Variation of deuterium excess in surface waters across a 5000m elevation gradient in the east-central Himalaya

The strong elevation gradient of the Himalaya allows investigation of altitude and orographic impacts on precipitation isotope values as captured in river samples. This study provides new high-elevation data along a 5000-m gradient collected from rain, snow, and glacial-sourced surface waters and time-series data from April to October 2016 to differentiate the time-variable contributions of source waters to the Arun River. We find nonlinear trends in 15 δO and δD lapse rates driven by samples collected at high elevations and a distinct seasonal signal indicative of moisture source influences the surface-water isotope values. Deuterium excess is correlated to snowpack and used to track melt events during the monsoon. Our analysis identifies contributions from snowpack to river discharge before the monsoon onset followed by a 5-week transition to Indian Summer Monsoon-sourced rainfall around mid-June 2016. 20

The manuscript is adequately organized and contains no grammatical or orthography errors. Text is concise and to the point. The English is fine and does not need any revision; but note that I am not a native speaker.
However, the manuscript suffers from many technical and scientific drawbacks. For the technical details on terminology please see my comments below, but I got the impression that the field of isotope hydrology is rather new for the first author with respect to conventions and expressions. Using percentage differences in isotope values as for d in Figure 3 is useless.
From the scientific point of view, I see a major flaw in using a mixture of various water samples from snow and ice, glaciers, surface water, river water, springs and groundwater, and lakes to determine the isotope altitude gradient. Note that this gradient relates to precipitation and not to the water types sampled in this study. River water can be used as a substitute under specific circumstances but the authors fail to explain how and why their samples can be used to determine the isotope altitude effect for the region. So, how can you use a suite of various sample types to conclude on a specific parameter (altitude effect) that, in fact, relates to a complete different type of samples (rain water)?
Overall, the study also lacks clear scientific objectives in the introduction. Some ideas are mentioned but it remains unclear (also in the conclusions) if the results could be used to e.g. better manage high mountain water resources on a local (see e.g. Königer et al. 2017Königer et al. , doi:10.1002 or regional scale. The authors need to clearly explain what were the aims of the study (Introduction), how does the sampling strategy was designed to fulfill these scientific goals (Methods) and how does the data (Results) helped to solve these scientific question (Discussion and Conclusions; the latter with an outlook to future work and recommendations) Unfortunately, I have to reject the study for publication in HESS at current stage but encourage the authors to re-submit a strongly revised version of this interesting data set.

Specific comments P2L14
The definition of deuterium excess is a bit awkward. I notice what you mean, but in fact deuterium excess itself is not a "deviation from the GMWL" but its definition, of course, used the dual-isotope equation of the GMWL. By definition, d is calculated with the slope of 8 from the GMWL and can be visualized as the intercept of a line with slope 8 which crosses the given isotope pair in a dual isotope plot of δ 18 O vs δ 2 H. Thus, the GMWL connects points of d = 10 ‰. Either rephrase here or simply delete "which….(GMWL)," to avoid confusion.
Also note that sea surface temperature (SST) also modifies the deuterium excess. Please this information to the sentence.

P1L39
The term stable isotope "lapse rates" is unfamiliar to me. In hydrology, to my knowledge, this is typically referred to as '(isotope) altitude effect' or 'altitudinal (isotope) gradient' (cf.

P2L16
In addition, the information that d relates to the moisture source (and more specifically to the rH and T at this point of origin during evaporation) is missing from the text. This makes this sentence hard to understand for the readers as you here directly move on to the identification of sources of precipitation by deuterium excess values. Please provide here a few more details.

P2L38
"Arun River catchment" (catchment/watershed/basin missing from text) P3L1 change to '8480 m in altitude' to make clear that you are talking about height (and not distances) here.
P4L10-25 and Figure 1: It is unclear to me how groundwater ( Figure 1) relates to the type of samples described in the text. I assume that the term "springs and surfacewater tributaries […] (N = 50)" relates to these points? Please bring text and figure labels/caption in line so that the reader can follow what samples are taken at which location. Also note that lakes might undergo strong evaporation processes, also in high and cold environments and that these samples might be not indicative of the original precipitation.
Overall, samples from river water are rather problematic with respect to the altitude effect as river water is affected by the 'catchment effect' (see Dutton et al. 2005). High altitude springs, gaining groundwater conditions along the river course and surface runoff change the isotope values with respect to the on-site precipitation. This can be of minor influence, but for rivers in high altitude regions this should definitely be considered and discussed.

