the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 31 Jul 2025
Importance of plant functional type, dynamic vegetation, and fire interactions for process-based modeling of gross carbon uptake across the drylands of western North America
Abstract. Drylands cover ~41 % of the Earth’s land surface and contribute more than one third of the global net primary productivity. Several studies have demonstrated that drylands play a crucial role in global carbon cycle interannual variability. However, drylands are vulnerable to the impacts of climate change. To predict changes in dryland productivity under climate change we depend on dynamic global vegetation models (DGVMs). Compared to more mesic ecosystems, DGVM carbon cycle dynamics have not been widely evaluated against data. Existing studies are mostly focused at site scale; rarely have these models been assessed or benchmarked against dryland carbon flux products at regional to global scales. Global gross primary productivity (GPP) products have poor performance in dryland regions. Only recently upscaled in situ flux products have been developed specifically for drylands. Here, we evaluated GPP inter-annual variability (IAV) simulated by 15 DGVMs from the TRENDY v11 model intercomparison project against theDryFlux GPP, which is newly developed upscaled GPP product that considers dryland- specific ecohydrological responses. Comparing model simulated GPP IAV to DryFlux, we identified two groups of models: a one group of models with generally lower GPP IAV than DryFlux (e.g., lower standard deviation in annual GPP than DryFlux and slope values of the linear regression between each model and the DryFlux product that are less than 1.0) and a second group of models with generally higher GPP IAV than DryFlux. We examined if including a representation of dynamic vegetation (i.e., changes in the spatial distribution of plant functional type (PFT) fractional cover) or fire in the models can explain the inter-model spread and model performance in comparison to DryFlux. Models that do not include a representation of fire and/or dynamic changes in plant functional type distribution over time generally have lower annual GPP variability compared to DryFlux (1st group of models), except for the eastern and southeastern region of the study area with high rainfall variability. We also found that models with dynamic vegetation exhibit high variability in grass fractional cover (that was higher than two independent reference fractional cover datasets), which was strongly correlated with high GPP IAV. Only some models that included fire simulated burnt area annual variability that correlated well with GPP IAV. Other models that included fire simulated low burnt area variability and therefore we did not find any strong relation between burnt area and GPP IAV. Finally, we examined the relationship between the dominant PFT and GPP IAV. We did not find a strong correlation between the spatial mean of the slope of the linear regression between each model and DryFlux annual GPP and their spatial mean woody, grass, or C3 grass fractional cover (although many models with generally low GPP IAV had higher woody plant cover). However, we did find a high correlation between the slope of the linear regression between each model and DryFlux annual GPP and spatial mean C4 grass cover. Therefore, our findings suggest that DGVMs inability to accurately represent the spatial distribution of herbaceous (specifically C4 grass) cover as well as processes controlling dynamically changing vegetation distributions over time (including fire) contribute to poor model performance in capturing annual variability in dryland productivity. Our findings can provide a roadmap for DGVM teams seeking to improve vegetation representation in sparsely vegetated dynamic dryland ecosystems.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Biogeosciences.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(4961 KB) - Metadata XML
-
Supplement
(12797 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2025-2841', Anonymous Referee #1, 30 Aug 2025
-
RC2: 'Comment on egusphere-2025-2841', Susanne Rolinski, 01 Mar 2026
Review of “Importance of plant functional type, dynamic vegetation, and fire-interactions for process-based modeling of gross carbon uptake across the drylands of western North America” by Pervin et al.
General assessment
The study investigates on the representation of the variability of vegetation cover and productivity in models that contribute to the global carbon budget activity TRENDY. The focus lies on a specific semi-arid region, the western part of the US, and properties are discussed not only of the models but also of the potential data products that are used for comparison. The paper is very technical with a strong leg in the methods but at the same time suffers from vague method descriptions. The presentation of the results is the part needing most improvements whereas the discussion is smart, concise and appropriate. The main critique in this review targets the presentation of the results which could be much more reader-friendly. This refers to:
- the naming and use of variables. Variability and standard deviation for GPP and vegetation cover are used but the repeating use of descriptive and long naming hinders the understanding. Please find for all the used variables (also coefficient of variation and the slope values of the linear regressions to the data) telling names and use them consequently in the text.
