Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This manuscript presents an interesting and useful modelling tool. Before considered for publication, the following main issues must be addressed. First, the technical details are not sufficiently clear; second, the uncertainty of the model is not appropriately addressed.

About technical details: the manuscript presents details of how to compute erosion due to different reasons but the hydrological part is not described in sufficient detail. Also , the description of the erosion module also have many places that need further clarification. About the uncertainty of the model: many empirical parameter values are not developed for the concerned area; are the good model performance a result of the mutual offsetting of uncertainties among different parameters or truly well-represented by the model parameters?

1. Moreover, Eq. (14) is very old (45 years ago). A short review on alternative relations should be included.
2. Is Drd in Eq. (2) the same as Dr in Eq. (3)?
3. Is Dcf in Eq. (2) the same as Df in Eq. (8)?
4. In Eq. (2), how is SD computed?
5. In Eq. (8), how is G computed?
6. Lines 164-174: is it rational to have raindrop detachment if the water depth at some cells is sufficiently large.
7. While Eq. (2) computes the sediment yield for each cell, how is this amount be transport? A relevant equation is required in the manuscript.
8. (3): How are K and C determined? Can you provide their distribution in the revised manuscript? Are they constant with time? Is the rainfall intensity P uniform across the space and constant with time? Does the grid size affect the results (note I suppose grid size affects the cell slope and thus may affect the final results)?
9. (4): how are the runoff volume Q, peak runoff rate computed? While P is explained as erosion control practice in Eq. (4), it was explained as rainfall intensity in Eq. (3). How is LSp computed? I think Eqs. (5, 6) should appear before Eq. (4), as C and K appears also in Eq. (3).
10. Line 212: How are the alpha and beta values in Italy applicable to the present concerned areas?

Author Response

Response to reviewer 1

Major Issue 1 – Insufficient Clarity of Technical Details

1. Comment:
   “The manuscript presents details of how to compute erosion due to different reasons but the hydrological part is not described in sufficient detail.”

Response:
Thank you for highlighting this concern. In Lines 155–156, we direct readers to the description of TOPMODEL, as its structure and application have been extensively documented in previous research, including our own prior work. Nevertheless, we have clarified the distinction between calibration and model operation with the erosion module, emphasising the different time resolutions used (Lines 156–162). Additionally, the complete source code is openly available on Zenodo (https://zenodo.org/records/15575730), ensuring transparency and reproducibility.

2. Comment:
   “The description of the erosion module also has many places that need further clarification. About the uncertainty of the model: many empirical parameter values are not developed for the concerned area; are the good model performance a result of the mutual offsetting of uncertainties among different parameters or truly well-represented by the model parameters?”

Response:
We appreciate your thoughtful observation regarding parameter uncertainty. Indeed, many empirical parameter values in the erosion component of TopEros were adopted from published literature and not specifically calibrated for the study area, due to limited local data (especially parameter alpha and beta in Eq. 5). As a result, we cannot definitively determine whether model performance is due to parameter interactions or the accuracy of the parameters themselves. To address this, we will (1) add explicit text describing the sources and limitations of parameter values, and (2) plan to conduct sensitivity analyses in future work, applying TopEros to data-rich catchments. We also note that, as observed during hydrological calibration, equifinality remains a possibility even in well-instrumented areas.

Specific Questions on Equations

3-1. Comment:
“Moreover, Eq. (14) is very old (45 years ago). A short review on alternative relations should be included.”

Response:
Thank you for this valuable suggestion. We acknowledge that Eq. (14) (Beasley et al., 1980) is an older, empirically derived formulation, and that more recent or process-explicit transport relations exist. We selected Eq. (14) for its discharge-based simplicity, compatibility with TOPMODEL outputs, and computational efficiency. In the revised manuscript, we have added a concise discussion of this choice (Lines 291–293). Due to time constraints and limited local data, we did not perform a full comparative sensitivity analysis, but we commit to targeted sensitivity tests in future work. We also note that Eq. (14) continues to be used in recent studies, such as Wang et al. (2010), Hydrological Processes.

3-2. Comment:
“Is Drd in Eq. (2) the same as Dr in Eq. (3)?”

