Dear Authors,

the reviews of your manuscript were extremely positive, I have decided that minor revisions are necessary before the manuscript can be published. I agree with the reviewers that this manuscript is very readable and will likely be well received by the community. In revising your manuscript to address the various issues highlighted by the reviewers can I ask you to consider a few points from me as well.

1. The algorithm descriptions covered in the appendix are extremely useful for the community. I note that it is referred to once in the introduction but it would be great if you could find at least one more place in the manuscript to refer to the appendix (perhaps the discussion?). I think what you've done here is very valuable and it would be a shame for a reader to miss this.

2. In 2.1 where you talk about the CMIP models having a T/ET ratio of 0.22-0.58 I think it would be valuable to offer some insight into why: (a) they disagree with each other; and (b) why this ratio is noticeably below other data-based estimates. This is an optional suggestion but I do think given this is a review it would be good to inform the reader. Perhaps they might consider citing Berg and Sheffield - Evapotranspiration Partitioning in CMIP5 Models: Uncertainties and Future Projections. Similarly, you might wish to more explicitly raise the issue of discrepancies amongst how models simulate LAI and the impact this has on the water cycle / ET partitioning.

3. Again feel free to ignore this, but this is one of the few papers I've seen raise this issue. The authors neatly raise the issue of interception. In our 2013 GCB paper on WUE (Forest water use and water use efficiency at elevated CO2: a model-data intercomparison at two contrasting temperate forest FACE sites), we found that the proportion of intercepted water varied among the models by between 2-14%. This was considerably below the field estimates for the sites (and the range you quote in 3.6). It was striking how data free the assumptions were than underpinned how interception is treated in models. I'm not suggesting you get into how models simulate interception, I just think you might consider highlighting this is a serious problem for models and may contribute to erroneous partitioning ratios - see above.

3. The paper didn't seem to make much of soil evaporation? I realise it is a minor component of total ET, but recently we noted how poorly this was simulated by models. In a water-limited, semi-arid ecosystem, some models thought soil evaporation was around 50-130 mm yr-1, whilst other models thought it was 2-3.5 times greater (Challenging terrestrial biosphere models with data from the long-term multifactor Prairie Heating and CO2 Enrichment experiment). I only raise this example because it suggests to me that this isn't a trivial process to model (otherwise there wouldn't be this disagreement). I was expecting to see some sub-section on soil evaporation, but this may simply be my personal bias on this issue, so ignore as you wish.

4. In table 2 where the variability in the exponential term is shown across models, I feel whilst interesting - without some context or explanation, it is a bit limited in terms of insight. Could the authors group the models by their stomatal assumptions (Ball-Berry, Leuning, etc). Does this help explain why they vary? Why is the BEPS model most similar to the observations?

5. Was there a reason the ECOSTRESS mission wasn't mentioned in section 5?

6. When the authors discuss partitioning via the use of GPP in the WUE approaches, it would be worth mentioning that GPP isn't strictly an observation (so any errors in GPP will propagate here).

Best wishes,

Martin

Dear Authors,

as discussed with Paul during the resubmission stage, I am going to send this out for a further independent review.

Many thanks,

Martin

Dear Authors,

I have now received two additional reviews of your manuscript, both are positive and I'm recommending minor revisions to address the points they raise.

Reviewer 1 was strongly felt that adding some text reviewing the state of modelling ET partitioning would enhance your paper. I realise that I also suggested two model-based studies (that I was involved with) that identified ET partitioning issues in the first revision of the manuscript. I would like to clearly restate my position here, I really was not suggesting you needed to include those citations, I was making a point aligned with reviewer 1, i.e. that perhaps the review could include a few more model-based studies, where relevant. I would like to see the authors consider this point, particularly as the reviewer has taken the time to highlight a number of potentially relevant papers. To be honest, I can't really see a downside to a few more sentences and citations in this direction.

Reviewer 2 also raised a number of minor points that should be carefully addressed in your revision.

Best wishes in your revision,

Martin De Kauwe

Dear Authors,

I am pleased to be able to recommend that your manuscript is accepted for publication. I think the strong positive reviews you received suggest it will be well received.

Best wishes for the future,

Martin