In this manuscript, the authors have addressed an important question, effectively how glacier fed streams change longitudinally, which I fully believe is in the interest of the Biogeosciences readership. The data themselves are very interesting (with some nice longitudinal patterns), as the analytes chosen are good proxies for glacier processes and for relevant biogeochemical cycles. The paper is overall also quite well written, and the figures well illustrated. However, in carefully reading through the methods and results sections, I have three major comments: 1) The statistical analysis can be improved. The issue to me is that there are three streams, sampled at different points in the hydrograph (premelt, melt, postmelt), at 3-4 different transects, over three years. This is already a lot of categories for the dataset size which makes interpretation tough. Also problematic is that the categories are also not neatly defined: the ‘longitudinal’ categories encompass quite some different distances, and there is uneven replication and temporal representation (mostly in melt versus pre- or post- melt, the distinction between each seems arbitrary). Another issue is it seems that the data for the three streams are lumped together in the analyses. Can this really be justified…..wouldn’t it be better to analyze each glacier stream individually? As a result, I think that there could be quite a problem with spatial and temporal autocorrelation, and I think data independence is an assumption with the utilized analyses. I have read through this paper several times now, and I must admit, I don’t have an obviously better solution given the sampling design, but one possibility could perhaps be some mixed models (ex GLM or GAMM etc) that account for the different categories, as well as the possibility of making some of the categorical variables continuous (i.e. distance from the source glaciers) to improve subsequent interpretations. 2) Another general issue is the justification for why these things are done in the first place….what do the authors hypothesize, particularly with regard to each of these tested categories? The stated rationale of seeing how they change with distance downstream can be much elaborated upon. 3) Finally, I was a bit disappointed in the lack of information regarding the microbiological analyses and results (in more detail below), which are really necessary to assess the quality of the data and interpret the results. Therefore, I challenge the authors address these three major points in their revision. Some more detailed points follow by line number.

Line 13: be careful here…..not sure that this OM is necessarily structuring water column communities if they were just exported there, and especially given the result that mass effects seems more prevalent than environmental filtering

line 18: microbial communities inhabiting what? There are a lot of habitats in streams, and I think this detail would be of interest to those reading the abstract. As I have already read the paper, I feel like these are probably not communities as all, but microbial assemblages in transport. I think this should be clearly stated here, and I make further remarks on this below

line 21: although I never really read any names of the putatively chemolithoautrophs or cold adapted taxa….mostly just phyla and other high-order taxonomic names

line 22: while I don’t dispute the role of glaciers in ‘seeding’ headwaters, im not sure the data really show this…..you don’t have glacier endmembers, and you cannot prove that the microbes were actually alive or living there, so there is not really a 'smoking gun'….just some precision may be needed in the language

line 23: probably all could be argued to be indicators of water source

line 29: ‘complex ecosystem responses’ sounds pretty vague….could you make any more specific predictions?

Line 43: ‘supraglacial’ and ‘subglacial’ would be related to glacier tops and bottoms, respectively. Marginal channels are not found on top or beneath the glacier (unless I am misunderstanding something).

Line 49: This process also helps to explain a lot of the differences in your analyzed variables, which should probably be discussed to justify their collection

line 56: the POC is from crushed rock? Couldn’t it also be from the overidden vegetation?

Line 64: agreed that these are important variables for benthic communities, but do you think the same is true for the organisms in the water column?

Line 73: this seems to be a specific number…...from my experience, glacier surfaces are remarkably heterogeneous, which may be good to note.

line 78: although there is only microbial data for the headwaters and far reaches, and none from Bow

line 81: how specifically can this work inform us about how glacier loss will impact microbial diversity?

line 84: this is repeated from the first paragraph, line 34 I think….

Line 90 to line 93: While this doesn’t bother me, it seems a bit strange to give the results of the paper in this last paragraph…..this could potentially be skipped. What would be nice to include, however, would be some specific hypotheses that could be tested. What exactly did you predict would happen to the OM and why? How should microbial communities change with space/time and why?

General comment for introduction: The introduction is written nicely, but many of the papers used as citations are from work on the Greenland Ice Sheet, which is really an enormous piece of ice with its own very special characteristics. I know nothing about the glaciers which were the basis for this study, but I encourage the authors to evaluate whether or not these citation are applicable in the context of their study system, and if some papers dealing with smaller glacier might be more relevant to cite in some cases.

