the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Optimization of a Picarro L2140-i cavity ring-down spectrometer for routine measurement of triple oxygen isotope ratios in meteoric waters
Jack A. Hutchings
Bronwen L. Konecky
Download
- Final revised paper (published on 29 Mar 2023)
- Preprint (discussion started on 10 Oct 2022)
Interactive discussion
Status: closed
-
RC1: 'Comment on amt-2022-230', Anonymous Referee #1, 01 Nov 2022
The manuscript "Optimization of the Picarro L2140-i Cavity Ring Down Spectrometer for Routine Measurement of Triple Oxygen Isotope Ratios in Meteoric Waters" by JA Hutchings and BL Konecky takes a close look at the way many of us use this instrument and the way we analyze the data. The paper is extremely thorough and takes the time to test many aspects that all users of this model, and really all IRIS and even IRMS instruments, should make note of. The manuscript is well written, well organized, sufficiently referenced and warrants publication in AMT after considering a few minor suggestions described below.
Specific Comments:
The permeg notation: "10^6 * delta" is awkward and confusing. I am not familiar with expressing permeg in this way. The authors define a delta value using equation 1 which includes the *1000 term. To me, then, this means delta values are in permil and have thus had 1000 multiplied by them. So if Delta17O is 10^6 * delta, then it is * 10^9. I know this is not what the authors mean, but perhaps you can see my point. I suggest simply using "per meg" and removing the "10^6 * delta" throughout. Equation 3 is sufficient for the authors definition of D17O.18O Laser Flag: The 18O Laser Flag seems like a great tool. I would like to see the authors make a short statement about how they came upon the "18O Laser Flag". The authors also need to describe in greater detail how the flag value is calculated, or is it simply the "SD_18_18" value from the h5 files? What specific values of the 18O Laser Flag constituted action? That is, what is the threshold value of the 18O Laser flag that, if exceeded, meant organic contamination? Why does this 18O Laser Flag work? Are the authors assuming the alcohol absorption will have a changing baseline from 7190 to 7200 /cm ? It seems to that the 18O Laser Flag is a way of comparing the the two lasers. 18O Laser Flag and D17O both appear to be good indicators of contamination and both make use of comparing the two disparate absorption bands. I suppose we could also use deuterium excess calculations that use 18O from laser 1 and 18O from laser 2. I wonder then if it is appropriate to extend a generalized statement about the utility and benefit of having two lasers to compare. I realize much of this is musing speculation but it would be great to have a sentence or two giving the reader insight into the authors thought process around this idea and this flag. As it is, the 18O Laser Flag is just thrown in without any introduction or elaboration. Lastly, and very minor, on line 285, the authors state "referred to hereafter as “18O Laser Flag”" but rarely if ever refer to this metric in that way. I see "18O-Laser Values", "18O Laser", "18O-Laser Metric", "18O-Laser spectral contamination metric".
What is the outcome or course of action after finding an aberrant point such as is shown in Figure S14? Does use of h5 level data allow you to recover without rerunning the water? Given the authors preference for high resolution data over coordinator data, further explaining this process will help their case. Minor comments related to this topic: Fig S14 (a) caption, I think should refer to injections, not vials, not because the statement isn't true as is, but because the authors are showing injection data, not vial data. Also, and again, minor, it may be worth pointing out in the caption of S14 that C and D data do not come from the h5 files referred to thus far in the manuscript, but rather from the very high resolution spectral data. I know the authors already say "spectral readings" but helping the readers who don't use any of the h5 files may be useful.
Technical Comments:
- Consider referring to the supplemental figures and tables as appendices. It seems as though AMT uses supplemental information to refer to a text based document and stand-alone figures and tables are appendices.
- Check for delta notation missing superscripts throughout the manuscript, for example line 216.
- Also, on occasion, I see "D-Excess" instead of "d-excess". Change to lowercase "d" throughout.