Response:
Yes, Drd and Dr refer to the same quantity (rain-drop detachment). We have harmonised the notation to Dr throughout the manuscript for consistency.

3-3. Comment:
“Is Dcf in Eq. (2) the same as Df in Eq. (8)?”

Response:
Yes, Dcf and Df both represent concentrated-flow detachment. We have standardised the notation to Df for clarity.

3-4. Comment:
“In Eq. (2), how is SD computed?”

Response:
SD (sediment deposition) is computed immediately after erosion in each cell by enforcing transport capacity: if the transported sediment exceeds the transport capacity (Eq. 14), the excess is deposited locally (SD), and the load is reduced to the transport capacity. This process occurs first on the hillslope, and then any remaining sediment is routed into the assumed channel, where further erosion and deposition may occur. We have added a detailed explanation of SD computation in Section 2.2.2 (Lines 186–194), with further discussion in Lines 288–295.

3-5. Comment:
“In Eq. (8), how is G (sediment load) computed?”

Response:
G represents the upstream sediment delivered to the cell’s assumed channel after hillslope erosion. It is calculated as the sum of Dr and Dse (G = Dr + Dse).

3-6. Comment:
“Lines 164–174: is it rational to have raindrop detachment if the water depth at some cells is sufficiently large?”

Response:
We agree that calculating raindrop detachment may not be rational in cells where saturation is attained quickly. However, in catchments or periods where saturation is delayed (e.g., following a prolonged dry season), raindrop detachment could be significant. Therefore, we retain this process in the model until further research confirms its insignificance across diverse catchment types.

3-7. Comment:
“While Eq. (2) computes the sediment yield for each cell, how is this amount transported? A relevant equation is required in the manuscript.”

Response:
For clarity, we have not implemented a specific routing algorithm, as distinct flow routing is generally unnecessary for small catchments at daily time steps. Instead, sediment discharge for the entire catchment is aggregated at the outlet. At the grid scale, sediment is assumed to move instantaneously from the hillslope to the assumed channel, where gully erosion may occur if the cell exceeds the topographic index thresholds.

3-8. Comment:
“How are K and C determined? Can you provide their distribution in the revised manuscript? Are they constant with time? Is the rainfall intensity P uniform across the space and constant with time? Does the grid size affect the results?”

Response:

• C: Derived from NDVI, which is derived from Sentinel-2 imagery taken during peak growing season. Particularly, using the van der Knijff et al. (1999) formulation (Eq. 5), with parameters α and β initially set as in van der Knijff (α=2, β=1) (Section 2.2.2).
• K: Computed from soil texture and organic carbon using Eq. 6 and HWSD edaphic inputs (Section 2.3.4, Lines 316–318, Eq. 6, and Lines 230–232).
• Temporal constancy: K is treated as time-invariant, reflecting static soil properties. C was varied year by year, based on annual NDVI averages (2015 and 2016), and we will clarify this explicitly.
• Spatial scale: Both K and C are calculated at the pixel scale.
• Rainfall intensity: Rainfall intensity is modelled at hourly resolution and assumed constant across the catchment.
• Grid size: We acknowledge that grid size affects cell slope and may influence results; this is an important consideration for future sensitivity analyses. For context, a 50m grid was chosen as compromise between high spatial resolution and economically available computing resources.

3-9. Comment:
“How are the runoff volume Q and peak runoff rate computed? While P is explained as erosion control practice in Eq. (4), it was explained as rainfall intensity in Eq. (3). How is LSp computed? I think Eqs. (5, 6) should appear before Eq. (4), as C and K appear also in Eq. (3).”

Response:

• Q: Surface runoff volume, clarified as the output from TOPMODEL at an hourly time scale (Line 227).
• qp: Peak runoff rate (m³ s⁻¹) (Line 228), with rainfall intensity assumed uniform within each hour due to data limitations.
• Labeling: P now consistently refers to soil erosion control practice, while precipitation intensity is labelled as I.
• LSp: Computed as shown in Eq. 7, using ArcGIS Pro with slope information from a DEM.
• Equation order:  We appreciate the reviewer’s suggestion regarding the logical flow of equations. However, after careful consideration, we have decided to retain the current order. This is because reordering the equations at this stage could inadvertently disrupt the internal consistency and flow of the manuscript, potentially introducing errors or confusion. We believe the present structure best preserves clarity and minimises the risk of unintended issues. Thank you for your understanding.