Line 96: I see that there is a hydrograph for one of the streams, but how do the discharges of the other two streams compare with this one? Could be much different given that different size of the icefields mentioned here, yet there is no way to tell. This is an important consideration given that all three rivers were lumped together in the analyses, though my feeling is that the effect of ‘stream identity’ should be taken into consideration.

Line 103: be careful with the word “evolve”, since it has quite a specific meaning in the biological sciences

Line 105: These glacier distance binnings seem arbitrary to me. Is there some ecological reason that we could expect changes with distances of these magnitudes, or are they indeed arbitrary? A concern for me is that the difference in the far sites (100-40=60km) is actually greater than the distance of the first three other categories (0 to 35 km), as well as 1/3 of the ‘far’ category itself. I guess my question is if these categories make ecological sense given this, or would it make more sense to use distances from the glacier as a continuous variable, in which case the ‘realistic nature’ of these categories would no longer matter?

Line 108: I think “stream” works, but its then a bit weird then that you used ‘river’ in the title

Line 113: how many of these samples were taken in December and January? There are not a whole lot of winter glacier fed stream data, so these could be of extra interest…. Where they different than the others? Unfortunately they are buried within the rest of the other data and its not possible to see them. They also don’t seem plotted on the hydrographs. Thus, despite reporting them here, they seem invisible in the paper.

Line 115: though the sample coverage over these three periods is very uneven, and most samples are from the melt period. Should put some sample numbers here to give the reader an idea of how many samples were collected at each period. Also, do you have a feel for how well the Athabasca Glacier hydrograph corresponds to the hydrographs of the other two streams?

Line 118: instead of arbitrarily using 1 m3, couldn’t you use your hydrological/chemical data you collected to determine if the subglacial channels have opened or not? By using arbitrary cutoffs as you have here, there is a good chance that you will get results from your analyses without the stated ecological relevance. Also, what kinds of differences might you expect to appear in your data as a function of season? These hypotheses are not clearly identified….

Line 125: thus sample year should probably also be accounted for in an eventual model

Line 126: would be nice to know how many samples were collected from each of these categories. I know that some of this information is in Table S1 (i.e. sites as a function of distance), but it would be really good to have some of these numbers here too. From looking at the plots, its seems that the pre-melt and post-melt periods are vastly undersampled compared to melt periods, yet there is no way to really know this except by looking at figures.

Line 133: should probably also put the charges on the trace metals since there are charges on the other ions (also on line 200). Also, why exactly were these measured? There is no hypothesis stated, and the data arent really discussed anywhere else? Same could be said for the major ions for that matter.

Line 134: I think its common to acid wash bottles for nutrients too, although I realize this wasnt the focus on the work, and the difference in results likely not large. What does the citranox do?

Line 150: how much water did you filter for the microbes, and what exactly does ‘prepared’ mean in terms of the bottle?

Line 200: calling dSi a nutrient is contentious. Also, what were the limits of detection for your nutrient analyses?

Line 204: could give the nutrient methods numbers here as well

Line 215: do you have a citation for the modifications of the protocol? Or do you have some data on what kind of improvement you can expect? Just that this might be of interest to others doing this kind of extractions with these kits

line 233 and elsewhere: I think that there should be quite a lot more information on the microbial data. For example, how much water was possible to pass through the sterivex in the end? How much DNA were you able to get from the filters following extraction? What kind of depth did you sequence to/how many reads per sample? Were they rarefied for the analyses? I think these are pretty important things to report given that these are often difficult habitat types to work with…..

line 235: what do you mean by “low abundance ASVs”? Like….you removed rare ASVs? At what threshold….why??

line 242: not a big deal, but seems that the DOM ab/fluor should go with the DOC methods above rather than the microbial data

line 255: why did you choose three way ANOVA rather than keeping these as continuous variables? If you made mixed models or similar, you could control for the effects of stream, season, etc, while also using the data directly as continuous variables rather than making arbitrary categories for downstream distance, discharge, etc

line 260: though I think deuterium can also differ from glacier to glacier…..would be possible that these numbers different between the headwaters of the three streams. Also, why plot this as a passive overlay on the ordination? Should explain this….

