Dear authors,

thank you again for the submission of your manuscript. Both reviewers generally seem positive and consider the manuscript suitable for publication. Your responses are generally appropriate and I invite you to implement them carefully in a revised version of the manuscript.

The two main concerns of the reviewers regard: (i) the motivation for the static/dynamic nature of the parameters; and (ii) the approach to validating the downscaled data. I would ask you to put particular emphais on adressing these two points, and I want to note that I agree particularly with the second one.

More specifically, I am very concerned that the 1 km data consistently outperform all finer-resultion data at all stations, both at CRNS and HOBE reference sites. I strongly disagree with the statement that "... the additional value of the downscaling lies more in the improved spatial resolution than in the better statistical fit of time series".

The (properly chosen) statistical fit does, in fact, quantify objectively whether or not spatial resolution is actually improved. Visual spatial detail alone does not prove higher (soil moisture) information content. Land cover patterns are clearly visible in the high-resolution maps, but it is not a given that these accurately reflect soil moisture patterns. In contrast, 1 km data consistently correlate better even with point-scale data, and also exhibit smaller RMSEs. That is, even in absolute terms the 1 km estimates are closer to the point measurements than the sub-km resolution estimates.

In essence, this suggests that the downscaling does not add, but in fact removes soil moisture information from the original measurements when progressing beyond 1 km resolution, so the only conclusion supported by the analyses shown is that downscaling beyond 1 km scale is not feasible with the methods tested.

Note also that I do not think that the reason why no better improvements are observed is a lack or scale mismatch with the reference data. I, personally, consider the two-step evaluation strategy with both CRNS and HOBE solid, although it might benefit from additionally evaluating spatial correlations (at least with the HOBE network) to actually quantify the degree to which spatial patterns are resolved.

I believe that this links back to the first main concern of both reviewers regarding the parameterization of the downscaling model... Could it be that the model and its parameterization converges to a limit in accuracy once approaching 1 km resolution due to all the factors pointed out in their comments, i.e., that the spatial and temporal variability in roughtness, vegetation, etc. is not (or perhaps even cannot be) properly resolved?

Best regards,
Alexander Gruber

Dear authors,

thank you for the revision of your manuscript which has greatly improved its quality. The referees and I consider all issues raised during the discussion properly addressed. I am therefore happy to accept the manuscript for publication in HESS.

Note, however, that Referee #1 is encouraging you to extend the literature with the most recent contributions. I leave this up to your discretion, but I agree with the referee that this would benefit the readership.

Kind regards,
Alexander Gruber