3-10. Comment:
“Line 212: How are the alpha and beta values in Italy applicable to the present concerned areas?”

Response:
We acknowledge this uncertainty. The applicability of α and β values from Italy to our study area cannot be guaranteed. A sensitivity analysis in diverse data-rich regions will be necessary to assess their suitability in future work.

Reviewer 2 Report

Comments and Suggestions for Authors

What I appreciate about this paper is the authors’ ambition to develop a model that simultaneously predicts surface erosion by water and gully erosion. Their approach—combining established components from existing models—means that the underlying mathematics are well tested and theoretically appropriate.
However, a major limitation lies in the lack of sufficient data for proper calibration and validation. As a result, several aspects of the model’s performance remain uncertain (see my specific comments below). I also have some questions concerning the model’s assumptions and implementation, which I outline next.

 

Model structure and components

The model aims to improve the accuracy of soil erosion predictions by integrating both sheet/surface erosion (typically captured by RUSLE or MUSLE) and rill and gully erosion. It is composed of three main modules:
• the TOPMODEL for runoff simulation;
• the sheet erosion and rainfall detachment modules;
• the gully erosion module;
and it also employs the FAO Penman–Monteith ET₀ equation for evapotranspiration.

Subgrid heterogeneity within the catchment is represented by threshold values of two topographic indices:

1. The Topographic Index (TI) = ln(α / tan β)
2. The Contributing-Area–Slope Index (Aᵢ tan β)

The TI, in TOPMODEL, is a hydrological similarity index expressing the tendency of a location to saturate, based on the upslope contributing area (α) and the local slope angle (β). However, the authors state that the TI “is also able to track the state of saturation of the soil surface” (line 138). That formulation is not accurate: the TI does not track saturation dynamically but rather indicates the propensity for saturation. The actual state of soil moisture depends on the water balance, which must be modeled explicitly.

Moreover, the assumption of hydrological similarity—that cells with the same TI have similar responses—is critical but may not hold across soils with contrasting properties (e.g., peat, kaolinitic clay, or smectitic clay). TOPMODEL does not distinguish between such differences, which may strongly affect runoff generation and erosion. This is an issue because, after checking the Harmonized World Soil Database version 2, the soils in the study area include Nitisols and Ferralsols – which are dominated by kaolinitic clays – while the lower part of the basin seems to have Vertisols, which are also very clayey but of a much stickier nature by the nature of smectite clays. Vertisols are known to be particularly sensitive to gully erosion, while Ferralsols and Nitisols are much less prone to it.

Still, to the authors’ credit, soil differences are reflected indirectly through the K-factor used in the MUSLE and gully-erosion components, where it influences sediment detachment (Eq. 3) and sediment yield (Eq. 4).

For gully erosion, the key variable is the critical shear stress (τₙᶜʳ).  This is calculated as function of Sand fraction, organic matter, and bulk density.  It's not clear to me where did you find the data for organic matter content and BD; to the best of my knowledge that's not standard available in the HWSD. Moreover, missing here the nature of clay seem critical kaolinite clay (Ferralsols) or smectite dominated soils (Vertisols), this critical shear stress will be very different.

Similarly, for sheet erosion, the authors use the C-factor from USLE, which they estimate as a function of NDVI using parameters α and β from Van der Knijff et al. (1999). However, those parameter values were derived for Mediterranean environments (Italy). The authors need to justify why these would be applicable to humid tropical conditions in Uganda, where vegetation dynamics and soil properties differ markedly. In addition, the method for determining NDVI (temporal averaging, seasonal window, or specific imagery date) should be clarified.

Please add for all the equations the units used ; sometimes you have them other times not.  So I could not check if it’s all consistent

 

Calibration and validation

The model was calibrated for both TI and Aᵢ tan β. However, the absence of continuous catchment-scale sediment data limits the robustness of the validation. The authors themselves note the challenge of equifinality, where different parameter combinations can yield similar outputs. This suggests that the model is under-constrained and possibly too complex relative to the available data.