Line 265: I think it should still be possible to compare and contrast headwaters and downstream samples with the middle sites included…..not sure I understand this justification….I would probably just include all of the data, no?

Line 266: maybe I am old fashioned, but I like to know the actual number of ASVs rather than using shannon for alpha diversity

line 268: is it necessary to square root transform AND calculate bray curtis distances?

Line 269: so you identified clusters on the figure and then tested the clusters with permanova? That sounds a bit self-fulfilling to me…..wouldnt it be better to test hypotheses using the data? Also, there are likely far too many parameters that were included in the dbRDA, and its not clear why most of them were included. Perhaps you can provide some justification for why some of these are here? For example the trace metals? How many factors were left after the highly correlated ones were omitted?

Line 278: Should probably correct for multiple testing, no? Also, what exactly will testing co-variation between indicator species and environmental parameters tell you? My feeling is that indicators of upstream will just be correlated with things more likely to be characteristic of upstream sites, like low temperature for example

line 280: well…..its the R statistical environment, which uses the R programming language

Line 287: what kind of test does this p value correspond with? Also, there is likely to be big differences in sample sizes between categories…..were there differences between rivers, and if so, can you justify lumping the data together?

Line 294: although there wasnt so much pre-melt data for the other years…..

line 297: ‘largely non-significant trend’…..probably it was just not significant

line 319: could be due to the high/low points on the hydrograph during the meltseason, which could be related to greater sub/supraglacial contributions

Line 338: These phyla are literally everywhere, thus making this sentence not very informative. Not that it is necessarily bad to mention these, but I would focus on lower taxonomic levels….

Line 343: How was the core defined? There are many many ways to do this, and it would be good to know what were your assumptions….probably in the methods. Also, there are two interesting things that come up with this: First, you call it a community, but im not sure this is an accurate term to apply to these microbes, given that almost all of them are probably being passively transported, and therefore not ‘residents’. Also, why do you think there should be a core…..what are your hypotheses here….how is this core maintained? In any case, a core of 1,409 ASVs tells us very little because we have no idea who they were, how many reads were generated per sample, and what the ASV richness looked like per sample.

Line 344: the 10 most abundant taxa belonged to 8 different families? This seems improbable. How are you defining taxa here?

Line 348: I would still like to see the number of ASVs

Line 350: This is really only reflecting differences between the two rivers, since Bow was excluded. However, if rivers are distinct from each other, and differ by year, should samples be merged? My feeling is that this would be a major reason that the separate effects of individual streams and years needs to be accounted for.

Line 354: To see if environmental variables explain variability is poor justification for conducting an analysis. Please expand on this. Also, how was the % variance adjusted?

line 375: I would rather see hypotheses led comparisons rather than throwing everything at it and seeing what sticks...it really makes a difference for the reader in terms of focusing on particular results. Also, Im not sure that these are 100% independent datapoints (they have spatial and temporal autocorrelation), yet there comparisons are assuming that they are. I think this really needs to be justified that it is the best approach to show what you want to show. My feeling is that many of the points are already obvious (e.g. headwater indicators being related to temperature. Etc)

line 377: can you say Microbacteriaceae sp? (like family sp.?) To be honest Ive never seen it before, but might be more clear to say its a species from the family Microbacteriaceae

line 382: speaking of autotrophs, were you able to quantify stream turbidity or otherwise estimate algal biomass? Could help with some of the interpretations….

Line 426: why is this paradoxical?

Line 430: I interpreted this slightly differently…..early in the season, the subglacial flowpaths are inefficient and poorly formed, whereas in the peak melt there is an efficient subglacial channel where most of the supraglacial melt is routed….. I think this needs to just be re-worded to make it more precise…..also, the lack of variability in this study might be due to an inadequate sample size from baseflow conditions?

line 436: Yes, I am wondering how relevant some of these papers on Greenland are for these smaller glaciers…..not that they are bad to cite, but maybe just good to include some smaller ones as well

line 468: although there are a lot of other photoautotrophs in the streams besides cyanobacteria, such as Hydrurus, that wouldnt show up in the 16S

line 490: I think priming in streams is a bit of a debate in general

Line 504: Indeed many of these are found in glacier ecosystems elsewhere, but again phyla are very broad categories. Better would be to identify groups at lower taxonomic levels where more relevant comparison can be made.

