the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 28 Oct 2025
Assessing Flood-Influenced Sediment Dynamics Using UAV Photogrammetry and Machine Learning: Insights from River Sense, Switzerland
Abstract. Gravel-bed rivers are shaped by complex interactions between hydrological forcing, sediment sorting, and channel morphology, yet fine-scale, spatially continuous observations of these processes remain rare. We combine UAV structure-from-motion photogrammetry with machine-learning grain segmentation to quantify flood-driven sediment redistribution in a minimally disturbed gravel-bed river (Sense River, Switzerland). Two surveys of four gravel bars (2021 and 2024) mapped individual clasts in images at centimetre resolution, allowing spatial and temporal analysis of grain-size patterns. We show clear intra-bar fining from crests to tails and a reach-scale morphology control on sorting: bend-associated bars are moderately to well sorted, while straight reaches are more poorly sorted. Grain-size distributions converge to self-similar forms across all bars, with analysis of ca. 1.86 million grains providing unprecedented empirical validation of scale-invariant sorting, an improvement by orders of magnitude over conventional pebble counts. To understand the detailed hydraulic controls during the moderately large flood captured between surveys (ca. 180 m³/s; ca. 2–10 years recurrence), we performed detailed hydraulic modelling for one bar, estimating spatial fields of shear stress, Shields parameter, and stability conditions during the flood. We also differentiated the topography between the two surveys to map the relative elevation change. The crest and margin armour remained largely stable, whereas the tail part was extensively reworked. A hydraulically driven mobilisation model reproduced observed mobility with ca. 65 % overall accuracy (up to 82 % in tails) but under-predicted movement on crests. We also show that where floods were large enough to mobilise the grains, coarse patches were rapidly buried or completely replaced, demonstrating that local hydraulic geometry can override patch stability. Overall, bar adjustment was deposition-dominated for that bar, consistent with the waning stage of the flood, during which reduced shear stresses promoted net deposition. Our data indicates that flows <200 m³/s can remobilise large bar areas, and analysis of gauging data for the Sense River since 1928 shows that such events are becoming more frequent. Our results highlight the important geomorphic role of moderate to moderately large floods in such rivers and demonstrate high-resolution, hydraulically informed grain mapping as a robust tool for predicting gravel-bed river response under changing flood regimes.
- Preprint
(3667 KB) - Metadata XML
-
Supplement
(809 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2025-5145', Anonymous Referee #1, 16 Dec 2025
-
RC2: 'Comment on egusphere-2025-5145', David Whitfield, 04 Feb 2026
The manuscript by Rezwan et al. aims to quantify the granular response to an approximately bankfull-magnitude flood, at the location of four depositional bars along the River Sense, Switzerland. Grain size distributions (GSD) are evaluated pre- and post-flood using Structure-from-Motion photogrammetry and a Machine Learning image segmentation approach. The resulting dataset is extensive and effective in quantifying surface change within each bar between the two sample dates, as well as downstream trends in bar surface composition. Hydraulic modelling is utilised to evaluate boundary shear stresses across one bar surface, which is used to interpret spatial patterns in surface stability and subsequently to account for topography change within the bar. The sediment mobility prediction aspect of the paper could benefit from clarification in terminology, and some additional justification and discussion around the assumptions and their potential impact on interpretations. Interpretations focus on a single event and would also benefit from further thought and discussion around the influence of smaller flows, for example, in controlling grain size sorting patterns and morphological change within the bar. Nonetheless, this paper presents a novel case-study application of the grain size analysis techniques which underpin the study, which is further contextualised to climate change sensitivity. Provided that the paper addresses the queries listed below, this work would be a valuable addition to the journal.
Main Comments:
(1) Validation of GSDs
In addition to the uncertainty estimation via the bootstrapping methodology, I think these results would benefit from a brief validation of the image segmentation. Some comparison of how the GSD obtained from the ML approach compares with reality would be really valuable; for example, were there any manual grain size measurements subsample taken in the field that could be used to confirm that the GSDs estimates are representative of reality?
How might grain embeddedness and/or orientation might affect GSD uncertainty? For example, could it be that changes in median grain size could be partly attributed to changes in grain arrangement or partial burial, particularly if a change in grain size sorting is observed. It would be useful to add a brief discussion around this.
