The authors have addressed most of the comments by the reviewers. However, some weaknesses remain unresolved. I am not convinced the authors present compelling evidence for a new climate change reinforcing feedback, because of the nature of the data, and the sort of analysis that the data allowed. I do see however opportunities for future improvements in this field, and therefore I would recommend the authors revise the paper, tune down their claims a bit and resubmit the manuscript to be considered for publication.
L17 Not clear what the authors call ‘species turnover’
L18 models were used to ‘estimate’ not to project
L19 the analyses ‘suggest’, not revealed
L20 …and correspondingly ‘may’ increase methane production
Please bear in mind that livestock do selective grazing, so at low stocking density, there may be a compensation for lower quality due to increased temperatures
L26 …[the suggested changes] ‘may’ reduce this additional source of …
L81-82 Please re-phrase: I don’t understand here the use of ‘where possible’ “Published statistical models were used to explore the effect of increased temperatures due to climate change to grass nutritive value and cattle methane production”
L110. This database only includes 1 dataset from the tropics (Brazil). However, the analysis of methane prediction is done at the global scale (Fig 5)
L175. The use of assumed constant nutritive values reduces the validity of models C, D and E (Table 1) to estimate future methane production. I wonder whether these models should have been excluded from the ‘global’ analysis
L232-236, were the ranges (Fig.2 up to 30 degrees) used in fitting models for NDF respected when making projections using future temperatures? It is not clear from Figures 3 and 4 whether there were ceilings to the temperature change.
L251-260. I wonder how useful it is to use models with more feed quality variables, when only NDF changes with temperature.
L275-281. These ‘projections’ are the weakest part of the manuscript, given all the assumptions and unbalanced nature of the datasets used in the statistical models, with the tropics and Asia heavily under-represented.
L330. What do the authors mean with turnover in species identity? I suspect the authors refer to a change in the composition of the plant community drive by increase in temperature – a sort of adaptation mechanism?
L338. I am not convinced the predictions are robust. The authors find a statistical significant relationship, not a causal relationship.
L349. I suggest to replace Projections by Explorations
L358. Please make sure that the 5 degrees’ increase doesn’t place the ‘predictions’ outside the range of validity of the statistical models.
I wonder whether the authors have seen the paper by Caro et al 2016 Mitigation of enteric methane emissions from global livestock systems through nutrition strategies Climatic Change August 2016, Volume 137, Issue 3, pp 467–480 – which is similar to their study
L403. It seems the authors don’t know the literature well. The mechanistic cattle model (RUMINANT) by Herrero et al 2013 (paper cited by the authors) simulates methane emissions for individual animals, of different breeds and for different regions if feeds are known. The LINGRA model maintained by the Plant Production Systems group of Wageningen University (see http://models.pps.wur.nl/node/958) simulates grass quality, and is driven by temperature, rainfall and radiation. It is likely that the combination of both mechanistic models could produce more robust estimates than the ones presented here by the authors.
L419 Conclusions should be tuned down in ambition. Preliminary evidence is presented here