Thank you to the ananymous reviewer for its detailed and constructive review. We have carefully taken it into account and believe it will improve the article. Below is a detailed response to all the points raised.

General Comments

This manuscript investigates the role of groundwater and surface runoff in controlling suspended sediment dynamics in a Mediterranean mountainous catchment using a chemically-based hydrograph separation approach. The study is supported by an extensive and valuable dataset covering multiple years and a large number of flood events, which represents a significant strength. The manuscript is generally well structured, clearly written, and appropriately referenced, reflecting a substantial amount of work.

The study addresses a relevant scientific question, and the proposed approach—particularly the attempt to account for temporal variability in end-member chemistry—has the potential to provide useful insights into hydro-sedimentary processes.

However, the manuscript presents, in its current form, a mismatch between the level of methodological complexity and the strength and clarity of the conclusions. In addition, some inconsistencies and methodological assumptions require clarification.

Overall, I consider the manuscript promising and potentially suitable for publication after moderate revisions, if the authors better justify their methodological choices and refine the interpretation of their results.

 Specific Comments

One of the strengths of the manuscript is the use of multiple groundwater end-members within the EMMA framework. However, it is later stated that distinguishing between groundwater sources is not critical for the study. This creates ambiguity regarding the actual role of groundwater heterogeneity in the analysis. The authors should clarify from the beginning whether the primary objective is (i) to separate surface runoff from total groundwater, or (ii) to characterize different groundwater sources. The level of methodological complexity should be more clearly aligned with this objective.

The aim of the study is to distinguish groundwater (including subsurface and deeper groundwater) from water that has run off hillslopes during rainfall events. However, since groundwater originates from different geological formations within the watershed (limestone, marl, molasse, gypsum), it is not possible to assign a single chemical signature to characterize it. Instead, we must account for the variability in the geochemical signatures of the different groundwater sources in order to isolate the runoff component. This is why a four end-member EMMA model was developed, even though the primary focus of the study is ultimately to isolate the runoff flow rate. The respective contribution of each of the three groundwater end-members are at the end gathered in one to focus on runoff flow rate and groundwater flow rate.

A line has been added line 293-294 to clarify this necessity of considering several GW EMs.

The dynamic adjustment of end-member chemistry is an interesting development, but its added value is not fully demonstrated. The manuscript would benefit from a clearer quantitative assessment of how much this approach improves hydrograph decomposition compared to simpler alternatives. More generally, the necessity of using multiple groundwater end-members should be better justified if the final interpretation focuses mainly on SR vs GW partitioning.

The interest of the approach of dynamically adjusting end-member chemistry is detailed in the appendix, but we summarize here its key quantitative benefits. First, our analysis shows that the EMMA implementation with adjusted end-member chemistry (MB_fit) yields in average lower SR contributions for samples collected outside flood periods compared to the non-adjusted approach (MB_mean). This behavior is more consistent with hydrological expectations, as surface runoff contributions should be equal to zero under baseflow conditions, which is not always the case when using the MB_mean approach.

Second, MB_fit results are closer to those obtained with MB_last (Fig. A4), which relies directly on tracer concentrations measured in tributaries during the most recent spatial survey performed out of flood events and can therefore be considered as an observation-based reference. This agreement supports the relevance of the dynamic adjustment.

Third, we provide a quantitative assessment of the impact during flood events (Fig. A5). Under dry antecedent conditions, MB_fit increases SR contributions by up to +10% on average relative to MB_mean, whereas under wet antecedent conditions it decreases them by up to –6%. This contrasting behavior is physically consistent with the exponential adjustment applied in MB_fit: at low discharge, increasing groundwater tracer concentrations enhances the relative contribution of SR in the mixing model, while at high discharge, reducing these concentrations brings groundwater signatures closer to stream water, thereby lowering the estimated SR contribution.

In addition, the uncertainty associated with SR estimates is reduced when using MB_fit compared to MB_mean.

The text in the manuscrit has been developped lines 314-332.

Regarding the use of multiple groundwater end-members, it is clarified in the modified manuscript (line 293-294) that this choice is necessary to represent the spatial variability of groundwater geochemistry within the catchment.

The manuscript shows that suspended sediment concentration correlates more strongly with surface runoff than with total discharge, which is an important result. However, the relationships remain highly scattered, even when using surface runoff. The discussion could be strengthened by more explicitly distinguishing the role of additional factors such as sediment supply limitation and transport capacity.