(2) Sediment mobility: assumptions and terminology
Some clarification of the hydraulic and sediment transport terminology would be beneficial to avoid potential confusion. I would recommend referring to the modelled dimensionless shear stress as tau* (rather than Shields parameter), and the critical entrainment threshold as tau*c throughout the manuscript. Personally, the use of the term ‘Shields parameter’ makes me think specifically about the critical threshold of particle motion. In parts of the manuscript, the distinction between modelled tau* and assumed tau*c becomes a little unclear.
The assumption of tau*c = 0.06 requires some additional consideration. In the context of this study, I think that assuming a spatially constant value of tau*c is perfectly reasonable. However, this assumption should be justified (e.g. was the value informed by reach-average channel slope?) and the associated implications for sediment transport interpretation should be further discussed. Sediment mobility is likely to vary in different parts of the bar due to hiding effects, grain sorting, bed armour, etc. The potential influence of armouring on differences between predicted vs observed sediment transport is briefly mentioned in the discussion, but a more detailed discussion would be beneficial.
(3) Bar adjustment during lower-magnitude flow events
The paper attributes all of the observed change in grain size and bar elevation to the November 2023 flood, however, there are also considerable flow peaks between the Nov 2023 peak and the second sample date that would presumably be reworking parts of the bar. It would be useful to include an elevation map of the bars, with some indication as to how much of the bar is inundated at various flood magnitudes (i.e. so that the reader can get some idea of which parts of the bar are more frequently wetted). The idea that some parts of the bar are disturbed more frequently than others might assist interpretations and/or account for regions that are more/less active than predicted. It would be really valuable to add a discussion around this.
(4) Within-bar trends in grain sorting
This links to comment (2) and (3). There is a focus on trends in grain sorting between bars, but it would be interesting to see whether there are patterns of grain sorting within the bar. E.g. an extra map which shows variation in sorting parameter in each of the grid squares. This would be a valuable contribution to the discussion around particle mobility assumptions (is there any link between sorting and observed mobility?). It might also give some insight into the role of the smaller floods (e.g. is sorting different in parts of the bar that are disturbed more frequently?).
(5) Figure Size
In most of the figures, text size (particularly legends and axis coordinate labels) would benefit from being made much larger. Some of the figures could be made to fit a whole page, rather than being squeezed in a landscape orientation.
Line-by-Line Comments:
L22: It might be worth introducing the idea that the paper will specifically quantify the response to the Nov 2023 flow event earlier on in the abstract – the focus on this particular event is underplayed in the abstract.
L25: ‘compared’ rather than ‘differentiated’?
L32-34: This is really insightful context, but the potential significance of climate change requires a little more emphasis as it is a valuable outcome of the work here.
L105-117: The final paragraph really sets up the structure of the manuscript nicely! The climate change context is introduced very quickly – I wonder whether it would be useful to add a bit more climate change context earlier in the introduction, so that it doesn’t come as a surprise.
Figure 2: How high would the stage/discharge have to be to fully inundate the bar you’ve done the hydraulic modelling on? It would be useful to indicate an approximate discharge which relates to a bar-top height on this timeseries to get an idea of how frequently it is wetted (or how frequently it is almost completely wetted). Presumably the lower magnitude floods will be doing some of the water-working too… if you survey after one of these smaller events, might the GSD be affected and is this observable across different bar elevations?
L205: Is this a generally used sharpness threshold, or one that you'd found to be ideal (e.g. for identifying your minimum grain sizes)? This would benefit from a brief justification or a citation.
L210: “Excessive Covariance”. Do you have a ballpark number for what you'd consider excessive? Just to be a bit more quantitative so that others can follow your workflow.
L220: Is there a particular reason why you chose a 5 m grid resolution? e.g. If others were to do the same, would you recommend adjusting grid resolution for something like D50 (e.g. so that the sample size of grains in each grid sample stays relatively constant). Does grid size relative to D50 affect GSD uncertainty within each section of the bar? It would be useful to add these details in case others wanted to follow your methods but for a much coarser/finer bed.
L229: Do you have any field observations to account for possible error due to embeddedness or grain orientation. How sure are you that these are reliable a- and b- axes and not affected by partial burial?
L230: Minimum grain size threshold of 30 mm seems reasonable. Do you have any examples that you could show of grains that are about 3 cm in diameter? It might be worth adding this somewhere just to show evidence that it's a reasonable minimum threshold, and that a grain can of this size be reasonably identified.
