the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Diurnal carbon monoxide observed from a geostationary infrared hyperspectral sounder: First result from GIIRS onboard FY-4B
Chengli Qi
Abstract. The Geostationary Interferometric Infrared Sounder (GIIRS) onboard FengYun-4 series satellites is the world’s first geostationary hyperspectral infrared sounder. With hyperspectral measurement covering the carbon monoxide (CO) absorption window around 2150 cm-1, GIIRS provides a unique opportunity for monitoring the diurnal variabilities of atmospheric CO over East Asia. In this study, we develop the FengYun Geostationary satellite Atmospheric Infrared Retrieval (FY-GeoAIR) algorithm to retrieve the CO profiles from FY-4B/GIIRS data and provide CO maps at a spatial resolution of 12 km and a temporal resolution of 2 hours. The performance of the algorithm is first evaluated by conducting retrieval experiments using simulated synthetic spectra. The result shows that the GIIRS data provide significant information for constraining CO profiles. The degree of freedom for signal (DOFS) and retrieval error are both significantly correlated with thermal contrast (TC), the temperature difference between the surface and the lower atmosphere. Retrieval results from one month of GIIRS spectra in July 2022 show that the DOFS for the majority is between 0.6 and 1.2 for the CO total column and between 0 and 0.25 for the bottom 0–1 km layer. Consistent with CO retrievals from low-earth-orbit (LEO) infrared sounders, the largest observation sensitivity, as quantified by the averaging kernel (AK), is in the free troposphere at around 3–6 km. The diurnal changes in DOFS and vertical sensitivity of observation are primarily driven by the diurnal TC variabilities. Finally, we compare the CO total columns between GIIRS and IASI and find that the two datasets show good consistency in capturing the daily variabilities. This study demonstrates the capability of GIIRS in observing the diurnal CO changes in East Asia, which will have great potential in improving local and global air quality and climate research.
Zhao-Cheng Zeng et al.
Status: closed
-
CC1: 'Comment on amt-2022-305', Mengqi Zhang, 21 Nov 2022
This article used the GIIRS, which is the fisrt geostationary Infrared hyperspectra Sounder over the world, to retrieve the diurnal carbon monoxide. CO is very important atmospheric pollutant and a tracer of CO2. This work is very meaningful and the paper is well-writen and well organized. This would be the sencond work on atmospheric trace gas retrieval after Lieven Clarisse(2021)(https://doi.org/10.1029/2021GL093010), and also is the fisrt work on GIIRS-FY4B and CO.
As a community comment, I highly recommend publication to raise awareness of thermal infrared detection of trace gases.
I also have the following suggestions and questions.
The colormap in Figure 10(a) and Figure 11(b) shoud be changed. The viridis colormap is hard for reading and knowing the spatial change. May jet ,rainbow, or some other colormap are suitable.
The GIIRS of FY4A has a certain degree of wavelength calibration offset. Is the GIIRS of FY4B better in wavelength calibration? How is this considered in the inversion? Should the wavelength be calibrated first, or should it be brought into the inversion model for optimization iterations?See the GIIRS FY4A wavelength calibration problem on: https://www-cdn.eumetsat.int/files/2021-01/8%20-%20Coheur%20-%2017h15%20-%20Preliminary%20results%20on%20NH3%20retrievals%20using%20GIIRS.pdf
P12,L320. The xa, xtrue in the formula shoud be differentiated from the previous formula (Eq 5). The same express xa and xa may lead some confusion. May CO_a or CO_true be better. This may be helpful for some readers. x_a in Eq 5 including CO_a and other state vectors.
Are there any plans to apply the algorithm to FY4A with data from 2019? (Although FY4B has better instrument performance.)
GIIRS completes a scan cycle in about 2 hours, so the data at a certain position within 0-2h is just an instantaneous value within the cycle. Although there is no difference in value, it may be better to remind readers to pay attention.