- Line 39 - "slope of 0.528 calculates values" is awkward. Reword.
- Line 59 - "isotopogues" should be "isotopologue"
- Line 66 - "H2H16O" should be "1H2H16O" with appropriate superscripts, of course.
- Table 1 caption ends prematurely. "...measurements by this study." perhaps?
- Line 154 - "To apply the drift correction, the slope of the drift regression was multiplied by an injection’s position and subtracted from the injection’s observed delta value." As worded, this sentence suggests that the slope is multiplied and then subtracted. I think the slope is multiplied by the injections position and then the product is subtracted?
- Line 182 - "experienced increased memory experienced", I think, should read "experienced increased memory"
- Line 246 - "account drift" should read "account for drift"
- Line 317 - "term", while correct, is slightly awkward with the two other "-term" uses in that sentence. Perhaps "factor"?
- Line 333 - The word "excess" seems out of place. "exceed" perhaps?
- Line 337 - "as measure" should perhaps read "as a measure"- Figure 5 - The average of the MSD should be zero, or not different from zero. I am not sure if it is worth showing this in the figure text. It would be like changing your equation 5 to x_i minus x_bar instead of x_hat. We expect the sum of (x_i minus x_bar) / n to equal zero or not different from zero. Also, and minor, I am not sure you need an additional y-axes label of "Histogram (counts)" since it is in the caption. Furthermore, Table S3 has exactly the same data presented as the text in Figure 5. I feel the same way about the mean MSD in Table S3 and since you have the same data in Fig 5, I suggest deleting Table S3.
- Line 345 - "an average of by" should perhaps read "by an average of"
- Figure 6 caption - "were differenced" should read "were subtracted"
- Line 500 - "sample" should be "samples"
- Line 628 - "only strongly" seem to contradict each other. Seems like "only" should be deleted- Figure S1 - The order of figures is inconsistent with other figures. All other figures are d18O, d17O, dD, d-excess, D17O.
- The data in Figure S3 seems like it may be better presented as histograms, which are made good use of elsewhere in the manuscript.
- Figure S13 caption - "differenced" should be "subtracted"
- Table S2 - missing peak 1, H218O. Also, consider removing this table entirely and simply cite Steig et al 2014.
Citation: https://doi.org/10.5194/amt-2022-230-RC1 -
AC1: 'Reply on RC1', Jack Hutchings, 10 Nov 2022
Thanks for the positive and helpful review! Here are a few thoughts for the open discussion period.
Briefly, you are correct that the '18O Laser Flag' was calculated as a way to compare disparities between the two lasers in a familiar notation (standard deviation of the two different calculations of d18O). We calculated d18O using either the 18-Oxygen spectral peak of laser 2 ('str11_offset' in the h5 files) or laser 1 ('str1_offset' in the h5 files). You are also correct that the inference we make is that organic interference leads to different baselines between the two lasers. During post-processing, we calculate the '18O Laser Flag' for standards use the maximum observed value plus a small amount approximately equivalent to precision (0.05 permil) as the testing threshold beyond which a sample is flagged. We'll elaborate a bit more are 'muse' some in the revised manuscript.
Aberrant points are simply removed. The h5 files, in this case, both allow the detection and the removal of these points and allows us to use that injection. The injection does have one fewer data point, but each injection's 'observed value' is the mean of ~560 data points so this is an essentially negligible amount of signal loss (in our view). We'll add a small amount of extra content clarifying this in the revised manuscript.
Citation: https://doi.org/10.5194/amt-2022-230-AC1 -
AC3: 'Reply on RC1', Jack Hutchings, 29 Nov 2022
- All author responses are bulleted. For completeness, the original reviewer text is included here in full.