L241-244: Great to include this GSD uncertainty analysis. Do you have any manual (either a Wolman count in the field, or measured visually from orthophoto) estimates of GSD to further validate the uncertainty estimates obtained from the bootstrapping approach?
L254: Was the distinction between crest/middle/tail just done visually, or did you use any quantitative information like slope, grain size? How did you decide your transition between each of the zones? Additional details of how each zone was classified would be important to add because it might inform interpretations, e.g. if you draw the line between bar crest and tail where there’s an obvious break of local slope, does the difference in slope between crest/tail influence bed mobility observations?
L275: How close was the gauging station to your sampled bar, and how much agreement is there between flow depths at the gauging station vs at the site? An idea of flow depth uncertainty would be useful, as it will influence your boundary shear stress (tau*) estimates.
L279: It would be useful to add some uncertainty analysis for tau* by quantifying uncertainty in flow depth and D50 within each grid. What is the uncertainty of tau* compared with your assumed value of tau*c?
Equation 1: Just to clarify that the slope used in calculating boundary shear stress in Eq 1 is the local bed slope within the grid, rather than the surface water slope, or reach-average slope? This probably requires a little more thought, because the slope parameter will incorporate uncertainty into estimates of tau and tau*. Realistically, the water surface (and therefore hydraulic slope) will not be equal to the local bed slope – I wonder if this is leading to overprediction in tau in parts where the bar is steepest? This is likely This will no doubt influence interpretations too. Would something like the energy slope of each grid cell would be more suitable? It is worth making this clear and adding further justification/acknowledgement of uncertainty.
L285: Please refer to main comment (2); I would recommend referring to dimensionless boundary shear stress as tau* throughout, as they usage of ‘Shields parameter’ might easily get confused with the critical threshold. Tau* and Tau*c notation ensures clarity.
Equation 2: Tau*c; the star is missing.
L288-291: This could read a bit more clearly: Tau* is the modelled dimensionless boundary shear stress, and will be compared against a critical threshold (Tau*c) which defines the point of incipient motion. It’s super important that distinction is made.
L291: Tau*c; The critical threshold is also dimensionless, so a star is missing again.
L291: See main comment (2). Can you explain why you've assumed the critical shields threshold as 0.06? E.g. did you estimate based on reach-average channel slope. While this is a reasonable value for gravel beds, tau*c can be quite variable (~0.02 to 0.07) which can have a big influence on sediment mobility estimates.
Secondly, if you've assumed that tau*c is spatially constant across the entire bar (which is totally reasonable for this application as it prevents any additional error), it is worth adding a brief justification. Tau*c can vary with grain size distribution (e.g. grain hiding effect), local bed slope, etc. I'd therefore recommend just acknowledging that as that assumption will be important in interpreting results and should be discussed.
L292: This could read a bit more clearly, e.g. "For each gird, we also predicted the maximum grain size which could be mobilised under the modelled flood condition".
L304: Good idea to scale as the numeric scale of all of these parameters are super contrasting. Could you briefly comment on whether this is likely to introduce any bias into your comparisons? If the range of one of your parameters is very low (not very much variability) and one is very high (loads of within-bar variability), might you be giving the one with very little range more statistical importance than it actually has, by bringing it onto an equivalent scale as the super variable one? Unsure what the answer to this is, but it is worth considering.
Figure 4: It would be helpful to add an arrow in each to show flow direction.
L394: See main comment (4). Would it be possible to calculate the sorting coefficient for each grid cell to (a) see whether there any changes in sorting within the bar pre- and post-flood; and (b) this might help to interpret bar stability predictions, e.g. are more sorted parts of the bar less active during the flood, e.g. realistically do they have a different Tau*c than 0.06? I think there would be something interesting to explore there.
L399-401: Super interesting! Though, I think this interpretation should be moved into the discussion.
Figure 5: Is it possible to plot the pre and post-flood GSDs on the same axes to make it easier to notice differences in their curve? (It might be too messy, but could be helpful).
Figure 6: The presence of woody debris might be interesting... do you think this might have any influence on grain size (e.g. trapping finer sediment during lower magnitude flows prior to the 2021 survey?)
L494-496: This could be more clear. Do you mean that the Tau* modelled (using depth and grain size) was compared with assumed tau*c?
L508-510: I think this interpretation should be in the discussion.
