General comments

Li and co-authors study the moisture sources of precipitation in 2 river basins for the (seemingly randomly chosen) 2022 July period with 2 moisture tracking algorithms WAM2layers and FLEXPART-WaterSip. They compare the methods and subsequently test sensitivities when changing certain parameters. The study is timely, relevant, generally easy-to-follow and substantiated with good figures and tables. However, I have two major comments:

1. The study is not at all reproducible as no detailed model settings are provided in relevant scripts. Moreover, people that use other moisture tracking models or settings would not be able to compare their results against that of the authors as no output data is provided. Only generic links to input scripts and data are available which are by far insufficient in this new era of FAIR and Open Science.
2. The authors make several strong statements and conclusions about the tracking models ability, which, in my opinion are mere hypotheses by lack of knowledge about an actual truth. These hypotheses should be substantiated by additional analysis and/or toned down.

Specific comments are attached.

Review of "Unravelling disparities in Eulerian and Lagrangian moisture tracking models in Monsoon- and Westerlies-dominated basins around the Tibetan Plateau" by Li et al., submitted to WCD

The authors perform a sensitivity study of two methods to identify moisture origin for one selected summer month over two regions in the Tibetan plateau. From the comparison between the two methods, the authors see differences with regard to moisture contributions from Eurasia and over coastal regions, that are explored in a sensitivity study. The authors then draw conclusions about the consistency and validity of the two methods. The manuscript is overall written coherently and in a well-readable manner. However, I find the conclusions are too general given the episodic evidence presented in the manuscript itself. The authors could consider changing this paper to a shorter, research letter format. I also have some comments about the structure of the manuscript, the precision of the language, reference to code and use of literature, and the presentation and interpretation of the results. I hope my comments will help the authors to prepare an improved version of their manuscript.