The manuscript "Optimization of the Picarro L2140-i Cavity Ring Down Spectrometer for Routine Measurement of Triple Oxygen Isotope Ratios in Meteoric Waters" by JA Hutchings and BL Konecky takes a close look at the way many of us use this instrument and the way we analyze the data. The paper is extremely thorough and takes the time to test many aspects that all users of this model, and really all IRIS and even IRMS instruments, should make note of. The manuscript is well written, well organized, sufficiently referenced and warrants publication in AMT after considering a few minor suggestions described below.
- Thanks again for the thoughtful and thorough review of our manuscript.
The permeg notation: "10^6 * delta" is awkward and confusing. I am not familiar with expressing permeg in this way. The authors define a delta value using equation 1 which includes the *1000 term. To me, then, this means delta values are in permil and have thus had 1000 multiplied by them. So if Delta17O is 10^6 * delta, then it is * 10^9. I know this is not what the authors mean, but perhaps you can see my point. I suggest simply using "per meg" and removing the "10^6 * delta" throughout. Equation 3 is sufficient for the authors definition of D17O.
- This is a fair point. We’ll go ahead and use ‘per meg’ going forward.
18O Laser Flag: The 18O Laser Flag seems like a great tool. I would like to see the authors make a short statement about how they came upon the "18O Laser Flag". The authors also need to describe in greater detail how the flag value is calculated, or is it simply the "SD_18_18" value from the h5 files? What specific values of the 18O Laser Flag constituted action? That is, what is the threshold value of the 18O Laser flag that, if exceeded, meant organic contamination? Why does this 18O Laser Flag work? Are the authors assuming the alcohol absorption will have a changing baseline from 7190 to 7200 /cm ? It seems to that the 18O Laser Flag is a way of comparing the the two lasers. 18O Laser Flag and D17O both appear to be good indicators of contamination and both make use of comparing the two disparate absorption bands. I suppose we could also use deuterium excess calculations that use 18O from laser 1 and 18O from laser 2. I wonder then if it is appropriate to extend a generalized statement about the utility and benefit of having two lasers to compare. I realize much of this is musing speculation but it would be great to have a sentence or two giving the reader insight into the authors thought process around this idea and this flag. As it is, the 18O Laser Flag is just thrown in without any introduction or elaboration. Lastly, and very minor, on line 285, the authors state "referred to hereafter as “18O Laser Flag”" but rarely if ever refer to this metric in that way. I see "18O-Laser Values", "18O Laser", "18O-Laser Metric", "18O-Laser spectral contamination metric".
- As mentioned in our open discussion response, our revised manuscript will expand on the ‘18O Laser Flag’ metric. We also see our sloppy reference in the text/figures to the metric, which will be rectified in the revised manuscript. Beyond what was already said in our open discussion response, we should clarify that the reason we left the concept somewhat hanging is that, although it appears to be a more powerful indicator of spectral contamination than the default metrics, the ‘18O Laser Flag’ is ultimately found to be insufficient to detect minor amounts of contamination that must be dealt with by removal of the interference. The interference must, then, be removed before the sample is analyzed. We accomplish this through online combustion using Picarro’s micro combustion module, but offline removal is possible, although efficiency of techniques such as solid phase extraction appear to be about 90% removal (Chang et al., 2016), which may not be sufficient for samples with high enough initial contamination. We will add detail on the inferred operating principle of this metric as well as its limitations in the revised manuscript.
What is the outcome or course of action after finding an aberrant point such as is shown in Figure S14? Does use of h5 level data allow you to recover without rerunning the water? Given the authors preference for high resolution data over coordinator data, further explaining this process will help their case. Minor comments related to this topic: Fig S14 (a) caption, I think should refer to injections, not vials, not because the statement isn't true as is, but because the authors are showing injection data, not vial data. Also, and again, minor, it may be worth pointing out in the caption of S14 that C and D data do not come from the h5 files referred to thus far in the manuscript, but rather from the very high resolution spectral data. I know the authors already say "spectral readings" but helping the readers who don't use any of the h5 files may be useful.
