A New Spectral Transformation Approach and Quantitative Analysis for MarSCoDe Laser-Induced Breakdown Spectroscopy (LIBS) Data
, Zhongchen Wu 1,*, Zongcheng Ling 1, Changqing Liu 1, Wang Liu 1, Wenxi Chen 1 and Li Zhang 2
Round 1
Reviewer 1 Report
‘A New Spectral Transformation Approach and Quantitative Analysis for MarSCoDe Laser-Induced Breakdown Spectroscopy (LIBS) Data’ by Jin et al., 2022, Remote Sensing
In this manuscript, the authors propose a 4-step spectral transformation approach, to unify the different LIBS datasets acquired by different Mars exploration missions. It’s a very interesting work. However, there are some important issues need to be resolved before consideration for publication:
Major comments:
1. The authors mentioned in line 86: due to the lack of data in MarSCoDe, it is difficult to implement data transformation. How can the 4-step spectral transformation approach solve this problem? Please clarify in the text. Because this is the significance of your work.
2. In this study, only a small amount of data is used for training. Are the results reliable enough? Can you increase the percentage of validation sets?
3. Figs.5-6: RMSE is not the best index to assess the accuracy, because the contents of different elements vary. R2 is a good choice. The points in Figure 6 are rather scattering.
4. Line 330-331 (Table 8): the estimated value of Fe is quite different from the ideal value, why does this phenomenon not appear in the SDU-LIBS data?
5. The key issue I am concerned about is the model test. The results derived from SDU-LIBS spectra don’t look good enough.
Minor comments:
Line 56: MVA—MLA?
Line 105-106: How to select the data? What is high SNR? Please provide a number. SN should be SNR, signal-to-noise ratio.
Line 161: Table 1--Table 2?
Which dataset did you use to derive the results in Figure 5? Please clarify.
Author Response
Dear Reviewer,
We would like to express our sincere appreciation for your comments concerning our MS entitled “A New Spectral Transformation Approach and Quantitative Analysis for MarSCoDe Laser-Induced Breakdown Spectroscopy (LIBS) Data”. These comments are valuable and helpful for us to revise and improve the quality of our paper. We have revised this MS based on your comments one by one. All the revised portions are marked in red in the revised MS which was uploaded to the website as the attachment. The main corrections in the paper and the responds to the comments are listed as following:
Major comments:
Point 1: The authors mentioned in line 86: due to the lack of data in MarSCoDe, it is difficult to implement data transformation. How can the 4-step spectral transformation approach solve this problem? Please clarify in the text. Because this is the significance of your work.
Response 1: The 4-step approach was done to unify the spectral data formats of ChemCam, MarSCoDe and SDU-LIBS which does not require a large amount of data for training. The spectra can be well transformed by our 4-step approach. Therefore, the small amount of MarSCoDe spectral data does not prevent the preformation of our 4-step approach. This explanation has been added to the revised MS (line 100– line 103).
Point 2: In this study, only a small amount of data is used for training. Are the results reliable enough? Can you increase the percentage of validation sets?
Response 2: In this study, the best way for model cross-validation is using one dataset (i.e., ChemCam LIBS standard database) to train the PLS model, and then using another data set (i.e., SDU-LIBS data set with their element abundance) to verify this model. The reason for only selecting 11 samples of SDU-LIBS set for model verification is that the 11 samples are the same calibration samples in ChemCam calibration data set. Those selected 11 samples are most representative for verification of this cross-validation model. As shown in Figure 5. The similar values of RMSEC and RMSEP illustrated the PLS model was acceptable.
Point 3: Figs.5-6: RMSE is not the best index to assess the accuracy, because the contents of different elements vary. R2 is a good choice. The points in Figure 6 are rather scattering.
Response 3: R2 values and the relevant discussions were added in Figs.5-6 and right places in text (line 281-311). Figure 6 was modified to make the points looking less scattering based on the facts. Now, R2 together with RMSE works well for assessing the accuracy of our method.
Point 4: Line 330-331 (Table 8): the estimated value of Fe is quite different from the ideal value, why does this phenomenon not appear in the SDU-LIBS data?
Response 4: As shown in Table 4 (line 242 – line 243 of the revised MS). The step size of UV band in ChemCam (~0.0488 nm) and SDU-LIBS (~0.0494 nm) systems is very similar. Therefore, the information loss of SDU-LIBS in the UV band was smaller after interpolating. Compared with SDU-LIBS systems, the step size of the UV band in ChemCam (~0.0488 nm) and MarSCoDe (~0.0667 nm) systems is larger. This caused more information loss from MarSCoDe in the UV band after interpolating. On the other hand, Fe emission lines were widely distributed in UV band. Therefore, the Fe derived value of the MarSCoDe system spectra was lower than the ideal value because of much information loss after interpolating. This phenomenon does not appear in the SDU-LIBS data.
Point 5: The key issue I am concerned about is the model test. The results derived from SDU-LIBS spectra don’t look good enough.
