This paper presents a broad analysis of grain size in streams in postglacial settings which builds on previous studies in non-glaciated landscapes that have been conducted using random forest and Spearman correlations (Snelder, 2011 and others). These methods are employed by the authors in a novel landscape setting and with an alternative approach use to data capture - a citizen science survey. The paper frames a clear hypothesis and is well structured and clearly written. I consider the paper will provide a valuable contribution to the field following some revision. I have three main points which I think should be addressed to enhance the work for publication. 

As outlined in the main points below, I feel the paper would benefit from further discussion of the limitations of the data captured by the citizen science survey approach, including the distribution of sites and the representativeness of the images of sediment characteristics at the reach scale. Additionally, (and if supported by the data), I think more could be made of the sites where multiple data points are available for a single stream, particularly to illustrate the model for sediment delivery established for post-glacial settings.

A more substantial point relates to the erodibility metric used, and I believe further consideration of the formulation of this metric may be required to ensure it adequately reflects the geological variability in Scotland. Although I suggest the metric needs revision, I don’t believe changes will alter the main conclusion of the paper. I look forward to seeing the final work in print. 

Best wishes,

Katie Whitbread

Main points

1.    Data distribution and representativeness:

The sites are quite clustered and some discussion of potential impact of clustering bias on the outcomes is needed. To what extent does the distribution of sites cover a reasonable range of values of your key parameters (erodibility, steepness etc.)? Are very big, or very small channels undersampled? Could you include some plots in the supplementary information?

Note that the distribution of data in Snelder et al. (2011) is a well-spread sampling pattern at the national scale which is pretty much ideal, so it’s important to consider the effect of clustering in relation to the lack of correlation in your study. 

In terms of representativeness of the sample, grain size may vary locally within reaches, e.g. between riffles and pools, inner channels and channel bars, even between adjacent bars. The Snelder (2011) study describes a process of visual assessment of the geomorphic components of a reach and areal proportions of different grain-size deposits, followed by selection of representative points for sampling. Are there ways you can assess the representativeness of the sample (in the context of the reach morphology)? 

The paper would benefit from further consideration of the distribution and representativeness of the data in the discussion as part of a more extended treatment of the limitations of the crowd-sourcing approach.

2.    Analysis of single-stream sample groups

The treatment of sites where multiple samples are available for the same stream is very limited. It would be interesting if more could be made of this data – for example a fuller treatment of the data for the River Feshie shown in figure 6 (b) to assess whether there are localised spatial trends and points where disruption occurs due to sediment inputs. Is there potential to include e.g. long profiles where data and key local features of influence could be plotted data could be projected on to a long profile? This could help to illustrate the spatial model in figure 7 and bolster the conclusion by demonstrating it operating through the local dynamics in the study area. 

3.    The erodibility metric

Metamorphic grade is encompassed by the lithology described on BGS maps (e.g. a gneiss is high metamorphic grade by definition, and a metamorphosed sandstone is either a “metasandstone” or a “wacke”). I’m therefore not convinced of the relevance of combining separate metrics reflecting lithology and metamorphic grade when using the BGS bedrock map. The metamorphic grade seems redundant. Also, what is the estimate of rock strength (LL) based on? Has a strength dataset been used?

The classification shown in the erodibility map in supplementary Figure S2 is different to what I would expect, and I think this points to the need for revision of the metric.

•    The moderate erodibility estimated for the Ordovician-Silurian age metasandstones (“wackes”) across the Southern Uplands appears similar to the Devonian-aged Caithness Flagstone (an unmetamorphosed lacustrine siltstone/fine sandstone sequence) in the far northeast, and to the Coal Measures of Carboniferous age in the Midland Valley. I’d expect the Southern Uplands ‘wackes’ to have lower erodibility.

•    The highest erodibility units (yellow) seem to be associated with small outcrops of Early to Middle Devonian conglomerates around the fringes of the Moray Firth and with units that occur along the Highland Boundary Fault zone and Southern Uplands Fault. It is not clear why these specific Devonian conglomerates should be more erodible than other Devonian units including conglomerates, sandstone and lacustrine siltstone and mudstone which are estimated with medium erodibility. 

•   I am unsure what would be giving rise to the high erodibility values along the Highland Boundary Fault – perhaps the Highland Border Complex, but these strata are metamorphosed.

