This study by Xu and others evaluates the suitability of five different turbulent flux models on a glacier on the Tibetan Plateau. This is done by comparing modeled and observed turbulent fluxes over a five-month period in 2023. In a next step, the turbulent flux models are included in a mass balance model to test sensitivity of the mass balance to the choice of turbulent flux model. The manuscript is well written, presents new data, and contains useful new insights in performance of different turbulent flux models on a glacier on the Tibetan Plateau. This information will be valuable e.g. for future applications of mass balance modelling in the area. In my opinion the manuscript needs major revisions though before it can be published. One concern I have is that the description of the methods is incomplete. For example, it remains entirely unclear how the calibration of the mass balance model with the different turbulent flux models was done. Furthermore, I think it is a missed opportunity that no attempt has been made to calibrate e.g. roughness lengths to improve agreement between modeled and observed turbulent fluxes. This would give useful insight in what parameter settings of the turbulent flux models are most suitable to use on this glacier (and possibly other glaciers on the Tibetan Plateau). The calibrated turbulent flux models could then be used in the mass balance model to assess mass balance sensitivity to the turbulent flux formulations. Maybe some of this was in fact done; if so, it just needs to be described more clearly. Detailed comments are given below.

Specific comments:

Missing parameter values & errors in equations:

• Section 3: For many model-specific and physical constants (e.g. Ctub, Ctub2, Ls/f, Cp, Ch, CH etc)  values are not specified in the manuscript (unless I missed it). Please add a table with all these parameters and their values, units and, where relevant, the source.
• Equation 1: Division by pressure is missing. Please check that it is not missing in your code. See for example Radic et al. (2017). I think the corresponding equation in Oerlemans and Grisogono (2002) may also miss this.
• Equation 2: The "0.622" should not be there and instead air density should be in the equation. Also here, see Radic et al. (2017), and please check that your code is correct. 

Discrepancies between modeled and observed turbulent fluxes:

• L139-140: "the roughness lengths are computed dynamically". Does this mean that roughness lengths are time-dependent? It is important to add a bit more information about how the roughness lengths are determined and add the relevant equations from Andreas (1987). The information is relevant in order to be able to judge whether some of the discrepancies between modeled and observed fluxes  (Fig. 4) could result from this.
• L308-309: "However, ... observed variability". It seems that the underestimation of variability of modelled turbulent fluxes applies over the whole period and for different turbulent flux models. Is this underestimation of temporal variability something that other studies also found (on other glaciers)? Could it be related to the calculation of the roughness lengths? How uncertain are the turbulent flux measurements themselves? Is it possible to assign a value to this uncertainty?
• L336-341: "All five methods ... turbulent exchange."  Since all methods seem to give roughly the same errors, I am inclined to think that it could also be related to the way roughness lengths are determined (since all methods depend on them). It could be worth looking into this further and check e.g. whether an alternative description (or calibration; see next point) of roughness lengths would strongly change the model results.
• Figure 5: Given the weaker intra-day variability of in particular the latent heat flux, maybe the roughness length for the latent heat flux is off? I think it could have been an interesting strategy to try to find optimum values for (constant) roughness lengths by trying to minimize the differences between modelled and observed turbulent fluxes. Have you considered doing this?

Mass balance modelling strategy:

• Section 5.2: I can see why the authors want to run a mass balance model with the different turbulent flux schemes as it may give insight in sensitivity of the mass balance to the turbulent flux descriptions. However, in its current form I do not think the mass balance modelling experiment is of very much use. To me it would have made more sense to first calibrate the individual turbulent flux models against the observed turbulent flux data (by optimizing e.g. roughness lengths). In a next step these calibrated turbulent flux models could then be used in the mass balance model and the resulting mass balance from the runs with different turbulent flux schemes can be compared. There may be (like now) some biases between modeled and observed mass balance, but at least such an experiment will give robust insight in the sensitivity of the mass balance to using different turbulent flux models. If the above approach is what was already done by the authors, then it is just a matter of describing it more clearly.

Selecting a best turbulent flux model:

• L327: "indicating the best accuracy among the tested methods". I think it should rather be emphasized that the performance of the five methods is very similar. Given uncertainty in the observations, and these small differences, it is not possible to argue (with statistical significance) that one method is better than the others based on these results.
• L425: "the best performance". It is only marginally better than the other models. So statistically speaking the performance is not significantly different across the models.
• L549-552: "Among the five schemes ... variability.". Again, this is a rather bold statement based on the presented results. There are clear benefits for using the simpler (easier to implement and quicker to run) models too, especially if their performance is similar to the MO model.
  