512: Do you think that increases in alpha diversity are a response to changing environmental conditions or reflecting a greater number of cell sources? Since the mixing and residence time of streams is relatively short, my guess would be that these are not actually communities but assemblages. It is unlikely from my opinion that they are responding to their environment per se, but more that the assemblage is being formed by the inputs from the surround adjacent landscape. I think the Wilhelm paper was primarily about benthic communities.

Line 522: what are the various lines of evidence? I think you should specifically give the argument….if these are the results from the RDA, keep in mind that this is correlative and not necessarily indicating causation….in this case that organisms prefer or are repelled by a given parameter

Line 552: I know that it can include any rank, but when I think of the word ‘taxa’, I am generally thinking of something at the species or ASV level. To me its really weird to refer to these much higher order taxonomies as ‘taxa’ when you could instead say ‘phyla’ ‘orders’ or ‘classes’. I also wonder how much biological sense it makes…..For some of these its almost like referring to ‘insects’ or ‘mammals’.

Line 565: I think a big difference is that lakes and soils are mostly stationary, while streams by definition are in motion….much more likely to get mass effects in the water column that way, and very difficult for communities to develop

Figure 1: Just like for the binning of longitudinal distances, the binning of the melt period also seems a bit arbitrary. For example, there is only one sampling event that seems clearly to be during a baseflow period. Also, the post melt sampling point in 2020 is associated with the third highest discharge in that given month…..why would this not be associated with the ‘melt season’? Furthermore, the sampling of pre/post season samples is really limited in comparison with the melt season samples. Also, how different is post melt to premelt…...could these just be lumped together? What is the rationale to keep them separate (like do you expect distinct patterns in the post melt vs pre melt?)? On the other hand, some of the pre-post melt samples seem that they could very easily belong to the “melt” season, yet are separated by seemingly arbitrary dotted lines in the hydrograph, as the dotted lines appear to be in a different position in 2021 than in 2019 and 2020. Thus, it seems like almost all the samples are taken during the ‘melt’ period based on the hydrographs. Might it make more sense to derive continuous hydrological variables rather than try to make three categories? Potentially here also continuous variables may also help with interpreting differences in the data, as I can imagine that peaks and troughs within the peak melt season are likely to also have important differences just like at different times of the year? I have to admit, I don’t have a better idea, but Im also not convinced that the current strategy is the best one

Figure 2: In the box and whisker plot, there is a big white gap where the samples are missing from the COVID period. While I think it is good to be transparent about missing data, it also seems weird to have a big empty spot in the middle of the figure. Would it be possible to just cut this portion out? Also, what happened to the post-melt samples…..were they combined with premelt samples?

Figure 3: Was deuterium excess different among the three streams as a function of distance? I am just wondering how reasonable it is to lump the sites together in the analyses and figures. Also, while I appreciated that the season was considered in the graphing, it is really hard to pick out any seasonal patterns in these figures based on season, given that the shapes all kinda blur together at some point, and most of the trend seems to be longitudinal.

This study tests how the loss of glaciers will change the composition of organic matter in downstream waters and subsequently microbial community structure.  The research question is important given the rapid rate at which glaciers are being lost globally, so should be of widespread interest.  The authors, somewhat unsurprisingly, discover clear shifts in the composition of organic matter along the river networks, though the magnitude of the effects on microbial community structure are small.  The latter though is somewhat concerning as a negative result and begs the question whether the “right” independent and dependent variables were measured.  Nonetheless, I think the paper is well put together.

The technical approaches are sound and well explained, especially the field sampling.  However, the data are not statistically independent.  There is spatial and temporal autocorrelation in the sampling design, and I do not think that the three-way ANOVAs consider these effects.  For example, when estimating responses across distance bins, bins can show more similar values just because they are closer together in space and that needs to be accounted for in the statistical models.  Although one could argue that including distance as a fixed factor would enable two closely related bins to have similar values the important point is that we can't disentangle if that effect is purely because of distance or autocorrelation.  In other words, assume headwaters and near sites have similar values – is that because each site downstream is similar to its nearest upstream site (spatial autocorrelation) or because there is something special about those distance bins?  The same arguments could be made for year and hydrological periods.  While the breadth of field sampling is impressive, it is quite complex statistically to analyse something of this nature correctly.