- We addressed this in our open discussion response and will add additional context in the manuscript. We’ll also update our reasoning for using high resolution processing in light of this feature as well as the ‘18O Laser Flag’.
Technical Comments:
- Consider referring to the supplemental figures and tables as appendices. It seems as though AMT uses supplemental information to refer to a text based document and stand-alone figures and tables are appendices.
- We had done this originally and were informed that appendices are, in AMT, appended to the end of the paper such that all the supplemental figures/tables would actually be part of the manuscript. Since we have a good number of additional figures/tables, we opted to treat them as supplemental so that the formatted manuscript would not be excessively long/large due to all the extra figures/tables.
- Check for delta notation missing superscripts throughout the manuscript, for example line 216.
- Yes, there were a few! We think we have them all now.
- Also, on occasion, I see "D-Excess" instead of "d-excess". Change to lowercase "d" throughout.
- Thanks – we’ll double check this throughout the manuscript.
- Line 39 - "slope of 0.528 calculates values" is awkward. Reword.
- Fixed, thanks!
- Line 59 - "isotopogues" should be "isotopologue"
- Fixed, thanks!
- Line 66 - "H2H16O" should be "1H2H16O" with appropriate superscripts, of course.
- Indeed, thanks!
- Table 1 caption ends prematurely. "...measurements by this study." perhaps?
- Yes, you are correct. That was cut off when we imported the table.
- Line 154 - "To apply the drift correction, the slope of the drift regression was multiplied by an injection’s position and subtracted from the injection’s observed delta value." As worded, this sentence suggests that the slope is multiplied and then subtracted. I think the slope is multiplied by the injections position and then the product is subtracted?
- Correct! Fixed that text.
- Line 182 - "experienced increased memory experienced", I think, should read "experienced increased memory"
- Fixed, thanks!
- Line 246 - "account drift" should read "account for drift"
- Fixed, thanks!
- Line 317 - "term", while correct, is slightly awkward with the two other "-term" uses in that sentence. Perhaps "factor"?
- Yes, thanks!
- Line 333 - The word "excess" seems out of place. "exceed" perhaps?
- Correct, thanks!
- Line 337 - "as measure" should perhaps read "as a measure"
- Correct, thanks!
- Figure 5 - The average of the MSD should be zero, or not different from zero. I am not sure if it is worth showing this in the figure text. It would be like changing your equation 5 to x_i minus x_bar instead of x_hat. We expect the sum of (x_i minus x_bar) / n to equal zero or not different from zero. Also, and minor, I am not sure you need an additional y-axes label of "Histogram (counts)" since it is in the caption. Furthermore, Table S3 has exactly the same data presented as the text in Figure 5. I feel the same way about the mean MSD in Table S3 and since you have the same data in Fig 5, I suggest deleting Table S3.
- We’re not completely sure we follow your reasoning. Are you saying that, ideally, our MSD values *should* be zero / indistinguishable from zero? And that we should highlight this feature of the metric? If so, then we have added some text to that effect. We also agree that table S3 is an exact reproduction of the text in Figure 5, but we reasoned that some readers would prefer tabular data and that there isn’t any harm in having it in the supplement.
- Line 345 - "an average of by" should perhaps read "by an average of"
- Correct, thanks!
- Figure 6 caption - "were differenced" should read "were subtracted"
- Fair enough, we’ve changed that for clarity.
- Line 500 - "sample" should be "samples"
- Correct, thanks!
- Line 628 - "only strongly" seem to contradict each other. Seems like "only" should be deleted
- Figure S1 - The order of figures is inconsistent with other figures. All other figures are d18O, d17O, dD, d-excess, D17O.
- This was an issue in a couple other figures (S10, S11) and they are all now in the same order as other figures.
- The data in Figure S3 seems like it may be better presented as histograms, which are made good use of elsewhere in the manuscript.
- We wanted to preserve the original individual data points in this figure to demonstrate that lack of extreme outliers that might otherwise be masked by a histogram.