The discussion highlights that the persistent seasonal variability observed in the SSC–Q_SR relationship indicates the influence of additional factors, such as sediment availability and transport capacity. Specifically, it is proposed (line 441) that this variability may be linked to changes in sediment supply, as reflected by variations in the coefficient a. It is also noted (line 461) that transport capacity may be limited during summer, likely promoting sediment deposition. These mechanisms not only explain the seasonal variability in the SSC–Q_SR relationship, but also contribute to the overall scatter observed in this relationship, even when seasonal dynamics are not explicitly considered. To clarify this point, a statement has been added (line 439-441) explicitly addressing the scatter in the SSC–Q_SR relationship, and the potential causes of this dispersion.

As I said before, the dataset is extensive and very valuable, taking into account the difficulties of field measuring. However, out of 86 flood events, only 27 are used for high-frequency hydrograph decomposition and detailed analysis. Since several key conclusions rely on this subset, its representativeness should be more explicitly discussed.

The restriction to 27 flood events, out of 86, results from our selection criterion: we retained only those events for which the Eckhardt filter yields results consistent with the MB_fit hydrograph separation (RMSE < 0.5 m³/s). This filtering step is essential to avoid interpreting results based on hydrograph separation methods that are inconsistent with the chemically derived decomposition.

The resulting subset includes 12 summer and 15 winter events, which we believe still constitutes a sufficiently large dataset for robust analysis. In addition, all correlation analyses include only statistically significant relationships (p-value < 0.05).

We should stress that not all analyses and interpretations in the manuscript are based on this restricted subset. The assessment of seasonal variability in SR and GW contributions (Figures 5c and 5d), as well as the SSC–Q and SSC–Q_SR relationships (Figure 7), rely on the full geochemical dataset. Furthermore, hysteresis analyses are conducted on all flood events recorded since 2019 (89 events), while only the Q_SRmax indicator is limited to the 27 selected events.

Regarding representativeness, the distribution of peak discharge and suspended sediment concentration (SSC) for the 12 summer events is 2.3 ± 2.5 m³/s and 27 ± 20 g/L, respectively, while for the 15 winter events it is 7.3 ± 7 m³/s and 13 ± 10 g/L. For comparison, the distributions of all events since 2008 show values of 1.3 ± 1.5 m³/s and 52 ± 60 g/L for 77 summer events, and 7 ± 8 m³/s and 12 ± 12 g/L for 69 winter events. These comparisons indicate that the 27-event subset falls within the overall distributions of flood characteristics observed since 2008.

To clarify this point, an additional sentence has been included in the manuscript specifying that we verified the representativeness of this subset with respect to the average seasonal dynamics of the catchment over the past decade. The added sentence line 334  is: "It has been verified that this subset is representative of the mean seasonal dynamics observed in the catchment since 2019."

Another concern is the hysteresis analysis. The analysis of hysteresis patterns is one of the key elements of the manuscript and is interpreted mainly in terms of sediment sources (hillslope versus riverbed processes). While this interpretation is consistent with the results, it would benefit from a more explicit discussion of the relative propagation of discharge and sediment waves. Differences in travel times and connectivity between water and sediment sources are likely fundamental drivers of the observed clockwise and anticlockwise hysteresis loops. Explicitly addressing this aspect would strengthen the physical interpretation of the results.

The focus of this study is on the seasonal variability of the hysteresis index, particularly on explaining the shift from negative hysteresis in winter to positive hysteresis in summer. It is true that differences in propagation velocities between flood waves and sediment waves can enhance anticlockwise (negative) hysteresis patterns, especially if sediment sources during summer are located farther from the outlet than those active in winter. While winter rainfall events tend to be widespread across the entire catchment area, summer rainfall events take the form of thunderstorms, which can be more localised (Legout et al., 2021). However, no seasonal trends in rainfall have been identified between the two rain gauges located upstream and downstream of the catchment area. Moreover, we consider that differences in travel times of the flood and sediment waves alone cannot account for the observed transition to positive hysteresis in winter. This shift likely requires a substantial change in sediment sources, such as an increased contribution from riverbed sediments. This is why this aspect was not emphasized in the original manuscript. To address this point, we have added a sentence to the discussion (line 460): “Velocity differences between sediment and flood wave propagation can favor anticlockwise hysteresis patterns when sediment sources are located far from the outlet.”

For instance, distributed monitoring along the river network (e.g., stage and turbidity sensors at multiple locations, not only at the outlet) could significantly strengthen future work. Such measurements would allow direct observation of the spatial propagation of flow and sediment waves, and would provide better constraints on hysteresis mechanisms, and improve understanding of sediment connectivity within the catchment.