Figure 8: It would be useful to also plot maps of (a) bar elevation and (b) local slope for each grid cell, as this would be important in interpreting why predictions might not match reality, justify why some parts of the bar are more active than others, and explain spatial variability in tau*. Mobility thresholds are theoretically dependant on channel slope (Lamb 2008), so it would be interesting to explore whether your predicted bed stability is largely influenced by your assumption that tau*c is spatially constant – mapping slope would really help to explore this.
L581-584: And don't forget patterns of sorting within the bars! I think this could help to explore mechanisms of sediment mobility and deposition patterns, and worth discussing.
L621: The discussion around bed armour can be developed here. More stable parts of the bar would have a higher Tau*c and therefore overprediction of the likelihood of transport is more likely. This is probably exacerbated by Tau*c being lower at higher bar elevations (due to shallower flows and potentially different local bed slopes).
Citation: https://doi.org/10.5194/egusphere-2025-5145-RC2 -
RC3: 'Comment on egusphere-2025-5145', Anonymous Referee #3, 06 Feb 2026
# GENERAL COMMENTS
I have carefully read this manuscript several times. It presents a granulometric and morphological analysis of a segment of the Sense River before and after a moderate flood, combining UAV photogrammetry and automatic image grain segmentation. The topic is highly relevant, the dataset is exceptional, and the technological approach is modern and promising for advancing fluvial geomorphology.
The manuscript is well structured and has several strong points: the integration of UAV imagery with automated grain segmentation, the unusually dense dataset (>1.8 million grains), and the detailed documentation of morphological changes associated with a moderate flood. The authors also make a clear effort to link hydraulics, grain size, and bar‑scale dynamics within a broader climate‑change context.
However, several key aspects require clarification or strengthening before the manuscript can be considered for publication. The most important issues concern the absence of a formulated working hypothesis (clearly stated sentence), the undefined use of key geomorphological and sediment‑transport concepts, the lack of some methodological justification (photogrammetric uncertainties, hydraulic estimation, choice of tile size), and the discussion would benefit from a more critical examination of methodological limitations, sensitivities, and applicability. The figures also require substantial improvement (scale bars, readability), For these reasons, the manuscript cannot be accepted in its current form. This does not call into question the quality or relevance of the underlying work: the dataset, methodological framework, and overall structure are solid. What is needed is a stronger justification of methodological choices, and a more focused discussion that helps readers understand the strengths, limitations, and appropriate use of the proposed approach. With these revisions, the manuscript has the potential to become a valuable contribution. The specific comments below detail the points that should be addressed.
# SPECIFIC COMMENTS
1. Lack of a clearly defined working hypothesis
You apply an innovative method, but perhaps without fully exploiting its potential. You describe processes, at a 25 m² resolution, that have been known for several decades using traditional methods. It gives the impression that the innovation (continuous grain segmentation) is applied, but without real innovation in the way these high‑resolution data are used. With such a method, one could test new hypotheses on processes that are still poorly understood. Here, no hypothesis is formulated regarding sediment transport or morphological changes.
What were your initial hypotheses?
2. Insufficient definitions of key concepts in sediment transport and fluvial geomorphology, creating confusion
Several terms used in the manuscript have a precise meaning in fluvial dynamics, but are neither defined nor seem to be used correctly:
- mobility, stability, immobility, mobilisation (commonly describing processes at the grain scale)
- selective transport, partial transport
- spatial scales (fine/small, medium, large)
- grain, patch, tile, unit, zone, bar: what are they and what scales do they represent?
- sorting, spatial granulometric variability
Then, two major issues arise:
(a) Confusion between spatial scales: Grains are segmented individually, but the analysis is then aggregated into 25 m² tiles. This erases the fine resolution of the raw dataset and prevents any grain‑scale analysis. Your analyses are in fact at the scale of the patch (tile) or the morphological unit (zone), not the grain.
(b) Possibly inappropriate use of the term “mobility”: The manuscript mentions evaluating mobility, remobilisation, clast‑level mobility, but your methodology does not allow measurement of actual grain mobility. The segmentation is static (two independent dates), you do not track grains, you do not quantify initiation of motion, trajectories, or the mobile fraction per grain size classes. DoDs reflect elevation changes (burial, exhumation, deposition, erosion), not the process‑based mobility of segmented grains. To avoid ambiguity, rather than grain mobility or remobilisation, it would be preferable to refer to: grain‑size variations and textural changes, surface changes (=morphological changes), spatial heterogeneity, patch/zone/unit dynamics, estimated hydraulic competence.
