The authors have brought together many analyses into a useful, integrated platform.  As they point out, a fully distributed model that can capture all of the processes represented would be unwieldy at best and unattainable at worst.  On this basis, I fully support the development of this tool.  My only concern is that this model, while an improvement, still requires many simplifications.  The authors have sufficient experience and expertise to know which module is best to use for specific conditions ... especially as they are familiar with the Walnut Gulch site and with modeling COSMOS data.  Could they provide more guidance for less well-informed users?  Which assumptions are most critical?  Under what ranges of conditions should this model not be used?  How could a user identify potential systematic errors from the model results?  In short, simpler models are better - but, only if they aren't significantly flawed.  Can the authors help to make sure that their model will be used appropriately and constructively?  If I have one other quibble, it is with the testing of the model.  It is true that spatially distributed numerical models are expensive to run and I wouldn't suggest that they need to build and run one to test the applicability of their simplified model.  But, I would have thought that someone, somewhere, would have a model that could be used as a basis for comparison to give a better estimate of the magnitude and distribution of errors.  It isn't my field, exactly, but I'm not sure that mass balance errors alone are sufficient to validate a model.

All of these comments should be seen as suggestions only.  It is entirely possible that I am out of touch with what is expected/required for models at this scale and especially for use in hydrometeorology.  I leave it entirely to the authors' discretion to decide to consider or ignore my comments.

Nice work!

Ty Ferre

P.S. I did see one spelling error on a plot axis ... cummulative.  I first thought that it might be a British spelling, but Google doesn't support that hypothesis.

Authors have developed a distributed parsimonious hydrologic model to represent surface water-groundwater interactions and channel transmission losses in dryland environments.  The modeling package is written in Python and provides some flexibility in terms of implementing a variable time step for different hydrologic processes and using different equations to represent infiltration. The new modeling package was set-up over the Walnut Gulch experimental watershed in Arizona and results were validated using observed streamflow, soil moisture and actual evapotranspiration data. Despite a relatively simple structure, the model was able to capture the main hydrologic processes of the studied catchment. The paper is well written and authors have performed robust model evaluation using synthetic and real numerical experiments. I provide a few comments below to further clarify some of the points discussed in the paper.

Line 75- Could you please explain which "hard-coded parameterization" you refer to here.

Line 145- How do you represent the streams? Is width of a stream equal to the width of a grid cell or do you have flexibility in representing the streams using different geometries? While this information is added later in the text, it would be more useful to state it earlier in the manuscript.

Line 335- Do you assume the subsurface is homogenous for both saturated and unsaturated zones?

In Equation 20- Do you assume a constant specific yield across the entire subsurface? 

How do you define the riparian unsaturated zone? does the extent of the riparian zone variable or fixed?

Do you implement an iterative scheme to consider the two-way coupling between the saturated and unsaturated zones? Please clarify. 

I suggest to move section 2.6 after section 2.1. to first describe the model inputs and then explain the equations.

What are the vegetation specific parameters in the model? It seems only the rooting depth is set depending on the landcover type.

Line 635- Since the model resolution is 300 m, it means the width of the stream cell is 300 m as well. This width is too large to represent the width of streams in arid regions.  

Figure 9- Why do you show uncertainty bounds for the simulated soil moisture and AET and not the runoff? does the “red” simulated line represent the best model run in Figure 9?

Please discuss the limitations of the current model and future development plans.

Minor Comments:

Line 21- Add “distributed” to the model to provide this important information to the reader upfront.

Line 118 – Remove “(a la” from the reference

Line 120- Replace “streambed” with “stream stage”.

Line 131 – Replace “an” with “a”

Line 275 – Add “runoff” to the next downstream cell

Line 355- Replace “relative water content” with  “volumetric water content” .

Line 823 – Change “Fig. 9a” to “Fig. 9b”.

Line 832 – Change “Fig. 9b” to “Fig. 9c”.

Line 841- Change “Fig. 9c” to “Fig. 9f”.