- the grouping of the models. Also here, consistent naming of the different groups and the use of these names could improve readability and make the text more concise.
My recommendation is ‘major revisions’.
Specific comments
title: This is very model-specific. Is it possible to adjust it and include a statement coming from your analysis and addressing a wider audience?
Abstract: much too long.
L48: ‘global net primary productivity’ – you mean the terrestrial contribution, right? Please clarify.
L49: also global versus terrestrial
L51: really? What do you think about upscaled Fluxnet data products like Jung et al. (2020, doi: 10.5194/bg-17-1343-2020)
L53: Since DGVM carbon cycles have been widely evaluated, please insert ‘from drylands’ at the end of the sentence to be more specific.
L53: The sentence starting with ‘Existing studies’ is somehow not complete. Please clarify.
L55: Which GPP products do you mean? Please be more specific.
L55: I would prefer a comma after ‘Only recently’ but this is personal choice for better readability.
L58: In case there are references, please include them after ‘intercomparison project’ and ‘DryFlux GPP’. When there is none for the latter, maybe inlcude ‘the here presented’ or something similar before.
L63: The sentence seems odd using ‘if including’. Please reformulate.
L68: ‘study area’ is mentioned here the first time without stating it. Of course, ‘western North America’ is part of the title but it would be nice to specify it before.
L71: Could you give some detail or a reason why it seems a wanted feature that the annual variability of simulated burnt area correlates with GPP IAV. This aim should be motivated.
L72: The term ‘any strong relation’ is misleading. Please clarify what you were looking for. Was the aim to find any relation or did you only accept significant correlations?
L74-77: Also here, it does not become clear which kind of correspondence you were looking for. In the abstract, it would be helpful to leave the technical terms aside and give an impression which property of the model simulations matches those of the observations and for which you were looking in vain. On the other side, be more specific when you characterize the correlation and omit terms like ‘strong’ or ‘high’.
L78: It remains unclear what is related to what because you have 3 datasets connected with ‘and’. Please clarify. Did you correlate annual GPP from each model with DryFlux? How does C4 grass cover come into play?
L93: ‘earth’ → ‘Earth’
L95: It would be good to include here the abbreviation ‘ESM’ after ‘earth system models’.
L100: Which ‘vertical structure’ do you refer to? Do you consider just plant compartments such as leaves, stem, or roots or height distributions of the biomass? Please be more specific.
L101: Start a new sentence with ‘However’.
L105: ‘ESM’ is used without definition. That’s why I suggest to include it in L95.
L106: There is a bracket missing after ‘(Bonan, 2019)’.
L111: Again start a new sentence with ‘However’. This applies also to many more lines.
L115: Please include ‘applications in’ or similar after ‘Compared to’.
L116: Here and at other places some white spaces are missing before brackets.
L121: Sentence with ‘;’ not easy to read.
L126: Please include ‘at the’ after ‘versus’.
L126-132: Please make two out of this very long sentence.
L143: You mention ‘exact causes’ for the model behaviour and I am not sure what this would be even after working for more than 30 years with models. Could you be more specific what you expect to find?
L149: Again you use a ‘;’ instead of beginning a new sentence. Please avoid this.
L149: Maybe better ‘the type of dryland vegetation’ instead of ‘dryland vegetation type’.
L157: You mention PFTs are not ‘well adapted’ to drylands. Please be more specific if they are not well parametrized or rather lack processes or linkages so that the kind of improvements becomes clearer.
L170: Spitfire reference Thonicke et al. 2010 (doi: 10.5194/bg-7-1991-2010) to be included?
L191: In this case, a new Sentence is more appropriate then “;”.
L201: I would have expected that the target is to match the variability of model results to those in the data. Surely, deviations from the data variability have to be specified and reasons have to be found. Please clarify and reformulate.