The temperature profile is a key physical quantity for CO inversion, and the ERA5 reanalysis data was used in this study. How sensitive is the algorithm to the temperature profile? The CO2 absorption band of GIIRS has the ability to invert temperature profiles. Would the inversion results for trace gases be better using their inversion temperature profiles?
P5, Figure 1c. .... (bottom) Jacobian for CO ..... May add the matrix would be better( Jacobian matrix).
Overall, this article is very valuable and meaningful. I highly recommend publication.
Citation: https://doi.org/10.5194/amt-2022-305-CC1 -
AC3: 'Reply on CC1', Zhao-Cheng Zeng, 01 Mar 2023
The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2022-305/amt-2022-305-AC3-supplement.pdf
-
AC3: 'Reply on CC1', Zhao-Cheng Zeng, 01 Mar 2023
-
RC1: 'Comment on amt-2022-305', Anonymous Referee #1, 13 Dec 2022
Review of Zeng et al., 2022
This paper describes CO retrievals from the geostationary hyperspectral infrared sounders GIIRS onboard FY-4B. This is the first publication presenting CO retrievals from a geostationary platform that could be valuable to document the diurnal cycle of this species in the lower troposphere. The paper is correctly writen and structured with some interesting information. It is therefore adapted to AMT.
Nevertheless I have some important concerns and questions that have to be adressed before publication. My most important concern is about the diurnal cycle itself which is the main topic of the paper and included in its title. It should be better documented and compared to other datasets to be validated at minima.
CO a priori profiles :
The advantages and inconvenients of a climatological a priori are mentioned in section 7.1 and the use of a single a priori as a way to improve the algorithm to detect anomalies. It should also be mentioned that using such an a priori makes the retrievals more complicated to interpret and to use for model validation.
The use of a 3 hourly profile climatology based on 5 years simulation is done to help to provide the correct diurnal cycle to the retrieval algorithm. But CO is a pollutant with a lifetime much larger than a day. The daily cycle for CO is not as important as for NOx. The authors should provide the plots of the daily variations of CO in the troposphere and lower troposphere in the 3 selected regions from the ECMWF CAMS for instance together with the surface and bottom air temperatures in Fig 2.
Figure 4 shows that the same low biases for high concentrations in the a priori are partially kept in the retrieval for North China Plain and Mongolia. The dissapearence of these biases when the AvKs are applied to the « true » profiles clearly indicates that these biases are linked to the a priori and the lack of sensitivity of the sensor/retrieval to the polluted BL. As stated by the authors, this problem could be related to the too tight a priori covariance matrix used with the climatological profiles but it is not sure. It would be very interesting to provide some results from a simple test using a single a priori profile and its more loose a priori covariance to verify this assumption. In that case the signal to noise ratio for the retrieval which has been tuned according to the a priori covariance matrices (section 6.1) should be lowered which could lead to a destabilisation of the retrieval and possible oscillations in the profiles.
What diurnal cycle ?
The problem to document the CO diurnal cycle with the GIIRS retrievals come from the fact that it could be linked to :
- the real CO cycle that is the objective
- the variability of the BL layer with probably a better detection of pollution in the afternoon when the BL is higher where the sensor is more sensitive
- the variability of the DOFS
In order to disantangle these different sources of diurnal CO cycles
- the diurnal cycles of CO total columns over the 3 selected zones should be provided clearly the same way as the DOFS in Fig 8. The plot of the DOFS for the 0-1 km could be removed as the retrieval for this layer provides no relevant information (see next comment) and as its diurnal variability is mostly similar to the total column.
- the diurnal cycle of the BL height could also be documented from ECMWF ERA5 data for instance.
- the diurnal cycle of CO from other sources such as local pollution networks in China/ Beijing area, ECMWF CAMS used for the a priori (see above) and some references to relevant publications should be provided to check wether or not GIIRS retrievals are sensitive to BL pollution diurnal variability.
I have some doubts about the diurnal variability displayed in fig 10 and 11 in the NCP : the maxima are detected between 16 and 22 UTC that is between midnight and 6AM Beijing time (If I understood correctly). So it does not correspond to the time of day (i) with the highest activity where we expect the largest emissions and concentrations (this should be highlighted by surface /CAMS data as proposed above) (ii) with the highest BL which is in the afternoon (iii) with the largerst DOFS which is the begining of the afternoon (see Fig 8 and Fig 10 b). The authors have to provide some explanations about this peculiar diurnal cycle.
Why 0-1 km layer ?
The 0-1 km could be interesting to document BL pollution but it is characterised by a very low DFS mostly below 0.1 to 0.15 (Fig 3 and 9) and below 0.125 on Fig 8 which means that there is almost no information about this layer in the retrieval whatever the thermal contrast. DOFS is even negative (what does that means?) in Fig 8 and 9 showing that this layer is absolutely not a good choice.
The sentence line 400 « The DOFS can be as large as 0.3 providing a strong constraint on the bottom 0-1 km » is a flagrant overstatement (just a couple of points at 0.3 in Fig 9!!!) and should be removed or changed. Even a DOFS of 0.3 would have meant that the information for this layer is low.
In Figure 7 that displays the AvKs we see that the AvKs peak at 800 or 700 hPa in the best cases.
I therefore do not see the relevance to display results about the 0-1 km layer in the different figures. As the DOFS for the total columns are roughly between 0.8 and 1.2, the authors should separate the atmosphere/troposphere in the two layers in which the information is equally provided and display results for the lowermost of those layers.
Comparisons with IASI :
The comparison with IASI data is made to partly validate the diurnal variations but some important information is missing :
- as there are only two overpasses of IASI daily at 9:30 LST AM and PM, the authors have to detail how they average the GIIRS data temporaly which is not clear at all.
- the correlation coefficients and rmse are given in the Fig 12 but a table with those figures and other basic statistics such as mean biases +/- rmsd should be added.
- the comparison methology to smooth IASI with GIIRS AvKs is assuming that IASI has a much better vertical resolution than GIIRS which is not the case (IASI has probably a DOFS of 1.5). In that sense it is worth to display IASI AvKs to compare with GIIRS (as in Fig 7) and to provide IASI’s DOFS. It would probably be better to avoid to apply equation 11 assuming that both sensors have similar vertical sensitivity or to use the more (too) complicated methodology detailed in Rodgers and Connor, JGR (2003).
- IASI CO "diurnal cycle" is mostly related to its decreased sensitivity at night. So the agreement with GIIRS for day and night described as good by the authors is just indicating that GIIRS has the same decreased sensitivity at night. There should be some statements about this issue.
Detailed comments :
Section 3 and 4 :
Some generalities about radiative transfer and retrieval methodology and basic known equations could be removed. Equations 3 to 8 have been largely documented such as in Rodgers (2000) and it is unnecessary to repeat this here.
Time : the time is given in UTC but it is not a correct choice to interpret diurnal cycles around China. Beijing local solar time would be much better. Furthermore the time is often given without the precision that it is UTC.
Figure 7 : precise the time system chosen.
Figure 8 : we suppose it is UTC !
Figure 10 : please provide hour in LST because UTC is not adapted to the geographical zone.
Fig 12 : Dayth => Day
Line 59 : The authors mention Kobayashi et al. (1999) as one of the first attempt to document CO from space with the japanese IMG ADEOS. Nevertheless, this paper do not present CO retrievals from IMG. The only retrievals of CO from this first spaceborne IR FTS have been published later by Barret et al. (2005).
refs:
C.D. Rodgers and B.J. Connor, Intercomparison of remote sounding instruments, JGR, 2003.
B. Barret, et al., Global carbon monoxide vertical distributions from spaceborne high-resolution FTIR nadir measurements, ACP, 2005.
Citation: https://doi.org/10.5194/amt-2022-305-RC1 -
AC1: 'Reply on RC1', Zhao-Cheng Zeng, 01 Mar 2023
The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2022-305/amt-2022-305-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Zhao-Cheng Zeng, 01 Mar 2023
-
RC2: 'Comment on amt-2022-305', Anonymous Referee #2, 14 Dec 2022
This study describes the CO retrieval algorithm for the geostationary GIIRS sounders on board FY-4B satellite. The paper is in general well explained and clear, particularly for the algorithm, however some information or clarification are missing.
Comments:
Ln 9: The year of when the sounder was launched should be indicated in the abstract to inform the reader.
Ln 10: how having hyperspectral measurements of CO provide diurnal observation of CO? I would think that having geostationary CO data allow observation of diurnal CO variability. Could you reformulate your sentence or give more precision?
Ln 34. I would maybe say “(GEO) orbit can provide contiguous coverage with similar or higher spatial resolution than LEO and a revisit time of 1-2 hours [...]” or a similar sentence. Because, GIIRS has the same spatial resolution than IASI (12km diameter at nadir).
Ln. 89: Figure A1. b and c, are these figures for the same day as Fig. A1.a?
Fig. 1b. The values on Fig. 1B are a too small.
Fig. 1c. It would be interesting to have, as well, the Jacobian by pressure (the ones used for the radiative transfer model). This would inform on the variability that GIIRS channels sensitivity have depending on atmospheric pressure. Additionally, there is no comment on this figure. What do you conclude with this figure in term of sensitivity for CO and H2O with GIIRS?
Ln. 171. You computed 3-hourly CO profile climatology for each month, which month are you talking about? The period of study has not been introduced yet, except result for July 2022 introduced in the abstract.
Ln. 174. Should be “2080 cm-1 to 2120 cm-1”.
Ln. 194. “The number of pressure grids in the forward RT model should be large enough to reduce the error [...]”. A reference is missing here regarding this remark.
Ln 262: To be consistent, I would write the title as “ Averaging kernel (AK) matrix and Degree of Freedom for Signal (DOFS)”
Ln. 276. How much data are removed after the quality filter? Similarly, how much data are removed before and after labeled clear sky for the period of your study, before the post-screening is done?
Ln. 289. Could you precise why you added a Gaussian white noise? Is the added noise mentioned Ln 291 referring to the Gaussian white noise. If yes, then Ln 291 should appear just after the white noise is mentioned Ln. 289.
Ln. 293. You could introduce a map to visualize the regions of interest.
Ln. 304. How can you conclude that from Figure 2?
Ln. 314.: The results of Figure 3 are only for North China Plain, but have you done it also for the two other regions? Mongolia has a more complex diurnal TC change than North China Plain but the surface pressure/topography is also different between the two regions. Would the results of Figure 3 be the same for Mongolia region or not?
Ln. 315. It is confusing, the “truth” is based on the ECMWF EAC4 results but it is also used as CO a priori profile in your retrieval algorithm, so what is the difference between the comparison of (1) and (2)? How can you compare the retrievals to ECMWF EAC4, if you already used this CO profile as a priori profile in your retrieval algorithm (see Ln. 291)?
Ln. 470-479: Talking about wildfires, during the month of July 2022, several wildfires occurred in Siberia. The transport and mixing of CO in the Northern Hemisphere might have bring CO concentrations from the Siberian fires to the regions of your study. Have you looked at that? The ECMWF EAC4 and ER5 simulation/reanalysis used in your algorithm do not include 2022 and so might not be representative of CO concentrations for the month of July 2022. It is difficult to determine if your retrievals are well representative of July 2022 consequently. It could be then interesting to have evaluation of your L2 retrievals to in situ data.
Appendix A: Figure A1.a is not used in the study. Considering the wildfires in July 2022, was there fire CO emissions included in CAMS model?
This study was only done for a summer month, but have you look at other season? The diurnal cycle might be associated to meteorological conditions and emissions patterns different by season. Additionally, having hourly data and comparing North China with Mongolia and East China Sea, I was wondering if you looked at the difference in CO concentration between these regions during the daytime. I would expect to observe highest concentration for North China than Mongolia during the morning time corresponding to rush hours, however this would depend on synoptic disturbances.
Citation: https://doi.org/10.5194/amt-2022-305-RC2 -
AC2: 'Reply on RC2', Zhao-Cheng Zeng, 01 Mar 2023
The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2022-305/amt-2022-305-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Zhao-Cheng Zeng, 01 Mar 2023
Status: closed
-
CC1: 'Comment on amt-2022-305', Mengqi Zhang, 21 Nov 2022
This article used the GIIRS, which is the fisrt geostationary Infrared hyperspectra Sounder over the world, to retrieve the diurnal carbon monoxide. CO is very important atmospheric pollutant and a tracer of CO2. This work is very meaningful and the paper is well-writen and well organized. This would be the sencond work on atmospheric trace gas retrieval after Lieven Clarisse(2021)(https://doi.org/10.1029/2021GL093010), and also is the fisrt work on GIIRS-FY4B and CO.
As a community comment, I highly recommend publication to raise awareness of thermal infrared detection of trace gases.
I also have the following suggestions and questions.
The colormap in Figure 10(a) and Figure 11(b) shoud be changed. The viridis colormap is hard for reading and knowing the spatial change. May jet ,rainbow, or some other colormap are suitable.
The GIIRS of FY4A has a certain degree of wavelength calibration offset. Is the GIIRS of FY4B better in wavelength calibration? How is this considered in the inversion? Should the wavelength be calibrated first, or should it be brought into the inversion model for optimization iterations?See the GIIRS FY4A wavelength calibration problem on: https://www-cdn.eumetsat.int/files/2021-01/8%20-%20Coheur%20-%2017h15%20-%20Preliminary%20results%20on%20NH3%20retrievals%20using%20GIIRS.pdf
P12,L320. The xa, xtrue in the formula shoud be differentiated from the previous formula (Eq 5). The same express xa and xa may lead some confusion. May CO_a or CO_true be better. This may be helpful for some readers. x_a in Eq 5 including CO_a and other state vectors.
Are there any plans to apply the algorithm to FY4A with data from 2019? (Although FY4B has better instrument performance.)
GIIRS completes a scan cycle in about 2 hours, so the data at a certain position within 0-2h is just an instantaneous value within the cycle. Although there is no difference in value, it may be better to remind readers to pay attention.
The temperature profile is a key physical quantity for CO inversion, and the ERA5 reanalysis data was used in this study. How sensitive is the algorithm to the temperature profile? The CO2 absorption band of GIIRS has the ability to invert temperature profiles. Would the inversion results for trace gases be better using their inversion temperature profiles?
P5, Figure 1c. .... (bottom) Jacobian for CO ..... May add the matrix would be better( Jacobian matrix).
Overall, this article is very valuable and meaningful. I highly recommend publication.
Citation: https://doi.org/10.5194/amt-2022-305-CC1 -
AC3: 'Reply on CC1', Zhao-Cheng Zeng, 01 Mar 2023
The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2022-305/amt-2022-305-AC3-supplement.pdf
-
AC3: 'Reply on CC1', Zhao-Cheng Zeng, 01 Mar 2023
-
RC1: 'Comment on amt-2022-305', Anonymous Referee #1, 13 Dec 2022
Review of Zeng et al., 2022
This paper describes CO retrievals from the geostationary hyperspectral infrared sounders GIIRS onboard FY-4B. This is the first publication presenting CO retrievals from a geostationary platform that could be valuable to document the diurnal cycle of this species in the lower troposphere. The paper is correctly writen and structured with some interesting information. It is therefore adapted to AMT.
Nevertheless I have some important concerns and questions that have to be adressed before publication. My most important concern is about the diurnal cycle itself which is the main topic of the paper and included in its title. It should be better documented and compared to other datasets to be validated at minima.
CO a priori profiles :
The advantages and inconvenients of a climatological a priori are mentioned in section 7.1 and the use of a single a priori as a way to improve the algorithm to detect anomalies. It should also be mentioned that using such an a priori makes the retrievals more complicated to interpret and to use for model validation.
The use of a 3 hourly profile climatology based on 5 years simulation is done to help to provide the correct diurnal cycle to the retrieval algorithm. But CO is a pollutant with a lifetime much larger than a day. The daily cycle for CO is not as important as for NOx. The authors should provide the plots of the daily variations of CO in the troposphere and lower troposphere in the 3 selected regions from the ECMWF CAMS for instance together with the surface and bottom air temperatures in Fig 2.
Figure 4 shows that the same low biases for high concentrations in the a priori are partially kept in the retrieval for North China Plain and Mongolia. The dissapearence of these biases when the AvKs are applied to the « true » profiles clearly indicates that these biases are linked to the a priori and the lack of sensitivity of the sensor/retrieval to the polluted BL. As stated by the authors, this problem could be related to the too tight a priori covariance matrix used with the climatological profiles but it is not sure. It would be very interesting to provide some results from a simple test using a single a priori profile and its more loose a priori covariance to verify this assumption. In that case the signal to noise ratio for the retrieval which has been tuned according to the a priori covariance matrices (section 6.1) should be lowered which could lead to a destabilisation of the retrieval and possible oscillations in the profiles.
What diurnal cycle ?
The problem to document the CO diurnal cycle with the GIIRS retrievals come from the fact that it could be linked to :
- the real CO cycle that is the objective
- the variability of the BL layer with probably a better detection of pollution in the afternoon when the BL is higher where the sensor is more sensitive
- the variability of the DOFS
In order to disantangle these different sources of diurnal CO cycles
- the diurnal cycles of CO total columns over the 3 selected zones should be provided clearly the same way as the DOFS in Fig 8. The plot of the DOFS for the 0-1 km could be removed as the retrieval for this layer provides no relevant information (see next comment) and as its diurnal variability is mostly similar to the total column.
- the diurnal cycle of the BL height could also be documented from ECMWF ERA5 data for instance.
- the diurnal cycle of CO from other sources such as local pollution networks in China/ Beijing area, ECMWF CAMS used for the a priori (see above) and some references to relevant publications should be provided to check wether or not GIIRS retrievals are sensitive to BL pollution diurnal variability.
I have some doubts about the diurnal variability displayed in fig 10 and 11 in the NCP : the maxima are detected between 16 and 22 UTC that is between midnight and 6AM Beijing time (If I understood correctly). So it does not correspond to the time of day (i) with the highest activity where we expect the largest emissions and concentrations (this should be highlighted by surface /CAMS data as proposed above) (ii) with the highest BL which is in the afternoon (iii) with the largerst DOFS which is the begining of the afternoon (see Fig 8 and Fig 10 b). The authors have to provide some explanations about this peculiar diurnal cycle.
Why 0-1 km layer ?
The 0-1 km could be interesting to document BL pollution but it is characterised by a very low DFS mostly below 0.1 to 0.15 (Fig 3 and 9) and below 0.125 on Fig 8 which means that there is almost no information about this layer in the retrieval whatever the thermal contrast. DOFS is even negative (what does that means?) in Fig 8 and 9 showing that this layer is absolutely not a good choice.
The sentence line 400 « The DOFS can be as large as 0.3 providing a strong constraint on the bottom 0-1 km » is a flagrant overstatement (just a couple of points at 0.3 in Fig 9!!!) and should be removed or changed. Even a DOFS of 0.3 would have meant that the information for this layer is low.
In Figure 7 that displays the AvKs we see that the AvKs peak at 800 or 700 hPa in the best cases.
I therefore do not see the relevance to display results about the 0-1 km layer in the different figures. As the DOFS for the total columns are roughly between 0.8 and 1.2, the authors should separate the atmosphere/troposphere in the two layers in which the information is equally provided and display results for the lowermost of those layers.
Comparisons with IASI :
The comparison with IASI data is made to partly validate the diurnal variations but some important information is missing :
- as there are only two overpasses of IASI daily at 9:30 LST AM and PM, the authors have to detail how they average the GIIRS data temporaly which is not clear at all.
- the correlation coefficients and rmse are given in the Fig 12 but a table with those figures and other basic statistics such as mean biases +/- rmsd should be added.
- the comparison methology to smooth IASI with GIIRS AvKs is assuming that IASI has a much better vertical resolution than GIIRS which is not the case (IASI has probably a DOFS of 1.5). In that sense it is worth to display IASI AvKs to compare with GIIRS (as in Fig 7) and to provide IASI’s DOFS. It would probably be better to avoid to apply equation 11 assuming that both sensors have similar vertical sensitivity or to use the more (too) complicated methodology detailed in Rodgers and Connor, JGR (2003).
- IASI CO "diurnal cycle" is mostly related to its decreased sensitivity at night. So the agreement with GIIRS for day and night described as good by the authors is just indicating that GIIRS has the same decreased sensitivity at night. There should be some statements about this issue.
Detailed comments :
Section 3 and 4 :
Some generalities about radiative transfer and retrieval methodology and basic known equations could be removed. Equations 3 to 8 have been largely documented such as in Rodgers (2000) and it is unnecessary to repeat this here.
Time : the time is given in UTC but it is not a correct choice to interpret diurnal cycles around China. Beijing local solar time would be much better. Furthermore the time is often given without the precision that it is UTC.
Figure 7 : precise the time system chosen.
Figure 8 : we suppose it is UTC !
Figure 10 : please provide hour in LST because UTC is not adapted to the geographical zone.
Fig 12 : Dayth => Day
Line 59 : The authors mention Kobayashi et al. (1999) as one of the first attempt to document CO from space with the japanese IMG ADEOS. Nevertheless, this paper do not present CO retrievals from IMG. The only retrievals of CO from this first spaceborne IR FTS have been published later by Barret et al. (2005).
refs:
C.D. Rodgers and B.J. Connor, Intercomparison of remote sounding instruments, JGR, 2003.
B. Barret, et al., Global carbon monoxide vertical distributions from spaceborne high-resolution FTIR nadir measurements, ACP, 2005.
Citation: https://doi.org/10.5194/amt-2022-305-RC1 -
AC1: 'Reply on RC1', Zhao-Cheng Zeng, 01 Mar 2023
The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2022-305/amt-2022-305-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Zhao-Cheng Zeng, 01 Mar 2023
-
RC2: 'Comment on amt-2022-305', Anonymous Referee #2, 14 Dec 2022
This study describes the CO retrieval algorithm for the geostationary GIIRS sounders on board FY-4B satellite. The paper is in general well explained and clear, particularly for the algorithm, however some information or clarification are missing.
Comments:
Ln 9: The year of when the sounder was launched should be indicated in the abstract to inform the reader.
Ln 10: how having hyperspectral measurements of CO provide diurnal observation of CO? I would think that having geostationary CO data allow observation of diurnal CO variability. Could you reformulate your sentence or give more precision?
Ln 34. I would maybe say “(GEO) orbit can provide contiguous coverage with similar or higher spatial resolution than LEO and a revisit time of 1-2 hours [...]” or a similar sentence. Because, GIIRS has the same spatial resolution than IASI (12km diameter at nadir).
Ln. 89: Figure A1. b and c, are these figures for the same day as Fig. A1.a?
