We deeply thank reviewer 1 for his/her supportive comments. We are glad they appreciated the manuscript.

We deeply thank reviewer 2 for his/her thorough reading and the detailed comments they provided. We are glad they found the paper of interest and hope we will satisfactorily address all the questions raised.

1. Comment 1 on Figure 1

This Figure is central and can help one reader to properly understand the method. We therefore very welcome all the comments and edits the reviewer provided.

The quality of the figure has been approved in line with recommendations.

HN₃ now replaces NaN₃

m/z ratios were added as well as N₂

The subfigure ‘size correction and calibration’ is the same as figure 5. We acknowledge that no clear legend is present on the figure. We made the following changes:

• Added legend
• Changed the colours to make it colour-blind friendly
• Wrote δ¹⁵N(N₂) instead of δ¹⁵N(NH₄⁺)

We also make this last change for Figure 5

2. Lines 117-118

The reviewer is absolutely right in the sense that Kaiser et al., 2007 present an online method to measure the isotopic composition of NO₃^- thanks to the conversion into N₂O and then into N₂. We note that our sentence was probably misleading and change it into:

“After NH₄⁺ is converted into N₂O, the latter follows an online method (Kaiser et al. 2007) to be reduced into N₂ whose δ¹⁵N isotopic composition is measured with a Finnigan MAT-253 IRMS.”

3. Lines 135-136

Thank you for this. We correct the sentence to consider this point:

“The overall system is suitable for ice core measurements with a limit of detection of (0.22 ± 0.03) ppb, a linear calibration down to 1 ppm, and a repeatability <0.7 %.”

4. Lines 144-145

We add these much needed details

“NH₄⁺ concentrations for these contamination tests were measured through ion chromatography device (IC) device (ThermoFisher INTEGRION), connected to an autosampler (ThermoFisher AS/AP), with separation on CG16-4 µm and CS16-4 µm 2 mm columns. The mobile phase is an isocratic manually mixed MSA 29 mM solution for the cations, applying a 0.16 mL min^-1 flow rate resulting in a 27-minute separation run with autoregenerated suppression and conductivity detection.”

5. Lines 148-149

We thank the author for the pertinent question, and we have clarified the original sentence from:

“To investigate this route of contamination, we created artificial ice sticks made from MQ-pure water and replicated our typical ice core cutting and handling process.”

to

“To investigate this route of contamination, we created artificial ice sticks made from cutting a block of frozen MQ-pure water and replicated our typical ice core cutting and handling process.”

6. Lines 180-181

Thank you for these questions to which we initially have also considered. The capacity of the resin is 2.1 mEq mL^-1. Since our method focuses on water and ice samples from environments with low ion concentrations, this volume of resin will always surpass its saturation by other cations. Additionally, we remind that the selectivity of NH₄⁺ to the resin is higher compared to all cations except H⁺, Li⁺, and Na⁺. Therefore, as our results show (Lines 280-285) the volume does not influence the isotopic composition. However, we would temper this statement by pointing out that these tests were carried out in environments with low concentrations (ppb range) and the question should therefore be studied for more concentrated environments (seawater, for example).

7. Lines 204-207

We understand the necessity for further details about the chemical conversion. We will therefore modify our paragraph accordingly:

“The measurement of δ¹⁵N(NH₄⁺) is performed according to the methods described by Zhang et al. (2007) and McIlvin & Altabet (2005). In details, 0.4 mL of hypobromite solution is added to the NH₄⁺-containing vial closed with septum and the reactions is run for 30 min while agitating. The hypobromite solution is prepared in a container rinsed with MQ water, into which 20 mL of MQ water is added, followed by 1 mL of a bromate/bromide solution (0.6 g of NaBrO₃ + 5 g of NaBr in 250 mL of MQ water), and then 1.05 mL of 6 M hydrochloric acid immediately after which the reaction is left for 5 minutes in the dark before the final addition of 20 mL of 10 M sodium hydroxide solution. Then, the excess hypobromite from its reaction with NH₄⁺ is quenched adding 0.1 mL of an arsenide solution (5.1 g of NaAsO₂ in 100 mL MQ-water). The solution is let to stir for 1 minute. In the same vial, 1 mL of azide/acetic acid solution (1:1 v/v) previously flushed with Helium is added and the reduction is run for 30 min before the final addition of 1 mL NaOH 10 M to quench the excessing reagent.”

Regarding the question whether the oxidation of NH₄⁺ is complete, following Zhang et al (2007) results, a 30-minute reaction was reported to be optimal for MQ-water and seawater samples, which we assume the former to be close to ice core matrix.

As wisely suggested by the reviewer, the size difference between what is measured with the IRMS and what was expected from sample preparation is a way to flag suspicious samples. We note a 6 % difference between these two sizes (expected size vs IRMS size) for the tests we conducted to develop and validate the method, suggesting our method to be reliable. When it comes to the ice core example (section 4.5), the difference is larger between the size calculated from the ice core and what is measured with the IRMS (26 %). This difference is due to the calculation of the sample size (n(NH₄⁺) = Concentration × Section × Length × density / Molar mass). Indeed, the fact to consider ice cores as perfect cylinders does not take into account the imperfections of the ice cores. Therefore a 7 % difference exists between the estimated density and the modelled density. This 7 % difference results in a 28 % difference for the sample sizes, in agreement with the 26 % difference. The difference between the IRMS and the CFA measurements is solely explained by calculation assumptions of the sample size. This also stresses the necessity to have a size correction model.

Because of the previously stated reasons, the nitrogen isotopic peak size is not an ideal precise marker of suspicious samples. However, we instead chose to flag samples whose sizes were extremely different (>2σ) from the mean peak size and followed up with examining the N/O peak ratio (since the oxygen isotopic composition is also measured for N₂O), which is seen as a good marker of the accuracy of the measurement.

8. Lines 309-310

We thank the reviewer for proposing an explanation of the large blank size. We try here to list the reasons why this step is unlikely to contribute to the blank:

• The hypobromite solution could be a source of contamination if in contact with atmospheric NH₃. This is why we prepare the hypobromite solution under the NH₃-scavenging fume hood.
• NO₂^- present in the analyte can alter the isotopic composition as suggested in Zhang et al 2007 which we assume the reviewer is referring to. Since we perform a cationic solid phase extraction, all anions present in the analyte (including NO₂^-) are removed and should therefore not alter the isotopic composition

We thank the reviewer for highlighting the necessity to clarify the chemical conversion protocol and we hope that the answer to the previous question also help to address his/her question whether the reaction is run in single vial or separate vials.

9. Lines 475-478

Absolutely! Thank you for pointing this out. It will be corrected.