3. Insufficiently justified methodology
3.1 Photogrammetry / Topography
- No ground control points (GCPs). This is not necessarily a problem, but it must be discussed.
- No discussion of the alignment quality between the two point clouds.
- The 75th‑percentile method is insufficiently explained. It is not mentioned again in the results or discussion. You should provide:
- the 75th‑percentile values for each survey,
- a map of elevations or at least a description of elevation ranges (mean crest, middle, tail elevations),
- a discussion of the limitations of the 75th‑percentile method.
- DSM uncertainties are not specified:
- spatially variable uncertainty?
- error propagation?
- justification of the ±0.06 m threshold?
3.2 Grain detection
How do you handle grains located on the borders of your tiles?
3.3 Hydraulics
- Confusing presentation of metrics, partly due to undefined use of mobility/stability terminology.
- It is unclear whether stability is evaluated using Dmax, D50, or D84.
- This makes the results difficult to interpret.
4. Analysis of spatial variability of sorting
Sorting should be clearly defined in the introduction. In sedimentology, sorting quantifies dispersion within a GSD. Here, you use sorting to refer to spatial variability of GSD across the bar, which is not the same. You did not calculate or produce a sorting map. For each tile, extracting a sorting index would have been informative. Perhaps in Figures 7 and 8, if you merge some curves and make space, you could add a sorting map.
- How does grain size vary within a 25 m² tile?
- How does it vary between morphological units?
- How does it change after the flood?
5. Questioning the representativeness of Bar 3
You focus on Bar 3, the only one of the four that changed the least. How is this site representative of your river if the other three bars (75%) underwent substantial change?
6. Lack of information regarding bar and reach configuration
Essential elements are missing to understand the morphological context. Differences in bar response also appear linked to large‑scale controls, but these are not described.
- low flow channel width,
- bankfull width,
- slope of the different reaches,
- Q50 or Qmean,
- description of the longitudinal profile.
7. Representativeness of 25 m² tiles and questionable use of the term “fine‑scale”
Your methodology is new, innovative, and very interesting, I am considering applying it myself. But be careful with the wording: depending on river size, 25 m² can be coarse.
It would be useful to discuss:
- how to choose tile size,
- whether it should vary with channel width, morphology, computational computer (or GPU) power, etc.,
- what you actually mean by “fine scale”,
- whether your results would change significantly with a different tile size?
8. Figures: numerous systematic issues that strongly hinder understanding
- Text too small (axes, legends, annotations).
- Scale bars inconsistent with tile size.
- Reuse of the same colours for different variables (e.g., green/purple for 2021/2024 and for other metrics) , visual confusion.
- Some figures use a basemap, others do not; in some cases the basemap reduces readability.
- Poorly positioned legends or unnecessary repetition of the same legend across panels.
- Missing elements: flow direction arrow, clear crest/middle/tail boundaries, precise location of the detailed Bar 3 study area.
- Boxplots grouped in unintuitive ways, forcing the reader to mentally reconstruct comparisons.
# LINE-BY-LINE COMMENTS
# 1 Introduction
L50: “restoration” Antinomic to: “constraining sediment dynamics” (L48). Restoration often aims to release or reactivate dynamics, not constrain them. Reformulate to clarify the intention.
L71: “selective” Transport is not only selective. If you wish to use this notion, you must first present the distinction between selective transport and equal mobility before using it.
L97: “A major unresolved issue involves…” Add a citation to support this statement.
# 2 Study area
L134: “display diverse morphologies and sorting characteristics” Please add some numbers in the sentence (overall D50 or at least a sorting coefficient).
Fig. 1:
- Panel B:
- Add national borders, or a semi‑transparent mask to other country to highlight Switzerland.
- Remove the names of towns, which are useless and unreadable.
- Instead of a small rectangle + arrow, display the full extent of the catchment, which will give the reader a better sense of its relative size.
- Panels C and D:
- Display the two patches but covering the same size area and at the same scale (e.g., two 1 m² patches).
- The scale bar seems incorrect: probably in mm rather than cm.
L156–157: “this flood … catastrophic floods.” You could add more numbers: well above annual peak (XX m³/s) but below rare catastrophic floods (XX m³/s).
⇒ It is necessary to mention the mean Q or Q50.
L158: “Despite its moderate magnitude,” This could be removed. Start with: “This flood …”
Fig. 2:
- No need to repeat the legend on both panels: one on top pannel is enough.