- Figure S13 caption - "differenced" should be "subtracted"
- Fair enough, we’ve changed that for clarity.
- Table S2 - missing peak 1, H218O. Also, consider removing this table entirely and simply cite Steig et al 2014.
- Good catch! For completeness we will leave the table in, especially since it is not in the main text and does not occupy more valuable space.
Citation: https://doi.org/10.5194/amt-2022-230-AC3
-
AC1: 'Reply on RC1', Jack Hutchings, 10 Nov 2022
-
RC2: 'Comment on amt-2022-230', Anonymous Referee #2, 05 Nov 2022
Review of Optimization of the Picarro L2140-i Cavity Ring Down Spectrometer for Routine Measurement of Triple Oxygen Isotope Ratios in Meteoric Waters” by Hutchings and Konecky, AMT.
General:
The manuscript presents an in-depth investigation of potential problems and an improvement in detecting and correcting these through optimizing the analysis of the raw data and adjusting the measurement series setup. This problem is particularly relevant for the secondary isotope parameter D17O, as this parameter is strongly influenced by spurious spectra of various organic molecules, as the authors show using water standards spiked with alcohol. I like their approach and was surprised that they even observed significant shifts of D17O in rainwater due to such influences. The manuscript should be published in AMT as it discusses important problems with this analytical method and suggests remedies that should be known in the scientific community.
Major points:
It would be interesting to know how the samples were stored. Which containers (material, colour), caps and membranes were used and how long were they stored before analysis? This could be important for the development of organic substances in the sample container.
Since the analysis time for high-precision measurements per sample - as suggested by the authors - is very long at 6.3 hours, it would make sense to first run the sample value range with a 1-injection sequence and order them according to the pre-measured values. This would minimize the memory effects and could therefore lead to a shorter total time.
Why was the BSM and not the ANT standard used as the second normalization standard, the latter standard would be most negative for both the d18O, d17O and the dD that you have available. Your choice requires extrapolation of your normalization for highly depleted samples, but I suspect that you have not measured any such samples other than the ANT standard.
It is not clear how the memory effect for D17O was calculated, as this cannot be calculated directly from a simple D17O balance, as it relates to a difference in logarithmic values. Please add a sentence on how this was done.
The memory effect has been calculated based on alternating the Kona and ANT standards, i.e. an enriched and a depleted isotope value from which percentage memory influences are calculated. This is correct, but this leads to different times for reaching the instrument noise level as the differences among the samples will be significantly smaller for all isotopes. Therefore, an ordering according to their pre-measured or estimated sample values could most probably lead to a significant reduction of the overall time investment without losing precision. See also a minor comment to Fig. 6 below.
I don't quite understand why they used a particular standard for drift correction. Would it be possible to use one of the other four standards you have in use?
Minor points:
P5 Table 1 legend: something is missing….by this….
P6 Is there a reason why the vial position is different for the standards but not for the MCM QAQC sample (vial position 1 for both the first and last measurements)?
PXX the formatting of d18O and d17O is not consistent, check the complete text.
P15, l324ff You have measured an unknown sample several times distributed over several series and months. Have you measured the sample water in the same vial? Have you recapped the vial after each measurement sequence? Or have you measured the same water sample by decanting it into individual vials every time you measured it? The procedure is not clear to me it but should be mentioned.
Fig. 6 I guess that the values shown depending on the previous sample or standard value measured or is this no issue? The mean absolute error may indicate that it is dependent on the various sample. The accepted value of the current sample is reached either from below or above this value. Therefore, one would have to indicate based on which jump (current to previous sample or standard) these values shown in Fig. 6 are based. The absolute error then depends on this jump. This should be mentioned.
Citation: https://doi.org/10.5194/amt-2022-230-RC2 -
AC2: 'Reply on RC2', Jack Hutchings, 10 Nov 2022
Thanks for the positive and helpful review! Here are a few thoughts for the open discussion period.