Response 5: Under present conditions, the copy of MarSCode is not ready to build the accurate LIBS standard database. Our 4-step approach had to be proposed for data format unification of different LIBS system in order to build a cross-validation model. Currently, our 4-step approach can only solve parts of the data transformation problems. Information lost during our transformation still existed. However, our 4-step approach is necessary for building the cross-validation model and the current results are acceptable.
Minor comments:
Point 6: Line 56: MVA—MLA? Line 161: Table 1--Table 2?
Response 6: Thanks for your reminding. Done as the as follows:
'
Line 57: “MVA” was modified to “MLA”
Line 172: “Table 1” was modified to “Table 2”
Line 114, line 130, line 194: “SN” was modified to “SNR”
Point 7: Line 105-106: How to select the data? What is high SNR? Please provide a number. SN should be SNR, signal-to-noise ratio.
Response 7: SNR is the ratio between the desired signal intensity and the background noise. The signal strength is the strongest peak intensity of the LIBS spectrum in the analysis band, and the noise level is the standard deviation of the same spectrum band which does not contain signal. SNR values for all spectra were calculated in this study. And spectra with the SNR higher than 400 were selected for further analysis. The “SN” was modified to “SNR”.
Point 8: Which dataset did you use to derive the results in Figure 5? Please clarify.
Response 8: Thanks. The data set has been clarified in the text (line 280). Figure 5 shows the derived values of the test set by PLS model against their actual values. The test set was made of the 80 spectra which was shown in Table 3 (line 225).
Author Response File: 
Author Response.docx
Reviewer 2 Report
Dear Author
I appreciate your proposal to evaluate this article and your interest in my participation.
The structure of the article seems correct to me and is: 1. Introduction, 2. Data and Methods 3. Reliability validation of spectral transformation approach 4. Analysis and discussion of MarSCoDe LIBS data 5. Conclusions
1. Introduction
The authors have done a good job of reviewing the state of the art in the literature referenced in References. The introduction seems to me to be well developed and substantiated with the bibliographical references being very up-to-date.
line 57. PLS, it should be defined. It appears defined in line 199, but it must appear before line 57.
2. Data and Methods
This chapter seems well developed.
line 103 and 105 where it says "in-suit" does it mean in-situ?
line 137. "PSO" should be defined.
line 121.."FeOT" should be defined. It does not explain what is its meaning Fe03? Fe2O3?
The figures are correct in dimensions and quality.
3. Reliability validation of spectral transformation approach
This chapter seems well developed.
4. Analysis and discussion of MarSCoDe LIBS data
This chapter seems well developed.
4. Conclusion
The conclusions are consistent with the proposed objectives.
references
References are correct, complete, and up-to-date.
Before recommending publication, authors must perform minor revisions.
Best regards
Author Response
Dear Reviewer,
We would like to express our sincere appreciation for your comments concerning our MS entitled “A New Spectral Transformation Approach and Quantitative Analysis for MarSCoDe Laser-Induced Breakdown Spectroscopy (LIBS) Data”. These comments are valuable and helpful for us to revise and improve the quality of our paper. We have revised this MS based on your comments one by one. All the revised portions are marked in red in the revised MS which was uploaded to the website as the attachment. The main corrections in the paper and the responds to the comments are listed as following:
Point 1:
line 57. PLS, it should be defined. It appears defined in line 199, but it must appear before line 57.
line 103 and 105 where it says "in-suit" does it mean in-situ?
line 121."FeOT" should be defined. It does not explain what is its meaning Fe03? Fe2O3?
Response 1: Thanks. Done. The list of changes is listed as follows:
Line 25: PLS was defined.
Line 112-line 115: “in-suit” was modified to “in-situ”
Line 142: FeOT was defined
Point 2: line 137. "PSO" should be defined.
Response 2: “PSO” has been defined when it first appears in line 68 of revised MS.
Author Response File: Author Response.docx
Reviewer 3 Report
This paper presents a method for spectral data calibration between ChemCam, MarSCoDe, and SDU-LIBS instruments. The primary purpose of this method development is to transform LIBS data from the Zhurong rover to make it compatible with the PLS model used for ChemCam data processing.
The 4-step transformation procedure is clearly presented and well structured.
It is clear, that authors are writing many papers dedicated to the Zhurong rover data, its processing, and calibration. The authors dived deep into the subject of the project etc., however, there is a lack of general information in the introduction making the whole picture of the project clear for new readers. I would recommend more accurate address readers to the previous authors' studies with the comments about what kind of information could be found there. For example, in the introduction, it is very difficult to recognize that SDU-LIBS is the laboratory analog instrument of the MarSCoDe.
I would like to kindly recommend to you the way of improvement of the transformation procedure presented here, in case I correctly understand the matter. Between steps 2 and 3 I would introduce one more step. Every spectra signal, recorded by your instruments, has a continuous model function. Before the interpolation step, I would make a fitting of MarSCoDe data with the model distribution function by the least square method and only then implement the interpolation procedure. It seems this way will help to escape from the inaccurate 'peak top loss' problem.
Minor comments:
1) There are abbreviations used without explanation:
- PLS line 25;
- SDU line 89;
- PSO - line 137;
2) Missed space in line 215;
3) Table 2 is titled Table 1;
4) There is no reference in the text to Table 3.
