper