The Comparison of Electron Density between CSES In Situ and Ground-Based Observations in China

Round 1

Reviewer 1 Report (New Reviewer)

A brief summary (one short paragraph) outlining the aim of the paper, its main contributions and strengths.

The paper compares CSES electron densities at 500 km with concurrent measurements by ISR and multiple ionosondes. The main result is that the CSES Ne are ~1.8 times smaller that those measured by ISR and ionosondes. Attempts have been made to identify spatial regions and time sectors where agreement is better. Overall, the paper looks as fairly comprehensive assessment of CSES data for the Chinese sector.

General concept comments

Perhaps the result on so much underestimation by CSES is under-stressed in the paper while it is the most important one.  The reader left with an impression that not only satellite malfunction but ground-based instruments are so deficient that perhaps CSES Is not so bad. I think this is not correct emphasis in narration in many instances. There are some uncertainties with ground-based instruments, but most of them are taken care of. Or, the authors cannot use them at all.    

The extension of ionosonde data to 500 km via the IRI model needs explanation and justification. To me, this kind of work is only good enough as a first attempt while the paper is worded as everything is perfect, and we know what ionosondes measure. In reality, the model also contributes and it Is not known to me how much. I wonder about the value of this work at all. I understand, the authors think differently, it is Ok, but then they have to justify with numbers that this is a legitimate procedure.

In several places, the authors seem to imply the low correlation coefficients, below 0.5-0.6, are Ok and indicate a good performance of CSES. I disagree with these type of statements;  they, on the contrary, show that the CSES instrument is hardly possible to use for serious research. I suggest the authors to take a look at such statements and state clearly the value of inferred “correlations”.   

 

Specific comments:

              The paper requires rewriting by a person fluent in English. I think I understood the content, but on a top of elementary grammatical errors in vocabulary and grammar, some statements are misleading and require reader’s interpretation.

              I suggest the authors include several footprints of CSES satellites on the map. It would help understanding where the data were collected. The meaning of the square in bottom left corner is not explained.

Lines 124-132. Justification for comparison a point measurements by ISR (wa sit?) and 20 degrees averaging for the satellites needs justification or at least explanations. What is the time integration for radar data? Details of radar operation mode, spatial and temporal resolution must be incorporated into the text.

Figure 2. Characters are too small to be properly visible. Need enlargement by a factor of~2

Line 149: “moderate correlation” contradict earlier statement in line 139 that variations are “almost similar”. Correlation <0.5 means no similarity at all.

Assessing the differences in % is a bit off what others in the field are presenting. Ratio of CSES/ISR would give direct and clear number, consistent with other publications. Please redo to be on par with others.

Figure 4. Fits look terribly bad. I recommend to remove this Figure, it brings no useful information and only adds negativity.

Section 4.1.  The spatial coincidence criteria and time coincidence criteria need to be described.

Figs. 5,6. Labels for Y axes are strange, it is ionosonde-based density projected with IRI, whatever you get.

Figs 5,6 and 7 are too many diagrams for the value of the outcome.  Having temporal variations for each station does not bring much information. Scatter plots and histograms are good enough to me.  Perhaps, present a case for one ionosonde and place the data into a Table for all.

Figs 8 and 9. I suggest to present one of the diagrams, and place the data inferred for others into Table for all.

Value of Figs. 10 and 11 is highly questionable. The reader does not know the coverage, how matching was done for individual stations. I suggest to remove the plots.  

Conclusion 3 is not clear in terms of its meaning

Conclusion 4. The last statement has  to be removed. It works without doing anything, it is applicable to every work. Span 50-80% is too big to judge on the quality of agreement.

Author Response

Please see the attachment.

Author Response File: Author Response.docx

Reviewer 2 Report (New Reviewer)

Comments for author File: Comments.pdf

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 3 Report (New Reviewer)

This paper is a validation study. The author employs CESE satellite Ne profile and compare with local ISR and ionosonde. The comparison results exhibit high and mid correlations between the satellite and local ground-based observations. After all, this study make up the blank part of the data validation of this satellite. I recommend this paper to have a minor revision before it is suitable for publication.

 

Line 15 magnitude, not amplitude

 

Line 34-35 this sentence is useless, please delete it 

After the satellite launch, not only the mission team but also many scientists have paid more attention to data quality of CSES observations.

Line 51 housekeeping?? What do you mean

In the introduction, the author shall also cite other papers about recent validation of satellite missions such as COSMIC, GOLD and ICON

Harding, B.J., Chau, J.L., He, M., Englert, C.R., Harlander, J.M., Marr, K.D., Makela, J.J., Clahsen, M., Li, G., Ratnam, M.V. and Bhaskar Rao, S.V., 2021. Validation of ICON‐MIGHTI thermospheric wind observations: 2. Green‐line comparisons to specular meteor radars. Journal of Geophysical Research: Space Physics, p.e2020JA028947. Public accesss. DOI: 10.1029/2020JA028947

Cai, X., Burns, A. G., Wang, W., Coster, A., Qian, L., & Liu, J., et al. (2020). Comparison of GOLD nighttime measurements with total electron content: Preliminary results. Journal of Geophysical Research: Space Physics, 125, e2019JA027767. https://doi.org/10.1029/2019JA027767

 

Pedatella, N. M., Zakharenkova, I., Braun, J. J., Cherniak, I., Hunt, D., Schreiner, W. S., et al. (2021). Processing and validation of FORMOSAT-7/COSMIC-2 GPS total electron content observations. Radio Science, 56, e2021RS007267. https://doi.org/10.1029/2021RS007267

 

Line 73-74 for science objectives, it means what the observation data can be utilized to do XX study. why the author present the mode of the satellite?? Since science objectives shall be the related science aspects.

 

Line 98 1.5 and 3 seconds

Line 111  25.6N, 103.7E??

 

Line 130-131 in the region between 93.8E-113.8E, 15.6N-35.6N

 

Line 183 were output

 

Line 269 relatively lower

 

Line 317-326 the author shall not only just overemphasize the effect of sun. The author shall mention that the electron density above F2 region is mainly determined by chemical production and loss, transport due to neutral wind, E cross B drift and ambipolar diffusion.

 

Line 360 separate the sentence here. In the nighttime, the value at the middle and mid-low lati-360 tude is higher than that at the low latitude.

 

Line 371 what do you mean by ionospheric phenomena

 

In the discussion or results part, the author shall also compare their results with previous studies. For example, there are also previous studies using low-earth orbiting satellite to compare with local ionosonde, and they also calculate the correlation coefficient. The author shall check these results and compare with theirs

 

 

Why the author provides the monthly mean of Ne over China? What are these used for? It appears that they cannot be compared with the local ground observations

Author Response

Please see the attachment.

Author Response File: Author Response.pdf

Reviewer 4 Report (Previous Reviewer 1)

The suggested revisions have been satisfactorily applied. Therefore, I consider the paper to be ready for the publication.

Author Response

We sincerely thank a lot for reviewer’s suggestion.

Round 2

Reviewer 1 Report (New Reviewer)

Review of Liu et al.

 

The authors attempted to improve the manuscript. I do not think it is more sound now. I am Ok with the paper being published although my firm opinion is that the paper is of low quality. Here is why I think so:

11.       Comparison with ISR is very strange. You have data with 5-km resolution, the satellite height is well known and I do not see the reason for averaging over height. Allowing 2 hours difference in time is too much to have the meaningful comparison. Perhaps this is the reason for large differences reported. I suspect, there were not too many good conjunctions. I see no problem with this as the comparison with ISR can be removed from the paper.

S   Splitting data on 2 groups is artificial and not justified. By this logic, one can select further groups by 3 points for which correlation coefficient would be even better. This splitting needs justification, or these data should not be presented/discussed at all.

 

 

22.    Regarding comparison with ionosondes. I would say that what has been done is that the IRI model output was compared with the satellite data while the model was updated for NmF2 from ionosondes. It is not a comparison with ionosondes per se, as the authors claim. The writing gives a wrong perception.

33.   Conclusion: “The trend of CSES Ne is mostly similar to ionosonde observations no matter the data in continuous time or in a certain month” is strange. Trend in what? If the instantaneous conjunctions show good agreement, then monthly averages will be automatically in agreement. Why “monthly” was added?

Reviewer 2 Report (New Reviewer)

The authors wasted time and effort on an absolutely wrong task. And now they're wasting my time too.

Comments for author File: Comments.doc

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

-       Brief summary:

The paper aims to validate the CSES Ne measurements of LAP by comparing these values with ground measurements obtained by ISR and ionosonde. The accuracy improvement is clearly identified as the main target of the work that, in general, is well written and well organized. Nevertheless, some points need to be better described together with the relevant figures and the Authors should try to explain the reasons of the identified discrepancies.

-       General comments:

1.     The comparison between Ne_CSES and Ne_QJISR is reported starting from Fig. 2 where the two time-series are displayed with different amplitude scales to better underline the simultaneous variations. The choice to use different scales should be remarked in the caption. In the Fig. 3 the “Empirical cdf” should be defined in the caption. In Fig. 4 the points in the scatter plot are supposed to be well fitted by an empirical formula but most of the points are very far from the red line. One could think the Chi-squared test could clarify such doubt. Actually, the points seem to be better represented by linear correlation once these are divided into two different population (see figure in attached pdf file).

If the dataset includes both day and night measurements, it could be interesting to separate them and check if the scatter plot could be analyzed separately. Otherwise, the two populations could be addressed to different geomagnetic activity level. Authors should try to explain in some way such twofold appearance.

2.     In the comparative analysis, Authors decided to use a 10° squared geographical superposition between ground antennas and S/C position. Is that value an output of any computation or is it taken from literature? Please explain why 10° should be the best value for the comparison.

3.     The comparative analysis has been separated in different sub-sections. The reason for separating ISR from Ionosonde dataset in quite clear, even if not clearly expressed, while is not clear for me why the ionosonde data set has been displayed with different markers in Fig.1 and analysed separately (Leshan, Tengchong and Puer first, then Lhasa and finally the other CRIRP stations). In particular, is not clear why Leshan, Tengchong and Puer are not used for the seasonal analysis reported in Fig. 9 and tables 1 and 2.

4.     Although the data and method are well described, the paper does not face the reasons of the difference between ground and in-situ measures. In particular, even if the ground-based measurements are model dependent, these are considered absolutely reliable, thus addressing the observed discrepancies only to LAP values. Particularly debatable is the sentence at line 292-295. In fact, some kind of Lp operates as relative probes (e.g. Swarm Lp) since these do not apply a complete voltage ramp to obtain the characteristic curve. On the contrary, CSES LAP operates as an absolute instrument able to identify the plasma potential thus retrieving the actual electron density with a very small uncertainty (mainly due to voltage ramp step amplitude). In order to better convince the reader, authors should try to investigate also the ground based measurements uncertainties and compare these with the LAP ones. Such analysis would provide a significant improvement of the paper and on the CSES data validation process.

Due to the above comments, the manuscript should be reconsidered after major revision mainly aiming to better identify the reasons of the differences between in situ and ground based measurements.

-       Specific comments:

Line 15; “according” sounds better for agreement meaning. It is better to start the sentence with something like “with reference to”.

Line 17; “obviously” means that is a well-known result. In such case a reference is needed. It is better to postpone such a comment in the discussion.

Line 33; add “team” after mission

Line 47; variation or variability are commonly used to express that a parameter modulation. It would be better to replace “change” with variation in the text.

Line 54; “Simultaneous” would be better than “Synchronous”. Rephrase the sentence with something like “Simultaneous variations with some events ….can also provide data validation improvement”.

Line 65; “as reported” in Yang et al.

Line 81; replace “including” with “called”

Line 87; rotate the sentence as follow, “LAP payload is equipped with two sensors”

Line 112; Please specify how many data points per day are collected

Line 114; what does it mean “working” in this context?

Line 120; “histograms”

Line 130; Ne is always positive so it is better to replace “absolute” with ”average”

Line 159; please control the relevant positions in Figs. 5 and 6

Figure 5, 6, 8, 9 labels and text are too small. Please use font and dimension of Fig. 3 (also for Figs. 10 to 13).

Line 258-259; subject is lost

Line 259-261; character dimension is larger than elsewhere in the text.

Comments for author File: Comments.pdf

Author Response

The paper aims to validate the CSES Ne measurements of LAP by comparing these values with ground measurements obtained by ISR and ionosonde. The accuracy improvement is clearly identified as the main target of the work that, in general, is well written and well organized. Nevertheless, some points need to be better described together with the relevant figures and the Authors should try to explain the reasons of the identified discrepancies.

-       General comments:

1. The comparison between Ne_CSES and Ne_QJISR is reported starting from Fig. 2 where the two time-series are displayed with different amplitude scales to better underline the simultaneous variations. The choice to use different scales should be remarked in the caption. In the Fig. 3 the “Empirical cdf” should be defined in the caption. In Fig. 4 the points in the scatter plot are supposed to be well fitted by an empirical formula but most of the points are very far from the red line. One could think the Chi-squared test could clarify such doubt. Actually, the points seem to be better represented by linear correlation once these are divided into two different population (see figure in attached pdf file).

A: As suggested, the illustrations for the choice of using different scales were added in lines 152-153 and the captions of Figures 2, 5, and 8 in the “Track Changes” version.

We tried to fit data with different orders and then accepted the empirical formula with third order. In the revised version, we gave out the detailed pictures (Figure 4 with new version) and illustrations (lines 180-186) why we made the decision. We agree with the reviewer’s opinion that the points seem to be divided into two different population, and we added the picture (Figure 12) and some discussions in lines 344-358.

2. If the dataset includes both day and night measurements, it could be interesting to separate them and check if the scatter plot could be analyzed separately. Otherwise, the two populations could be addressed to different geomagnetic activity level. Authors should try to explain in some way such twofold appearance.

A: The data of QJISR are just in the daytime corresponding to the CSES local time, which has been illustrated in lines 140-141.

As suggested, we divided the data into two groups and fitted the data separately. which was added in Figure 12 and text in lines 344-358.

2. In the comparative analysis, Authors decided to use a 10° squared geographical superposition between ground antennas and S/C position. Is that value an output of any computation or is it taken from literature? Please explain why 10° should be the best value for the comparison.

A: With the burst mode of CSES, the distance at latitude for each sample is almost 0.1°, so there are about 200 data points per orbit to obtain the M and std. The reasons of selecting ±10° as the range are to insure enough samples to obtain the M and std, and avoid the ionospheric variation in large scale. The illustrations were added in lines 147-149 and 202-204.

3. The comparative analysis has been separated in different sub-sections. The reason for separating ISR from Ionosonde dataset in quite clear, even if not clearly expressed, while is not clear for me why the ionosonde data set has been displayed with different markers in Fig.1 and analysed separately (Leshan, Tengchong and Puer first, then Lhasa and finally the other CRIRP stations). In particular, is not clear why Leshan, Tengchong and Puer are not used for the seasonal analysis reported in Fig. 9 and tables 1 and 2.

A: Sorry for the confusion. As illustrated in the text, Leshan, Tengchong and Puer stations belong to IEF, and other stations marked by red triangles in Figure 1 belong to CRIRP. The data of IEF are continuous in time since the station was constructed, while the locations are just in the southwest China. Science 1940s, CRIRP have established several ionosondes from lower (20.0°N) to middle latitudes (49.6°N) in the mainland of China. Through the cooperation, we obtained the data of stations at different latitudes in March, June, September and December in 2019. Since the time range is different for the two data, we separated the analyses. The locations of Leshan, Tengchong and Puer are nearly to Chongqing and Kunming of CRIRP, and the data process is not same for two types data, so the data of Leshan, Tengchong and Puer were not used in the seasonal analysis reported in Fig. 9 and tables 1 and 2. The illustrations were added in lines 124-131, 193-198, and two sub-titles were added in section 4.

4. Although the data and method are well described, the paper does not face the reasons of the difference between ground and in-situ measures. In particular, even if the ground-based measurements are model dependent, these are considered absolutely reliable, thus addressing the observed discrepancies only to LAP values. Particularly debatable is the sentence at line 292-295. In fact, some kind of Lp operates as relative probes (e.g. Swarm Lp) since these do not apply a complete voltage ramp to obtain the characteristic curve. On the contrary, CSES LAP operates as an absolute instrument able to identify the plasma potential thus retrieving the actual electron density with a very small uncertainty (mainly due to voltage ramp step amplitude). In order to better convince the reader, authors should try to investigate also the ground based measurements uncertainties and compare these with the LAP ones. Such analysis would provide a significant improvement of the paper and on the CSES data validation process.

A: As suggested, some discussions and illustrations were added in lines 380-397, 424-450 of “Discussions” section.

Due to the above comments, the manuscript should be reconsidered after major revision mainly aiming to better identify the reasons of the differences between in situ and ground based measurements.

-       Specific comments:

Line 15; “according” sounds better for agreement meaning. It is better to start the sentence with something like “with reference to”.

A: Changed as suggested in lines 16-17.

Line 17; “obviously” means that is a well-known result. In such case a reference is needed. It is better to postpone such a comment in the discussion.

A: “obviously” is changed to “relatively” in line 19. The discussion of this phenomenon is in lines 362-379.

Line 33; add “team” after mission

A: Changed as suggested in line 36.

Line 47; variation or variability are commonly used to express that a parameter modulation. It would be better to replace “change” with variation in the text.

A: Changed as suggested.

Line 54; “Simultaneous” would be better than “Synchronous”. Rephrase the sentence with something like “Simultaneous variations with some events ….can also provide data validation improvement”.

A: Changed as suggested in lines 59-60.

Line 65; “as reported” in Yang et al.

A: Changed as suggested in line 71.

Line 81; replace “including” with “called”

A: Changed as suggested in line 77.

Line 87; rotate the sentence as follow, “LAP payload is equipped with two sensors”

A: Changed as suggested in line 100.

Line 112; Please specify how many data points per day are collected

A: With the burst mode of CSES, the resolution of latitude is almost 0.1°, so there are about 200 data points per orbit to obtain the M and std. The illustrations were added in lines 147-149.

Line 114; what does it mean “working” in this context?

A: The “working” means QJISR has data in these days. Some modifications were made in line 151.   

Line 120; “histograms”

A: Changed as suggested in line 158.

Line 130; Ne is always positive so it is better to replace “absolute” with ”average”

A: Changed as suggested in line 173.

Line 159; please control the relevant positions in Figs. 5 and 6

A: Changed as suggested.

Figure 5, 6, 8, 9 labels and text are too small. Please use font and dimension of Fig. 3 (also for Figs. 10 to 13).

A: As suggested, we enlarged the font of labels and texts of these figures, while restricted by space, the dimension of some figures is not same as Figure 3.

Line 258-259; subject is lost

A: The subject was added, and the sentence is like this “In the ionosphere, the processes of production, loss and transportation are mainly affected by the sun.”

Line 259-261; character dimension is larger than elsewhere in the text.

A: Sorry for this mistake, and we modified it in the same dimension.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments on the paper entitled “ The comparison of electron density between CSES in-situ and ground-based observations in China” submitted by Jing Liu, Tong Xu, Zonghua Ding and Xuemin Zhang to Remote Sensing Journal (2022)

 

General comments:

 

This paper presents validation results of in situ Electron density (Ne) measurements made using Langmuir probe at ~500 km altitude using China Seismo-Electromagnetic Satellite (CSES) with that of electron density measured by Qujing Incoherent scatter radar and electron density extrapolated to 500 km using combined IRI and ionosonde foF2 weighted density over China for the years 2018-2021. It is mentioned that the CSES was launched with 8 scientific payloads to measure manyionospheric parameters. Since validation of the electron density or any other parameter for any satellite is very important, it is important that thesecomparativestudies are done meticulously and extensively evaluated. While already several validation results have been made in the past using CSES, still the results showed in this paper are importantfor validation purpose and useful for the community.

 

The results suggest that the trend of Ne measured using CSES is consistent with that of Ne measured using ISRwith the correlation coefficient of 0.4638 based on 33 events. The absolute value of Ne from CSES is lower than that of Ne from ISR. There are nearly 10 ionosondes which are operated have been utilized in this comparison. For correlation, ionosondes located at Leshan, Tengchong and Puer have been used. The satellite observations during the months of March, June, September and December 2019 have been used to correlate with the ionosonde observations. In terms of latitude, the correlation between CSES and ionosonde is better at the low latitude than that at the middle latitude during the day and night. However, the value at the middle and mid-low latitude is higher than that at the low latitude. In terms of time, the correlation coefficient in June and September is generally better than that in March and December.The distribution of Ne of CSES and correlation coefficients at different latitudes show that Ne of CSES with the relative high value exhibits a good consistency to the ionosonde observation. Systematic biases have been identified between Ne of CSES and ionosonde data in the daytime and nighttime, respectively.

The following explanation is offered for the day and night differences in correlations: Since solar activity plays a significant role for the changes in Ne detected by CSES and NmF2 observed by ionosonde in the daytime, Ne in the daytime may have a connection with the strong solar effect at the dayside. However, since chemical, dynamic, and electrodynamic processes are the important factors in the nighttime, which may have characteristics of latitude, longitude and altitude.

 

It is suggested that the density measured using CSES satellite is relatively low as compared with the ISR data and ionosonde. It is mentioned that systematic biases of Ne between CSES and ground-based observations could be due to differences in data processing techniques. It is suggested that such biases are also found in the comparison of electron density between CSES and Swarm, DEMETER and IRI model. However, I believe that the validation is not done properly and it needs to be done systematically. I suggest that authors to do thorough revision and address the following before the manuscript can be considered for publication.

 

The following are main comments:

 

 

 

 

1.      The general trend of Ne of CSES seems to be agreeing with that of ISR and Ionosonde on large temporal scales, while on shorter scales they differ. Also there is no meaningful explanation for this has been offered. This needs to be examined.

2.      Why Ne measured from CSES is lower than the ISR density?

3.      Here, authors have only correlated the Ne with that of the ground based ISR and ionosondes over China. However, why authors didn’t do outside China? If any specific reason is there, it should be mentioned. I believe it could be due to burst mode of data collection over China? But this point should be clearly mentioned.

4.      All the figures, the Y-axis (right) and Y-axis (left) scales are not same. Without maintaining the scales same, it is impossible to compare them. So, I suggest them to revise the figures to maintain uniform scales for all the figures.

5.      No separation of Quiet and Disturbed day activity.

6.      Why daytime and nighttime correlations can differ significantly. Since same instrument is used, there shouldn’t be any difference in day and night.

7.      I am not able to understand how and why the density has been extrapolated to satellite altitude from NmF2. If any reference is added for this, it will be nice. It is possible that topside density profile can be obtained using Chapman function. Why authors have done weighting factor fitting using ionosonde and IRI2016. Since it is fitted with IRI model, the data is more weighted towards model than ionosonde observations. If only IRI 2016 is used, how the correlation looks like?

8.      Figure-4 shows curve fitting between Ne of CSES and ISR. But the fitting line is not linear. Why it is fitted with non-linear curve. But in contrast, the figures-5 &6 shows the fitting done with linear curve. Please explain usage of different curve fittings for the same data for two different independent instruments. 

9.      The correlation coefficient is very low. It is ~0.5 which is low.

10.  How the correlation coefficient is calculated for the Figure-10&11. The time interval and how they are calculated is not described. This should be explained.

11.  I find many important aspects are not discussed completely. However, they are important for the reader to understand the criticality of the paper. 

12.  While the manuscript is readable and understandable, but it needs further improvement in English usage. This is general comment.  

 

Comments for author File: Comments.pdf

Author Response

The following are main comments:

1. The general trend of Ne of CSES seems to be agreeing with that of ISR and Ionosonde on large temporal scales, while on shorter scales they differ. Also there is no meaningful explanation for this has been offered. This needs to be examined.

A: The day-to-day variation, bias caused by ionogram interpretation, geomagnetic storm and ionospheric model can bring the differences. Some discussions were added in lines 380-397, 424-450 of “Discussions” section in the “Track Changes” version.

2. Why Ne measured from CSES is lower than the ISR density?

A: About the differences of absolute values, we added the discussion in lines 424-450.

3. Here, authors have only correlated the Ne with that of the ground based ISR and ionosondes over China. However, why authors didn’t do outside China? If any specific reason is there, it should be mentioned. I believe it could be due to burst mode of data collection over China? But this point should be clearly mentioned.

A: Yes, the reason is the burst mode. As suggested, some illustrations were added in lines 77-83.

4. All the figures, the Y-axis (right) and Y-axis (left) scales are not same. Without maintaining the scales same, it is impossible to compare them. So, I suggest them to revise the figures to maintain uniform scales for all the figures.

A: Different amplitude scales can better underline their own variations for different data. If we use the same scale, the data with low amplitude will represent like a line and cannot exhibit the variability, seeing the picture in PDF. As suggested, same scale was used in Figures 10 and 11.

5. No separation of Quiet and Disturbed day activity.

A: Since the space disturbed events can affect the variation of electron density at each altitude, we did not separate the quiet and disturbed day in the comparison between CSES Ne and ground-based observations. As suggested, the data in geomagnetically quiet day were re-analysed, and the illustrations and discussions were added in lines 391-397 and Table 3. For the limitation of sample numbers, the separation analysis of QJISR and ionosonde at different latitudes was not included.

6. Why daytime and nighttime correlations can differ significantly. Since same instrument is used, there shouldn’t be any difference in day and night.

A: This question is also puzzled us, and we have tried to discuss it in lines 362-379.

7. I am not able to understand how and why the density has been extrapolated to satellite altitude from NmF2. If any reference is added for this, it will be nice. It is possible that topside density profile can be obtained using Chapman function. Why authors have done weighting factor fitting using ionosonde and IRI2016. Since it is fitted with IRI model, the data is more weighted towards model than ionosonde observations. If only IRI 2016 is used, how the correlation looks like?

A: We completely agree with the reviewer that topside density profile can be obtained using Chapman function, while the selection of ionospheric scale height can also affect the fitting result (Lei et al., 2005; Liu et al., 2007a). At the beginning, we planned to obtain the Ne at orbit altitude using the formula (seeing in the PDF) like Liu et al. (2004), where(h) is the density at orbit height, (400km) can be the NmF2, and H is the corresponding scale height calculated from MSIS90 which is now updated to NRLMSIS 2.0 (Emmert et al., 2021). Unfortunately, due to the limited permissions, we cannot download the NRLMSIS 2.0 from website in China. As the electron density is assumed to be mostly linear dependence on altitude above the hmF2 [Lei et al., 2006; Liu et al, 2007b], we finally tried to use the ratio outputted by IRI model to obtain the Ne at satellite orbit.

There are 3 steps to obtain the comparison dataset for ionosonde data, which has been illustrated in the text (lines 204-210). Firstly, the ratio of Ne at 507km and NmF2 at the ionosonde location was output with a 15-minute time resolution based on IRI 2016 model. Secondly, the NmF2 observed by ionosonde was multiplied by the above ratio to obtain Ne data at the satellite altitude. At last, the mean value and standard deviation of above Ne during 01:00-03:00 (LT) and 13:00-15:00 (LT) were acquired separately to represent the data in the nighttime and daytime. In this process, we just used the ratio between NmF2 and Ne at 507km, and the data of NmF2 are the real observations, so we cannot consider the Ne data is more weighted towards model than ionosonde observations. We have compared the Ne data between IRI model and CSES in previous work (Liu et al., 2021). If just using the IRI model, the data will be smooth and there is no much perturbances like real observations. The correlation coefficient between CSES and IRI Ne is almost same as or lower than that of this study in the region of China, seeing the picture in PDF. The correlation between the two data is also better at the low latitude than that at the middle latitude in the daytime (right panel).

Emmert, J. T.;  Drob, D. P.;  Picone, J. M.;  Siskind, D. E.;  Jones, M.;  Mlynczak, M. G.;  Bernath, P. F.;  Chu, X.;  Doornbos, E.;  Funke, B.;  Goncharenko, L. P.;  Hervig, M. E.;  Schwartz, M. J.;  Sheese, P. E.;  Vargas, F.;  Williams, B. P.; Yuan, T. NRLMSIS 2.0: A Whole‐Atmosphere Empirical Model of Temperature and Neutral Species Densities. Earth and Space Science 2021, 8 (3).

Lei, J.; Liu, L.; Wan, W.; Zhang, S.-R. Variations of electron density based on long-term incoherent scatter radar and ionosonde measurements over Millstone Hill. Radio Science 2005, 40 (2), RS2008.

Lei, J.; Liu, L.; Wan, W.; Zhang, S.-R.; Van Eyken, A. P. Comparison of the first long-duration IS experiment measurements over Millstone Hill and EISCAT Svalbard radar with IRI2001. Advances in Space Research 2006, 37 (5), 1102-1107.

Liu, H.; Lühr, H.; Henize, V.; Köhler, W. Global distribution of the thermospheric total mass density derived from CHAMP. Journal of Geophysical Research: Space Physics 2005, 110 (A4).

Liu, L.; Wan, W.; Zhang, M.-L.; Ning, B.; Zhang, S.-R.; Holt, J. M. Variations of topside ionospheric scale heights over Millstone Hill during the 30-day incoherent scatter radar experiment. Annales Geophysicae 2007a, 25 (9), 2019-2027.

Liu, L.; Le, H.; Wan, W.; Sulzer, M. P.; Lei, J.; Zhang, M.-L. An analysis of the scale heights in the lower topside ionosphere based on the Arecibo incoherent scatter radar measurements. Journal of Geophysical Research: Space Physics 2007b, 112 (A6), A06307.

Liu, J.; Guan, Y.; Zhang, X.; Shen, X. The data comparison of electron density between CSES and DEMETER satellite, Swarm constellation and IRI model. Earth and Space Science 2021, 8, e2020EA001475.

8. Figure-4 shows curve fitting between Ne of CSES and ISR. But the fitting line is not linear. Why it is fitted with non-linear curve. But in contrast, the figures-5 &6 shows the fitting done with linear curve. Please explain usage of different curve fittings for the same data for two different independent instruments. 

A: The fitting curves with different orders were tried and the third order was finally selected. The explanations were added in lines 180-186 and Figure 4 (new version). Furthermore, we found that the linear fitting was better when the data of Figure 2 were divided into two groups, seeing the discussion in lines 344-358 and Figure 12. The linear fitting in Figures 5 and 6 was used to check the correlation between the two data.

9. The correlation coefficient is very low. It is ~0.5 which is low.

A: We tried to discuss the reason for the relative low values. For the comparison between CSES and QJISR Ne, if the data were divided into two groups seemly according to the amplitude, the correlation coefficient can reach to 0.9250 and 0.8884. For the ionosonde data, if the day-to-day variation was ignored, the correlation coefficient can increase. The disturbed events can also affect the correlation between CSES Ne and ionosonde observations. The detailed illustrations were added in lines 344-358, 380-397, Figure 12 and Table 3.

Furthermore, we did not consider ~0.5 is very low. To our knowledges, the correlation is defined like the following principle based on the correlation coefficient.

0.8-1.0, Very strong correlation;

0.6-0.8, Strong correlation;

0.4-0.6, Moderate correlation;

0.2-0.4, Week correlation;

0.0-0.2, Very weak or no correlation.

10. How the correlation coefficient is calculated for the Figure-10&11. The time interval and how they are calculated is not described. This should be explained.

A: Sorry for this confusion. The correlation coefficient was calculated according to Equation 2 for each station in March, June, September and December of 2019, just like the example of Lhasa station. Figures 10 and 11 are contour maps based on the data in Table 1. As suggested, illustrations were added in lines 299-304.

11. I find many important aspects are not discussed completely. However, they are important for the reader to understand the criticality of the paper. 

A：As suggested, more discussions were added in the revised version, including the discussion on the correlation coefficient and absolute value, seeing the added illustrations in “Discussion” section.

12. While the manuscript is readable and understandable, but it needs further improvement in English usage. This is general comment.  

A: As suggested, we carefully made the improvement in English writing.

Author Response File: Author Response.pdf

Reviewer 3 Report

Nice article and very good results in view of data comparison 

Author Response

Nice article and very good results in view of data comparison.

A: Thank you for your comments, and we improved the English writing in revised version.

Author Response File: Author Response.pdf

Reviewer 4 Report

The manuscript compares Ne measurements from Langmuir Probe (LAP) onboard China’s  Seismo-Electromagnetic Satellite (CSES) with an Incoherent Scatter Radar (ISR) measurements, as well as three Ionosonde measurements in the mainland of China. The agreements between LAP measurements and the other two instruments are very poor and it seems that the results are impacted by systematic basis. The authors conclude that although systematic bias exists, LAP Ne data can be used for ionospheric characteristics and also for other scientific research. 

 

Major comments:

 

1) In the case of ISR, only 33 days of data are available over the period of 3 years. The ISR and CSES LAP Ne measurements differ by an order of magnitude as shown in Figure 2. 

 

2) Authors then tried to establish an empirical formula to establish an agreement between the two data sets, and the agreement is very poor as shown in Figure 4.

 

3) In the case of ionosonde comparison, NmF2 is obtained from ionosonde, and then Ne values at satellite orbital altitude are estimated with the help of IRI2016. The agreement seems very poor again. In most cases, the agreements differ by an order of magnitude. 

 

4) There is no description of how the Ne measurements are made other than a couple of sentences about measurement resolution. Also, no information is given as to whether any quality control measures were taken or not, especially after noticing significant level of differences with other instruments.

 

5) Figures are not made with proper scales (figure 12 is an example of this). It is also very hard to read the scales in figures. 

 

6) Considering the fact that a serious disagreement among these three instruments, the manuscript would have been better if the authors simply compare the CSES LAP Ne measurements directly with IRI2016 however; the first author already published a comparison work for the LAP Ne measurements with IRI-2016 but the results did not agree. 

 

7) While the LAP Ne measurements already showed a systematic bias with IRI-2016 (in the previous work), how would the authors expect a better result in this work when they compare LAP Ne against the combination of ionosonde and IRI-2016 without making any attempt to investigate the root cause for the bias? Isn’t it very obvious that the authors will get the same results again as it is evident in this manuscript? 

 

8) Instead of publishing a very similar but not agreeing results again and again, the authors should seriously consider in figuring out the root cause for the disagreements with other instruments so that this data can be meaningfully used in research. 

Author Response

The manuscript compares Ne measurements from Langmuir Probe (LAP) onboard China’s  Seismo-Electromagnetic Satellite (CSES) with an Incoherent Scatter Radar (ISR) measurements, as well as three Ionosonde measurements in the mainland of China. The agreements between LAP measurements and the other two instruments are very poor and it seems that the results are impacted by systematic basis. The authors conclude that although systematic bias exists, LAP Ne data can be used for ionospheric characteristics and also for other scientific research. 

 

Major comments:

 

1. In the case of ISR, only 33 days of data are available over the period of 3 years. The ISR and CSES LAP Ne measurements differ by an order of magnitude as shown in Figure 2. 

A: We know the samples are not so many, while it can also exhibit the relationship between CSES and QJISR Ne. If the data were divided into two groups seemly based on the amplitude, the correlation coefficient can reach to 0.9250 and 0.8884, seeing the discussion in lines 344-358 and Figure 12 of “Track Changes” version.

2. Authors then tried to establish an empirical formula to establish an agreement between the two data sets, and the agreement is very poor as shown in Figure 4.

 A: The fitting curves with different orders were tried and the third order was finally selected. The explanations were added in lines180-186 and Figure 4 (new version). Furthermore, we found that the linear fitting was better when the data of Figure 2 were divided into two groups, seeing the discussion in lines 344-358 and Figure 12. 

3. In the case of ionosonde comparison, NmF2 is obtained from ionosonde, and then Ne values at satellite orbital altitude are estimated with the help of IRI2016. The agreement seems very poor again. In most cases, the agreements differ by an order of magnitude. 

A: The differences between CSES Ne and ionosonde observations can be caused by some reasons, such as day-to-day variation, the disturbance from geomagnetic storm, bias from the ionogram interpretation, and ionospheric models, which was extensively discussed in the revised version, such as illustrations in lines 380-397, 424-450, and Table 3. If the day-to-day variation and geomagnetic disturbance were ignored, the correlation coefficient can increase. The agreement between the two data is not so good in the nighttime, while we consider the agreement in the daytime is ok, and some correlation coefficients can exceed 0.7.

Furthermore, we did not consider some correlation coefficients are very low. To our knowledges, the correlation is defined like the following principle based on the correlation coefficient.

0.8-1.0, Very strong correlation;

0.6-0.8, Strong correlation;

0.4-0.6, Moderate correlation;

0.2-0.4, Week correlation;

0.0-0.2, Very weak or no correlation.

4. There is no description of how the Ne measurements are made other than a couple of sentences about measurement resolution. Also, no information is given as to whether any quality control measures were taken or not, especially after noticing significant level of differences with other instruments.

A: The description of the measurement resolution is used to illustrate the number of samples that are applied to calculate the mean value and standard deviation, which was added in lines 147-149. More illustrations about LAP measurement were also added in lines 100-106. The CSES mission do not stamp the quality flag to LAP Ne. In our analysis, Ne samples with absolute value exceeding the measurement range were omitted, which may represent some mistakes in the fitting process based on the raw data, which was added in lines 113-116.

5. Figures are not made with proper scales (figure 12 is an example of this). It is also very hard to read the scales in figures. 

A: We want to see the amplitude of Ne in each month, so the same scale was used in Figures 12 and 13. Here we pasted the same picture with different scale in Reply Figures 1 and 2 (showing in the PDF). Could the reviewer give us some suggestions which figure will be better. Furthermore, the change of Ne at middle latitude is small, so although we changed the scale, the variation of pattern is faint. As suggested, we enlarged some texts for figures.   

6. Considering the fact that a serious disagreement among these three instruments, the manuscript would have been better if the authors simply compare the CSES LAP Ne measurements directly with IRI2016 however; the first author already published a comparison work for the LAP Ne measurements with IRI-2016 but the results did not agree. 

 A: Yes, we have compared the LAP Ne of CSES directly with IRI2016 model in an already published paper. We consider that there is no conflict between this work and previous paper. The value of correlation coefficient between CSES and IRI Ne is almost same as or lower than that of this study in the region of China, the figure shown in PDF. The correlation between the two data is also better at the low latitude than that at the middle latitude in the daytime (right panel shown in PDF).

7. While the LAP Ne measurements already showed a systematic bias with IRI-2016 (in the previous work), how would the authors expect a better result in this work when they compare LAP Ne against the combination of ionosonde and IRI-2016 without making any attempt to investigate the root cause for the bias? Isn’t it very obvious that the authors will get the same results again as it is evident in this manuscript? 

A: To compare with CSES Ne, the ionosonde data are needed to infer to the Ne at satellite altitude. The topside density profile can be obtained by fitting data to a Chapman-α layer, while the selection of ionospheric scale height (H) can also affect the fitting results [Lei et al., 2005; Liu et al., 2007a]. As electron density is assumed to be mostly linear dependence on altitude above the hmF2 [Lei et al., 2006; Liu et al, 2007b], the ratio outputted by IRI model can be applied to obtain the Ne at satellite orbit. In this process, we just used the ratio between NmF2 and Ne at 507km altitude, and the data of NmF2 are the real observations, so we consider that the inferred Ne can fully exhibit the characteristics of the NmF2.

Just like the above illustrations, some discussions were added in the “Discussions” section in the revised version to discuss the differences between the two data.

Lei, J.; Liu, L.; Wan, W.; Zhang, S.-R. Variations of electron density based on long-term incoherent scatter radar and ionosonde measurements over Millstone Hill. Radio Science 2005, 40 (2), RS2008.

Lei, J.; Liu, L.; Wan, W.; Zhang, S.-R.; Van Eyken, A. P. Comparison of the first long-duration IS experiment measurements over Millstone Hill and EISCAT Svalbard radar with IRI2001. Advances in Space Research 2006, 37 (5), 1102-1107.

Liu, L.; Wan, W.; Zhang, M.-L.; Ning, B.; Zhang, S.-R.; Holt, J. M. Variations of topside ionospheric scale heights over Millstone Hill during the 30-day incoherent scatter radar experiment. Annales Geophysicae 2007a, 25 (9), 2019-2027.

Liu, L.; Le, H.; Wan, W.; Sulzer, M. P.; Lei, J.; Zhang, M.-L. An analysis of the scale heights in the lower topside ionosphere based on the Arecibo incoherent scatter radar measurements. Journal of Geophysical Research: Space Physics 2007b, 112 (A6), A06307.

