General comments: The manuscript “Three-stage evolution of particle shape in headwater streams” investigates the shape evolution of sediment particles along the Mosawa River in NE Japan. Both basalt and shale particles are studied, with a focus on the evolution of their circularity and roundness. The paper’s topic and findings definitely fit into Earth Surface Dynamics. Still, the results' presentation and discussion need improvement before the final publication.

The most crucial issue is that both roundness and circularity are computed for 2D images taken of spatial particles. Please discuss the limitations of this approach. Especially the sentence in line 169 “…we aligned the particles with their maximum projected areas parallel to the tray…” should be justified: one would expect, that the particles sit on one of their stable balancing points (a.k.a. equilibrium point), which is not needed to yield their maximum projected area. Another questionable approach of the manuscript lies in using the transport distance x as a main variable in equations (2)-(4) because x lacks information about the vertical section of the river. It seems, mass loss would be a more natural measure to describe the stage of shape evolution. Please discuss these issues.

Specific comments:

1. Line 65 on page 3: “… field studies have shown that the rate of change in particle shape decreases with the transport distance….” - this sentence should be clarified: the rate of change of some shape measure (like the axis ratio, the volume… etc.) can decrease, not the shape itself. Furthermore, a reference to Sternberg’s law stating exponential decay of the particle mass/volume is missing here.
2. Line 71 on page 3: The overview of image analysis is not complete; a recently invented scanning technique for sedimentary particles has been published here:

Fehér Eszter, Havasi-Tóth Balázs, Ludmány Balázs: Fully spherical 3D datasets on sedimentary particles: Fast measurement and evaluation CENTRAL EUROPEAN GEOLOGY 65:2 pp. 111-121. (2022)

1. In the first paragraph in section 2.2 on page 5: the formula defining the roundness should be provided, similarly to the formula of the isoperimetric ratio. How is the ‘curvature of the corners’ defined exactly?
2. In line 185 on page 7, the workflow in software Rgrains is cited. What about placing it in an appendix or supplementary material to have a complete overview of your work?
3. Line 378 on page 18: In the discussion authors state “…the differential histograms in the most upstream sections are…that the effect of the rapid removal of sharp corners and edges…”. This observation agrees well with (Domokos et. al., 2014) and (Novák-Szabó et. al., 2018); here, a quantitative comparison to their results (based on the average mass loss) would be insightful. Actually, it seems that this first phase observed along the upstream section coincides with Phase-1 introduced by (Domokos et. al., 2014) for individual particles in curvature-driven abrasion.
4. Line 385 on page 18: The statement “…a decrease in shape parameters with transport distance has been reported in fluvial, beach, and aeolian environments…” is vague: which parameters are you talking about? Actually, fluvial and aeolian environments are rather different; for instance, the isoparametric ratio typically increases in fluvial, but decreases in aeolian environments.
5. Lines 460-465 on page 20: These statements, I think, are not supported by the paper. The evolution of R and IR versus the evolution of the mass/volume is a very delicate question, especially because the measurements are carried out on 2D images. Either show a simple model that supports your claims or rephrase this last paragraph.
6. Lines 545-550 on page 23: It would be nice to have some estimate of where the fragmentation ends along the river. Does it coincide with the end of the intermediate phase?

Minor Comments:

1. ‘Roundness’ and ‘circularity’ are not defined in the abstract. I think a short sentence about their difference would improve the text, e.g., a hint that circularity is measured using the entire perimeter, and roundness focuses on the vicinity perimeter segments with high curvature (i.e., edges).
2. Figure 1: What is the meaning of the black dots on panel (a)? The sampling sites should be denoted by the same signs in panels (a) and (b). The starting point of the distance measurement (i.e., the 0 of the horizontal axis on panel (b)) should also be marked on panel (a).
3. Before eq. (3) The correct equation reference reads eq. (2).
4. Writing real fractions (instead of power (-1)) would increase the readability of the paper (e.g., in equations (4) and (5)).
5. Line 468 on page 20: “particle durability” is not well-defined. The material composition, which determines the hardness and the brittleness of the material, seems to be the right framework here.
6. Figure 9: It seems the vertical axis is not on a logarithmic scale, as opposed to the caption of the figure.
7. Among the references: (Szabó et. al., 2018) should read (Novák-Szabó et. al., 2018).