Main comments

1. In their introduction, the authors set forth a basic distinction into Eulerian and Lagrangian methods for "moisture tracking". I find this distinction too coarse with regard to the results presented in this study. The two methods that are being compared are broadly seen part of the respective categories, but there are many (other) approaches within the Lagrangian category (see for example the discussions in Keune et al., 2022), and many other within the Eulerian category, that are not compared here. For example, moisture tagging in a regional model (Yoshimura et al., 2004), or the E-P Lagrangian approach of Stohl and James (2004), and so on. The authors claim that the two methods they compare are most widely used - I think this is debatable, plus they are focussing here on the Tibetan Plateau only.
2. The study now only compares one month (July 2022) and two specific catchment areas of the Tibetan Plateau. It remains thus unclear if the findings here can be generalised, or are rather coincidental. Therefore, it would be adviseable to tune down the quite authoritative/concluding language and formulate more modestly, such that it be in agreement with the somewhat anecdotal evidence that is actually investigated and presented here. This concerns both the Abstract, Introduction, and Conclusions.
3. The authors state that they use the FLEXPART-WaterSip method. I don't think this is correct, since the WaterSip code is a specific implementation of the Sodemann et al. (2008) moisture source diagnostic in C++ language which is currently not yet available publicly. The WaterSip code has first been used by Sodemann and Stohl (2009) and later my many other studies (Bonne et al., 2014; Läderach and Sodemann, 2016; Sodemann 2020 to name a few). The authors also state that all original codes are available from the official websites - this is not correct for the WaterSip method. A separate publication on this actual "WaterSip" code is in preparation by this reviewer. My impression is that the authors have written their own implementation of the algorithm of Sodemann et al. (2008), which they then use for this study. This must be stated clearly and correctly, and the authors' own code should be linked to in the Code availability section. In any case, the reference to the website at University of Bergen is no proper code reference to the WaterSip method.
4. The immense literature review presented in Table 1 is never properly described and hardly used in the manuscript. I also note that a similar table has been presented already in the supplement material of Li et al. (2022), a study by the same authors that is not cited in this manuscript. I do appreciate the effort put into this table. Currently, however, there are just two sentences in the introduction that make general remarks about this table. A more systematic discussion of what was found during the literature review would be needed to justify including this table in the main manuscript. In addition, it would be useful to tie the results from this study up agains the reviewed literature in a Discussion section in the end.
5. Section 2 discusses the generalities of the two selected methods. I think the broad description of these two examples as Eulerian and Lagrangian methods in general does not fit the two specific methods that are applied here. Also, how these specific methods work are described sufficiently elsewhere in the literature. Instead, the authors would need to describe more clearly how exactly the respective simulations have been set up. Specifically regarding the FLEXPART-WaterSip like method, was a domain-filling setup selected in FLEXPART? Was the calculation run in forward mode? Has convection parameterisation been used? What domain has been used? All these details are important. Furthermore, the WaterSip code is currently not available publicly, and the website pointed out in the data section only provides a manual. What code has then been used to diagnose the moisture sources from the FLEXPART particle trajectories, and where is this code accessible? How were Lagrangian moisture sources gridded? What output interval and humidity thresholds were used? These aspects are all essential aspects for reproducibility of the work, and to understand the preconditions of this comparison.
6. The difference in moisture source contribution from Eurasia between the two methods is quite interesting. We don't know what is the truth from the two approaches, but a gridded map of air parcel location density for trajectories arriving in the study domains could help indicate if FLEXPART (based on ERA5) does identify transport pathways from Europe. In this context, I find the sensitivity of the WAM2layer method to finer resolution quite striking. What is possibly going on that leads to such a strong senstivity to grid resolution in the results? Maybe be there is numerical diffusion at coarser resolution (see Sodemann 2020, Sec. 7)? Additional sensitivity experiments or analyses of different time snapshots could be useful.
7. The sensitivity study in Sec. 5 is quite interesting, but does not really include the most important sensitive parameters of this approach, as discussed widely in the literature. Instead of number of particles (Fremme et al., 2023), it would be more important to test the threshold of specific humidity (dqc in Sodemann et al., 2008) as well as the relative humidity at arrival (RHc in Fremme and Sodemann, 2019). The areal source-receptor attribution method comes a bit out of the blue here. It is an entirely different method of the Lagrangian category. The difference between this method and the others should be described in the methods.
8. I am puzzled that the authors do not discuss nor cite their own study in NHESS about the spatial distribution of moisture sources for the Tibetan Plateau using the WAM2layer model (Li et al., 2022). In the supplementary material of that paper, they show a map with Eurasian moisture sources, just as discussed here from the two methods. What could possibly be the reason that you do not discuss this previous work done with the WAM2layers method? Is this not a golden opportunity to balance or rectify any conclusions drawn in Li et al. (2022) in the light of new evidence? I also note that Li et al. (2022) contains a table similar to Table 1 presented here. A discussion of the relation between this work and your own previous work is definitely required.

Detailed comments

1. Figure 2: The gridding of the FLEXPART-WaterSip results in Fig. 2 looks more spotty than the WAM2layers - I would argue that either a larger grid spacing or larger gridding radius of the identified sources should be used, or the number of particles increased to mute these distracting artifacts. Maybe just show the same resolution as used in Fig. 3 where the same grid was used for both models?
2. Figure 6: I find panels a and b hard to interpret objectively, as there are subjective/conceptual arrows superimposed on the panels. Are these two panels adding new information compared to the trajectory examples shown in panels c-f?
3. Figure 7: Why do you show 300hPa vertical velocity in panel b? Maybe it would be more useful to add a figure that shows the average/median vertical air motion as a view of trajectory (pressure) altitude vs time arriving at the two selected regions. These vertical pathways seem to be quite different.
4. Figure 10: These two examples from a set of 5 million trajectories can hardly be considered representative. What is really the value of discussing exactly these two examples? It does not become entirely clear to me what to take away from these examples, and I think i is not justified to draw as general conclusions about the weaknesses of the Lagrangian diagnostics (L. 399 onward) as the authors do on this basis alone. Also, I got confused by the time axis at first, it should be made clear where the arrival point is. Winschall et al. (2014) have discussed with similar examples before that (deep) convection can contribute to moistening at upper levels that is not captured by motion of individual trajectories. Is this the case here as well? Do you use a convection parameterisation in FLEXPART? Are these locations over land or ocean? It would also be helpful to indicate the specific humidity threshold adopted in this study, and maybe include specific humidity and relative humidity in addition.