8. Instead of publishing a very similar but not agreeing results again and again, the authors should seriously consider in figuring out the root cause for the disagreements with other instruments so that this data can be meaningfully used in research. 

A: The ground-based observations in China were used to carry out comparison in this manuscript, the data source of which is different from that of previous papers. Although the method of correlation analysis is somewhat similar, the data sources and conclusions are different between this manuscript and previous paper. Since the data validation of satellite is very important, these comparative studies are trying to give more extensive evaluation from different data source.

The day-to-day variation, bias caused by ionogram interpretation, geomagnetic storm and ionospheric model can bring the differences. As suggested, more analyses and some discussions were added in the “Discussions” section in the revised version.

Author Response File: Author Response.pdf

Round 2

Reviewer 1 Report

The paper has been significantly improved, nevertheless some specific points need clarification or small changes.

Specific comments

Line 15 which are the different group features?

Line 78 object, CSES designs two operation modes, objectives, CSES payload operate in two different modes, 

Line 100, “spherical” instead of “sphericity”

Line 101-102 “long stub” instead of “extension bar”

Line 103-104 better modify as follow; The plasma parameters can be obtained by fitting I-V characteristic curve of  LAP

Line 108, 0.1° and 0.2° should be related to 1.5 and 3 s respectively (the  smaller is time the shorter is the crossed space) 

Line 110-113 explain better which data have been removed and why

Line 112 use simply value instead of “absolute value” since Ne cannot be negative

Line 141-142 and 173 it is still not clear the meaning of “working day”

Line 156 please define and quantify the small time scales

Line 307 the name of each station is used the abbreviation. Better “the name of each station is reported together with its acronym”.

Line 360 start the sentence as follow “since the dataset available for the comparison of Ne for CSES and QJISR is still limited …..”

Line 370 weak instead of week.

Line 372-373 subject of the sentence is still missing. The production of what?

Line 385 in general instead of in generally

Line 385-387 the sentence is not clear. Please rephrase or remove if not necessary

Line 387 Figures 5 and 6 shows show that the day-to-day daily variation 

Line 396-397 remove “magnetic storm, substorm and et al.” Solar activity is enough

Line 400 please specify which are the two datasets mentioned in the sentence

Line 429 average instead absolute

Line 438 dependent instead of dependence

Line 468 relationship of the two data characteristics of the two datasets

Line 470-471 no matter the 470 data in continuous time or in a certain month even if the dataset of the comparison is not continuous 

Line 473-475 please rephrase in more direct way (e.g. higher latitude observations show a better correlation between CSES and ionosonde)

Reviewer 2 Report

Comments on the paper entitled “ The comparison of electron density between CSES in-situ and ground-based observations in China” submitted by Jing Liu, Tong Xu, Zonghua Ding and Xuemin Zhang to Remote Sensing Journal (2022)

 

General comments:

I have gone through the revised manuscript that presents the validation results of in situ Electron density (Ne) measurements made using Langmuir probe at ~500 km altitude using China Seismo-Electromagnetic Satellite (CSES) with that of electron density measured by Qujing Incoherent scatter radar and electron density extrapolated to 500 km using combined IRI and ionosonde foF2 weighted density over China for the years 2018-2021. The results suggest that the trend of Ne measured using CSES is consistent with that of Ne measured using ISR with the correlation coefficient of ~0.5. For correlation with ionosonde, ionosondes located at Leshan, Tengchong and Puer have been used. The satellite observations during the months of March, June, September and December 2019 have been used to correlate with the ionosonde observations. In terms of latitude, the correlation between CSES and ionosonde is better at the low latitude than that at the middle latitude during the day and night. However, the value at the middle and mid-low latitude is higher than that at the low latitude. In terms of time, the correlation coefficient in June and September is generally better than that in March and December.The distribution of Ne of CSES and correlation coefficients at different latitudes show that Ne of CSES with the relative high value exhibits a good consistency to the ionosonde observation. Systematic biases have been identified between Ne of CSES and ionosonde data in the daytime and nighttime, respectively. It is suggested that Ne in the daytime may be affected strongly by solar effect as it plays a significant role for the changes in Ne. However, since chemical, dynamic, and electrodynamic processes are the important factors in the nighttime, which may have characteristics of latitude, longitude and altitude.  Still, I believe that the validation is not completely done and it needs to be done systematically. I suggest to authors that they have to revise the manuscript addressing the issues mentioned here so that it can be improved further.

 

The following are my concerns/comments:

 

1.       The general trend of Ne during daytime seems to be agreeing with that of ISR and Ionosonde on large temporal scales, while on shorter scales they differ. However, there are consistent differences exists in the night time. This is suggested to be due to weak solar effect during nighttime. But I can’t go with this argument as the explanations are not making any sense.

2.       In all the figures, the Y-axis (right) and Y-axis (left) scales are not same. Without maintaining the scales same, it is impossible to compare them. So, I suggest them to revise the figures to maintain uniform scales for all the figures. If log scales looks better, they should show them and make uniform scales for better comparison.

3.       The correlation coefficient is very low. It is ~0.5 which is low. How one can trust the results when the correlations are very poor. Re-grouping is done to improve it without any scientific reasons.

4.       I find many important aspects are not discussed completely. However, they are important for the reader to understand the criticality of the paper. 

5.       If possible, authors should show complete diurnal variation of Ne from CSES and their comparison with ISR/Ionosonde. This may provide better sense.

6.       Figure-13 also should be compared with TEC maps to see the general trend.

7.       Whether different probes are used for Day and night which are having different sensitivities during day and night? It should be ascertained. 

8.       Similarly, it is mentioned that higher the density in CSES, better the correlation and vice versa when the density is lower. Why? These are important and critical for design aspects.

9.       Though manuscript looks alright in general, still there are difficulties in reading the manuscript and understanding the sentences due to poor usage of English. It must be further improved for smooth reading the manuscript. 

Comments for author File: Comments.pdf