General comments:

This manuscript presents new data on particle roundness and circularity from a small catchment in the Tsugaru Mountains of north-eastern Japan. The study applies previously established roundness and circularity metrics to analyse the downstream evolution of sediment shape over a 6 km transport distance along the main trunk stream. Using multiple shape metrics, the study evaluates which parameters show significant downstream changes. In addition, the authors compare the shape distribution of fine sediment fractions (2–4 mm) between talus-derived crushed material and downstream fluvial samples. The structure and title of the manuscript suggest that the main finding is a three-stage evolution of particle shape within the headwater stream: an initial rapid increase in roundness, a middle phase of minimal or no change, followed by a gradual increase downstream.

Previous theoretical models supported by field evidence (e.g., Krumblein 1941; Domokos et al., 2014; Miller et al., 2014; Roussillon et al., 2009; Pokhrel et al., 2024) have consistently demonstrated a non-linear, two-phase downstream evolution of roundness that reflects the exponential nature of the abrasion process. These models predict a rapid increase in roundness in the upstream reaches followed by a slower change downstream, with the rate of change dominantly controlled by lithology.

In contrast, the present study presents a three-stage model based on field data from six locations spanning approximately 6 km (roughly one sample per kilometer). However, this is a relatively short segment compared to previous studies that examined roundness evolution over tens to hundreds or even thousands of kilometers. It is likely that the observed three-stage pattern in this study reflects local variability, sampling resolution, or natural scatter in the roundness data—possibly due to heterogeneity in source material—rather than a distinct or generalizable evolutionary model.

Importantly, the manuscript does not provide a functional explanation or mathematical formulation for the proposed three-stage trend, nor does it convincingly argue for its broader applicability. The manuscript lacks a clear explanation of the mathematical expressions or functional combinations used to fit the trend line to the field data, which forms the basis for the author's conclusion of a three-stage evolution in the headwater region. As a result, the primary conclusion—that of a three-stage evolution—is not sufficiently supported by the data or linked to a broader theoretical framework.

Therefore, I do not find adequate scientific justification for the main claim of a distinct three-stage evolution in this study. In its current form, the manuscript does not meet the threshold for publication.

That said, the study does provide useful insights into:

• initial distribution of shape parameters from the talus sediment
• which shape parameters change most significantly over short distances, and
• how particle shape distributions differ when influenced by different mechanisms (e.g., talus vs. fluvial transport).

These findings have potential value and could form the basis of a new manuscript focused more clearly on shape parameter comparison and the mechanics of particle abrasion.

Specific comments:

Abstract

• Line 14: The phrase "each process" is used without having mentioned any specific processes earlier. Please define or list the processes first before referring to them.
• Line 16: It would be clearer to include the term "downstream" to specify the direction or context of evolution being discussed.

Introduction

• Several references are missing where key information is provided. For example, the statement in lines 55–56 requires a proper citation.
• Line 87: Instead of saying "various processes", it would be clearer to name the specific processes and indicate the shape metrics or indices used as proxies to assess them.
• Line 127: The manuscript mentions the use of Wadell’s roundness and briefly explains its definition but fails to provide the mathematical expression. Readers would expect this expression in Equation 1, rather than introducing the circularity index first. It is recommended to present all shape metrics in sequence, accompanied by their respective mathematical expressions.
• Line 141 / Equation 2: The metric 'R' is used (citing Krumbein, 1941a), but there is no explanation of what 'R' represents or how it is measured. If 'R' refers to Wadell’s roundness, this should be explicitly stated and described using the proper formula, similar to the explanation given for IR in Equation 1.
• Figures 1a and 1b: Assign unique labels to each sampling site and use these consistently throughout the results and discussion sections.
• Figure 1c: This subplot shows slope versus contributing area, but the relevance of this plot is unclear. Please explain its purpose and where it is discussed in the main text.

Methods

Sampling

• Line 155: The purpose of collecting samples from the trunk stream and talus is explained, but the rationale for sampling from tributaries and downstream of the dam is lacking and should be elaborated.
• The number of samples collected at each sampling site for each size fraction should be clearly described in this section, instead of mentioning it later in the image analysis section.

