the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 08 Dec 2025
Snow Mass Components Analysis: Greater Contribution to Atmospheric Water Vapor than to Water Resources on the Tibetan and Pamir Plateaus
Abstract. Snow in the high-altitude and high-latitude regions is essential for water resources and climate regulation. However, studies on snow mass balance components in alpine areas like the Tibetan and Pamir Plateaus (TPP) are limited. To fill the gap, a novel snow simulation framework was developed, combining in-situ snow depth, satellite snow cover, and point- and grid-scale modelling, supported by sensitivity analysis, automatic calibration, and deep learning. Key snow components—snowfall, snow water equivalent (SWE), refrozen snow, sublimation, evaporation, and snowmelt—were simulated across the TPP from 1962 to 2019 with reliable accuracy. Regionally averaged annual snowfall and refrozen snow—together representing snow pack input—were 70.67 ± 17.32 mm and 16.56 ± 3.85 mm, respectively. On average, 38 % of this input is converted into SWE and snowmelt that contributes 12–19 % of total river discharge over the TPP, while the remaining 62 % is lost to the atmosphere through sublimation and evaporation. Snow contributes less to water resources than to atmospheric moisture over the TPP on annual average. Seasonal snow patterns vary by region: in the Pamirs snow accumulates throughout the winter, making March–April SWE a key water resource indicator; while in the Tibetan Plateau, limited snow accumulation means total annual snowmelt better representing snow water resources. Significant regional declines have been simulated for key snow components though the trends vary spatially, potentially greatly influencing weather and climate both locally and remotely. Precipitation drives SWE changes in the north and west of the TPP, while temperature and wind speed play greater roles in the center and south.
- Preprint
(9788 KB) - Metadata XML
-
Supplement
(7198 KB) - BibTeX
- EndNote
Status: open (until 25 Jun 2026)
- RC1: 'Comment on egusphere-2025-4849', Anonymous Referee #1, 20 Mar 2026 reply
-
RC2: 'Comment on egusphere-2025-4849', Anonymous Referee #2, 24 May 2026
reply
GENERAL COMMENTS
This study presents a simulation-based analysis of the climatological snow mass budget and its trends in the years 1962-2019 in the vast Tibetan and Pamir Plateaus in high-mountain Asia. The main findings include the overall smallness of the snow water equivalent SWE (annual area mean only 0.85 mm), the dominance of sublimation over snowmelt as the primary sink of snow (about 62 % of the total), and a near-zero annual and area mean trend in SWE in 1962-2019, although more than two thirds of the area has experienced a decrease in SWE.
Due to the very limited coverage of in-situ observations and the difficulties in remote sensing of snow conditions in rugged terrain, modelling is necessary for deriving credible estimates of SWE and the snow mass budget in the Tibetan and Pamir Plateaus. Modelling always includes uncertainties, both from the input data used and from the parameters and the structural choices in the models used. The latter are especially difficult to estimate without repeating the research with a completely different models, which is unfeasible in practice. Nevertheless, this study makes a commendable effort in deriving input data that are us free as possible from systematic errors (in particular, the undercatch of precipitation), in taking into account the large sub-grid scale topographic variations (by running the VIC model separately for each 100-m elevation band in each grid cell) and in optimizing the parameters of the model using machine-learning methods. The resulting framework is somewhat intractable, but the results appear to be in reasonable agreement with available station observations and earlier studies for individual locations. Some sensitivity analysis focusing on the input data of the VIC model is also provided.
Despite inevitable uncertainties (among which a 62 % change in simulated SWE in response to a 10 % perturbation of precipitation is a striking example), I find this research well-done, important and worth publication. However, I still have many comments on the details, mostly on the presentation but also on the science.
DETAILED COMMENTS
- L34-64. The beginning of the Introduction is very general. You could reduce the general text in the interest of conciseness and focus on topics that are directly relevant to your study.
- Winter Olympics is somewhat too much a “tip of an iceberg” example.
- L53-54. The suppression of outgoing longwave radiation is a consequence of low surface temperature, not a cause of it.
- Start a new paragraph from "The sensitivity".
- L68-69. This list is not internally consistent. Snowpack depth and SWE are measures of the amount of snow, all the rest budget terms that change the amount of snow.
- A better reference for GlobSnow: Luojus, K., Pulliainen, J., Takala, M. et al. GlobSnow v3.0 Northern Hemisphere snow water equivalent dataset. Sci Data 8, 163 (2021). https://doi.org/10.1038/s41597-021-00939-2
- L364-365. Two interannual standard deviations strongly overestimates the sampling uncertainty in long-term mean values. If snow conditions in individual years can be considered independent (which should be a reasonable assumption in areas where snow does not survive over the summer), then the sampling uncertainty in the long-term mean is inversely proportional to the square root of the number of years.
- Figure 4. Add headings above the figure panels to facilitate faster understanding (“Annual snowfall” in (a) etc). In addition, the area means over the entire TPP could be added to the maps, to make it easier to see the magnitude differences that are hidden by different colour scales.
- Figure 5. Same comments as for Figure 4 above.
- This is a surprising definition. SWE only contributes to runoff when it melts. For what purpose does it count as an available water resource before melting? Also, if the same snow contributes to the resource both when on ground and when melting, it is counted twice.
- L559-560. Refer to Figure S6 where these results are shown.