  
  
  

Minor comments / technical corrections:

• L30: "a negative total ... climate condition". Replace with "the commonly observed negative total turbulent heat flux".
• L32: "extreme weather and climate event". Replace with "extreme event".
• L47: "frequencies ... events". Replace with "frequency and intensity of extreme weather events".
• L48: Not only enhanced turbulent fluxes are responsible for increased glacier mass loss but also e.g. increased incoming LW radiation. Please reformulate.
• L63: "climatic mechanisms". Please note that climate refers to much longer timescales than considered here. Maybe use "weather patterns" or similar instead.
• L64: "has become" --> "is".
• L65: "are" --> "have been".
• L68: "intensive" --> "expensive".
• L76-80: "Taking the ... 2006-2011". This comparison requires a bit more info to be useful. E.g. what methods were used (modelling or observations)? How large were uncertainties in the two studies? Could the different periods considered explain the differences?
• L88: "... on the TP.". Add "has previously been done".
• L90-92: "These uncertainties hinder ... between glaciers and climate.". This becomes a bit repetitive now. I suggest to remove it here.
• L93: "systematically analyze" --> "systematic analysis".
• L98: "the model robustness" --> "performance of the turbulent flux models".
• Section 2: I suppose surface temperature is estimated from the observed outgoing longwave radiation. This should be described somewhere. Was an emissivity of 1 assumed?
• L119: "using" --> "using a".
• L190: "pressure Ps". Please note that air pressure was already defined as "P" in L165.
• L201: Please note that the 𝛜 constant is the same as the "0.622" that is used in several equations (e.g. Eqs (1) and (3)).
• Sect. 4.1: Please make sure that italic font is used for variables.
• L224-242: This is a lot of text about the temperature time-series, which are also shown in a figure and do not contain too many surprising features. I suggest to shorten this.
• L243-268: Also this can be shortened. Essentially, everything mentioned here is also visible in Figure 2 so just a short summary would work.
• Figure 2: Units on the y-axes should be in brackets. Additionally, please use the same variable names as in the main text.
• Figure 2: You could consider adding a panel with Ta-Ts (the temperature deficit), which is an important parameter for the turbulent fluxes and their direction. Furthermore, albedo could be shown. It is discussed in connection to the extreme event in July 2023.
• L286 (for example): With an observation period starting in May and finishing in October it maybe is better to not refer to "spring" and "autumn" but rather the months in question, e.g. "May-Jun" or "Sep-Oct".
• L287-288: "with June having the lowest monthly average". I suppose that coincides with the period with the lowest temperature deficit at the surface(?). If so, that could be mentioned here.
• L291: "through turbulent heat exchange". So sublimation / evaporation are large. Related question, do you find penitentes on the glacier surface? If so, it could be worth highlighting in the introduction as it is a visible confirmation of the significance of sublimation and evaporation.
• Figure 3: Please add the type of flux in the y-axis labels.
• Figure 3: Since there are data gaps I suppose some filtering of the data was done. Unless I missed it, I do not think this is currently described. If so, please add this.
• Figure 4: Since the modeled data are unique in every graph, I suggest to plot these lines on top of the lines with the measured values. Currently it is the other way around.
• L462-463: "while CMB estimates ... -304.19 mm.". It is not surprising that there are biases as the mass balance model (EBFM), with standard values for model parameters, is not optimized for this region/glacier. Besides turbulent flux parameters, e.g. correct albedo parameters are at least equally important for the simulated mass balance. This should be acknowledged.
• Figure 7: Why not zoom in only the heat wave period? The rest of the timeseries are also shown elsewhere, so no need to include them again here.
• L506-507: "Such positive LE values ... maritime glaciers.". I agree, but also in the high Arctic where surface temperatures are typically much lower than air temperatures due to less incoming solar radiation.
• L509: "18.53". Please be consistent in the use of significant digits (e.g. I recommend to use "18.5" here).
• L512: "surface albedo decreased". I suppose the surface was already ice-covered at the start of the heat wave(?). If so, I do not see how the albedo dropped further during the heatwave. It would by the way be nice to see albedo time-series in the manuscript, e.g. in Fig. 2.
• L526: "warrant further investigation". I am not sure what is the cause either, but since the even the sign is off, I suppose it in some way must be related to the humidity gradient (which in turn depends on the temperature gradient).
• L528-531: "All calculated values ... to glacier melting.". This applies to the uncalibrated turbulent flux models. It would be great to see if that still applies after calibrating the turbulent heat fluxes (e.g. by tuning the roughness lengths).
• L535-536: "to increasingly ... increased H. ". It could be worth emphasizing that it is primarily that the surface temperature cannot exceed 0 ^oC that causes positive LE and higher H during such extreme events.

See attached comments.