X-ray Fluorescence Core Scanning for High-Resolution Geochemical Characterisation of Soils

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The authors report on an experimental-chemometrics study to assess the applicability of x-ray fluorescence (XRF) core scanning for high-resolution geochemical characterization of soils. More specifically, using a chemometrics approach (calibration regression) they investigated whether higher resolution element concentrations could be inferred from lower resolution data on reference samples. They showed successful calibration for elements calcium, phosphorus, zinc, strontium, and sulfur, with high R-squared values.

The manuscript discusses a study with important practical implications especially regarding high-resolution analysis. The authors present the results and correlations well, the conclusions are clear. Due to the crucial role of soil mineralogy in soil biogeochemistry, and the lack of well-developed tools for fast, core-scale chemical imaging, this is an important study, especially if the potential for concentration predictions is considered. While I have no significant issues with the report, I have the following requests/recommendations that I believe would improve the manuscript:

-          From the elements, where good calibration was achieved, I would say that Ca, P and S are important for soil analysis. C and N would be critically important, but I understand that those are not measurable with this technique. However, Fe is even more important than Ca, P and S for soil redox chemistry, organo-mineral interactions, and it is a pity that no useful results could be obtained for it. I would like some discussion and possible explanation for that result, and possibly a suggestion how to improve that capability.

-          On a similar note, it is stated in section 3.4 that “generalizibilty of calibration models for new cores can vary greatly depending on the target element”; please discuss what would be a potential solution to this problem?

-          Potassium is of similar importance to Ca, and it is another element where no good calibration results could be obtained. Please include a brief discussion of that as well (see above).

In summary, while the method has novelty and high relevance, there is a lot left to be done for a useful XRF core scanning tool and methodology, in particular for soil biogeochemistry. The manuscript should include a discussion as to how the measurement of critical elements like Fe, K could be improved and whether they have any suggestion for C and N.

Author Response

Dear reviewer,

Thank you very much for your interest and helpful feedback. Please find the replies to your points below.

"From the elements, where good calibration was achieved, I would say that Ca, P and S are important for soil analysis. C and N would be critically important, but I understand that those are not measurable with this technique. However, Fe is even more important than Ca, P and S for soil redox chemistry, organo-mineral interactions, and it is a pity that no useful results could be obtained for it. I would like some discussion and possible explanation for that result, and possibly a suggestion how to improve that capability." 

• The reason Fe didn't work well in our soils is that there is very little variation in our soils' Fe both between the cores and down the core profile. So, all the numbers are essentially just the mean value and a calibration curve won't fit well. In circumstances where there is a variation down a core Fe should work very well. C & N will not work, but could be measured by an alternative technique to provide similar spatial resolution. As an aside, we are pursuing hyperspectral imaging for this reason. An explanation was added to L364 of the revised manuscript.

“On a similar note, it is stated in section 3.4 that “generalizibilty of calibration models for new cores can vary greatly depending on the target element”; please discuss what would be a potential solution to this problem?”.

• I believe this can be improved by having bigger datasets, as is the case for most modeling efforts. However, some elements might never improve to an adequate level due to peak interference or low concentration.  General suggestions were added to L386 of the revised manuscript.

"Potassium is of similar importance to Ca, and it is another element where no good calibration results could be obtained. Please include a brief discussion of that as well (see above)." 

• Potassium although very well measured by the XRF core scanner, suffers from very bad precision with ICP. A better calibration is only possible when a better reference taking method is used. An explanation was added in L342 of the revised manuscript.
   

Best regards,

Corresponding author

Reviewer 2 Report

Comments and Suggestions for Authors

Dear Authors.

Your article presents interesting and not usual application of XRF core scanner to the investigation of soils. I have just some minor comments

I recommend to add values of "high R2 values" to abstract.

I recommend to clarify equipment of ICP Ms and ICP OES analysis.

Tables 1 and 2 are identical and contain empty column (the last one).

Line 184. Please clarify "5 PCA components"

Line 199 and Figure 2. Ar? Please check it. It is doubtfull that you determined Ar by XRF core scanner.

I also recommend to change elements' names to symbols.

Author Response

Dear reviewer,

Thank you very much for your interest and helpful feedback. Please find the replies to your points below.

"I recommend to add values of "high R2 values" to abstract."

• added in the revised manuscript L14.

"I recommend to clarify equipment of ICP Ms and ICP OES analysis.”

• added in the revised manuscript L118.

"Tables 1 and 2 are identical and contain empty column (the last one)"

• fixed in the revised manuscript, Corr. Coef. with core count as a single column.

"Line 184. Please clarify "5 PCA components"

• fixed in the revised manuscript L184.

"Line 199 and Figure 2. Ar? Please check it. It is doubtfull that you determined Ar by XRF core scanner.".

• The XRF core scanner technically does measure Ar, but it is most likely just the Ar in the air. Incidentally, it was dropped in the feature selection process based on low quality of measurement so it has no bearing on the calibration. 

"I also recommend to change elements' names to symbols.",

• corrected in the revised manuscript. 

Best regards,

Corresponding author