- 578-581. What do you mean by "each snow event" in the calculation of the annual accumulated maximum SWE? If SWE decreases after a snowfall event but does not drop to zero before the next snowfall event, then summing the two SWE maxima without subtracting the minimum SWE in-between will overestimate the total water input from the two snowfall events.
- L594-595. Explain the meaning of the first column numbers (1-11) also in the table caption.
- L629-630. Please indicate the section of the Supplementary material where the details are given.
- L632-633. SWE responses to 10 % perturbations in precipitation … The word “changes” should be avoided because it is easily mixed with the long-term trends discussed in the previous sentence.
- L637-639. The relative importance of the different variables cannot be evaluated from the idealized 10 % sensitivity tests alone, without knowing the actual magnitude of uncertainty / varibaility in the four perturbed variables. More representative values (regarding the contributions to interannual variability of SWE) could be derived by multiplying the sensitivity ratios from the 10 % tests with the observed interannual standard deviation of each four input parameters.
- Figure 7 is confusing. Panel (a) shows sensitivities of area mean SWE to long-term trends in different parameters, whereas (b)-(e) show the spatial distribution of the sensitivity to the idealized 10 % perturbations in these variables. It would seem more logical to show in (b)-(e) the sensitivities of SWE to the local precipitation, Tmax, Tmin and wind speed trends instead. If the figure is retained as it is, at the very least the difference in interpretation between panels (a) and (b)-(e) should be stated explicitly.
- Table 3. Please also give the contributions of the four variables to the SWE trends in the different basins, either in this table or in a separate table.
- L704-705. Probably true but not rigorously quantified (cf. comment 16 above).
- Figure S3. Please explain the meaning of the red lines.
- Figure S7. Please use the same colour scale in all three columns. Furthermore, the scale for Pbias should be symmetric with respect to zero, that is, the change from red to blue should occur at zero bias.
- Figure S10d. Absolute contributions of precipitation, temperature and wind speed to the SWE trends would be easier to understand than the map of relative contributions.
- Figure S11. Add area mean values to the maps.
- Table S2. For Daily maximum, minimum and mean temperature, bias in absolute (C) units would be more informative than percent bias.
TECHNICAL COMMENTS
- L202 and later. Replace “warming-up period” with “spin-up period”.
- The sensitivity of VIC to the values of snow parameters was investigated?
- L280 and later. The impactful snow parameters / the most important parameters? “A sensitive parameter” means a parameter whose value is strongly sensitive to assumptions etc., rather than a parameter whose value has a large impact on the model results.
- were
- L431 and later. decreasing trend
- There is no evaporation in Figure 4 (and probably should not be, since only liquid water evaporates).
- Supplementary text S1, just above Equation (4): SWE is converted from snow depth?
- Supplementary text S3, just above Equation (5). spin-up period
- Caption of Figure S4, L2-3. Snow band numbers of 11-20 dominate, with an average snow band number of 18
Citation: https://doi.org/10.5194/egusphere-2025-4849-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 1,344 | 1,679 | 122 | 3,145 | 413 | 208 | 244 |
- HTML: 1,344
- PDF: 1,679
- XML: 122
- Total: 3,145
- Supplement: 413
- BibTeX: 208
- EndNote: 244
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This manuscript attempts to quantify snow mass balance components over the Tibetan and Pamir Plateaus using a combined modeling and machine learning framework, and to assess the relative contributions of snow to water resources and atmospheric moisture. The topic is relevant, and the integration of modeling with parameter regionalization is of some interest.
However, the manuscript has several fundamental issues. The central conclusion—that snow contributes more to atmospheric moisture than to water resources—is not robust. It relies heavily on simulated sublimation, which is highly uncertain, lacks proper validation, and is not accompanied by any uncertainty analysis.
More broadly, the credibility of the modeling framework is insufficient. The evaluation relies largely on model-derived SWE rather than independent observations, and key processes such as sublimation and snowmelt are not rigorously assessed. This weak observational constraint substantially limits confidence in the results.
A critical concern lies in snowfall, which is the fundamental input to the entire snow mass balance. As acknowledged by the authors, there are no direct snowfall observations, and precipitation data themselves carry considerable uncertainties over the TPP. These uncertainties are further amplified by the rain–snow partitioning parameterizations. Given the compounded uncertainties in this key input, it is difficult to see how the study can robustly constrain snow mass balance partitioning or support quantitative conclusions.
Furthermore, this study addresses a large and important scientific question, and the conclusion is potentially far-reaching. However, relying primarily on limited regional observations and a complex modeling chain is insufficient to support such a strong claim. The applicability of model parameterization schemes over the TPP requires careful, component-wise validation. Parameterizations developed and tested in other regions cannot be assumed to be directly transferable, and thus the reported partitioning results cannot be considered reliable.
The overall presentation of the manuscript also requires substantial improvement. The figures and writing are difficult to follow, with redundant descriptions and unnecessary abbreviations (e.g., HAL). The introduction reads more like a literature summary than a clear formulation of the scientific question. In addition, several terms and expressions are unclear (e.g., “standstill snowpack” versus SWE, and “warming-up period”), which further reduces readability. If AI tools are used in the writing process, greater attention should be paid to clarity and precision.
That said, the study shows some technical progress in snow process representation. The work may be more suitable for a model-oriented journal if it is reframed to focus on methodological development rather than broad hydroclimatic conclusions.