- Move the annotation rectangles to the upper panel only, top left, to avoid hiding the studied flood.
- Legend text could be slightly bigger.
- Caption suggestion: Figure 2: (a) Historical continuous hourly discharge time series from 1928–2025, with the blue shaded band representing the study period between the two UAV field surveys from September 30, 2021, to May 17, 2024, including the November 2023 flood. The red dashed line represents the moderate‑large flood threshold of 180 m³/s. (b) Hydrograph of the time portion corresponding to the light‑blue shading in (a).
# 3 Methods
L166–167: “representative field sites based on their accessibility” Accessibility is a logistical criterion, not a scientific argument for representativeness. Reformulate.
L180–181: “overlap between surveys …” Add this focused extent in one of your figures.
## 3.1 UAV‑based photogrammetry
### 3.1.1 Field survey
Any GCPs?
### 3.1.2 Photogrammetric processing
L211–213: “We improved … robust reconstructions.” Seems very similar to lines L207–210 above. Perhaps you could merge them.
## 3.2 Grain segmentation and size measurement
L220: “5 m² image tiles”
- How do you manage grains located at the borders of your tiles?
- 5 m² or 25 m², since the tiles measure 5 m × 5 m?
L227: “(5 m × 5 m)” 5 m × 5 m = 25 m², inconsistent with “5 m²” above.
L240: “96th (D96)” D95 instead of D96?
L254: “spatial position within the bar” A figure showing the boundaries of these three areas superimposed on the orthophoto (without the tiles) might be useful to allow the reader to clearly visualise your segmentation of the bars.
## 3.3 Hydraulic parameter estimation
L260–263: “To account … dataset,” Not clear to me:
- No GCPs?
- Is the 75th percentile calculated per tile or over the entire bar?
- How does this value represent an elevation uncertainty?
- What is the elevation range between crest and tail on Bar 3?
- What is the difference between the 75th percentiles of 2021 and 2024?
- How can you be sure that the crests were not submerged during a ~7‑year return period flood?
- This method and its limitations must be discussed in the Discussion section.
L285 Equation (2): “τ” You forgot the asterisk.
L289: “Finally” Remove “Finally”, a few lines later you describe yet another parameter (stability). Here we do not know how many parameters you are examining, which is confusing. You should state this clearly at the beginning of the section.
Moreover, the logical order should be: estimate shear stress,s hear stress equation, definition of h and S, method for estimating h and S, calculation of derived parameters (Shields, stability, etc.).
Currently, you present the steps in the chronological order of your calculations, which makes the reading difficult.
L289: “sediment mobilisation” Mobility of the D50? It should be specified.
L291–294: “We further … hydraulic conditions.” You do not explain how you calculate this variable, nor what it is physically based on. This makes the results difficult to understand. See Point 3.
## 3.4 Zone‑wise bar analysis
L296: “Zone‑wise” Could be Unit bar.
L302: “local stability.” Again, how is this assessed?
## 3.5 Hydrological and climate trend analysis
L329: “required to mobilise the grains” Which grains? A specific size? D50, D84? A certain proportion of grains?
# 4 Results
## 4.1 Discharge scenario
L348: “approximately 2–10 years” Is it 2 or 10 years? 2 years corresponds to a 50% probability of occurrence, while 10 years corresponds to 10%. That is not the same. What return period does 180 m³/s correspond to?
## 4.2 Physical characteristics of sediment grain size
L368: “averaged at a spatial resolution of 5 m²” Ambiguous formulation. Clarify. Is it:
- simply the D84 extracted from the GSD of each tile?
- the average of D84 within a 5×5 m tile?
- or a grouping of several 5 m² tiles?
Fig. 4:
- The text of the colour‑class legends is too small.
- A single common colour scale for all panels, enlarged.
- Major inconsistency: tiles supposedly measuring 5×5 m do not have the same size across panels. This suggests a scale error: Panel A: 20 m ≈ 3 tiles, Panel B: 20 m ≈ 2 tiles, Panel C: 20 m ≈ 2.5 tiles, Panel D: 20 m ≈ 5.5 tiles.
- The transparency of the orthophoto overlay on the D84 makes reading difficult.
- Some maps have a background, others do not: this must be standardised.
- Add a flow‑direction arrow.
- Add the extent of Bar 3 used in Figures 7 and 8.
Figure 5:
- A lot of colour; the frames on the left and right do not need colour.