Our samples were stored in glass vials with polyethylene-cone cap liners. Parafilm was additionally applied to each vial to seal the gap between the cap and the neck of the vial. We have no reason to believe that organics were desorbed off the containers, but this could certainly occur with other vials. While we did not test for this explicitly, we store many waters in these vials and have never found, after storage, that a sample would 'develop' organic contamination.
We used BSM as our second normalization because, during the period of this paper, we were measuring tropical precipitation samples that are not particularly depleted.
We didn't actually calculate any memory effect for D17O in the same sense as the primary isotope measurements. Instead, when investigating whether or not memory correction of the primary isotope parameters improved secondary parameters (d-excess and D17O), we calculated errors with or without primary isotope memory correction and looked for improvement due to memory correction. This is shown in Fig S1, but we'll definitely clarify that a bit in the main text.
Our usage of Kona for drift correction is largely arbitrary. Any other standard would also work fine. The main selection criteria is that the drift standard is somewhat similiar in isotopic composition as samples to further minimize any impact of memory on the drift standard measurements so that any 'signal' we observe is due to instrument drift.
Your comments on sample ordering are interesting, but I think I'll need to dwell on them a bit to give a proper response. Very briefly, even doing a 1-injection sequence would 'compromise' the vials such that additional vials would need preparing. Some considerations (larger vials with more water in them) could be made, but that would involve additional pros/cons. I'm also not quite sure how time would be saved from ordering sample vials isotopically - do you mean that we could get away with doing less than 6 injections? If so, we would need to consider the actual sample-to-sample isotope differences and come up with thresholds of differences that justify 3, 4, 5, 6, etc. injections... but varying injection number on the fly is quite limited by the autosampler software.
Citation: https://doi.org/10.5194/amt-2022-230-AC2 -
AC4: 'Reply on RC2', Jack Hutchings, 29 Nov 2022
- All author responses are bulleted. For completeness, the original reviewer text is included here in full.
The manuscript presents an in-depth investigation of potential problems and an improvement in detecting and correcting these through optimizing the analysis of the raw data and adjusting the measurement series setup. This problem is particularly relevant for the secondary isotope parameter D17O, as this parameter is strongly influenced by spurious spectra of various organic molecules, as the authors show using water standards spiked with alcohol. I like their approach and was surprised that they even observed significant shifts of D17O in rainwater due to such influences. The manuscript should be published in AMT as it discusses important problems with this analytical method and suggests remedies that should be known in the scientific community.
- Thanks again for the thoughtful and thorough review of our manuscript.
Major points:
It would be interesting to know how the samples were stored. Which containers (material, colour), caps and membranes were used and how long were they stored before analysis? This could be important for the development of organic substances in the sample container.
- We note this in the manuscript (around line 335 in section 3.4): 4 mL glass vials with polyethylene ‘PolyCone’ caps wrapped with Parafilm and stored in a 4°C refrigerator. We’ll clarify that our vials are made of glass in the revised version. This is, to our knowledge, a common approach and should not appreciably leach any organics, although we suppose it may be somewhat possible given that the liner is made of plastic.
Since the analysis time for high-precision measurements per sample - as suggested by the authors - is very long at 6.3 hours, it would make sense to first run the sample value range with a 1-injection sequence and order them according to the pre-measured values. This would minimize the memory effects and could therefore lead to a shorter total time.
- We addressed this a bit during open discussion. To expand: We do not believe we would see any benefit for our use-case by doing this. The samples are typically analyzed in triplicate only because we seek to minimize the estimated error around the D17O measurement, which itself is largely unaffected by memory correction (Fig. S1).
Why was the BSM and not the ANT standard used as the second normalization standard, the latter standard would be most negative for both the d18O, d17O and the dD that you have available. Your choice requires extrapolation of your normalization for highly depleted samples, but I suspect that you have not measured any such samples other than the ANT standard.
