The temperature index method is a convenient and widely used method in snow simulation. This study estimates the important parameters in the temperature index method based on the published SWE and climate datasets, and analyzes the influence factors of these parameters. The results can provide insights into the temperature index method, making this paper worth publishing in HESS. Having said that, I would like to point out some major concerns that should be addressed before publication.

1. Potential Circular Logic: The temperature threshold and melt factors are estimated based on the SWE dataset, which is subsequently used to evaluate the performance of the temperature index model. This approach risks circular reasoning. A more rigorous method would involve dividing the dataset into two subsets—one for parameter estimation and another for model validation.
2. Subjectivity in Determining the Melt Threshold: In Section 4.1.2, the melt threshold appears to be assumed as 0°C, with supporting analyses provided. However, if the threshold were slightly adjusted around 0°C, the conclusions in Section 4.1.2 would still hold. To minimize subjectivity, a quantitative approach should be employed. While the melting process is expected to occur at 0°C from a physical standpoint, the temperature data used do not precisely reflect the conditions at the exact location where phase changes occur (e.g., the snow surface for melting and the atmosphere for precipitation partitioning).
3. Clarity in Time Scale: The study computes temperature thresholds and melt factors at the daily scale in some instances, while averaging them in others. This inconsistency makes the methodology difficult to follow. A clearer distinction between different time scales should be provided.
4. Effectiveness of a Single Parameter Set: The model employing a common single parameter set outperforms the other two models in certain aspects. This raises the question of whether complex parameter estimation methods are necessary. A discussion on the added value of these methods compared to a simpler approach would be beneficial.

Minor issues:

• L106: Provide the full name of SNOTEL
• L123: Ensure consistency in terminology (e.g., CHCN-d, CHCNd, and CHCN-Daily).
• 1: Make it clear what you are specifically referring to here. There are more than two indices in the Table 1.
• L162: “toe” should be “to”
• L163-166: Difficult to understand. Please rephrase and explain it more clearly.
• 4: Consider merge sections 3.2 and 3.4, moving the descriptions of the model to the beginning of 3.4 section.
• Figure 3d: what does the dark orange mean?
• L257: Why is a threshold lower than 0℃ not allowed?
• In some figures, the text is overlapped. Please check and modify them.

Temperature index models can provide valuable estimates of snowpack characteristics and hydrological variables. One such model is examined in the present preprint. While there are compelling reasons to use simplified modelling, it must be acknowledged that state of the art snow modelling today can be much more complex, explicitly including multiple snow layers, snowpack energy balance, and more processes (wind redistribution, snowpack temperature gradients, liquid water in snowpack, refreeze). I found interesting ideas in this study, including the attempt to find a relationship for the spatial variability of the melt factor. However, I am concerned with the authors' claim to "comprehensively assess temperature index model assumptions, parameterizations, and performance across a range of snow climates". The study's scope is narrower than this and there are some issues that I believe need to be addressed.

1. Suitability of the datasets. This SWE dataset, derived from snow depth observations and co-located with precipitation and temperature observations, covers extensive spatial and temporal domains. This is a great strength. However, the temperature index model linking precipitation, temperature, and SWE relies on a balance between accumulation and melt processes. Therefore, the fact that the observational datasets cannot detect when both accumulation and melt have happened within one time step is problematic. This is addressed as a limitation for snow decreases in L161, but the same problem could affect the assumption in L151. Findings presented in L198-203 could also be affected by days with both melt and accumulation occurring, in addition to the explanations offered by the authors. L210-220 suggest that some SWE changes are not detectable in this dataset, but once again this could be confounded by a mixture of processes occurring in one time step. While small SWE changes may indeed be missed in the dataset, this means there is added uncertainty on derived estimates such as onset of snow, peak SWE, onset of melt, etc. L244-246 highlight more errors implicit in using this dataset for this work. These issues could be examined by using some other dataset to examine the magnitude of errors introduced by some of the provided plausible explanations.

2. Structural uncertainty introduced by binary thresholds, some thresholds not varied. The snow accumulation and snow melt thresholds are both assumed to be strict cutoffs for the respective processes. However, observations (e.g. Dai, 2008) suggest that there is a smooth rain-snow phase transition, though much uncertainty and fundamental difficulties remain (e.g. Jennings et al., 2025) . What is the consequence of this structural assumption for the temperature index model? Furthermore, while 0C is a reasonable guess for a melt threshold, and the data does not contradict it, there is no test provided to show that it is the best choice and there is even some indication that another choice would be more suitable (as described in section 4.2, 5.1, and 5.3). Further tests could examine the melt threshold's effect on the relative performance of the three simulations. 

3. Needs more focus on value added. There are interesting ideas in this paper, despite some data and methodological challenges. I believe that refocusing on the regional results is more critical than summary statistics covering the large regions. We see in Figure 10 that many regions remain poorly represented (description L314-318), but the reasons are left mostly unexplored. Further, while the median performance for most snow metrics is good (Figure 9), the box and whiskers show a large range. Which regions specifically contribute to this wide spread? How do all of these results comparable to observational uncertainty or model uncertainty from state-of-the-art snow models? Re-focusing the study on new findings and placing the results in context is needed to justify the conclusion that the model performs satisfactorily. While particularly intriguing because it could offer a way to extrapolate melt factors from the point-scale, the multi-linear regression model for melt factor also appears minimally predictive. In addition to the good practice of preserving a random subset of data from the regression in order to use it for testing, other variables could be added (longitude, mean SWE amount, perhaps a typical "snow type" (Sturm & Liston, 2021).

Other questions and minor points:

• L62 typo "toe" should be "to"
• L112-115/L445-453: How much missing data is there that needs gap-filling? Could this affect any of the results (e.g. giving deltaSWE=0)?
• L170: It could be possible for melt to occur on days with negative temperatures (e.g. if some period of the day exceeded 0C). Could it also be explained by the snow depth decreasing due to wind compaction?
• L235: Would equation (5) be sensitive to changes in the binary structure of equation (3)?
• L242 typo "is only be" should be "is only"
• L250 typo "blue and read" should be "blue and red"
• L260-261, L375-376: It would be good to really highlight these spatial results which show where temperature index models can or can't work robustly.
• L386-388: It is somewhat misleading to present hemispheric scale results when there are such big regional differences. Could focus the summary text onto the range of results in Figure 10, perhaps then aggregating results (e.g. for Western North America, Arctic, Europe, and Scandinavia) to give median summary statistics with regional grouping.
• L404-407: I think this is a key aspect for discussion. Should be revisited in the Conclusions section.
• L416-418: Could reference other data products that use simple snow modelling and data assimilation of snow observations already (e.g. CMC daily snow depth product; Brown and Brasnett, 2010), or else clarify if "assimilation" used here means something different.
  
  L420: Could the SWE decreases below freezing be due to some other explanation, including observational errors? If such errors are potentially present, then could they affect your other findings?
• L419-423: There were no tests showing 0C to be more valid than another choice of melt threshold, or another method of partitioning precipitation into rain and snow. It was fixed throughout the whole study.
• L424-427: How do these melt factors compare to those found in the literature? With respect to the values of the two rates, this is still a large range. While it has been diagnosed that there are regions where there is high interannual variability in melt factors, what might cause this variability? Does it disqualify temperature index modelling from being used in those regions?
• L429: Modelling peak SWE requires the right balance of accumulation and ablation processes. Which of these is most contributing to these challenges?