The study by Nijzink et al examined a vegetation optimality model is predicting water and carbon fluxes as well as vegetation properties across five Australian savanna sites along a precipitation gradient. In general, I like the motivation of the study. I am a big fan of optimality theories and I also agree with the authors that the optimality models are promising tools (and may be better tools than diagnostic models) to predicting and understanding the interactions between biosphere and climate changes. The current study did a comprehensive evaluation of one optimality model that is based on the maximization of net carbon profit across environmental gradient, which is an important step towards better understanding of the optimality theory itself and developing related optimality models. In this light, I am supportive of publication of this study.

However, into details, I am not satisfied with the manuscript organization and the writing itself. Overall, I found there are many information currently included in the manuscript is unnecessary. The introduction is too long and contains many individual studies, which should be largely shorten with more highly-summarized findings/conclusion from existing individual studies. The detailed site description is also not needed, simply summarize the five sites with their specific properties listed in Table1. Table 2 is suggested to move to supplementary.

On to content, I think the part of dealing with water transport cost parameter is more or less deviates from the main line. I would suggest remove the second hypothesis but describe how this parameter was chosen (either prescribed following previous studies or locally parameterized) in the manuscript. Then, the overall structure of the manuscript become: 1) test the VOM using site observations and compare it with TBMs; 2) what happened if remotely sensed vegetation cover was used? 3) what happened if prescribed rooting depth was used. Followed by discussion. I understand that the water transport cost parameter is also related to the overall performance of the model, but if that is included, why not other model parameters? And also you will need to describe you model in detail to allow readers who do not familiar with the model understand the role of this parameter in the model.

Another question is why not include a scenario that consider both prescribed vegetation cover and rooting depth, in comparison to the scenario with both vegetation properties optimized.

Other comments:

Line 215：infiltrate -> infiltration

Line 216: Why 30m? Is this the defined soil depth in the model? Not sure if the choice of this depth impacts the modelling results.

Line 225-230. This may present a source of uncertainty, as the observed fluxes are directly linked to the observed meteorological forcing at the sites, whereas the SILO data was used here to inform the model. Suggest to at least evaluate the used SILO data at each site against site-observed meteorological variables during their overlapping periods.

Line 223 Is there any published paper supporting this? Otherwise, simply states this information is measured at each site.

Line 260-263 and the following sections. If I understood correctly, the last two hypotheses are related to replacing vegetation properties with prescribed values and the second hypothesis is about water transport cost parameter. Please check.

Line 246 and throughout: evapotranspiration is often written without a hyphen.

Line 316-317: Where is the evidence for this? Figure 3. Please indicate where needed. In addition, in figure 2 at Howard Springs (but not for all other sites), there is a light green curve indicating the results of Schymanski 2015. What is the difference is model configuration between Schymanski 2015 and this study? And what is this for? It is not introduced.

Line 400. This is an overstatement. Looking at Figure 2, the VOM considerably overestimates GPP from observation and even compared with other TBMs.

Line 500. I do not agree with this hypothesis/statement. This may simply caused by the fact that the adopted VOM was not able to reproduce the actual rooting depth using the embedded optimality principles. Many previous studies have already demonstrated the importance of accurately representing rooting depth in the hydrological model to improve the modelled fluxes, for example, those by Kleidon and Heimann (1998) and more recently by Wang et al. (2016) and Yang et al. (2016).

Kleidon, A., and M. Heimann (1998), A method of determining rooting depth from a terrestrial biosphere model and its impacts on the global water and carbon cycle, Global Change Biol., 4(3), 275–286.

Wang-Erlandsson, L., W. G. M. Bastiaanssen, H. Gao, J. J€agermeyr, G. B. Senay, A. I. J. M. van Dijk, J. P. Guerschman, P. W. Keys, L. J. Gordon, and H. H. G. Savenije (2016), Global root zone storage capacity from satellite-based evaporation, Hydrol. Earth Syst. Sci., 20(4), 1459–1481.

Yang, Y., R. J. Donohue, T. R McVicar (2016), Global estimation of effective plant rooting depth: Implications for hydrological modeling. Water Resources Research, 52, 8260-8276.

With great interests, I have read through this interesting paper to understand vegetation dynamics along a preciptiation gradiant. I think it is publishable but subject to some revisions. Here I list some of my major concerns:

1. Hypothesis 1: One of my major concern concern it that the TBMs (or LSMs) in Whitley et al. (2016), which are designed for global-scale, long-term climate projectoin models, are optimized to make a fair comparison to VOM, though they have more specific vegetatioin parameters.

2.  Hypothesis 2 is too model specific. Does any other TEMs use this paramter? or what would other models learn from this? or does this hypothesis have any implication for plant adaptation or optimality? 

3. Hypothises 3 is also model specific. It depends on model representations of the links between LAI, vegetation cover, light transfer and absorption, assimilated carbon allocation. This mansusript does not clear describe how these processes are represented in VOM.

4. Hypothesis 4 shoul be the focus of this manuscript. If this would have been well tested, it is enough to be a good paper. However, this would be greatly affected by the rerepsentation of subsurface soil moisture profile and further likined to capillary fringe. I am wondering why the authors did not use the results of Schymanski et al. (2015). So I strongly suggest to include groundwater effects.