References

Bonne, J.-L., Masson-Delmotte, V., Cattani, O., Delmotte, M., Risi, C., Sodemann, H., and Steen-Larsen, H. C., 2014: The isotopic composition of water vapour and precipitation in Ivittuut, southern Greenland, Atmos. Chem. Phys., 14, 4419–4439, https://doi.org/10.5194/acp-14-4419-2014.

Fremme, A., Hezel, P. J., Seland, Ø., and Sodemann, H., 2023: Model-simulated hydroclimate in the East Asian summer monsoon region during past and future climate: a pilot study with a moisture source perspective, Weather Clim. Dynam., 4, 449–470, https://doi.org/10.5194/wcd-4-449-2023.

Fremme, A., and H. Sodemann, 2019: The role of land and ocean evaporation on the variability of precipitation in the Yangtze River valley. Hydrol. Earth Syst. Sci., 23, 2525–2540, https:// doi.org/10.5194/hess-23-2525-2019.

Keune, J., Schumacher, D. L., and Miralles, D. G., 2022: A unified framework to estimate the origins of atmospheric moisture and heat using Lagrangian models, Geosci. Model Dev., 15, 1875–1898, https://doi.org/10.5194/gmd-15-1875-2022.

Läderach, and H. Sodemann, 2016: A revised picture of the atmo- spheric residence time of water vapor. Geophys. Res. Lett., 121, 3040–3061, https://doi.org/10.1002/2015GL067449.

Li, Y., Wang, C., Huang, R., Yan, D., Peng, H., and Xiao, S., 2022: Spatial distribution of oceanic moisture contributions to precipitation over the Tibetan Plateau, Hydrol. Earth Syst. Sci., 26, 6413–6426, https://doi.org/10.5194/hess-26-6413-2022.

Sodemann, H., C. Schwierz, and H. Wernli, 2008: Interannual variability of Greenland winter precipitation sources: Lagrangian moisture diagnostic and North Atlantic Oscillation influ- ence. J. Geophys. Res., 113, D03107, https://doi.org/10.1029/ 2007JD008503.

Sodemann, H., and A. Stohl, 2009: Asymmetries in the moisture origin of Antarctic precipitation. Geophys. Res. Lett., 36, L22803, https://doi.org/10.1029/2009GL040242.

Sodemann, H., 2020: Beyond turnover time: constraining the lifetime distribution of water vapor from simple and complex approaches. J. Atmos. Sci. 77, 413–433.

Stohl, A., and P. James, 2004: A Lagrangian analysis of the atmo- spheric branch of the global water cycle. Part I: Method description, validation, and demonstration for the August 2002 flooding in central Europe. J. Hydrometeor., 5, 656–678, https://doi.org/10.1175/1525-7541(2004)005,0656:ALAOTA. 2.0.CO;2.

Winschall, A., S. Pfahl, H. Sodemann, and H. Wernli, 2014: Comparison of Eulerian and Lagrangian moisture source diagnostics—The flood event in eastern Europe in May 2010. Atmos. Chem. Phys., 14, 6605–6619, https://doi.org/10.5194/ acp-14-6605-2014.

Yoshimura, K., Oki, T., Ohte, N., and Kanae, S, 2004: Colored moisture Analysis estimates of variations in 1998 Asian Monsoon water sources, J. Meteorol. Soc. Japan, 82, 1315--1329.