In general, the authors' proposed revisions appear adequate to address the concerns and comments raised by the reviewers, and I encourage the authors to adjust the manuscript along the lines proposed and resubmit for my further review.

In addition, I wish to add the following comments for the authors to consider in their revision.

Major comment:
I'm struck by the relative arbitrariness of the approach in which HICs are drawn from a high-res land surface model to initialize seasonal ensemble forecasts that are propagated forward with a conceptual scheme at the opposite end of the spectrum in terms of complexity (ie the WBM). Are there examples of this type of approach being used elsewhere that could be referenced? The given rationale for the strategy is that there's no appreciable value to downscaling climate forecasts to the scale of the G2G model, yet there have been numerous examples of seasonal forecasting in which coarse climate forecasts are in fact downscaled to resolutions on the order of 10K or finer, and used to drive flow forecasts. Is another practical rationale that the computational cost of running the G2G model ensembles with downscaled forecasts is not justified by any added skill from to computing a full-resolution seasonal flow forecast after downscaling? In any case, I think this somewhat bespoke/uncommon approach to seasonal regional forecast construction is an interesting choice that warrants a bit more motivation. And is there is any way to benchmark it against any alternatives, that would provide practical insights? For instance, for regional prediction of flow anomalies, is there added value from taking HICs from G2G versus from a long-term run of a simple conceptual monthly model -- could that be tested? And is there any way to assess the marginal value of this approach, versus other alternatives (such as the far more common use of one model for both HICs *and* forecast, such as at SMHI or in the US NWS), against actual obs flows -- perhaps for a handful of case study basins where such flow estimates are available? I realize that this may all lie beyond the scope of this paper. In any case a broader discussion of the rationale for this particular prediction scheme, and acknowledgement of other alternatives, could add value to the paper.

I find the paper's attribution of forecast skill using the component-on/component-off approach very useful and effective, as a technique that is quite common in seasonal climate prediction (ie, use a climatological Indian Ocean or continental SM, versus a dynamic one), but less common in hydrologic prediction studies. It provides an slightly different angle amidst the recent interest in the attribution of seasonal hydrologic prediction skill.


Minor comments:
p8, l8 -- 'Clim' doesn't match the 'Hist' term later used in Fig 2 and elsewhere.
p6, l10 – the text states: “The coarse spatial resolution of the input rainfall forecasts has discouraged the development of river flow forecasts at a 1km resolution, and production of regional scale forecasts (in preference to national scale) is viewed as a pragmatic compromise. “ Is that the sole reason or is the computational cost of running a 1 km resolution model for seasonal ensembles also a limiting factor?
p6, l5 – when/why does it arise that the Qm_bar/Sm_bar >= 1? More explanation or insight would be helpful – especially as these are climatological averages. The monthly runoff ratio should almost always be between 0 and 1 unless a snowpack-related delay in runoff timing is involved.

Although the review process raised several suggestions for additional scope to strengthen the paper (e.g., benchmarking the hi-res/low-res hybrid approach versus more traditional variations), and these were not taken, the paper in its current form nonetheless presents useful analyses and appears to be suitable for publication.

I suggest a careful proofreading before submitting a final version to ensure clarity in all of the passages highlighted by the reviewers, after the changes that have been added in this latest revision. It is also customary to label colorbars and provide their units in a plot (eg Figs 3 & 4), rather than in the caption.