Image Analysis

• The manuscript lists the use of RGrains, MATLAB v4.11, Adobe Photoshop, and Affinity Photo—none of which are open-source tools. To improve transparency and reproducibility, the authors are encouraged to include a clear workflow diagram outlining each step from the raw field photograph to the final shape parameter extraction. This would clarify the role and purpose of each software used. For instance, line 168 states, “We used MATLAB application v4.11”, but does not specify the function or task it was used for.
• Line 171: The statement may be accurate, but please provide a reference or example to support it.
• Line 192: Authors mention converting pixels to millimeter. Was any correction applied for pixelisation effects? If so, please describe the method used.

Definition of shape parameters and downstream evolution model

• This subsection may better fit in the method section.
• Lines 196–209 describe the calculation of rate changes and histogram comparisons under the “Image Analysis” subsection. However, these analyses are not part of image processing per se. It would be more appropriate to move this content to a new or existing section—such as “Definition of Shape Parameters and Downstream Evolution Model”.
• Line 196: The manuscript states that median values of R and IR were used. Please justify how the median represents the sampling population, especially given lateral sediment inputs from tributaries and hillslopes.
• The exponential model used to estimate rates of change does not account for the three-stage evolution proposed in the manuscript. If the model is not designed to capture this pattern, please clarify its suitability for your objectives.

Results

• The section begins more like a figure caption. It would be better to start with a paragraph summarizing key findings and referencing specific subplots.
• Figures 2a and 2b appear to be key to the manuscript, as they form the basis for the proposed three-stage evolution. It is therefore recommended that these two plots be presented as a full-page figure to allow readers to evaluate the data more easily. While the data points in Figure 2b are legible, those in Figure 2a are overlapping, making it difficult to discern the trends for each grain size fraction.
• It is also recommended to use a distinct symbol for samples collected from the tributary, as their roundness values should ideally align with the trend observed for the main trunk stream based on transport distance.
• There is no strict standard for using upstream or downstream distance, it is generally preferable to use downstream distance (downstream transport distance), setting 0 km at the channel head, and to maintain consistency by referring to this convention throughout the text when describing key observations.
• Figure descriptions should follow a logical sequence. For example, Figure 2a and 2b are discussed in line 231, but subplots 2c and 2d are skipped entirely. Likewise, Figure 3 has four subplots, none of which are clearly referenced in the text.
• In Figure 3, the individual parameters used to compute the roundness metric ‘R’ are shown. If there is a specific reason for plotting them separately, it should be explained in the methods section where these metrics are introduced.
• Table 1 and 2 present rates of change derived from Equations 2 and 3. Please explain how these rates—calculated from a single exponential function—support the concept of a three-stage evolution.
• Line 282: Less rounded particles from tributaries are expected due to shorter transport distances. Consider using percentile distributions to better evaluate sediment sources within the mixed sampling population.
• Figure captions should be self-contained and informative. For example, the caption "(c, f) Comparison between the crushed and non-crushed 2–4 mm talus samples" implies no difference between panels c and f. Also, figure subplots and captions should follow sequential order. For instance, Figure 7 starts with subplots a and d instead of a and b.
• Line 348: A hypothesis is introduced in the results section, but hypotheses and statistical tests should be defined earlier—in the introduction or methods section.

Discussion

• Lines 364–368: While the observed rates of change may align with previous field observations, the proposed three-stage evolution is inconsistent with the theoretical and experimental models cited. Most prior work supports a two-stage evolution described by a single (often exponential) function, e.g., Krumbein (1941), Domokos et al. (2014), Miller et al. (2014), Roussillon et al. (2009), Pokhrel et al. (2024).
• The observation that coarser basaltic fractions are more rounded than finer ones is interesting and author mentions can be explained by weathering behavior in igneous rocks. However, this may not apply universally—for instance, granitic sediments often show more rounded pebble size particles compared to cobble size. Consider discussing whether this pattern is rock-specific or grain-size dependent.

Conclusion

• The current data and analyses do not provide sufficient scientific support to robustly conclude a three-stage evolution of particle shape in the headwater region.