Fig. 1b. The values on Fig. 1B are a too small.
Fig. 1c. It would be interesting to have, as well, the Jacobian by pressure (the ones used for the radiative transfer model). This would inform on the variability that GIIRS channels sensitivity have depending on atmospheric pressure. Additionally, there is no comment on this figure. What do you conclude with this figure in term of sensitivity for CO and H2O with GIIRS?
Ln. 171. You computed 3-hourly CO profile climatology for each month, which month are you talking about? The period of study has not been introduced yet, except result for July 2022 introduced in the abstract.
Ln. 174. Should be “2080 cm-1 to 2120 cm-1”.
Ln. 194. “The number of pressure grids in the forward RT model should be large enough to reduce the error [...]”. A reference is missing here regarding this remark.
Ln 262: To be consistent, I would write the title as “ Averaging kernel (AK) matrix and Degree of Freedom for Signal (DOFS)”
Ln. 276. How much data are removed after the quality filter? Similarly, how much data are removed before and after labeled clear sky for the period of your study, before the post-screening is done?
Ln. 289. Could you precise why you added a Gaussian white noise? Is the added noise mentioned Ln 291 referring to the Gaussian white noise. If yes, then Ln 291 should appear just after the white noise is mentioned Ln. 289.
Ln. 293. You could introduce a map to visualize the regions of interest.
Ln. 304. How can you conclude that from Figure 2?
Ln. 314.: The results of Figure 3 are only for North China Plain, but have you done it also for the two other regions? Mongolia has a more complex diurnal TC change than North China Plain but the surface pressure/topography is also different between the two regions. Would the results of Figure 3 be the same for Mongolia region or not?
Ln. 315. It is confusing, the “truth” is based on the ECMWF EAC4 results but it is also used as CO a priori profile in your retrieval algorithm, so what is the difference between the comparison of (1) and (2)? How can you compare the retrievals to ECMWF EAC4, if you already used this CO profile as a priori profile in your retrieval algorithm (see Ln. 291)?
Ln. 470-479: Talking about wildfires, during the month of July 2022, several wildfires occurred in Siberia. The transport and mixing of CO in the Northern Hemisphere might have bring CO concentrations from the Siberian fires to the regions of your study. Have you looked at that? The ECMWF EAC4 and ER5 simulation/reanalysis used in your algorithm do not include 2022 and so might not be representative of CO concentrations for the month of July 2022. It is difficult to determine if your retrievals are well representative of July 2022 consequently. It could be then interesting to have evaluation of your L2 retrievals to in situ data.
Appendix A: Figure A1.a is not used in the study. Considering the wildfires in July 2022, was there fire CO emissions included in CAMS model?
This study was only done for a summer month, but have you look at other season? The diurnal cycle might be associated to meteorological conditions and emissions patterns different by season. Additionally, having hourly data and comparing North China with Mongolia and East China Sea, I was wondering if you looked at the difference in CO concentration between these regions during the daytime. I would expect to observe highest concentration for North China than Mongolia during the morning time corresponding to rush hours, however this would depend on synoptic disturbances.
Citation: https://doi.org/10.5194/amt-2022-305-RC2 -
AC2: 'Reply on RC2', Zhao-Cheng Zeng, 01 Mar 2023
The comment was uploaded in the form of a supplement: https://amt.copernicus.org/preprints/amt-2022-305/amt-2022-305-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Zhao-Cheng Zeng, 01 Mar 2023
Zhao-Cheng Zeng et al.
Data sets
Carbon Monoxide (CO) retrievals from GIIRS onboard FY-4B Zhao-Cheng Zeng https://doi.org/10.18170/DVN/M7DKKL
Zhao-Cheng Zeng et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 280 | 153 | 14 | 447 | 9 | 7 |
- HTML: 280
- PDF: 153
- XML: 14
- Total: 447
- BibTeX: 9
- EndNote: 7
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1