| 06 Nov 2025
Correction and Calibration Protocol for Isotope Data via CRDS: A Study Case for N2O and Other Isotope Systems
Abstract. Advances in laser spectroscopy have significantly simplified the measurement of N2O isotopologues (14N15N16O, 15N14N16O, 14N14N18O), but the raw data require extensive post-processing. This problem arises from the complexity of spectral fitting, which is controlled by an intricate interplay between the physics of vibrational spectroscopy, gas composition, fitting algorithm, and instrumental parameters. Following the general principles of identical treatment, the highest precision and accuracy is achieved when reference gases mimic the sample composition, which underpins our correction and calibration protocol.
This study presents a comprehensive and detailed correction and calibration protocol to post-process N2O isotopic data, exemplified by data obtained from three commercial cavity ring-down spectroscopy (CRDS) analysers (G5131-i, Picarro Inc.). Experimental correction functions for delta values on changes in N2O, CH4, CO2 and O2 concentrations were determined for individual analysers to derive a mathematical framework, which was verified with spectral simulations. We confirm that the apparent δ-values scale inversely with the N2O concentration, with the slope being analyser-specific and highly variable over short time intervals. Consequently, any instrument must be routinely characterised to maintain high-quality data. Furthermore, when CH4 and CO2 concentrations vary simultaneously, their combined spectral interference displays a non-additive interaction. We strongly advise removing CO2 from the sample gas before analysis to ensure optimal data quality unless CH4 / CO2 variations are very small such as for N2O emissions from upland soils).
We provide an end-to-end, stand-alone MATLAB application with a user-friendly interface for standardised data reduction, which was validated by analysis of several known target gases but with different gas compositions. This protocol/MATLAB application aims to support researchers in efficiently obtaining high-quality and reliable N2O isotope data from the tested CRDS analyser model, while also providing a study case for data correction for other analyser models and detection schemes. Therefore, the code can be readily adapted to any isotope system for routine application.
Juliues et.al. have provided an excellent analysis of the performance of the N2O isotope analyzer, Picarro G5131-I, and the methods needed to accurately measure isotope data for N2O at ambient levels. They correctly account for some of the more common use-cases for this type of analysis and the associated challenges for the relevant environments, e.g. changes in N2O, O2 and CH4. The paper is expanding on the instrument analysis done in the mentioned paper Harris et.al 2020, showing that long-term drift and performance of the instrument necessitate a diligent calibration protocol.
Finally they provide access to a free MATLAB code written to incorporate the found corrections and guide users on what variables to measure and determine for their instruments before they can account for the instrument and matrix effects. This is a valuable addition to the community as it allows for standardization within the field and will help informing people of the potential pitfalls using the instrument-reported values uncritically.
I support the publication of this paper, though I have comments that I should like addressed to improve clarity and impact of the paper.
General comments:
Regarding the determining of the effect of changes in N2O concentration, I am rather surprised at the scale and direction of these changes in comparison to the calculated values. It stands in contrast to the levels previously measured by CRDS-III (as shown in table 4), which while not correct are on the same order of the calculated offset. What is your best hypothesis for why the three instruments behave so differently, and have discussed potential explanations with the manufacturer? Because unless there has been relevant software updates for CRDS-III since Harris. 2020, I am under the assumption that the built-in post-correction for non-linearity should be close to the same, and if not it would appear as if the updated correction performs worse?
A final question in relation to this topic is; whether you also tried diluting below the reported detection limit, because the non-linear behavior of the calculated values first appear pronounced at 1/N2O >0.003.
While access to MATLAB is common within the field, it would have been great and more inclusive if the developed MATLAB code functions were available in a free software such as R or python. I understand that it is quite the endeavor to change programing language, but I would be hopeful if it is something that you would consider within outlook and future work. Because as is, I unfortunately cannot test the code as I would otherwise have liked to do.
The uncertainty calculations in the appendix appear to be correct, so I believe it would improve the impact of the paper if you could provide an example of the scale of error one could expect. A suggestion would be correcting for the matrix effect of Cal 2.1 330ppb, where the value and error could nicely be compared to the known value?
Supplement comments:
In the tables S8-10 you have calculated the mixing ratio of oxygen, though for table S9 and S10 I believe that you have incorrectly copied the concentration from CO2 from the previous set of tables. A further correction to the correct oxygen values of table S8, is that the matrix of the Cal1 and Cal2 include oxygen, which means that you have a base concentration of oxygen at about 770 ppm rather than 0 for N2O 0.33 ppm and double and triple that for 0.66 and 0.99 ppm N2O. I understand that those can be considered negligible levels for the relevant purpose, but I would still prefer the correct values.
I appreciate the inclusion of experiment 6 and find it to be a worthwhile aspect to investigate, but I find the description of it difficult to follow. In part because the text refers to gas mixes not shown in the figure S6 or described in table S13, leaving me unsure how they are introduced and diluted. Additionally the many gasses with very similar names led me to mix up which gas was being discussed. I would suggest perhaps expanding on table S13 to show the dilution scheme for all discussed dilutions. And maybe update the figure to better indicate channel 1 and 2.
Minor comments:
Line 165: You have written 14N16O17O and 14N16O18O, where I believe you meant to write 14N14N 17O and 14N14N 18O
Line 462: There is a misplaced 6 following the (990 ppb)
Line 523: I harbor suspicions of the non-linearity measurement of the 10/4-2025, as it shows a significantly different behavior for both CRDS II & III.
Line 574: Did you conduct validation measurements for the inclusion of CO2 removal to compare the found values to your validation experiment?
Minor supplement comments:
Line 17: Missing lower case 4 for methane.
Table S1: You have in the accompanying text written that the change was in the range of 0.33-1.20 ppm, but the lowest concentration I see in the figure is 0.35 ppm.