Please see attachment for full review.

The manuscript presents a comprehensive study of INP concentration and size in soil and streams around Zackenberg, eastern Greenland. Further investigations point out members of the soil microbial community potentially having formed these INPs. Overall, Jensen et al. have managed to analyse and interpret their diverse results in a coherent, logically consistent way. I agree with Anonymous Referee #1 that the manuscript is interesting and should be published. In addition to their detailed review, I have two thoughts the authors may consider when revising their manuscript.

1.) When looking at Figures 1 and 2 placed next to each other, I get the impression that shorter streams on steep terrain tend to carry lower concentrations of INPs (e.g.: West 4, West 5) as compared with longer streams (e.g.: West 1, West 3). Longer streams also tend to have sections on less steep terrain, where drainage water likely percolates slowly through soil and, therefore, has time to accumulate INPs. For proper quantitative analysis one would have to estimate the time snowmelt or rain water has spent in soil before entering a stream. Of course, such an estimate is well beyond the scope of this already comprehensive study. More easily, the length or average slope of streams could be derived from Figure 1 and used to put this idea to the test.  

2.) Although there is a general association of stream INP concentration with soil INP concentration, there is a striking difference in INP spectra between soil (Figure 2) and stream water (Figure 6). Whereas the former spectra are shallow above -10°C and mostly extend to above -5°C, the latter spectra show a steep decrease above -10°C and none extends to above -5°C. In other words, the most efficient INPs do not seem to be transferred from soil to stream water. One explanation could be that such INPs are too large to pass with draining water through the soil matrix. Another, that they lose their efficiency quickly after having been produced, more quickly than they are transferred to the stream.

Lines 483 and 493: Instead of "p  > 0.05" I would prefer to the exact p-value (e.g., p = 0.08).

Jensen et al. present a novel and interesting study of soils and streamwater from northeast Greenland and probe the biological composition of their samples using filtration and a number of DNA analysis techniques, finding correlations between bacterial and primarily fungal species with INP concentrations. The manuscript is well written and the work is generally very thorough, while the results add to our knowledge of INPs in the environment and open up new avenues of exploration. This manuscript is suitable for publication pending a handful of major questions and comments below.

Major comments:

1)    Line 253 (and throughout the paper when referring to onset temperatures): “Our results showed higher onset temperatures (between -1.5 °C and -4.7 °C) compared to previous studies of Arctic soils (Fig. 3)” and the sentences thereafter – This could be a result of using larger droplet volumes (30 ul) in the droplet freezing assays compared to the literature, allowing the rarer particles to be detected (i.e. better sensitivity) rather than the soils being more active here than elsewhere. This needs to be discussed, at least as a caveat. Generally speaking, the use of onset temperatures or T50 values can only be compared in “like-for-like” experiments, and are not necessarily suitable for literature comparisons.

2)    Figure 2 and 6: Some of the data goes below -25oC, but it is not written or shown anywhere in the paper or Supporting Information what the freezing temperatures of the pure water controls were. In what range do the pure water droplets freeze and does it overlap with the sample data? It would be helpful to include the pure water data in the fraction frozen plots in the Supporting Information. If there is overlap between the control and sample data then background-corrections may need to be applied, as per Vali 2019 (https://amt.copernicus.org/articles/12/1219/2019/).

3)    The streams are all defined as freshwater, but was there any formal analysis of their salinity, even if low? Did salinity levels vary at all across the streams (some samples appear to be from near the coastline) and could this be reflected in the INP spectra, e.g. do the INP concentrations decrease with salinity?

4)    Supplementary figure 8 should ideally be in the main paper, particularly being that it is the equivalent of Figure 5 for the stream samples. Most of the factors shown in the figure are not discussed in the main paper but should also be mentioned.

5)    I was expecting more of a clear discussion about the links or their absence between the soil and water studies, for example what portion of microbial species were found in the soils that then also appeared in the water samples, and how did their relative amounts change, particularly for those that were ice nucleating. A weak positive correlation of INP concentration is mentioned in part 3.5, but there is not much discussion about the nature of the INPs between the two samples (unless I have missed it). This is a little lacking considering the abstract and introduction point to this link, e.g. “In addition, the transfer of bioINPs from soils into freshwater and marine systems has not been quantified. This study aimed at addressing these open questions…”. Perhaps there is a reason why it may not be suitable to discuss, but it is not clear, and so if possible I would like to see at least some discussion of the potential links between the soils and streams, or otherwise make clear that this is not one of the points of the manuscript.

6)    The plots should have error bars where possible, for example the INP plots.