Important clarifications are also required for the microbial analysis.  First, were there negative and positive controls in the sequencing?  These controls are particularly important given the finding that the same taxa seem to dominate the composition.  There should be more evidence given to rule out that lab/field contamination could be a source for the homogenization.  Second, what was the read depth and how much did it vary across samples, i.e. are normalization/rarefaction techniques required?  Please add this information.

I also have some specific comments:

Line 29:  I don't follow how this is necessarily "complex" given the previous conclusions of a core set of species that overwhelm (mass effect) environmental gradients, suggesting it is a simple predictable outcome.

Line 66:  shift"s"

Line 90:  The term "mass effects" is jargon and should be defined on first use, especially for biogeochemists that may be less familiar with this "ecological" term.

Line 153:  What volume of water was sampled for the microbes?

Line 155:  Not sure I follow the logic for why the microbes would change in the Bow River samples but not the carbon...  They're linked…that’s the argument of this entire paper.

Line 265:  Please can you explain why this is necessary for this comparison.

Line 267:  Beta-diversity is calculated from the Bray Curtis index not the NMDS.  NMDS is simply a visualization technique.

Line 269:  How was the perMANOVA performed?  And how were the clusters identified?

Line 270:  Why have you performed the RDA?  Please explain the biogeochemical question you are trying to test.

Line 287 and throughout:  The test statistics associated with the p-values must be reported to be reproducible.  I presume here you should have some F statistic from the ANOVAs with some degrees of freedom?

Line 335:  I don't follow what is meant by "passive overlay".

Line 341:  Aren't these 10 phyla dominant in most rivers?  It would be useful to contextualise these results, such as through comparison with the Earth Microbiome Project.

Line 356:  Please cite evidence showing that these parameters were highly inter-correlated.

Line 374:  There are no correlation statistics given anywhere to support this claim, i.e. of a trend in the clouds shown in Fig. 10.

Line 419:  But is there enough of this material in a mass-balance sense to matter?

Line 474:  A mixing model would really be the way to get at this question and the Discussion could at the very least point to its utility.

Line 498:  I don't think this paper tests this relationship as it cannot disentangle create from consumption of OM.  The rest of this paragraph also says little about this question and just reviews the composition of bacterial families.

Line 524:  I think this statement overstretches.  These were statistically significant but explained very little variation.  A total of like 9% all together, so how important was each variable?  I think the discussion that follows on lines 538 is much fairer.

Line 525:  Again, what is the biological significance and effect size?

Line 544:  Or we’re not measuring the "right" drivers.

Line 546:  Again, lack of strong “control” given the variables that were measured.  That's the problem with a negative result – is it the truth or the study?

Line 591:  Please cite some evidence to support this statement.  It does depend on the functional redundancy within these communities.

Figure 4:  I don't follow which end member the grey box corresponds with.  It looks like only part of the grey box corresponds with different end members.

Figure 6:  I think you should add to the caption that the percentages along the axes are explained variance.

Figure 8:  I don't follow why there are two circles if there are three groups for distance and season –

which of these variables correspond to the circles and why is being omitted?  I think you are looking at distance and grouping near and headwater together but the figure should be self-contained with its caption.

Figure 9:  I find it confusing to have three values for the variance explained, none of which match each other.  So the first two axes of the RDA explain 14%, all axes together explain 25%, and all axes when adjusted for the number of predictors explain 9%.  Is that correct?  Is it possible to focus on one number and just be more forthright that none of the environmental predictors do a particularly good job here?  As for the crosses in the centre, how come the near melt sites in the bottom-left are so far away from the taxa?  Are there no unique taxa associated with them?  It would be informative to see the indicator species labelled on here.

Figure 11:  There are ca. 85 correlations here.  Are you not worried about false positives, especially given that some of these Rho values look small?  Also, I don't follow how the Microbactericae - temperature correlation can be more statistically significant than the one between Beggiatoceae and temperature but have a smaller absolute rho value.  I haven’t checked all the other columns for similar problems.