5. It is not clear to me (I have to read also Schymanski et al., (2015)) that how VOM is optimized. "Maximizing the NCP"? what is the maximum NCP? how do we know the maximum NCP? Please expand Section 2.2.4 a bit to describe this process in model detail.

6. Section 2.3.3, the cost factor for water transport (crv) should be described here. If not reading Schymanski et al., (2008), I did not understand its meaning.

7. Conclusions: please generalize these conclussions through discussions. I do not care much about how VOM is better or not or how to improve it but more about how to implement improved understandings through VOM studies into the current TBMS that are widely used in IPCC climate projections.

The study discusses the approach of vegetation optimality models to predict the hydrological and carbon cycle. The paper evaluates one optimality model in detail and contrasts the results with TBMs which allows for an interesting discussion of the benefits and downsides of the optimality model used. The introduction (and especially the abstract) describes the motivation for the study clearly and support the usefulness of the study.

In general, I felt the authors could have been a bit more careful with their writing. For example, they confuse their hypotheses throughout the manuscript (see below specific comments) which sometimes makes it hard to follow their argumentation. The introduction is a bit too long and, in parts, not very well linked. I liked how the authors explicitly point out the four hypotheses they aim to explore, however, it is not quite clear to me how hypotheses 2 and 3 emerge from the introduction. It further would be nice to have more details on the model description – you could move table 2 to the supplement for example to make more room. Your discussion is detailed but I'd like to see more about possible future directions.

Specific comments

Line 41-44 The sentence is very long and hard to follow

Line 61 Doesn't the default version of LPJ-GUESS have more than five plant functional types?

Line 91-92 The contents of the sentence are not linked very well

Line 98 Increase or decrease in annual rainfall?

Line 118 ‘optimizing vegetation properties to maximize the NCP’?

Line 136 Do timescales of precipitation matter? I.e. is annual PPT driving the rooting depth or are seasonal timescales more important?

Line 136 Therefore instead of but? ‘[…] therefore is likely to change over […]’

Line 162 In table 1 it’s AU-How

Line 173 In table 1 it’s AU-DaS

Line 199 ‘[…] is defined by maintenance respiration, projected cover to the turnover and maintenance of leaf area’ – I find this sentence a bit unclear

Line 208 How can seasonal vegetation cover vary on a daily basis? Maybe rephrase

Line 225-230 Does SILO provide point data or are the site met data derived from a spatial dataset (if yes which resolution?) I understand the argument that a longer timeseries helps to run the model, but it would be nice to see any sort of comparison between the observed met data at the site and the SILO dataset. I guess in general I would just like to have more information about the input forcing to get an idea about the uncertainty. Do the models from the Whitley paper run with the same meteorological forcing or do they use the data collected at the site?

Line 235-236 Can you provide a bit more detail about the water retention model? It was never mentioned before

Line 247 Can you specify what the ‘usual energy fluxes’ are

Line 249 LE has already been introduced two sentences earlier

Line 264 Isn’t the last hypothesis about rooting depth?

Line 265-276 I might have just overlooked in your submission – but can you describe in more detail what the model set up is for the model intercomparison you use from the Whitley et al paper? Surely there will be more detail in the Whitley paper to help understand but while reading your submission I was for example wondering whether there are changes to some of the parameters in the models to capture the site specifics better or whether they 'just' ran in their original configuration with the meteorological forcing from the sites [...]

Line 277 Third and fourth hypotheses?

Line 294 In the introduction it says -0.1-0.1 for the cost factor for water transport (but I might have understood?)

Line 297 second hypothesis (also Line 298)

Line 300 ‘Regardless of the result here’ can you explain why you make this decision?

Figure 2: Maybe include shaded areas for dry and wet season, but also include dry and wet season months in caption. Can you include what the ensemble years are too? Panel e says Daly Uncleared but it was referred to as Daly River before

Line 306 Can you define dry/ wet season (which months)?

Line 327 Not sure, it looks like the minimum is quite similar for the models but for the maximum values, LPJ-GUESS and MAESPA seem to be too low

Line 400 I’m not sure I agree with this. It may be true when you look at the annual values in figure 3 but based on figure 2 you can’t really reach this conclusion.

Figure 3: Why are the data points connected in panel a and b? It already is hard to distinguish the data points, the lines make it even harder. You also do not connect them in the other panels – it might be nicer to be consistent. Further, it might be helpful to offset the points in a and b (like in c-f). You could also increase the size of the observation marker, it gets lost in all the other points. Lastly, it could be helpful to include an arrow below the lower x-axes indicating whether the sites go from dry to wet or the other way around (but a lot of this is personal reference of course)

Figure 4 and 7: Maybe use a global legend and remove obs legend from panels. Are you ever using the information of Qflag in the results or discussion? If it never comes up you might as well delete it.

Figure 6 the text on the x- and y-axis is too small

Figure 8 It would be nice to stay consistent in the color choice for the models

Line 569 Aren’t some of the models (all?) processed based and not empirical?

Figure S2.1, S2.4, S2.7, S2.10, S2.13, panel h – can you maybe add padding between 100% projected cover and the figure edge, the way it is now it looks like you’re cutting off at 100%

In general, you have a lot of supplementary figures but don't refer to all of them in the manuscript.