- On suggestion: show the GSDs and self‑similarities for 2024 and 2021 on the same graph for each bar. Perhaps grey curves for 2021 and coloured curves for 2024.
- No need to repeat the titles above each graph (“Cumulative Grain Size Distribution” and “Self‑Similar Grain Size Distribution”). You are wasting space.
- The text is too small.
## 4.3 Sensitivity of grain‑size distribution to floods
Figure 7:
- Text unreadable.
- The colour scale on panel C is quite confusing compared to panels A and B (perhaps too few classes).
- Panels E and F could be merged into the same plot to facilitate comparison and maintain the green/purple colours for consistency with Figure 6 (2021 vs 2024).
- In panel i, instead of three facets, only two would be needed. Put the orange ones (D50) together and the purple ones (D84) together. Currently, the graph shows, for the same area, the difference size between D50 and D84, not the evolution of percentiles by area.
- Orthophoto backgrounds are unnecessary and reduce readability.
- Add a flow‑direction arrow.
- This time, the scale bars seem to indicate 4 tiles for 20 m.
## 4.4 Hydraulic parameters and sediment mobility
L481: “±0.06 m” Where does this value come from? Do you take into account the cumulative uncertainty of the two DSMs?
L484–485: “Although … near the edges.” Are you sure this erosion is significant near the edges? These are generally steep areas. It is common practice to apply higher significance thresholds in steep zones. Poor alignment between DSMs will create large differences in these areas.
L493: “sediment mobilisation.” See Point 2. The correct term could be elevation changes.
L494: “1 to 200 Pa.” Why do the histograms in panels f, g, and h not use the same classes as the images above? This would make interpretation easier.
L499: “values remained below 0.2,” If >0.6 means red = mobility (mobility of what? see Point 3.3), then why does the map show more red than green, but panel G shows a distribution skewed toward low values (green)? Text almost unreadable.L502: “approximately 65% accuracy” Where does this accuracy calculation come from? How do you compute it? Not mentioned it in the methodology.
L511–516: opposing concepts: mobility and stability. You use many opposing words, which makes the paragraph difficult to understand. The metric is poorly described in the methodology. Also, the name of this metric may not be appropriate (see Point 2).
L511–512: “Further, spatial … conditions.” Poorly worded. We do not calculate a grain “required for immobility”. We calculate either:
- the maximum mobilisable size, or
- the minimum stable size.
L514: Change to: “with minimum estimates of stable grain sizes exceeding 350 mm in many parts of the tail zone to maintain sediment”
L515–516: Change to: “In contrast, the crest presents smaller estimates of minimum stable grain sizes (<90 mm), aligning with observed sediment stability and minimal grain‑size change.”
Figure 8:
- You could add crest/middle/tail segmentation.
- Why not use “bar head” instead of crest?
- Text unreadable.
- Panel C: for which percentile?
- Does panel D indicate potentially mobilisable D50?
- Does panel C show tiles according to 2021 D50, which is either immobile or mobile depending on panel D?
- Panels F, G, and H do not correspond to elevation change. Please check the caption.
- Why not follow panel A and use the same class boundaries as panels B, C, and D?
Figure 9:
- helpful if the variables in the legend appeared in the same order as you describe them, because:
- it is difficult to follow your explanation and find the colour of the box,
- the label text is too small.
- Suggestion: you describe the evolution by variable, but the figure is organised by zone. Could group the boxplots by variable (three boxes per variable: head/middle/tail).
- Figures do not need internal titles: everything should be in the caption.
# 5. Discussion
## 5.1 UAV and machine learning for grain size characterisation
L563: “sorting” See Point 2 and 4.
L573: “area‑by‑number grain‑size statistics” Why is area‑by‑number a strength? Explain why.
L575–576: “patch boundaries, and partial‑transport responses” You use the terms “patch” and “partial transport” without defining them or analysing them in your study. For example, partial transport requires mobility measurements (initiation of motion) by size class and an analysis of the mobile/immobile fraction.
585–590: “Here, … proxies alone.”Seems inccorect use of “mobility” and “grain scale”. See Point 2
L594–595: “Consequently, … river management.” The claim of scalability and applicability to diverse geomorphic contexts is not supported by the results. Please either provide cross‑site validation or temper this claim.
## 5.2 Linking hydraulic conditions with sediment mobility
L601: “between sediment mobility and local hydraulic conditions” Change to: between bed textural and morphological changes with local hydraulic condition.