Therefore, while the conceptual design may indeed be superior to existing models, the presented case study does not yet provide sufficient evidence to support that claim. The practical applicability of the model is also uncertain, since few real-world catchments will have the data density required for proper calibration and validation. Nonetheless, if future studies can demonstrate consistent performance across well-instrumented basins, these concerns could be alleviated.

 

Overall assessment

The paper presents a conceptually strong and innovative framework, but its implementation and empirical support remain limited. I encourage the authors to:
• Reword certain claims (e.g., regarding TI and soil saturation).
• Provide stronger justification for parameter transferability (NDVI–C-factor relation).
• Clarify the derivation and role of Aᵢ tan β.
• Discuss more explicitly the implications of limited calibration data and equifinality.
• Consider the important consequences that contrasting soil groups such as Ferralsols and Vertisols will have on soil erosion; at least address in the discussion how this could be taken into account.
• I strongly recommend presenting the results first and having the discussion in a separate section; your first paragraph of the results and discussion (lines 303–313) contains discussion points for which no supporting results have yet been shown.

 

Minor comments

Line 103: better to talk about “wetland rice” rather than “paddy field” — “paddy” is an Indonesian word for rice.
Line 114 (Fig. 1): we don’t need the spaghetti of administrative boundaries as the inset map — it would be more informative to show the topography of the catchment and its surroundings. You have the important city of Mbale just outside the catchment; show it, as it provides an important reference point.
Line 154: and many others — avoid using “respectively”; it makes the sentence harder to read. Your sentence
“We calibrated and validated TOPMODEL on the 2015 and 2016 hydrological season respectively.”
is easier to read as:
“We calibrated TOPMODEL with the 2015 hydrological season and validated it with data from the 2016 season.”
If you still doubt about “respectively”, next time you go to the bakery ask for bread rolls, muffins, and a loaf of bread — 3, 6, and 2 respectively.
Figs. 6 and 7: it would be much more informative to present “sediment deposition” in discrete classes with equal areas (use quantiles); the same applies to Fig. 8 – use equal-area discrete classes.
Check if all references are complete – [46] doesn’t seem to be complete; I think it should be:
Flanagan, D.C., and S.J. Livingston (eds.). 1995. WEPP User Summary. NSERL Report No. 11, West Lafayette, IN: National Soil Erosion Research Laboratory. 131 pp.

Author Response

Response to reviewer 2

Model Assumptions and Terminology

1. Comment: On TI and saturation.
   Response: Thank you for your observation. We agree that our initial wording may have been unclear. We have revised the manuscript to clarify that the Topographic Index (TI) indicates the likelihood of saturation, rather than actual saturation, in a grid cell. TOPMODEL calculates the water balance for each cell, with TI serving as an important variable (see Lines 140–143).
2. Comment: Hydrological similarity assumption across contrasting soils.
   Response: We appreciate this important point. The hydrological similarity assumption is a core aspect of TOPMODEL, but we recognise its limitations at the catchment scale, especially in areas with contrasting soil properties. In our study, this assumption was applied only during calibration of the hydrological component. When the erosion module was activated, we calculated runoff and erosion for each cell individually, abandoning the hydrological similarity approach. This distinction is now clearly stated in the updated methods section (Lines 156–162).
3. Comment: Data for organic matter and bulk density in τₙᶜʳ (Eq. 10)?
   Response: Thank you for raising this point. The values for organic carbon content and bulk density are reported in the Harmonized World Soil Database (HWSD). For our analysis, we adopted values from the topsoil layer.
4. Comment: Justification of α/β from Van der Knijff for Uganda; method of NDVI estimation.
   Response: The α and β values from van der Knijff et al. were used as a starting point, as they provided reasonable estimates for the C factor. We acknowledge that further testing in data-rich areas will be necessary to refine these parameters for local conditions. NDVI was computed from Sentinel-2 imagery, and we have added an explanation of this process in Lines 231–234.
5. Comment: Add units to all equations.
   Response: We have addressed this and ensured that all equations now include appropriate units.