Minor comments:

1)    Line 304: “The gradual loss of INA during filtration at the different locations suggests a mixture of different-sized INPs, predominantly originating from fungi.” – What would suggest that they cannot originate from bacteria?

2)    Was there any consideration of using heat treatments or peroxide treatments of the samples followed by reanalysis of the droplet freezing temperatures? These treatments have high uncertainties in that they do not necessarily “prove” the presence of biological (or entities produced by biological species), but can be a useful indication. On the other hand, DNA analysis allows direct detection of biological species including identification and even quantification, but appears to suffer from other issues, for example PCR would be used for the identification of specific known INP species (but could miss others), while sequencing informs on the identification of populations but not whether they are INPs or produce INpro. While not ideal, heat or peroxide treatments would at least allow an indication of the potential impact of INpro versus the mineral or clay particles that would presumably form the “background” signal.

3)    Lines 420-429: While the soil results were compared to the literature in Figure 3, there is no such figure for the stream water results despite a description of several relevant datasets. While not essential, this would be easier to follow in a visual format rather than trying to compare numbers.

4)    Could the fraction frozen and Nm/Nv data for the filtered samples be included in the Supplementary figures? Only the T50 values are discussed but this does not show whether there were any other influences on the INP populations, for example changes in the shape of the Nm curves upon filtering.

5)    What is the temperature uncertainty of the micro-PINGUIN technique?

6)    Lines 368-370: How does this compare to ice nucleating phyla found in other soil INP studies? Likewise Line 375 for fungi.

7)    Line 126: How many pure water droplets were analysed per experiment? And were the control experiments performed in the same plate as the samples?

8)    Line 152: Why is 5% of the droplets freezing used as the onset freezing value?

9)    Line 174: Add the word “respectively” after discussing bacteria and fungi to make clear that the 16S and ITS sequencing refers to specifically to one or the other.

10)    Line 181: Please define BSA.

11)    Line 184: Missing degree symbols in temperatures.

12)    Line 187: What are V3 and V4?

13)    Lines 189-190: The description of the PCR mix is a little confusing. What were the volumes of the components, and what is meant by “2 template DNA” (e.g. should this be 2 ul?) and “2 x KAPA…..”?

14)    Lines 203-204: Why are the products re-quantified after pooling? Due to losses when transferring between vials? Is the total concentration required for the sequencing?

15)    Line 214: Define ASVs.

16)    Figures 1 and 2: It would help the reader to color code the locations in Figure 1 with the same colors as used in Figure 2, especially when trying to determine whether there are regional grouping of INP concentrations etc. Ideally the same color coding would be used throughout (e.g. in Supplementary Figures 2, 5 and 6, although it is less important for the Supporting Information compared to the main paper).

17)    During the Introduction or Results sections, the authors may want to consider the recent work of Herbert et al. using fertile soil representations of INPs rather than simply desert dust: https://egusphere.copernicus.org/preprints/2024/egusphere-2024-1538/

18)    Line 244: Data for total carbon (TC) is discussed, but this TC data is not shown for the samples (likewise for nitrogen).

19)    Line 251: The sampling of Barry is discussed since they used bulk soil rather than sieved soil, but how was the soil in Tobo and Conen sampled/treated.

20)    Lines 376-377: “While this phylum is known to be encompass many different lifestyles, only saprotrophic, pathogenic and lichenized fungi are known to produce INPs.” – Are there appropriate references that could be used to support this statement?

21)    Line 390: Consider citing Meinander 2022 (https://acp.copernicus.org/articles/22/11889/2022/) and Bullard 2016 (https://pubs.usgs.gov/publication/70190769) when discussing emission of dusts and soils from these high Arctic locations.

22)    Figure 5 and caption: Is the concentration in terms of Nm (g-1 of particles) as in Figure 2? If so, please make this clear and have the formatting of parameters/units across the plots be more consistent.

23)    Line 440: Could biofouling be another possible mechanism for the loss of proteinaceous material via non-specific adsorption to the membrane material?

24)    Line 520: “…while at other locations they are present in solution” – what is meant by this?

25)    Supplementary Figure 1: Is this data for the 63 um sieved samples? Please provide further details about the samples in the caption. Also, provided it does not make the plot too busy, please add the fraction frozen plots for the filtered samples too.

26)    Supplementary Figure 2: It would be helpful to the reader to show more sizes on the x-axis, e.g. steps of 20 um or 50 um. The samples all seem to peak at around the 50-60 um region, please note the number in the caption.

27)    Page 6 of the Supporting Information is blank.

28)    Supplementary Figure 8: Please include in the caption the definitions of the various parameters shown in the plot.

29)    There are occasional minor spelling and grammatical errors throughout that can be picked up on a thorough proofread.