- In fact, we have begun using the ANT standard as our depleted normalization standard as we have moved into measuring more temperate-climate waters with our device. However, for the mostly low-latitude samples discussed in this paper, the less-depleted BSM standard was more appropriate.
It is not clear how the memory effect for D17O was calculated, as this cannot be calculated directly from a simple D17O balance, as it relates to a difference in logarithmic values. Please add a sentence on how this was done.
- This is addressed in the open discussion and we’ll clarify the point in the main text in section 2.2.
The memory effect has been calculated based on alternating the Kona and ANT standards, i.e. an enriched and a depleted isotope value from which percentage memory influences are calculated. This is correct, but this leads to different times for reaching the instrument noise level as the differences among the samples will be significantly smaller for all isotopes. Therefore, an ordering according to their pre-measured or estimated sample values could most probably lead to a significant reduction of the overall time investment without losing precision. See also a minor comment to Fig. 6 below.
- We addressed this partially in our open discussion response. We agree that, if injection amounts could be changed dynamically, then isotopic ordering could be used adjust the number of injections to save time needed to achieve a targeted level of performance. Practically speaking, even if we could vary injection numbers on a vial-to-vial basis, we would need to empirically determine the injection count needed to achieve some threshold of ‘memory-freeness’ for a given vial-to-vial isotopic difference. The main concern here compared to our current approach is that our standards currently have the same number of injections as our samples. This brings us slightly closer to an ideal Identical Treatment scenario, which we prefer over the idea you propose.
I don't quite understand why they used a particular standard for drift correction. Would it be possible to use one of the other four standards you have in use?
- This is addressed in the open discussion. We’ve added some context in the main text (section 2.2).
Minor points:
P5 Table 1 legend: something is missing….by this….
- Correct, that text got cut off! Thanks!
P6 Is there a reason why the vial position is different for the standards but not for the MCM QAQC sample (vial position 1 for both the first and last measurements)?
- The standards are all in separate vials because they begin to evaporate and isotopically fractionate after the septum has been pierced. We are not concerned with the isotopic integrity of the MCM QAQC sample, but rather are interested in whether or not it has highly elevated D17O values due to MCM cartridge failure.
PXX the formatting of d18O and d17O is not consistent, check the complete text.
- We think we’ve caught all of these throughout the manuscript.
P15, l324ff You have measured an unknown sample several times distributed over several series and months. Have you measured the sample water in the same vial? Have you recapped the vial after each measurement sequence? Or have you measured the same water sample by decanting it into individual vials every time you measured it? The procedure is not clear to me it but should be mentioned.
- We added text in that section to clarify our method. Samples were stored at 4 °C in 4 mL glass vials with polyethylene ‘PolyCone’ caps wrapped in Parafilm. During measurement, a sample was opened and a 200 μL aliquot was transferred to a measurement vial. Storage vials were then recapped, fresh Parafilm applied, and stored again at 4 °C.
Fig. 6 I guess that the values shown depending on the previous sample or standard value measured or is this no issue? The mean absolute error may indicate that it is dependent on the various sample. The accepted value of the current sample is reached either from below or above this value. Therefore, one would have to indicate based on which jump (current to previous sample or standard) these values shown in Fig. 6 are based. The absolute error then depends on this jump. This should be mentioned.
- Figure 6 is an analysis that converts all the final, fully-calibrated standard values into absolute errors from their true value. The analysis assumes that the behavior of each standard is similar. This appears to be at least roughly true given the similar performance of each standard when summarized separately (Table 3). Of course, a possibly bigger assumption with this figure is that sample performance is similar to standard performance. We strive towards the Identical Treatment principle with our standards to ensure as best as possible that they are a faithful indicator of instrument error and, thus, sample performance.
Citation: https://doi.org/10.5194/amt-2022-230-AC4
-
AC2: 'Reply on RC2', Jack Hutchings, 10 Nov 2022