L606: “High‑resolution” 25 m² , is that considered high resolution?
L607–610: “These results … processes (Ashworth et al., 2000; Nicholas, 2013).” The sentence suggests that a moderate flood is generally erosive, which strongly depends on:
- basin configuration,
- sediment supply,
- presence of tributaries,
- sediment availability.
As written, it sounds like an unfounded generalisation. Clarify or reformulate.
L656–660: “Importantly, active reworking (cf. Buffington & Montgomery, 1997; Wilcock & Crowe, 2003).” See Point 2. You could rework L657 as: “crests remain relatively stable, while significant changes in the bed have occurred downstream, with…”
## 5.3 Potential influence of climate change on gravel‑river morphodynamics
L668: “5.7 years” This information should be provided before the discussion, in the study site description, the hydrology section, or the first part of the Results, rather than mentioning 2/10 years each time.
L705: “even” Remove “even”. Moderately large floods such as the one studied here (2023, return period of 5 years) are expected to cause these changes. Not only extreme 100‑year floods. It is therefore likely that even weaker floods can also mobilise a significant area of the riverbed.
L709: “altered” Change to “modification”.
L715–716: “altering … river management.” Conversely, a more dynamic system can create greater habitat diversity, stimulating populations. Your sentences only presents negative effects.
# 6 Conclusions
L734: “unprecedented” Your methodology is indeed innovative and interesting, but the wording is too strong. Depending on the size of the river, a 25 m² tile can be very coarse, and may not constitute an “unprecedented” resolution. It would be useful to discuss:
- how to choose tile size,
- whether this size should vary depending on channel width or morphology,
- and to what extent your results would change with a different tile size. See Point 7 (tile scale).
L735: “fine‑scale spatial organisation” Same issue. To avoid problems, you need to define in the introduction what is meant by fine, medium, and large scale. What are patches and zones, and what scale do they correspond to? Once these scales are clearly defined, the discussion becomes much more coherent.
L735: “sorting” In sedimentology, sorting is used to quantify dispersion within a GSD. Here, you are talking about the spatial variability of the GSD.
L744–749: “To examine how these sorting patterns respond to hydrological forcing at grain scale,” this is not at the grain scale, but at the spatial scale of your tiles, or even your zones. You have not presented your results at the grain scale.
L754–757: “These floods a.. extremes,” Perhaps you forgot to add a superlative, but as written, the sentence seems incorrect. These moderate floods are capable of reshaping the landscape, as you have shown, and, by definition, they are more frequent than rare extreme floods. They are therefore already a main factor driving changes in the river landscape.Citation: https://doi.org/10.5194/egusphere-2025-5145-RC3
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 323 | 166 | 35 | 524 | 48 | 45 | 46 |
- HTML: 323
- PDF: 166
- XML: 35
- Total: 524
- Supplement: 48
- BibTeX: 45
- EndNote: 46
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The primary focus of this manuscript is to understand how grain sizes vary across gravel-bed river bars as a consequence of a moderate magnitude high-flow event. A set of two high-resolution datasets of grain size variation across a number of river bars are used. These datasets were derived from automated grain size mapping from aerial images acquired using SfM photogrammetry. The DEMs associated with SfM photogrammetry were used to analyse hydraulic gradients, which were input to a highly simplified hydraulic modelling approach to estimate bed shear stress. Overall, the manuscript demonstrates how reach-scale, multi-temporal datasets of grain size, coupled with analysis of sediment mobility, can be used to gain process-inferred insight into sediment transport and resulting changes in surface sedimentology. In doing so, it represents a good case study and methodological example of the application of high spatial resolution datasets (with several time periods) to investigate sedimentary dynamics. However, although this is demonstrated I do have some methodological queries with respect to the approach and a variety of other significant comments.
MINOR COMMENTS
REFERENCES
Reid, H. E., Williams, R. D., Brierley, G. J., Coleman, S. E., Lamb, R., Rennie, C. D., & Tancock, M. J. (2019). Geomorphological effectiveness of floods to rework gravel bars: Insight from hyperscale topography and hydraulic modelling. Earth Surface Processes and Landforms, 44(2), 595-613.
Williams, R. D., Reid, H. E., & Brierley, G. (2019). Stuck at the bar: Larger-than-average grain lag deposits and the spectrum of particle mobility. Journal of Geophysical Research: Earth Surface, 124, 2751–2756. https://doi.org/10.1029/2019JF005137