L203: Somehow, the sentence is not complete. Or do you mean that the second part should be read “causes in differences in model…”? That would become clearer, when both objectives are formulated in single full sentences and not as a listing.
L210: I would prefer “subsequent to a disturbance” instead of “following disturbance”.
L217: I would prefer “hypothesized” instead of “predicted”.
L228: Maybe better “simulate” than “predict”.
L262: The bracket is missing after October and the whole sentence is hardly readable with the two “are”. Please use two sentences.
L267: Do you mean “bi-modal”?
Fig. 1: Please bring the numbers at the box of the legend instead of the lables “semi-arid (0.2-0.5)”.
L282: Do you mean “selected” or really “elected”?
L284: The data are usually named “CRU JRA”.
L285-287: Please delete “what” and “included” to make it a sentence.
Table 1: I would shorten the caption so that it only describes the entries. Include explanations in the text.
L308: Maybe better “have a combination of” instead of “have combine”.
L315: The sentence starts a bit unfortunate with “If a model”. Please reformulate.
L322: This statement is a bit weird. Please either talk to the people directly, check the website or use other means to clarify this.
L328: Include the missing reference.
L341: Please include also a reference for the SPEI (and how you calculated it).
L344: I assume, you mean “seasons” instead of “reasons”?
L351: Delete the hyphen between ground and from.
L354: Also the hyphen after indices seems misplaced.
L365-366: definition of trees with height above 5m is doubled. Please reformulate.
L373: Please give download location, reference and date of download for MODIS data.
L384: There should be brackets around 2022.
L388-391: Please reformulate to make it better readable maybe using some like “out of two reasons: first ...main focus is natveg … and second … no crop flux tower sites”.
L394: Maybe “included in” instead of “common to”.
L395ff: When you introduce the term IAV for interannual variability for GPP and other variables, please define it (standard deviation of annual values?) and also use the term consequently. You can also define special terms or characters for specific variables that reoccur which would increase readability.
L396: Maybe begin the sentence with “For comparing the interannual GPP between…” and express it more simply.
L399: The closing bracket is missing (at the end of the sentence?).
L407: The paragraph is motivated by “Data analysis” but just plotting the data is not an evaluation. Please give reference and state the metrics used for the evaluation.
L409-412: Please reformulate and simplify the sentence.
L413: Maybe better “within the model ensemble” instead of “between models”.
L416: Could you use two sentences maybe dividing after “soil”.
L420: Specify how you did the comparison. Did you only compare visually or used a metric?
L424: Your sentence suggests that the linear regression statistics show the ratio of variance. Could you explain and specify this in the text.
L427: Either give a reason why CV is more appropriate or just state that you used the relative variation.
L432: Since “benchmarking” is a wide field, please specify how your benchmarking is performed.
L435: Delete the beginning until “fire”.
L437-441. The sentence is too long and the punctuation mark at the end missing.
L439 and 442: twice stating the threshold values. Please clarify.
L443: Again “plotting” is used for evaluating which should state the method or metric.
L447: It would be interesting to read how you “assess” the relationships. Please specify.
L452: Maybe this is only a personal preference but it would be nice to start the paragraph with a statement and avoid this “Fig. X shows…”. What is the most striking and interesting information you want to convey for the time series of the area sum of GPP?
L453 and 454: “curves” instead of “curve”
L456: Insert a space after tDryFlux.
L457 and 459: You specified these colors for a certain reason. Better use the content-related grouping and specify that these are denoted by bluish or brownish colors.
L467-468: The statement should be moved to the methods section.
L469-475: Please simplify this long sentence.
L479: Please clarify that your target value is 1 and that the values give information on over- and underestimations.
L481: To omit the data is completely fine but you could give some values for mean +- sd inside and outside of the region (and the respective source).
L488: The part of the sentence after “which are” is better placed in the methods section. I would recommend first to describe the target, the KDE of data and then the models. The whole section is very technical.
L490, 493, 495: Using the colors in this way is not so helpful. See my comment below.
