This manuscript presents the development and validation of the GRAnada Ice Nuclei Spectrometer (GRAINS) and uses it to compare different filter substrates and extraction methods for offline INP analysis. The question of how sampling substrates and extraction procedures influence INP quantification is highly relevant to the community. In particular, the assessment of quartz filters is important and potentially valuable, as they are widely used in routine aerosol monitoring networks but rarely applied in INP studies. Identifying and validating new sampling substrates could have far-reaching implications for the INP community, where polycarbonate filters have long been the standard choice. Because of the potential impact, it is essential to examine the results with exceptional care. Overall, the study fits well within the scope of AMT, and it could become suitable for publication after the following comments are addressed.

Major Comments:

A substantial portion of the manuscript describing the instrument characterization and intercomparison is well written and technically solid. My main concerns focus on the key scientific question of the paper. Can quartz filters replace polycarbonate filters for offline INP sampling? Under what conditions would such a substitution be valid? And how should the two quartz-based extraction methods (quartz punch and punch-washed) be interpreted relative to each other? Unfortunately, in its current form, the presented dataset does not convincingly answer these questions. Several interpretations appear speculative or insufficiently supported by data or physical theory. I strongly encourage the authors to revisit these sections, refine the language, introduce clear limitations and uncertainties, and tone down the strength of the conclusions.

A first conceptual issue is that the filter-method evaluation relies almost entirely on ambient aerosol samples. While ambient samples are valuable for method demonstration, a rigorous assessment of sampling and extraction methods should begin with controlled tests using laboratory-generated standard INPs (e.g., mineral dust, biological particles). The suitability of each method depends on INP type and the freezing temperature range, and this dependence cannot be separated using ambient mixtures alone. At minimum, the ambient dataset should span a wider range of atmospheric conditions (e.g., clean days, dust events, heavy pollution, coastal influence). The manuscript should clearly acknowledge these limitations and specify the conditions under which the quartz-filter methods are applicable.

A second major concern relates to the extraction methodology. For a methods paper, key parameters such as droplet volume, extracted filter area, extraction time, and extraction technique require thorough justification. Many immersion-freezing instruments, including CSU-IS and INSEKT, use 50 μL droplets. The authors use 100 μL droplets, but the implications of this choice are not sufficiently discussed. Likewise, the manuscript states that particles were extracted using “manual agitation” for 60 seconds, but the procedure is not described in sufficient detail to evaluate its reproducibility or effectiveness. Such a short extraction time and uncontrollable extraction method is particularly problematic for quartz fiber filters, whose porous structure retains particles much more strongly than smooth PC filters. A clearer description, supporting evidence, and discussion of potential biases are needed.

Related to this, extraction efficiency is a central consideration in comparing sampling substrates. PC filters are widely used because they allow collected particles to be washed off efficiently. Quartz filters, in contrast, consist of a fibrous matrix in which particles can become embedded, resulting in lower extraction efficiency. The manuscript should discuss how this intrinsic structural difference may influence INP recovery, especially for the punch-washed method. In that method, lots of quartz fibers are inevitably transferred into the suspension and can themselves act as INPs at lower temperatures (Conen et al. 2012; Harrison et al., 2019). Their contribution must be carefully considered.

These methodological issues directly affect the interpretation of the key results. For example, below approximately -12 °C, the manuscript reports higher INP concentrations for the two quartz-based methods than for the PC method. The explanation offered in the manuscript remains speculative and lacks supporting evidence. A more plausible interpretation, consistent with Conen et al. (2012) and subsequent studies, is that the quartz filters introduce additional quartz fibers that act as INPs at colder temperatures. This naturally leads to increasing discrepancies among methods as temperature decreases. Indeed, Conen et al. (2012) concluded that quartz punch method is a better choice and should be restricted to temperatures ≥ -12 °C. This important limitation should be explicitly discussed. In this context, it is also worth noting that the quartz 96-punch method is likely the cleanest configuration, with minimal fiber contamination; the very short 60-second agitation step may, however, lead to insufficient particle extraction. This distinction is important because the 96-punch method is the one recommended by Conen et al. (2012) and Wex et al. (2019). Therefore, statements in the manuscript suggesting that the punch-washed method may be preferable are potentially misleading and require substantial reconsideration or additional solid experimental evidence.

In summary, while the instrument development and the general methodological framework are strong, several core conclusions regarding the equivalence and relative performance of the filter methods are currently not yet supported by the available data. I encourage the authors to substantially revise the manuscript, explicitly state the limitations, incorporate a more cautious interpretation of the results, and restrict conclusions to the conditions actually tested.

Specific comments:

1. L42: “ice-nucleating particles (INPs)”

References:

Conen, F.; Henne, S.; Morris, C. E.; Alewell, C., Atmospheric ice nucleators active ≥ −12 °C can be quantified on PM10 filters. Atmospheric Measurement Techniques 2012, 5, (2), 321-327.

Harrison, A. D.; Lever, K.; Sanchez-Marroquin, A.; Holden, M. A.; Whale, T. F.; Tarn, M. D.; McQuaid, J. B.; Murray, B. J., The ice-nucleating ability of quartz immersed in water and its atmospheric importance compared to K-feldspar. Atmospheric Chemistry and Physics 2019, 19, (17), 11343-11361.