Author Response
Dear Reviewer,
We would like to express our sincere appreciation for your comments concerning our MS entitled “A New Spectral Transformation Approach and Quantitative Analysis for MarSCoDe Laser-Induced Breakdown Spectroscopy (LIBS) Data”. These comments are valuable and helpful for us to revise and improve the quality of our paper. We have revised this MS based on your comments one by one. All the revised portions are marked in red in the revised MS which was uploaded to the website as the attachment. The main corrections in the paper and the responds to the comments are listed as following:
Major comments:
Point 1: It is clear, that authors are writing many papers dedicated to the Zhurong rover data, its processing, and calibration. The authors dived deep into the subject of the project etc., however, there is a lack of general information in the introduction making the whole picture of the project clear for new readers. I would recommend more accurate address readers to the previous authors' studies with the comments about what kind of information could be found there. For example, in the introduction, it is very difficult to recognize that SDU-LIBS is the laboratory analog instrument of the MarSCoDe.
Response 1: Thanks. The more information was added in the introduction section and the relevant references were also cited in the right place (line 89 – line 93) of the revised MS.
Point 2: I would like to kindly recommend to you the way of improvement of the transformation procedure presented here, in case I correctly understand the matter. Between steps 2 and 3 I would introduce one more step. Every spectra signal, recorded by your instruments, has a continuous model function. Before the interpolation step, I would make a fitting of MarSCoDe data with the model distribution function by the least square method and only then implement the interpolation procedure. It seems this way will help to escape from the inaccurate 'peak top loss' problem.
Response 2: Thanks for reminding me. Wavelength calibration and spectral radiance calibration have been done before LIBS data recordation. After the two calibration, the influence of continuous model function of the SDU-LIBS system should be ignored. The relative description about the spectral radiance calibration was added in text (line 123- line 124).
Minor comments:
Point 3:
1) There are abbreviations used without explanation:
- PLS line 25;
- SDU line 89;
- PSO - line 137;
2) Missed space in line 215;
3) Table 2 is titled Table 1;
Response 3: Thanks very much. Done. The changes are listed as follows:
Line 25: “PLS” was defined.
Line 227: one space was added in the right place.
Line 172: The title was modified.
“SDU” is the part of “SDU-LIBS”.
“SDU-LIBS” is the name of our LIBS instrument. SDU is the abbreviation of Shandong University.
“PSO” has been defined when it first appeared in line 68.
Point 4:
4) There is no reference in the text to Table 3.
Response 4: The references of Table 3 were added in right place of the revised MS (line 214).
Author Response File: Author Response.docx
Round 2
Reviewer 1 Report
line 283: "The similar values of RMSEC and RMSEP illustrate the PLS model is available." This is unsufficient to conclude that the PLS model is available.
Figure 5, Figure 6, line 299: The plots of Ti don't show good relationship. R2 is too low.
Author Response
Dear Reviewer,
We would like to express our sincere appreciation again for your comments concerning our MS entitled “A New Spectral Transformation Approach and Quantitative Analysis for MarSCoDe Laser-Induced Breakdown Spectroscopy (LIBS) Data”. These comments are helpful for us to improve the quality of our paper. We have revised this MS based on your two comments one by one. All the revised portions are marked in red in the revised MS which was uploaded to the website as the attachment. The main corrections in the paper and the responds to the comments are listed as following:
Point 1: line 283: "The similar values of RMSEC and RMSEP illustrate the PLS model is available." This is unsufficient to conclude that the PLS model is available.
Response 1: The R2 values of the PLS model of training set are 0.78 (SiO2), 0.56 (TiO2), 0.75 (Al2O3), 0.54 (FeOT), 0.95 (MgO), 0.84 (CaO), 0.85 (Na2O), and 0.84 (K2O), respectively. And all R2 values of training set except TiO2 (0.56) and FeOT (0.54) are greater than 0.75. Because of that the actual values of TiO2 in training set are mostly distributed in the smaller and narrower range of 0-1.5 wt.%, the calculated R2 and RMSE values will be much lower as long as the derived values of TiO2 are deviated from the actual values. This is the main reason for the lower R2 of TiO2. Similarly, the actual values of FeOT in ChemCam standard database are also mostly distributed in the smaller and narrower range of 0-10 wt.%, which causes an overfitted model and a lower R2 of FeOT. In addition, the RMSEC and RMSEP of these models in Figure 5 are similar, and mostly points are tightly distributed around the 1:1 line. Therefore, the PLS model built in this study is available. These discussions were added to the revised MS (line 283– line 294).
Point 2: Figure 5, Figure 6, line 299: The plots of Ti don't show good relationship. R2 is too low.
Response 2: Because of that the actual values of TiO2 in used data are mostly distributed in the smaller and narrower range of 0-1.5 wt.%, the calculated R2 and RMSE values will be much lower and the relationship shown in plots will be much worse as long as the derived values of TiO2 are deviated from the actual values. As analyzed above, this is the main reason for the lower R2 of TiO2. However, the worse relationship shown in plots and the lower R2 values of TiO2 are acceptable.
Author Response File: Author Response.docx
