General comments:

I enjoyed reading this elaborated work on disentangling the hydrological and hydraulic controls on streamflow variability in E3SM V2. It is clearly written and organized, however missing some consistency in e.g. naming of experiments, and some figures could use a revision. Those adjustments in the text and figures would make it even more easy to follow the story in the text.

Xu et al. performed several experiments with the E3SM coupled with MOSART. The topic of studying parameter values of critical parameters of a land surface model is crucial, and especially the validation of model output against different observation products is important. It is a valuable contribution to the land surface modelling community.

The case study area of the Pantanal Region is interesting as the streamflow show a shift of about five months in the seasonality compared to precipitation.

I would like some more information about the calibration. I like the simplicity in the random sampling. Are 1000/2000 simulations an appropriate number? Why?

In my world calibration involves a mathematical optimizer, but your approach of doing several experiments with random sampling is an easy and simple approach to get an idea op optimal parameter values. It is not a demand, but maybe you could give some information on how the objective function distribute, and the performance of the “best” solution compared to the default? I would also like to know the parameter values of the “best” calibration. Those numbers would be valuable for other modelers.

Specific comments

Page 7 line 169: please explain why it is an acceptable assumption.

Figure 4: It is very hard to see difference in the size of the circles. The plot with rSD at xaxis: what is on the yaxis? Probably Manning coef., but this it not obvious. Please improve figure.

Page 11 line 238: dosen´t it say some other numbers on figure?

Figure 6: I suggest to improve the readability of the figure (only a suggestion, not a need):

• (b): I am confused about what is on the left yaxis
• perhaps place the three plots with identical xaxis below each other. It would make it easy for the reader to get a quick overview.
• If there is no secondary yaxis, then always place the yaxis to the left.
• Be consistent about using the term observation and the actual name of the observation product in the legend

Page 13 line 270: “The experiment of f_drain…”. It would be easier to read and understand the text if the naming of the experiments were consistent. This apply to the whole paper.

Figure 7: Please explain the term “ET trend” in the text. As I understand, you use the term “trend” regarding runoff in the paper. Please clarify in the text which trend you are referring to.

Figure 8: There is something with the naming, why c and d?

• I suggest making the figure in the same way as figure 6, and with the same order of the plots.

Figure 9: regarding legend in (a) and c): Be consistent with naming of the experiments in relation to text and other figures. The whole article would be much more readable if you were consistent with the naming.

Page 16, line 320: There is no eq. 7

Figure 10: c): the dotted line is missing in legend  

General comments

Thank you for the well written and organized manuscript. 

Calibration of parameters for hydrological and hydraulic processes is very important for river discharge, and I believe that calibration should be conducted carefully when applying to the global scale. It is difficult to apply the current results to the global scale. Further analysis of the current results is necessary.

Although it may not be the purpose of this study, specifically, an analysis of the relationship between the parameters and the characteristics of the target river basin (precipitation, soil, topography/geology, etc.) would increase its applicability to the global scale. Furthermore, we have difficulty in determining whether this two-step calibration is a good idea.The reason is that we do not know what kind of changes the first and second steps brought about in the river discharge, respectively.

Since this is a two-step calibration, there should be an answer for the first step, which can be shown in Fig. 12 to indicate how the river discharge changed from the default. It would be necessary to show how the river discharge changed and at what locations.

Specific comments

How about describing the characteristics of the sub basin observation points in 2.2? Currently, the entire river basin is described with a focus on outlet points, but it would help to reinforce the last part of the discussion. In particular, SB#3 has a high river discharge for the area of the upstream river basin. It would be nice to have a comparison between precipitation and runoff height, even if it is shown on the supplement.

What is the reason why you chose these three components for multi object funcition in 2.5, you wrote that you didn't include streamflow variability in 3.5 because we didn't know how much it would be affected by runoff process only, but why did you include SWF? I think the SWF is also affected by river routine model, MOSART.

In the comparison between 3.4 and 3.5 (Fig. 9(a)), why is the river discharge lower in the two-step case than in the case where hydrology and hydraulics are calibrated separately?

Regarding discussion 3.5, you list three factors for underestimation of river discharge, but I think we need to be sure that these three factors are really contributing to the problem. For example, if we consider average evapotranspiration, increase the area of the river basin by 5%, and increase precipitation, how much runoff will be generated and whether this runoff can represent an underestimation of the river discharge, we can check this at the order level. Alternatively, we could implement the results for other precipitation products. After that, I think it is necessary to consider the factors that lead to an overestimation of evapotranspiration.