•    I would expect the highest erodibility in strata in the unmetamorphosed Permian sandstones which occur in basins within the Southern Uplands and small exposures of Triassic/Jurassic/Permian strata around the Moray Firth Coast - but currently these seem to have intermediate erodibility values.

Finally – this approach to characterising rock erodibility was developed to configure models of river incision into rock rather than as a control on grain-size distributions supplied to channels. It would be worth noting, in the discussion at least, that grain-sizes supplied to channels are significantly influenced by the nature of discontinuities (joints and faults) – see overview in Sklar (2024). Whilst in general terms, lithology/rock strength influences fracturing and is therefore somewhat included in your method, fracture density can be very locally variable (e.g.  Neely & DiBiase, 2020; Whitbread et al., 2024) and the degree to which this may influence the supply of material to your sites isn’t known. There has been a fair amount of recent work on grain-size distributions on hillslopes and I think it is important to note the emerging literature in this area and consider it in relation to your conclusions.

Specific comments:

Figure 1: Increasing the size of the maps and photos would be helpful – these are very small in the preprint.  Add labels showing key locations mentioned in the text would also be useful (see comment below about image 3). If addressing the point above about the use of multiple points on key streams, could you highlight grouped data for key streams on this plot? Or perhaps show those more clearly on Figure 3). Use of a hillshade terrain model as the base map may also help illustrate the range of topographic settings associated with the data distribution. 

Figure 2: the segmentation of the photograph in c) is very difficult to see – increasing the photo size and upping the contrast of the lines would be helpful. 

Figure 3: Labelling is included on the map, but not very clearly – increasing the image size and using lines to show the association of the label and relevant points more clearly would be helpful.

Figure 7: Could you include an illustration of this model from one or more of your catchments where you have multiple points on the same stream? See notes on the text below. (But if you leave the figure as is, can you increase the size of the photos as they are too small to see clearly.) 

Supplementary figures S1 and S2: Increase the size of the maps 

Line 186 / Table 1: What scale and version was the bedrock geological map used? E.g. was it 1:50,000 Bedrock v8.

Line 198 / Table 1: What scale and version was the superficial geological map used?

Line 242 / Equation 6: The inclusion of Q and W in equation 6 means that it isn’t just rearranged from equation 5. Could you explain the inclusion or cite a relevant paper?

Line 264: There is very cursory treatment of the data for rivers with multiple samples. How many rivers had multiple samples? How many samples and over what range of drainage areas? Could you show the variation of D84 and potentially other variables on a long profile of the stream? This may be particularly useful if it helps illustrate your model in Figure 7.

Line 291 / Figure 6 (b): See main point 2 above – is there more in this data for the Feshie? It seems like there are perhaps two clusters of data, one with a trend and one without. Are there parts of the catchment with localised trends in flow competence, vs. parts disrupted by sediment inputs? Could you use local relationships in the Feshie or other streams to illustrate your model in Figure 7?  

Lines 345-352: The discussion of the citizen science approach should also address limitations of the approach and consider how these could be mitigated in future studies. As noted above, the clustered distribution of sampling sites is a key limitation – I think you should include some discussion of the potential impact of this on the outcome of your study.

References:

Neely, A.B. and DiBiase, R.A., 2020. Drainage area, bedrock fracture spacing, and weathering controls on landscape‐scale patterns in surface sediment grain size. Journal of Geophysical Research: Earth Surface, 125(10), p.e2020JF005560.

Sklar, L.S., 2024. Grain Size in Landscapes. Annual Review of Earth and Planetary Sciences, 52.

Whitbread, K., Thomas, C. and Finlayson, A., 2024. The influence of bedrock faulting and fracturing on sediment availability and Quaternary slope systems, Talla, Southern Uplands, Scotland, UK. Proceedings of the Geologists' Association, 135(1), pp.61-77.

Review of "Controls on fluvial grain sizes in post-glacial landscapes", submitted by Towers et al et Earth Surface Dynamics.

This manuscripts presents a very interesting and well-designed study on the relationships between grain sizes in rivers and environmental parameters, in the post-glacial landscapes of Scotland. The paper is well written and supported by nice data, however, I think it still requires some work before publication. In fact, I identify several points that could easily be a bit more developed to better root the study and the results. Please find my main comments below with some more minor points. I don’t see any issue in addressing my comments and none of them should modify the conclusions of this study so I look forward to reading the published version. 