L518: Is it right that Fig. 4 shows the same information as in Fig. 3d but in a different order? And the difference between Figs 4a and 4b is only the order, right? The order is determined either by grouping the models according to their dynamic vegetation and/or fire module inclusion. Maybe, the reader would benefit from a different presentation in a table-like style. You have a group of models without fire/dynveg modules, a group with both and a group without dynveg with fire (no model with dynveg and without fire). Thus, one dimension of the table would be natveg yes/no, the second dimension would be fire yes/no. The KDEs of the models could be placed into the respective quadrants. Please think about this possibility and maybe you have an even better idea.
L520: Please exchange “whether fire was included” with “the inclusion of a fire module”.
L532: Find an alternative to ”predicted” like “hypothesized”. This applies also for many other appearances of this word which is not appropriate in this context. You do not predict but you assume, hypothesize or suppose.
L543: You could help the reader by using your abbreviations like IAV and put in more to the beginning of the title like “Spatial patterns in IAV of PFT fractional…”.
L545: “section 2.2” is not a results section and should not be referred to as such. Then, you describe (again) the grouping of the models by their dynveg and fire module inclusion. It would be much easier to read when you find a name for the three groups (see comment to L518) and include this in table 1.
L546: Is this standard deviation the same as IAV? If not, specify it in the methods.
L548: “vegetation” occurs twice, please delete one.
L550: Here you explain the coefficient of variation which should be done in the methods, get an appropriate name there which is then used here.
L554: I understand that you want to mention LPX-Bern in this sentence but it seems to be done as a second thought. Please reformulate.
L555: You describe an “increased spread” of GPP IAV. Do you mean that there is no distinct mode of the distribution? Please be more specific.
L557-558: This is more interpretation and could be moved to the discussion.
L563: Again, a definition – of the coefficient of variation – which should be moved to the methods.
Fig. 5: Could you explain why the target is the perfect fit between the CV or did I misunderstand something?
Fig. 8: Also include the meaning of the red line into the caption.
L626: Similar to comment for L543.
L628: Also here, it would be good to find short names for both variables compared and use them consequently. This would shorten the text tremendously and focus better on the relevant parts.
L638: This relationship is not really visible. Better omit this.
L652: Somehow an “s” appears between GPP and (). Please delete.
L653: Include the reference to Fig. S1e after “fCover”.
L655: Close the bracket at the end of the sentence.
Fig. 9: Also include the color code for the models in the caption because you use them in the text (L647 and L648). Also the horizontal line for slope=1 should be included.
L670 and 680: The specifics of data sets RAP and MODIS should better be included in the methods sections 2.3.2.1 and 2.3.2.2 because the characteristics are properties of the data and not your results.
L684: Please reformulate for better readability.
L708: The term “this study” refers to the unpublished paper which is misleading. Please omit this part which goes beyond the announcement that there will be a further study.
L781: Do I get it right that the RAP data does not contain a specific crop layer but the crop contribution is included in the total signal? Please be more specific.
L786: Your paragraph does not only discuss the “inaccurate” contributions of the modules. So maybe better omit this word in the title or take it bit broader with “different” (or something similar).
L817: There are a bit too much brackets here. Please clean up the list.
L835: Please exchange “if it was better represented” e.g. with “in case that representation is improved”.
L838: Maybe the model names in brackets?
L855: Maybe replace the part after “question” with “how shrubs survive when grasses are...”.
L897: Although this subsection is interesting and relevant, please omit it because it relies on an unpublished paper.
L941: The word “accurate” does not really capture your discussion. Maybe use something with improvement or ecologically-based.
Citation: https://doi.org/10.5194/egusphere-2025-2841-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 1,488 | 131 | 38 | 1,657 | 117 | 54 | 70 |
- HTML: 1,488
- PDF: 131
- XML: 38
- Total: 1,657
- Supplement: 117
- BibTeX: 54
- EndNote: 70
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
General comments
The manuscript ‘Importance of plant functional type, dynamic vegetation, and fire interactions for process-based modeling of gross carbon uptake across the drylands of western North America’ analyses the simulated variability in vegetation carbon uptake by terrestrial biosphere models, and contextualises the results with observational datasets. The authors identify the representation of C4 grasses as a key reason for failure in reproducing interannual variability reported in observation datasets. Understanding interannual variability in the land carbon cycle (here done by analysing GPP) is very topical and relevant, and is becoming increasingly important for monitoring and successfully projecting ecosystem responses to climate variability and extremes.