We acknowledge that additional distributed measurements along the river network would greatly improve the understanding and constraint of these processes. Such data are not available for this study but could strengthen future work.

The manuscript is generally well written and structured. However, some methodological sections are dense and could be streamlined to improve readability. Clarifying the logic of the EMMA workflow and emphasizing the key steps relevant to the main conclusions would be beneficial.

Minor (Technical) corrections

Introduction

Line 55:  For coherence with the previous statements please change “catchment scale Goodrich et al. (1997); Ke and Zhang (2024); Mayor et al. (2011)” by “catchment scale (Goodrich et al., 1997; Ke and Zhang, 2024; Mayor et al., 2011),”

It has been modified line 55

Line 130-133. Same than before. Give coherence in referencing

It has been modified line 130

Data and Methods

Line 180: It is mentioned that there are two local weather stations, but in Figure 1, there is only one

Yes, there are two weather stations on the catchment, but the data of only one of them is used in this study, because there are some gaps in the data of the other one (due to technical problem). The manuscrit was modified line 180 to speak about only one weather station to avoid confusion.

Line 183: Are all the variables measured in the hydrological station?

Continuous measurements of flow rate, SSC, and electrical conductivity from 2019 to 2025 at a 10-minute time step are measured at the hydrological station (it has been specified line 182). Among the 667 water samples collected for chemical analyses, 579 were collected at the hydrological station, and the other ones were collected across the catchment.

Line 190: Locations indicated by crosses in Fig. 1. In the pdf, they are stars.

Yes, it has been modified by « stars » line 191.

Line 302: (tributariesArbresandGautier). Change form italics to regular and separate

Yes, it has been modified line 191.

Figure 7: Why in figure 7B there are uncertainty bars and not in the 7A?

Figure 7A represents the relationship SSC/Q, whereas figure 7B represents the relationship SSC/Q_SR. There are uncertainty bars on figure 7B because these uncertainties refer to uncertainties associated with the hydrograph decomposition. This is specified in the legend of the figure.

Line 370-371: If I am not wrong, the higher sediment volumes are exported in autumn-winter floods

The text in the manuscript is: “with higher sediment exports observed in spring/summer compared to autumn/winter for a given V_SR.” This interpretation is supported by Figure 6b, which shows that for a given value of V_SR, sediment export (V_s) is higher in spring/summer than in autumn/winter, as indicated by the blue points plotting above the green points.

It is also true that total sediment export is generally higher in winter than in summer. However, this is primarily because V_SR is also higher in winter. In other words, while absolute sediment export may be greater in winter, the sediment export efficiency (i.e., V_s relative to V_SR) is higher in spring/summer.

Figure 8: There are less points in figure 8D then in the previous (A, B, C). Transparent points (8B) are not marked in Figure 8D

It was a mistake in figure 8D, the figure has been modified.

Table 5: Clarify that the correlation matrix is presented in a reduced (non-symmetric) form.

It has been clarified in the legend of Table 5.

Line 395: (Table ??)

It has been modified by Table 7 line 402.

Line 400: as stated by Vereecken et al. (2019). If not (Vereecken et al., 2019)

It has been modified by (Vereecken et al., 2019)  line 407.

Line 456: Did you analysed GSD of the samples? If it is done, you should mention it.

Grain size distribution analyses were not conducted as part of this study. However, such analyses have been performed on badland hillslopes in a nearby catchment (the Laval basin) in Ariagno (2023). The manuscript has been revised to explicitly refer to Ariagno (2023) (line 465-467) when discussing the seasonal evolution of sediment grain size on hillslopes.

Line 461: You mention longer durations (average 46h), but this variable it has not been analysed, and under my point of view is interesting to understand sediment volumes transported in autumn-winter, and spring-summer floods.

 A line has been added in Table 4 to present seasonal variability of flood duration. A remark on seasonal variations of flood durations has been added in the result part line 345 and 347.

Recommendation

Overall, this is a valuable and data-rich study addressing an important scientific question. The manuscript demonstrates a substantial effort in data collection and analysis, and it is generally well structured and referenced.

However, the current version would benefit from a clearer alignment between methodological complexity and the strength of the conclusions, as well as from a more physically grounded interpretation of the results. Addressing the points above would significantly improve the clarity and robustness of the manuscript and make it suitable for publication.

I therefore recommend publication after moderate revisions.

We thank the reviewer for these constructive remarks that helped us to improve the article We answered all remarks in the document attached.