Introduction (and where relevant)

I would have expect reference to the recent review of Sklar "Grain size in landscapes"

Sklar, 2024, Annual Reviews https://doi.org/10.1146/annurev-earth-052623-075856

l. 30 "primarily by abrasion": deposition and sorting are also first-order parameters in size reduction during transport. This should be addressed in the Introduction. As grain size fining will modify the size of the sediments, I was also expecting something about the rate of fining and how the position along the river could affect the size. I think this should be mentioned here or in the Methods section as this is, I think, an important question: how the location of your samples with respect to the river network can affect your results?

l. 36-48 in this paragraph, there is a mix of grain production (l. 39) and transport (l. 41), and in grain size and flux (l. 47). This should be clarify by adding a few sentences and/or separating sentences to avoir mixing of concepts.

I really appreciate the final paragraphes, they state in a very clear way the motivation and purposes of the study. 

Methods

Data collection is a mix of automatically and manually segmented pictures with the addition of a few manual measurements. I totally understand the method, however, this part needs some clarification. In fact, it is well known (among other, Kellerhals and Bray, 1971 or Bunte and Abt, 2001) that grain-size distributions are not equivalent when the sediments are sampled by grid (Wolman counts) or by area. How did you account for this effect in your data set? 

In addition, it seems that you used a grid for the manuel counts on pictures with an imposed number of nodes, rather than a fix distance. How would this affect the distributions? 

PebbleCounts does not segment all the grains so that the distribution is not by area neither by grid. Did you check that the results from the automatic approach are consistent with the manual one? 

Finally, due to the overlapping of the grains, samples from pictures tend to underestimate the diameters. If imbrication or grain shape are very different from one site to another, one could expect some differences related to the method and not the grains themselves. This is very difficult to quantify but it should at least be acknowledged.

The selection, definitions and calculations of the environnemental variables should be further developed and illustrated. In the current form, it is a bit difficult to understand the choices, the calculations. The way erodibility is defined could be a bit more developed too, as it is not that commun. I would appreciate a bit more context and values (what do the landscapes look like, do we have some diversity or are they very similar, etc).

Random forest regressor and Spearman’s correlation might not be familiar to all readers, please explain them a bit more the Methods.

l. 210 "all grain sizes are available for transport": this is a very strong hypothesis. How true and universal can it be? If not true, how could it alter your interpretations? 

l. 214 supply-limited is often defined as a lack of sediments with no reference to a specific size. This could be mentioned as your definition might not be the most commun one. 

Equation 6 is not just a simple rearrangement of Equation 5. Please add the intermediate steps and write explicitly the final equation mentioned l. 244. 

Reading the abstract, I was really interested in the citizen science approach, yet, it is barely addressed in the Methods. How did you set up the survey? How many participants? What is the quality of the data? I think this is innovative and it should be better explained so that other groups can use the same approach. 

Results 

Figure 3 is a bit difficult to read because it is quite small. Please consider increasing the size. For comparison, it could be great to add a scale in mm on panel b (or to add a panel c with the distribution of D84 in mm). 

Figure 4 is barely used. In addition to the correlations with D84, you could explain the other correlations (for example, slope and aridity) and built on it either in the Results or in the Discussion.

l. 289 it is not clear to me how you can conclude that the river has the potential but not the sediments. I think this is an important aspect fo this study that could be further developed (see also my comment on line 210).

Discussion

l. 297 substrate cover is not mentioned previously and not explored in the Results. Please consider removing this point or addressing it in the whole manuscript. 

Here again, I was a bit disappointed by the way the citizen science survey is discussed. I guess there is more to say about it as it is not a classic method. For example, how long did it take to collect the data ? Did you stop or not the survey, when and why? How time consuming it is to deal with the variety of data with respect to "classic" field work? 

I would have appreciate some discussion about grain size variability along a stream, or at the same position if you have such data. A more focused analysis from rivers with similar variables could also be interesting, ie, the whole data set shows no trend but maybe at smaller scales, there are some. Did you explore this?

There are a few typos (parenthesis with references, l, 239-240)