While I like the overall scope of the study, the paper will need revising before it is suitable to be published. I would encourage the authors to shorten the paper significantly, which is, with 45 pages excluding references!, very long and at times verbose and repetitive. I also wonder whether the authors would consider changing the title to clearly communicate their focus on GPP variability but that is just a suggestion and I am also happy with the current title.
The methods the authors chose are overall relatively easy to understand, but I think they need to be more careful about the interpretation of the results. I especially found the use of the slope to assess the ability of models to capture interannual variability problematic, because the authors did not consider whether those relationships were meaningful, for example by testing the p-value. They also did not consider the correlation coefficient or coefficient of determination to fully understand the relationship between observations and simulated GPP anomalies. This is important because slopes can have values close to one, while not being significant and having low correlation and R2 values when one time series is more variable than another, which is a key aspect of this analysis!
Lastly, I appreciate that it is challenging to test the impact of PFT distribution, dynamic vegetation and fire on carbon fluxes in the TRENDY dataset, and in consequence results based on the grouping of models according to the (non-)inclusion of dynamic vegetation / fire were at times very speculative. I wonder whether the authors could have come to more robust conclusions if they had picked outputs from the ISIMIP fire sector where a (smaller) number of modelling groups have in fact run simulations with fire switched on and off. I am also curious about the choice of ecosystem variable: This paper is motivated with the importance of drylands for interannual variability in the net carbon balance. Especially in the context of fire, the more natural choice in key variable to me would be therefore net biome productivity (NBP) to represent the net carbon uptake. Why did you go with GPP?
Specific comments
In the following I have also pointed out a few typos - I think the manuscript could use more proofreading before re-submission.
L52-56 I think this is too much detail for an abstract
L58 Two typos - missing space between ‘theDryFlux’ and missing ‘a’ before newly developed
L60 Typo ‘a one’
L60-62 I would suggest to drop the text in the parentheses here
L63-64 Again, drop the text in the parentheses
L73-83 In this section too there is a lot of detail on the results but the abstract should only highlight key conclusions
L85 Would suggest ‘water demand (potential evapotranspiration; PET)’
L88 Remove white space between parenthesis and full stop
L93 Capital E for ‘earth’
L106-114 I think the description of TRENDY can be shortened
L116, 118, 122 Missing white space before citation
L189/190 Given the authors are mentioning an ongoing debate on which biome contributes the most the carbon cycle variability I would suggest to include an additional reference to Ahlstroem et al. that supports the dominant contribution of tropical forests
L201 The transition into the objectives is a bit abrupt
L201 Couldn’t a third hypothesis be that the models match the DryFlux product well? :)
L211 Here you say a potential outcome is a ‘better’ match with DryFlux GPP while before (see my previous comment) you say that models can only be too high or too low
L250 Was the aridity index calculated in this study, if so which method for the calculation of PET was chosen, if no, where is the dataset coming from [...]
L256 Would suggest to start a new paragraph
L257-258 You never use MAT or MAP again, so no need to define those abbreviations
L262 missing bracket after October
L256-269 I think this description is too detailed. Currently your Figure 1 is relatively large with a single variable, and given the importance of the regional climate I wonder whether Figure 1 could be changed into a multi-panel plot also showing MAT, MAP, and IAV of the relevant climate variables
L272 Have you calculated the Aridity index yourself - if so you should mention which reanalysis dataset it is based on, if you have taken someone else’s dataset it needs to be cited here.
L283 In line 255 you say you explicitly want to focus on natural vegetation, yet you take S3 with transient land-use change. Why?