Calibration and Validation

6. Comment: Limited sediment data, equifinality under-constrained.
   Response: We agree that limited sediment data constrains model validation. Application of TopEros in data-rich basins will help address this issue. However, as documented by Beven and Binley (2014) and others, equifinality can persist even in well-instrumented catchments. Overfitting is more likely related to the length and quality of observation data used for calibration.

Structure and Presentation

7. Comment: Reword claims about TI and saturation.
   Response: The manuscript has been revised accordingly for clarity.
8. Comment: Justify NDVI–C transferability.
   Response: We have addressed this by extending parameter discussion and noting the need for caution in transferability.
9. Comment: Clarify Aᵢ tan β derivation and role.
   Response: A brief explanation of the ai.tanβ index is provided in Lines 75–78, with further clarification of its role and meaning in Lines 258–266.
10. Comment: Discuss limited calibration data and equifinality.
    Response: This has been addressed in the revised manuscript.
11. Comment: Address contrasting soils in discussion.
    Response: This point has been incorporated into the discussion section.
12. Comment: Results and Discussion separation.
    Response: Thank you for this suggestion. While we recognise the value of separating results and discussion, we are mindful of maintaining the manuscript’s overall coherence. We have adopted your advice to present results first, followed by interpretation and claims, as much as possible.

Minor Comments

13. Comment: Use “wetland rice” not “paddy field” (line 103).
    Response: Terminology has been updated as suggested.
14. Comment: Inset map to show topography and Mbale (Fig. 1).
    Response: Figure 1 has been revised to show a major city (Kampala) and hill-shade. Since Mbale and the catchment location star overlap, we have used Kampala as a reference city, considering its status as the capital of Uganda.
15. Comment: Avoid “respectively” (line 154 and elsewhere).
    Response: We appreciate the stylistic suggestion. While we have aimed to minimise the use of “respectively,” we have also sought to preserve the manuscript’s overall flow and readability. We remain open to further specific adjustments if necessary.
16. Comment: Use quantile classes for Figs. 6–8.
    Response: We respectfully retain our use of continuous scales for Fig. 6 to illustrate the transition of erosion severity. For Fig. 8, we use fixed bounds to communicate erosion risk, as these values are defined by established ranges, e.g. Jiang. Et al. 2014’s “very low”, “low”, “moderate”, “high” and “very high” ranges.
17. Comment: Complete reference [46].
    Response: Reference [46] has been updated to the full citation for Flanagan & Livingston (1995) WEPP User Summary. Thank you for your attention to detail.

Author Response File: Author Response.docx

Reviewer 3 Report

Comments and Suggestions for Authors

An interesting apporach to soil erosion modelling, diversyfying the yield by three distinct erosion sources: sheet erosion, gulley erosion, and raindrop-impact. 

My main consern is lack of validation against real sedment yield/flow data. You validated hydrological part of the study, but using quite bold sentences stating TopEros quality without validating it against erosion yield data is certainly premature.

I see two options to make it much clearer for the reader:

1. You clearly indicate approach to validations as full hydrology plus cross-model validation against MUSLE and at least 1  (better 2) other model - starting form method chapter on. Then the rest of the paper may remain as is, watching the statements about TopEros performance.
2. You start with model full validation on a well data-equiped catchment eg. Southern EU, validate the mdoe fully against both data and MUSEL. After that I think you are free to excercise TopEros on any ther catchment in the region. 

Comments for author File: Comments.pdf

Author Response

Response to reviewer 3

Validation against sediment yield/flow data

1. Comment: Validation on hydrology only is premature.
   Response:
   Thank you for this important observation. We agree that validating only the hydrological component is not sufficient to fully establish the reliability of TopEros. To avoid overstating our results, we have revised the manuscript title to better reflect the current uncertainties associated with applying TopEros in a data-scarce catchment. We are actively working to test the model in data-rich environments and hope to report these results soon, or encourage other researchers to do so as well.
   To strengthen our current validation, we have added a multi-model cross-validation in Section 3.2.7, where TopEros outputs are compared against:

• MUSLE applied at the catchment scale,
• The empirical Vanoni sediment delivery ratio (SDR),
• Published RUSLE results from similar catchments.

