It was a please to read the manuscript on “Modelling runoff in a glacierized catchment: the role of forcing product and spatial model resolution”. The study analyzes model performance as a function of spatial resolution of the modeling domain and the choice of the precipitation products. The findings of the study are essential for finding optima between computational effort and minimum resolution needed for accurate glacio-hydrological simulations. The manuscript is well-structured and is well-written.

I appreciate that the authors have distinguished between spatial resolution of input data (i.e., precipitation) and resolution of model elements. However, I find that the effect of precipitation resolution is not well isolated in this study as it compares two things simultaneously, namely different source of precipitation (i.e., interpolated gauged and reanalysis) and different spatial resolution associated with each of the selected dataset. I think this aspect can be easily addressed by running additional simulations. Please find my detailed comments below.

Kind regards,

Larisa Tarasova

General comments

Choice of the precipitation products for the comparison: The rationale for selecting exactly these datasets (interpolated gauge-based dataset and two reanalysis ERA5 and ERA5 Land) is not clear to me. Particularly, it is not clear why two reanalysis products are compared, while the satellite and hybrid products are not selected Moreover, the Section 2.2.1 does not provide any information whether their performance was tested with the in-situ observations in the region. Please revise and clarify

Spatial resolution of precipitation: The narrative of the manuscript indicates that the goal is to investigate the effect of spatial resolution of precipitation input. However, in the experiments it is not only the resolution changes, but also the source of precipitation. In Figure 2 it is clearly visible that datasets are associated with different seasonality of precipitation among interpolated and reanalysis products. Given how different are the sources of precipitation, the effect of spatial resolution cannot be isolated. I think this can be easily fixed by upscaling (i.e., artificially increasing the resolution) of the same product (e.g., interpolated gauge-based precipitation) by several factors.

Specific comments

Line 12-13: At this point in the manuscript, it is not quite clear what is meant here by the constat precipitation adjustment. Please revise and clarify this part.

Line 40-45: It is important to mention here that gridded datasets are not always interpolated products, but can also be reanalysis and satellite data.

Line 47: It might be worth mentioning here the work of Pena-Guerrero et al. 2022 (doi: 10.1002/joc.7548) that compares the performance of different global precipitation products over complex terrain.

Line 119-120: Please explain this method in more detail and provide the corresponding reference.

Figure 1: Please explain acronym ELA in the caption

Line 142-145: Please explain this method in more detail and provide the corresponding reference.

Line 151: Please explain how the extrapolation is done.

Line 175-180: Please clarify how the lapse rates are computed and whether or not they are recomputed for different spatial resolutions. Please provide the estimates.

Line 183: It is not clear how this is done. Please clarify.

Line 228: It is not clear why precipitation correction factor represents accumulation parameter. Please clarify.

Table 3: Please clarify if these are best calibrated parameters.

Von der Eschet et al. present an important and interesting work in terms of modelling, which aims to simulate the glaciological and hydrological functioning of a catchment area of 39.4 km2, of which 16.7 km2 (44%) are glaciated. However, both the novelty and the relevance of their conclusions are not immediately obvious. The conclusions that the model simulates the runoff better when it is calibrated against this runoff, that reducing the model resolution reduces its capacity, and that a model resolution should be adapted to the size of the simulated object seem so trivial that more information is needed to convince the reader that this is not the case. To improve the manuscript, some important issues need to be addressed (Major comments 1, 2 and 3), and some specific comments should be considered (see below).

Major comments:

MC 1: Ability of the model to reproduce the hydro-glaciological functioning of the catchment

Since the model used here is a glacio-hydrological model, and that less than 50% of the simulated catchment is glacierized, hydrological conditions simulated for the non-glacierized part of the catchment are important on a daily time scale.

1. Non glacierized part of the model

• The description of the model, how it works and how it is calibrated is completely lacking for this non-glaciarized part. For example, what are the runoff coefficients chosen, how is the subterranean compartment considered, etc.... This can be important in term of hydrological functioning, particularly during summer rainfall events or during low flows periods.
• In addition, evaporation is low in such a mountainous environment, except in summer when it reduces the contribution of precipitation to runoff. How are the meteorological forcings applied to this part of the catchment, and how they differ from the glacier model part? How do these forcings compare with local observations (e.g. André Bernath has made precipitation and evaporation measurements in this catchment; and the Hydrological Atlas gives an estimate of the evaporation term)?
• Finally, this non-glacial part will have an impact on the separation of the types of flow (surface, underground, ice melt and snow melt). For the moment this is noted on lines 284 to 286 so there is a need to provide much more information.

1. Glacierized part of the model

The model chosen is a good choice as well as the methodology for investigating the sensitivity to the resolution of the input meteorological data and the multi-objective calibration based on mass balances and flow rates.