L285-287 I’m not sure it is necessary to explain why some simulation types were NOT chosen
L287-288 This doesn’t add much, I would just cite Sitch et al., 2024 at the end of L285 like ‘(see Sitch et al., 2024 for further details on simulation protocol)’ or similar
L288 Which method did you choose to regrid your data?
L304-315 Again, this is very wordy. Could an illustration or a table help to summarise the key differences in PFTs in addition to a few sentences?
L317 Which ‘certain analyses’?
L319 Presumably the peat graminoid would be zero in your study region anyway? I’m not sure I would recommend to just drop PFTs that are a bit annoying based on their non-traditional definition if they might have a strong contribution to your variable of interest
L321-322 Did you test this?
L323-324 I don’t think ‘visually inspecting with Panoply’ is a robust method
L328 I find the soilcrust PFT argument a bit random. Would this not rather be an extra layer of the soil column?
L328 You are also missing a reference here
L330 Why was nearest neighbour interpolation used here? Isn’t this more commonly used for categorical datasets?
L334-347 I think this could be shortened, e.g. it is mentioned multiple times that it is an upscaled dataset. Once should be enough to get that point across :)
L356-358 Is it really necessary to give that much detail on different errors?
L361 How did you regrid your dataset?
L370-371 Again, is it really necessary to provide detail such as the exact errors for this dataset? The methods are already very long
L376 I would suggest, instead of mentioning it separately for each dataset, to have a summary statement on how the regridding is done in this section. Presumably all regridding has been done with the same method?
L382 Already mentioned earlier
L383-384 I hesitate to suggest adding yet another paragraph in the methods, but i) why does the aridity dataset not deserve its own paragraph like the other datasets, and ii) why was the aridity index not derived from the climate forcing used to drive the TRENDY models? Was it used to train DryFlux?
L388 Why 1970-2000? Wouldn’t it make sense to match the study time period, i.e. 2001-2016?
L393 I would drop ‘That means [...] for any model’, this is already clear from the description
L395 In this section I would suggest to have different subheadings for each of the statistical methods that were applied
L398 It is not clear here how anomalies were derived / whether some normalisation method was applied and if so, which one
L399-400 Was the data detrended for the pixel-wise comparison of time series with a Pearson correlation?
L423-425 Is this referring to the coefficient of determination? In the results it was not obvious this was considered for the analysis of the slope, at least not in all of the analyses
L442 I don’t understand what ‘typical uncertainty’ means, and I’m not sure how including the 10% threshold helps grouping your data according to PFTs
L435-445 I found this description a bit convoluted and hard to follow so my comment might reflect my lacking understanding - but would it not make sense to use the PFT-level GPP output to partition directly which PFT or PFT group contributes most to IAV in GPP?
L456 Typo - ‘tDryFlux’
L462/ Figure 2: I think it would be interesting to also plot the standard deviation or CV as an extra panel here
L464 I would suggest to rewrite to ‘The ensemble mean’
L467-469 See my general comments - I am not convinced that the slope of the linear aggression alone is very robust.
L469 I don’t find it obvious that the standard deviation is generally lower in Figure 2, perhaps consider my suggestion to include this metric as an extra panel
L481 Why is the data not shown - might make a nice addition in form of a multi-panel plot in Figure 1
L507 Would it not make sense to show the differences in standard deviation between models and DryFlux? I think it would also be interesting to also include the ensemble mean, also in panel c. In panel c, the colorbar is hard to read and it is hard to see where values are negative (is this a frequent occurrence? And if so what does a negative slope between the observation and simulation of the same variable mean?) It would also be nice to have a high resolution figure here, there is a lot of granular detail in this figure that is hard to see in the low resolution
L525-532 I find the conclusion a bit speculative here, and this is also reflected in the wording (e.g. L531 ‘possibly with the exception of SDGVM’). Given the LPJ model family is singled out here, it might as well be any other process this model family does differently compared to the other models. How can you conclude that your hypothesis around dynamic vegetation is partially right, when you basically have only two model types in the dynamic vegetation which show different responses?