2. Comment: Option: cross-model validation vs. data-rich catchment.
   Response: We appreciate the suggestion to consider alternative validation strategies. In this study, we have chosen cross-model validation as the most feasible approach given current data limitations.

Comments on the pdf file

1. Comment: What’s the data for sediment calibration?
   Response: Basically, we did not have data to validate the sediment yield for the catchment.
2. Comment: Please provide more info on the plot setup and methodology of sample collection.
   Response: Thank you for this comment. This data is insufficient to validate the erosion model. We only mentioned it since it was part of the data collection efforts during the study. We describe this further in Lines 311-319.
3. Comment: Please provide more info on the quality of this crucial dataset. Was it pit-filled? Did you check it for flaws eg. extreme slope pixels or flat pixels - fake terraces?
   Response: Thank you for this comment. This DEM data is pit filled (Lines 321-322). We did a manual inspection of the DEM and its derivatives, e.g., slope and topographic index. One important strategy was to confirm continuity of major streamlines (This was confirmed, as can be seen in Extended Figure 4, panel (a). flow accumulation). We have added spatial information of the same in Extended data Figure 4 and Extended data Figure 5.
4. Comment: Could you compare its resolution/scale to the real soil variability in the assessed area?
   Response: All raster datasets were harmonised to the final DEM resolution: 50m.
5. Comment: Without calibrating the sediment part of the overall model the unique approach can’t be really validated. Consider at least cross-validation with MUSLE model oraz 1-2 other models, if you lack the sediment concentrations data. I see you compared the results to MUSLE in further results, good. :)

Response: Thank you for this comment. Indeed, we cross-validated with MUSLE. We also toned down the language to not overstate our findings.

6. Comment: please provide some more statistics for LS and K values: mean and median at least

Response: Thank you for this comment. Statistics for LS and K values have been provided (Lines 405-409). Further, we have provided spatial distributions of the same and boxplots depicting their summary statistics in Extended Data Figures 4 and Extended Data Figures 5 respectively.

7. Comment: comparable to toher models rather than realistic. For realistic you need more solid evidence - comparied to real life sedminet flow data or at least some plot results - that you have mentioned in the first chapters.

Response: Thank you for this comment. We again note that the plot results were insufficient to perform a reasonable validation. We have therefore toned down the language, to emphasise that the results of TopEros were only compared to the empirical MUSLE , Vanoni’s empirical relationship of sediment delivery ratio and results from previous studies in the tropics (Lines 464-466).

8. 8. Comment: Again, you are basing on another model not measurement. be careful

Response: Thank you for this comment, again. We again clarified that TopEros was compared to another model (the empirical relationship of Vanoni for sediment delivery ration (Lines 476-477)

9. 9. Comment: Nice comparisons at regional and geoclimate zonal scale, but this is not the proof that TopEros is better off with predicting REAL values. If you do not have any data on sediment flows in rivers or yields i controlled conditions I'd suggest to state it clearly already in methodology and point at comparisons with MUSLE and othe rmodels as validation. This will be clearly honest and will also automatically indicate limitations of the model. If I where you I will write an another paper trying Toperos on a catchment with all datasets required for full calibration - one of Southern EU catchments for example. That would truly validate the approach. Then you can use the already validated tools region-wide.

Response: Thank you for this comment, again. Again, we now avoid overstating our results. We now clearly mention in the Methods that TopEros was compared to MUSLE, other empirical relationships and studies in other tropical basins, through a new sect. 2.4 (Validation of the erosion model) (Lines 329-338). The entire manuscript has also been tweaked to intentionally reflect this honest fact.

10. Comment: I think the title of the paper is a bit too early. It will be valid if you have had full validated model. This paper could be rephrased as comparison of new approach TopEros against MUSLE, USLE etc. But again - be very cautious with formulating conclusions and statements approving TopEros without the real validation against sediment data.

Response: Thank you for this comment, again. We have updated the title to reflect this fact. The title is now “TopEros: an integrated hydrology and multi-process erosion 2 model—a comparison with MUSLE”

11. Comment: Link to reference [24]

Response: Thank you very much . This has been updated

 

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

Thank you for tkaing my suggestions into consideration. The title and conetnt now corresponds to your findings, considering that the models has not been calibrated strictly to real-life measurements.