• However, many parameters are not detailed and are not evaluated through a sensitivity study. This is the case for temperature and precipitation lapse rates (see also the next comment). The values of all the parameters should be given and the sensitivity tests carried out should be indicated, showing the ranges of consecutive values for simulated mass balances and flow rates.
• It appears that no spin-up was performed to bring the Rhone Glacier into equilibrium with the simulated mass balance (since all simulations started with the same area: Fig.5C). As the annual mass balance varies between simulations, part of the area change (Fig.5) is due to the initial imbalance. The simulated daily discharge is mostly a function of the daily melt rate applied to the glacier surface. Since the Rhône glacier has a time response of several decades, its surface area (and volume) is due to the initial simulation conditions and not to the prescribed accumulation rate, unless a long spin-up run has been applied to equilibrate the glacier with the prescribed forcing. This problem is mentioned very briefly (pp. 348-349) but not discussed.
• For the whole model, the modelling strategy for calibration and validation (or evaluation) is not well explained. There are numerous methods of data set selection (e.g. split sample tests).

MC 2. Impact of meteorological forcing and spatial model resolution on the accuracy of glacio-hydrological simulations

A first question is what is meant by "accuracy" or "reliability" of a simulation? This depends entirely on the context. For operational forecasting of e.g. hydropower, these daily simulations are far too coarse, whereas for centennial simulations even the weakest resolution is sufficient (since the annual mass balance is correct).

As raised in the previous comment we can ask the following question: are really the meteorological forcing and the spatial resolution responsible of the accuracy differences among simulations? Objectively, the Grimsel and MSgrid meteorological series are more accurate than the ERA5 at 30km resolution. Objectively, the 25m resolution model describes much more accurately the catchment than the 3km model. However, the meteorological series have been independently corrected, and the model calibrated differently for each setup, so that the link between each forcing or resolution and the corresponding simulation is not obvious. Especially, the elevation correction applied to precipitation is crucial. It is well known that lapse rates are not consistent in the Alps. Hence, the basis and magnitude of these corrections, their interplay with the model Cprec, are important questions here. Further, the precipitation correction factor, Cprec, is exactly 1 for simulations with a varying resolution (Table 3: 100-1000m), and much lower than 1 for lower resolutions: this seems at odd with precipitation being too low compared to glacier accumulation (as it is generally noticed). In fact, Fig.5B-E shows winter accumulation of 2m, hence annual rate of precipitation of more than 3m, not found in the precipitation products (Fig.2). Even in Switzerland, which has the best observational network and the best knowledge, the question of snow measurements underestimation has been in debate for decades (the Boris Sevruk version of the Swiss precipitation Atlas had a correction by +20-30%, whereas the more recent Ch. Frei version has not.). Also, looking at Figs.6 to 8it is not obvious that the objectively more accurate forcings and resolutions lead to 'more accurate' simulations?

So, some clarification is required on the magnitude of the precipitation correction, and how corrected precipitation compares with estimates. (The Gletsch catchment has been extensively studied, see Bernath 1989; Klok et al. and references therein). Some clarification is also required to understand how a 3km-resolution catchment could 'work so well', indeed. Especially, Fig.3 shows that the area of the catchment varies with its resolution, so that simulated and observed runoffs should not compare in absolute unit (in m3/s; as in Figures 6 and 8), but only in specific unit (mm/d). Some correction of the area has been obviously done?

MC 3. Uncertainty on runoff measurements and Nash-Sutcliffe criterion choice

The caption to figure 7 states that "...the grey shaded areas indicate the months considered in this study", but this fact is not specified in the text. This choice to evaluate only the summer months is highly questionable and more details are needed.

The choice of the Nash-S parameter to evaluate the model is highly controversial. The study by Althoff and Rodrigues, JoH, 2021, shows that this coefficient should be avoided. Other options exist, such as the KGE. Could you please provide other metrics to evaluate the model?

Lines 306 to 309, it is written: 'For most resolutions (except 3000 m), the NSE decreases from April to June, probably due to delayed runoff timing, ....' This conclusion is questionable as the model is run at a daily time step. An hourly time step should be used to draw this conclusion.

To compare simulated and measured runoff, the uncertainty on measurements should be accounted for. Measuring runoff in this highly variable environment is difficult. Also, the question of a potential water underflow not measured at the Gletsch gauge station was discussed by Bernath (1989).

Finally, line 340 rightly mentions the concept and definition of equifinality, and this principle should guide this study by testing most of the parameters.

Specific comments:

SP1: daily time scale needs to be specified more clearly (abstract, introduction, etc...)

SP2: L75-77: please specify the name of the river/catchment

SP3: caption of table 1: please specify the name of the glacier

SP4: figure 1: please add the river more clearly

SP5: Figure 2: it is not clear how the box plot is made (temporal vs. spatial aggregation) please give more details

SP6: line 156: please add the calibration and validation periods

SP7: lines 159-163: please add the land cover areas

SP8: line 178: please give the values for the lapse rates (evolving in time or not?)

SP9: line 189: please give a reference used to select the T° values.

SP10: figure 4: please redo it more readable (two small font).

SP11: Line 226: How are the values chosen?

SP12: Table 3: please add the values of NSE (and other metrics, see MC3)

SP13: Lines 285-286: ...'shows that ice melt may be underestimated...' How could you conclude that? Indeed it is not possible to quantify this term 'ice melt' on the basis of observed runoff alone.

SP14: L294 : 0.6 and 0.8 for NSE are not ‘good’ , please moderate.

SP15: figure caption of figure 7, it is not possible to select only a selected period to draw conclusion. One can have some doubts about the hydro-glaciological model with NSE below 0.2 for some months.

SP16: figure 8. Please add the value for 2011 (which should be 91.9 million m-3).