L552 It is not obvious to me how LPJ-GUESS captures the spatial patterns of GPP IAV better than any other model? Did you run any statistical test to support this claim?
L553-554 Please reference the appropriate figure here (referring to the R2 values)
L555-558 Would a non-linear fit maybe produce a better fit? Also for LPJ-GUESS (?) Where is the value 0.05 coming from?
L561 It would be useful to be consistent with the variable naming, the x-axis shows fCover? Defined as 100% - bare soil? Why were remotely sensed datasets not included in this analysis?
L593 Abrupt transition into new paragraph
L597 Again, why are those maps not shown (in the supplement?)
L599 I think this comparison is hard to make visually at least because all maps in Figure 7 have the same colorbar but very different magnitudes. I’m not saying it’s not right to plot the map like this but could it be that some models have a similar pattern which just get lost in the value range?
L619 Why isn’t there a reference dataset for remotely sensed burned area in this figure?
L623 Typo ‘showing for all the’ - please remove ‘for’. Again, would it not be interesting to also include what the comparison of the remotely sensed datasets would look like to get a sense for the ‘ideal’ relationship in the models?
L692-713 This section is very thoughtful and covers a lot of potential shortcomings, but 1 page is quite long for something that can be boiled down to ‘there is nothing wrong with the DryFlux dataset’. I assume this is also not meant to be the key result of this study so I would suggest to shorten that section and move it towards the end of the discussion where potential shortcomings in the study set up might be discussed (although clearly this dataset was not a shortcoming in this study)
L703 Typo, ‘..’
L725-742 This section kind of lumps together different aspects of PFT distribution, and is a very long description of defining different bioclimatic limits which can be summarized to the fact that all models discussed here use a temperature threshold to define where C3 and C4 grasses are allowed to grow
L755 Here it would be interesting to know (briefly) how this new distribution map was derived
L757 NEON hasn’t been defined
L760-761 Then why wasn’t this tested in this study?
L788 Wasn’t one conclusion that the (non-)inclusion of dynamic vegetation and/or fire could not be *really* be related to GPP IAV??
L792 Again, speculative wording ‘slight exception’
L801 But the ‘reality’ wasn’t shown in this analysis
L799 I wonder whether it would be better to refer to the models in TRENDY as terrestrial biosphere models (TBMs), because, as stated in the paper, not all TRENDY models simulate vegetation dynamically, which the name DGVM (dynamic global vegetation model) implies
L807 This statement is not really supported by the analysis as ‘only’ the IAV is studied. Dominance of a PFT is also linked to long term averages, and if a region experiences on average high fire activity (even with low IAV), I’d expect grass dominance for this case too
L808-809 ‘based on our knowledge of the study area’ is not an appropriate reference
L809 Why is this analysis not included?
L828 This is nothing you can do about in this experiment set-up, but I would say it is inherently difficult to derive a meaningful relationship between fire and other ecosystem processes in any model using the GlobFIRM model which has been shown to neither capture mean or interannual variability in burned area anywhere in the world. The strong focus on fire does make me wonder whether it would have been more useful to make use of the ISIMIP simulations, with a similar experiment protocol compared to TRENDY AND an additional sensitivity experiment where fire is switched off.
L832-835 This is purely a model result. Is this supported by observations? It might as well be an artefact where in those fire models burned area and GPP are too tightly coupled
L836 Not clear - does different fire parametrisations refer to different fire model ‘groups’, e.g. grouping all TBMs using GlobFIRM into one sub-group?
L837 Typo - two full stops
L840 As I mentioned above, these simulations actually already exist in a different MIP (ISIMIP)
L847 Is the Walter hypothesis in the majority of TBMs or why is it highlighted here? What does it entail? If this detail is not relevant this mention should be removed
L898-920 In section 4.5, the only reference to other literature is Bogucki et al. in prep. I noticed that this section was also written by L. Bogucki (according to the author contribution statement). I understand that it is tempting to rely on ongoing work but given this work is not even submitted yet, I would suggest the authors make an effort in supporting and validating their statements with published studies.