Author Response

Response to editor

We thank the editor for the clear and constructive guidance. We appreciate the acknowledgment that we have addressed the most serious criticisms regarding validation and are grateful for these final, specific instructions to bring the manuscript to a publishable standard. Accordingly, we have revised the manuscript to address the outstanding issues. Find our response below:

Comment one: "Could you compare its resolution/scale to the real soil variability in the assessed area? How was this harmonization achieved? How is the soil data resolution related to the 50 m grid size?"

Response to comment one:
Thank you for the important follow-up question. This must the concern raised by reviewer 3. We apologise that the previous response was quite inadequate. We have added a full clarification to the Methods section (Sec 2.3.4) and address the points as below

1. How harmonisation was achieved: The final resolution for implementing TopEros was 50m x 50m. This was a harmonisation of two key data sources at different scales:

• Topography (DEM): The source topographic data was a 12.5m-resolution DEM. This fine-resolution DEM was aggregated to the 50m grid to align with our final analytical scale, which was chosen to balance detail with computational constraints.
• Soil properties: The edaphic parameters (sand, silt, clay, organic carbon) were sourced from the Harmonized World Soil Database (HWSD) v1.1. This is a global, coarse-resolution dataset (~1km) that provides generalised soil properties. These coarse-scale properties were resampled (disaggregated) to our 50m analytical grid. This means that each 50m grid cell was assigned the soil properties of the single, large HWSD map unit that it fell within.

2. Relationship to real soil variability:

The reviewers are correct to infer a significant scale mismatch. Our 50m grid serves as the computational scale/framework, but it does not imply we have 50m-scale soil information.

• As evidenced by Extended Data Figure 4e and Extended Data Figure 5e, the HWSD provided generalised, near-uniform soil properties for the entire catchment.
• Therefore, the spatial variability in our simulated erosion results is not majorly driven by fine-scale soil heterogeneity. It is almost entirely driven by the high-resolution topographic variability and the model's sub-grid zoning logic that partitions flow into sheet and gully domains based on topographic indices.

We have added Lines 333– 341 accordingly. Further, we have also added a sentence to the Limitations section (3.3) to reinforce that the use of coarse global soil data is a key limitation and that future work would benefit from local, high-resolution soil mapping (Lines 575-577).

 

Comment two: " The authors responded to a comment about the reasonableness of including raindrop detachment where water depth is relatively deep: 'We agree that calculating raindrop detachment may not be rational in cells where saturation is attained quickly. However, in catchments or periods where saturation is delayed (e.g., following a prolonged dry season), raindrop detachment could be significant. Therefore, we retain this process in the model until further research confirms its insignificance across diverse catchment types." However, in lines 457-8 of the revised manuscript, they specifically comment on the possible reduction of raindrop impact during saturation. Thus it seems reasonable to at least mention this issue earlier, or to link the observation in 457-8 to issues of representing this process in the model."

Response to comment two:
Thank you for pointing out and following up on this issue. The editor is correct: our Methods section should better align with our findings in the Results. We have now added a qualifying statement to the Methods (Lines 217 – 221). This new text directly links to our later observation (in Section 3.2.4) that the process was negligible in our specific catchment.

Comment three: " Comment 3-7 addresses sediment routing, and the response addresses the issue raised. But the manuscript should include an explicit statement that sediment is assumed to be routed instantly to an assumed channel.."

Response to comment three:
We agree that this assumption must be stated explicitly. We have added a sentence to Section 2.2.2 to clarify the two-step "instantaneous" assumption: 1) from the hillslope to the assumed channel, and 2) from the channel to the catchment outlet (due to the lack of a routing algorithm) (Lines 192-196).

Comment four: " Comment 3-10 (and other reviewer comments) refers to the use of alpha and beta values from Italy. The manuscript should make the same acknowledgement of uncertainty as in the authors' response."

Response to comment four:
This is an excellent point. We have taken the acknowledgment of uncertainty from our reviewer response and inserted it directly into the manuscript in Section 2.2.2, where the parameters are introduced (Lines 238 – 241).

Author Response File: Author Response.docx