P4L38
This sentence makes no sense. Stable isotope analytical uncertainty can never be reported in percent, as it is not a concentration but a relative deviation. Therefore, standard deviations must always be given in the same unit as the value (in this case in permil that serves as a unit). Change to 'Analytical uncertainties for the stable isotope are reported as standard deviation in the data repository.' Note also that the term "water stable isotope" does not exist. Water has no stable isotopes but oxygen and hydrogen of the water molecule does. First, have all water types been mixed (river, snow, groundwater, spring, lakes)? As each of these hydrological compartments have their own behavior with respect to isotopic fractionation, it makes not really sense to mix them up (or at least not to use different symbols to separate them). Second, what is shown on the x-axis? Is this the sampling site altitude or (as stated in the caption) the "mean catchment elevation of the drainage areas of the samples". If the latter -how was that elevations derived? From a GIS model of the sub-catchments? Please clarify here, not only in the Figure caption bus also in the text. Third, the rather stable isotope values in the lower regions of the watershed (<3000masl) could be the result of a sampling artifact as most samples of the lower region seems to be influences by groundwater (Figure 1b, Sabha Khola), which tend to have rather stable isotope values with low seasonality while the steeper gradients relies on samples that were sampled close or along the river course ( Figure 1a) P7L5 See my comment on precision of stable isotope values. Simply state that error is smaller than symbol size and skip percent values here.

P7L22
What is the 'anomaly'? Do you refer the sample in Figure 3 that plots closer to the glacier-melt dominated regions (x-axis)? Please clarify. Figure 3 As already mentioned above, I don't think that percentage can be used here to express differences on d-values on the y-axis. Use Δd values here (the difference in deuterium excess between pre and post monsoon) and not percent values, which makes no sense here.
P10L19/20 You have to consider that the sampling of river water always gives you an integrated signal from the regions above your sampling point as tributaries are continuously entering your stream (in contrast to precipitation samples). As a consequence you will never measure the real (or true) "precipitation isotope signal" in a river, especially not in high relief terrains as in your study. Thus, you cannot use a river water sample to calculate the (precipitation) elevation isotope gradient. Your river water signal will be always more negative than any on-site precipitation as the river water represents a mixture of high altitude water and water derived from sources near your the sampling site. This may led to a significant underestimation of the real isotope altitudinal gradient with increasing uncertainty further downstream. Please include a brief discussion about the limitations of river water samples with respect to rainfall-derived lapse rates. I agree that river water data can be used to estimate on altitudinal isotope gradients but be careful to clearly state that your values relate to river water (and not to rainfall).

P16L16
I doubt that you can cite a manuscript "under review" (van der Veen). Check with the Editor but either it is published at time of final acceptance of this manuscript or delete.

Technical comments
P1L16 and throughout the manuscript: please change δD to δ 2 H (note: delta-symbol italic) to follow the actual conventions for the notation of stable isotopes (Coplen 2011;Brand et. al. 2014) Brand, W., A., Coplen, T., B., Vogl, J., Rosner, M. and Prohaska, T. (2014) P4L19 space characters missing (N=6 -> N = 6); check also line 24. Table 1 In headline change to permil-symbol (‰); not 0/00. To what does the "std. dev. for each sample" refer to? To the number of injections in the laser instrument? Then specify the number of injections in Table caption and methods. Also check superscript of 18 O (also in in Tables 2 and 3). Further: The samples have no date/time? Please add this information as it essential to the reader; in particular for the river water samples taken from the streams. Please also provide a table with the exact GPS locations for each sample in the supplementary material (or include here). The sampling locations cannot be identified from Figure 1 and need further details provided either in the tables or in the supplementary material. Table 3 What was the weighing parameter for the regression equation? And why was it not used or discussed in the text? Please specify. Figure 2 Lines in figure appear grey in my copy. I suggest using clear black and white colors for axis, lines and also symbols (no grey filling, make them black and put the white symbols on top layer) In (c) change y-axis label to 'd-excess' (not D-Excess). Also for Figure 5.  Change rainfall data to bar graphs with suitable resolution. The 'spikes' does not give a correct impression about the rainfall data collection interval (daily?). Same for the weekly T data.