This paper is excellently written and focuses on a very relevant issue, concerning the lack of representation of savanna ecosystems in global models. The work proposes a parameterisation of Brazilian savannas in LPJmL DGVMs, with great potential to be further developed for application to other savannas globally. All my comments are minor and can be found directly in the pdf. 

The paper is very clearly written and structured and focuses on an important knowledge gap by implementing a specific plant functional type for tropical savannas in the LPJml_VR-SPITFIRE model based on observational and literature constraints. The authors find that the inclusion of this PFT increases the model performance for vegetation distribution, biomass and fire. The results are promising and open up new interesting and relevant research lines among which are described in the discussion section. I only have very minor comments as are described below:

line 109, trailing e

line 165 It is unclear what fire disturbance is here

line 429 why is era interim used for evaporation and not the more recent era5? This is only a small comment and if the authors have good reason there is no need to do additional work to change this as it does not impact the results in a meaningful sense (only figure S9 I suppose)

line 444-445 not clear to me what is meant here, is it same thing as what is explained in line 446-448?

line 462-466: the fraction of raingreen trees also seems reduced in the savanna simulation, it might not be very important as it only covers very small fractions in the default run as well but is there any reason for this?

line 523-524 this phrase was quite unclear to me when I first read it, after reading the discussion it became clear to me but please avoid vague statements such as referring to 'real motivations' here.  

line 556-557: unclear what spatial burned area patterns implies here, could you clarify? I would assume the model has more spatially concentrated and intense fires than observations, is that was is implied here, please be more specific.

This manuscript reports on adding savanna biome Cerrado as a separate PFT in the model LPJml-VR-SPITFIRE and testing the results against various applications. The manuscript is well-written, and can be published after the following comments are taken into account:

Sec. 2.2, water stress mortality. I lack explanation why it appears as a separate topic in Methods if no results are reported later related to that, no comparison of results with this type of mortality included and without it.

Sec. 4.2 While discussion on Amazon part is clear, I think one cannot expect an improvement here from the inclusion of Cerrado type. For the reader, it would be easier to follow the discussion if it was kept closer to the results, consequences of implementation of this PFT. For example, it would be interesting to discuss why the regional distribution of burned areas within Cerrado domain has improved in some areas but worsened in others. Also L647-648, this is not related to discussion of these results, while important for the model in general and can be moved to conclusion

L498-499: ET and GPP, the authors report no improvement, but could some numbers please still be put there in the sentence, not sending the reader to supplementary? It would be also nice to have some other numbers in the discussion (more quantitative comparison with previous studies).

Minor comments:

L126-130 report numbers that are good to have in mind thinking of the impact of the work done: this is a large biome, and it is not represented in the models. I suggest moving this information to the Introduction.

L 134: ‘classification system, characterized by a rainy season’ – please rewrite

L147-150: I suggest moving these two sentences to the end of this section. There you already mention impact of fire on the vegetation formations, which you define later, in the next sentence.

L192 phen should always be written with small letters even if it starts the sentence.

L283: ‘earlier in their lifecycle’: I see from the figure at smaller heights. Fig. 2 does not show how quickly the trees grow, so I am not sure if ‘earlier’ is the correct word

L328-329: a bit awkward definition of VPD, it is just the difference between actual and saturation water vapour pressure. Also VPD was mentioned earlier but is explained only now.

L337: this is easy to read as alpha multiplied by VPD. I'd suggest make ‘VPD’ a subscript to avoid it

L444: ‘we weighed the burned area’: can you provide the formula or describe better how?

L488-492: these two sentences largely repeat each other.

L 492: Fig. 5 should be referenced in the next sentence, which discusses the ratio.

L575: not sure ‘enhances’ is the right word here, ‘impacts’ could fit better.

L576-578 about improved simulations: I am not sure I agree with the whole statement. Total burned area yes, improved, also its temporal dynamics but spatial distribution of burned area has become somewhere better somewhere worse.

L609 ‘future trajectories’ of what?

L620 remove ‘Nevertheless’ it does not fit these sentences