Wex, H.; Huang, L.; Zhang, W.; Hung, H.; Traversi, R.; Becagli, S.; Sheesley, R. J.; Moffett, C. E.; Barrett, T. E.; Bossi, R.; Skov, H.; Hünerbein, A.; Lubitz, J.; Löffler, M.; Linke, O.; Hartmann, M.; Herenz, P.; Stratmann, F., Annual variability of ice-nucleating particle concentrations at different Arctic locations. Atmospheric Chemistry and Physics 2019, 19, (7), 5293-5311.

Bazo et al. demonstrate and validate a new immersion freezing analysis platform for the measurement of INPs, the GRAINS, based on the CSU-Ice Spectrometer. The authors performed a comprehensive set of tests against other INP instruments to confirm its comparability and accuracy. Some of these tests also highlighted discrepancies between some instruments for some materials, such as between PINE and immersion freezing instruments for K-feldspar. Importantly, the authors demonstrated that the use of three filter sampling techniques, including traditional polycarbonate washing and quartz punch techniques, yield similar results, although there were some discrepancies at colder temperatures. This work highlights the need to compare different sample collection techniques to ensure consistency in INP measurements. These findings warrant publication in AMT pending the following revisions and comments.

Section 4.1: Often, new instruments are validated using a range of lab prepared samples that cover a wide temperature, e.g. Snomax, NX illite, K-feldspar. I would have expected to see a few examples, particularly given the discrepancies that can be encountered with NX illite, while acknowledging as the authors state that there are no reference standards for immersion freezing and the materials that the community uses all have some variability. This is perhaps more surprising given that multiple other samples (including K-feldspar and ATD) are tested as part of intercomparisons – in principle it should be possible to compare K-feldspar and ATD to literature data, although it may not be quite a 1-to-1 comparison since these were aerosolised rather than being prepared as suspensions directly from lab samples.

Minor comments:

1. Line 28: Were the randomly punched quartz filters also washed?
2. Line 35: Any indication of groupings? Were minerals more affected than biogenic materials (if analyzed) or vice versa?
3. Line 37: Is the quartz 96 punch method a standard method?
4. Line 42: If “ice-nucleating particles” are being discussed in plural, then the abbreviation should be “INPs”.
5. Lines 56-59: CFDCs and PINE are mentioned (I would recommend explicitly mentioning PINE here, particularly since it is used in the study), alongside “INP spectrometers” – a very brief description of these modes of operation should be provided here, particularly of the “spectrometers” and especially since this is not necessarily a term that is used consistently throughout the community.
6. Line 61: They are also require far simpler and much cheaper instrumentation, and often easier to use and easier data to process.
7. Line 64: As in the above comment, there has been no description of how the technique operates in general (i.e. cooling droplets of aqueous suspensions of INPs on a cold stage until they freeze) so this discussion of different droplet volumes does not make much sense.
8. Line 77: For the layperson, it might be confusing that there is not consistency across the instruments being used, e.g. why doesn’t everyone use the same instrument, or why are new but similar instruments still being developed? Within the community, we know that this is because no such “standard” cold stage instrument exists, and so every group must build their own and test it. It would be worth noting this here.
9. Line 86: Are INP spectrometers specifically defined as being those that use PCR plates? As opposed to other cold stage techniques?
10. Line 93: Replace “the first one” with “the polycarbonate filter washing method” or similar.
11. Line 96: What is meant by “a bigger punch”? What size is the punch?
12. Figure 1 caption: Some more description is needed. Why are some of the wells red in the image?
13. Line 109: The citation for Perez Fogwill 2024 is not in the reference list and I cannot immediately find it online.
14. Section 2.1: I am a little confused by the description of the setup. Are the PCR plates partially immersed in the ethanol in the aluminium block? Or is the aluminium block an enclosure for the ethanol but has a shaped top that allows the plates to be inserted? A photograph of the setup in the SI would be incredibly useful.
15. Section 2.2: The authors may want to provide an explanation as to why the grayscale value, after the initial nucleation event, drops and then rises again as this may confuse those unfamiliar with the freezing characteristics of droplets.
16. Section 2.3: The LAUDA chiller is ultimately used as the temperature reference for all experiments, but there is no description of mention of how the temperature of the LAUDA chiller itself is calibrated or checked. If the Pt100 is checked against the LAUDA chiller annually then this suggests the chiller has been in use for years, and therefore may have experienced a drift in its temperature measurements.
17. Line 203: The gap between the aluminium block and the plates is described as being filled with ethanol here, but this is not mentioned during the description of the standard operation of GRAINS, so is not an accurate representation of the temperature of the wells in relation to the aluminium block during a typical droplet freezing assay – I now see that this is addressed at the end of section 2.3, but this part of the procedure (adding ethanol between the plate and the block) should be included in Section 2.1 too since it is part of the standard procedure.
18. Figures S1 and S2: Given the wide spread in temperatures between the wells and blocks depending on position (assuming these are reproducible, I do not see any uncertainty values, presumably for clarity), could the temperature uncertainty be reduced by having bespoke temperature corrections for each well rather than a “blanket” correction?
19. Line 260: There should be a space between the “g” and “L-1” in the units of g L-1. Likewise between the m2 and g-1 in line 261, i.e. m2 g-1. I also see this throughout the manuscript, all units should have spaces between the different components of the units.
20. Line 271: Please note that some surface area will not be accounted for since particles smaller than 0.5 um are not measured.
21. Line 278: Are the original sample and dilutions all performed in one GRAINS experiment by distributing them across the two plates? Or are they run one after another? How many repeats were performed?
22. Line 318: What particular sample of K-feldspar was used, e.g. FS02?
23. Line 347: The collection of field blank filters is excellent practice. Was this also done for the other filter measurements described in the previous sections?
24. Line 365: Clarify that the punched filter parts were added to each well, it is implied but not actually stated.
25. Lines 368-370: Can the authors clarify why this method was employed? Is it a standard method used in the literature, in which case can citations be added (as for the other filter techniques)? For a more direct comparison to polycarbonate filters, could the quartz filters also be washed as a full filter (i.e. without punching)? Or were the 1 cm punches taken from the same filter as the other 1 mm punches?
26. Line 373: “where 𝑉punch is the volume of air that passed through one punch of the filter” – is this calculated as a fraction of the total amount of air that passed through the full filter? In which case, does this also take into account the “unused” portion of a filter where the O-ring is seated?
27. When analysing multiple 1 mm punches with GRAINS, are the freezing patterns consistent with homogeneous dispersion of aerosols across the filter surface? Or can discrepancies be seen, e.g. stronger signal in punches taken from the centre of the filter compared to nearer the edge?
28. Line 390: It is mentioned in a comment above, but without a measurement of the population of smaller particles, e.g. with an SMPS, how confident are the authors of their total surface area measurements.
29. Figure 3: The authors might also consider including the WSIDOM data from Reicher et al. 2018: https://amt.copernicus.org/articles/11/233/2018/. Further, it may be useful to show the droplet volume for each technique in the legend since GRAINS uses a very large volume compared to most other immersion freezing techniques shown here.
30. Figure 3 caption: Mention that A13 is for K-feldspar. Also, it is not entirely clear in the text why A13 is shown; is the scaling factor applied here for A13 related to the amount of K-feldspar found in NX illite? Is there a reference for this?
31. Figure 3 caption: The authors state that uncertainties are as per Agresti and Coull (1998), but there is no indication of what these uncertainties are or how they are calculated. Further, there is no citation for Agresti and Coull in the reference list.
32. Figure 4: The authors should comment on some apparent systematic biases since the comparisons to FrESH are above the 1:1 line while those with INSEKT are below the 1:1 line.
33. Section 4.2: This might work better divided into sub-sections for each of the sets of intercomparisons.
34. Figure 5: The large discrepancy between GRAINS/INSEKT and PINE is concerning. Are there any previous studies of K-feldspar with PINE during other intercomparisons to check whether this is a common problem? That GRAINS and INSEKT compare well is good.
35. Section 4.2: Given that there is a lot of literature data available for K-feldspar, I am surprised to see no literature comparisons here. Can the authors provide some comparisons to the literature data? This would actually fit better with Section 4.1. The same could also be done with ATD.
36. Section 4.3: Can the authors explain why different cut-offs were used for the filter samplers during this comparison? That they are not 1:1 suggests there could be differences in results, although this may only affect the “tail ends” of the data rather than the “main body” of the INP curves.
37. Figure 6 and lines 553-555: The authors use standard error of the mean here, whereas throughout the rest of the paper they calculate the error based on Agresti and Coull (which is never explained). There should be consistency in treatment of the uncertainties throughout the manuscript and dataset.
38. Lines 571-572: Does Figure 6 shows the average of all 27 samples? Or an individual sample? I am very surprised to not see the data for the 27 samples in the SI, I feel that should be included.
39. Section 4.3: An issue with this set of experiments is that the aerosol composition is unknown since they are ambient samples. Therefore, it is not possible to say whether differences in INP concentrations at temperatures colder than around -13 oC could be due to the differences in collection efficiency of different types of INPs across the sampling techniques. This comparison would have been better incorporated into controlled lab studies such as those in Section 4.2.
40. Line 641: “detail” rather than “detailed”.
41. Line 642: This begs the question of why this wasn’t done as part of the AIDAd intercomparison described earlier.
42. Line 644-645: Rather than punching a 1 cm hole, could the entire filter not be washed as in the polycarbonate method, thus reducing preparation further? Or is the rest of the quartz filter used for further analyses?
43. Line 675: The authors should reiterate here that there is a discrepancy with K-feldspar between the PINE and the cold stage instruments that warrants further study.
44. What is the nominal pore size of the quartz filters? Could their capture efficiency of smaller particles be greater than for the polycarbonate filters, which could potentially help to explain some of the discrepancies are colder temperatures?
45. Line 706: Do the authors mean “immersing” or “washing” rather than “dissolving”?
46. Lines 707-710: This is an excellent point and should be reiterated in the discussion of Section 4.3.