Dear Authors,

many thanks for the revised version. Based on your responses to the reviewer comments and my own reading of the manuscript I see the capability of this manuscript to become publishable in Biogeosciences after you have taken on-board the below major opportunities to improve it.

Editor comments:

Title: The title of a manuscript should confer with its content, not a potential future application. The model that you present is a typical offline model. It may be coupled to an HESM, but since it is currently not, the title of the manuscript is misleading. Please provide a title that corresponds to the manuscript.

Abstract:

Please limit the abstract to a concise description of your research. The first nine lines of it are pure introduction, without any direct link to your research. They may appear in the Introduction Section, but not here. Please shorten this motivation section of the abstract to one or two sentences.

Abstract P2 L3-4. “in boreal regions and xeric ecosystems, as well as fire size distribution” (otherwise the listing does not work out properly).

Abstract P2 L 4-7. The manuscript does not provide any evidence for the capacity of this model to make projections, and thus this expectation should not be raised in the abstract. The sensitivity of fire activity to what?

Keywords: This paper does not deal with the coupled human-Earth system (even if this is your ambition in the long-term). It simply describes a standalone fire model. Please remove “human-Earth system” from the key-word list.

Introduction:

paragraphs 3 and 4. Given that in your paper you do not address future scenarios, this section could be shortened.

P 5 L 13: I do not see the need to include these references. Neither is this list a complete appreciation of the vegetation patterns simulated by current DGVMs (the papers are more than 10 years old!), nor do the quoted figures make any direct link of this vegetation distribution to fire. The sentence is ok without any other reference than the Scott one.

P5 L22: Reconsider the use of “fire ecology” (here and hereafter), as in your case this only seems to refer to the length of fires. Your model does not account for any e.g. biodiversity or stand density or size related aspect of fire.

Methods:

P7 L5: Remove text in brackets. Please also note that the address you supplied is incorrect. The code is not available when following this link.

P8 L7: Add location of the fraction of no-fire-land here. A reader should be able to understand the model without having to go back and forth.

P8 L9: Irrelevant in the Methods section. Remove this sentence.

Eq 3: Something is wrong here: I’ve got a hard time understanding why you have an integral in this equation. The term in the integral (according to my understanding) is not a function of LU, but a scalar. There is a second error in this equation, because according to eq 1, LUign should be the result, not the input to this equation. And then, what is Humign? I appreciate that this is a heuristic function, but it’s probably worth while plotting this function as supplementary figure, such that one can understand what’s going on.

Eq 4 Max_rate is undefined.

P10 L 5-8 Figure S1 does not allow to constrain any of these parameters

P 11 L 10: why is the termination stochastic, if the rules laid out in line 6ff appear to be deterministic?

Eq 9: Please provide an equation for Weather_termp

Eq 10. I share reviewer #2 concern regarding the derivation of this equation. In particular, I cannot follow why fire should respond to rainfall 15 months prior to today - that seems a fairly large memory effect, which according to your response is not (as such) included in any of the cited studies. Please clarify in the manuscript why these values were chosen.

P12 L5: The replacement of Fuel_exp through information from litter and other carbon pools is non-trivial. Either explain in detail how you would do this, or remove here and outline in the discussion.

P12 L9f: How does the model allow for fuel build up? I thought fuel was not explicitly represented?

P13 L 14: What do you need Eq 3 and Section 2.2.1.2 for if LUign is calibrated (and used in Eq 1)?

P14 L4: The text says that RH is scaled between 30 and 80%, but Figure 2 suggests that this was scaled between 30 and 70%. Please correct and outline the implications of this error.

P14 L 13: Certainly there would also be convergence problems if reality would be strongly non-monotonic and you try to fit a monotonic function to it?

P 15 L 10: Is it relevant or not? relevant according to which objective?

P15 L 19: I don’t follow this argument. The model does not know about continents, but about temperature, precipitation etc. If you wanted to demonstrate that the model works outside the fitted driver data combinations, then you would need to demonstrate that in the space of the data that you provided the model with. How can you exclude that the drivers of fire in South America are indeed not captured by your selection of points?

P16 L2-6: Unless you provide your insight from these different configurations etc. in the discussion, this section is irrelevant and can be removed.

P 16 L 9: give years

P 16 L10: I assume you refer to area burnt, not the size of a fire?

P 16 L 13: Why do you not take account of the information from the optimisation procedure and use the standard deviation of these parameters? This would be a much better way to not only talk about model sensitivity, but model uncertainty. A missed opportunity.

P26 L 18-P27 L 6: Given that this is outlook and not discussion of your research per se, please shorten. I also don’t see the need for this to have a separate sub-heading.

P28 L 16: Whether or not this piece of software is a relevant tool should be judged about by the people who will now use it - I do not think that it is an appropriate statement to make here. Please rephrase. Also “certain aspects” is very weak - please be more precise.

Figure 1: To avoid confusion with the cut on the right-hand side of the figure, remove all boxes that are not fully displayed.


Reviewer #1 responses

#1: Please clarify precisely, how you have addressed this comment (I cannot identify this from e.g. “see below” - where precisely? Your response seems to suggest that there are too many factors, which are relevant, that the model does not take account of. Possibly these factors can be included, but you do not seem to provide any suggestion as to how you would actually do this consistently. So, with this lacking strategy in mind, how can you then conclude that the model’s predictions will be relevant also in the future (other than simple sensitivity tests that the available models are equally capable of)?

#2: I agree with the reviewers comment. The title of a manuscript should confer with its content, not a potential future application. There is no such thing as a one-way interaction (contradictio in adjecto). If a affects b, but b does not affect a, then this typically called effect. The model that you present is a typical offline model. It may be coupled to an ESM, but since it is currently not, the title of the manuscript is misleading.

Please revise the manuscript wrt to this. Please also propose an alternative title, which reflects the content of this manuscript, such that the readers get an appropriate impression of the manuscripts topic (the development and evaluation of an offline fire model).


#6 I suggest to plot the precipitation-temperature space as an additional supplementary figure, to demonstrate that you’ve captured the dominant climate regions. I am confused by the omission of South America - given that your objective is to develop an Earth system model - why omitting an entire continent?

#8 Help the editor and identify the sections you change - makes it easier for the editor to believe that what you state has entered the revisions.

#9 Please provide this information in the revised manuscript (the fact, not the figure). See comment on Figure 1 below.

Reviewer #2 responses:

#2 Maybe 1° is not coarse for a fire model, but certainly relative to the size of the average fire. Regarding the model-data discrepancy, I would expect that the linear averaging over climate conditions, which have non-linear consequences for the process you model will certainly have implications for its performance. Given that, as you rightly state, you will not be able to increase the resolution in an Earth system modelling context to avoid scale issues, identifying and discussing principle limitation of the model performance against Earth observations is important.

#3 Even is this term is renamed, it is still important to justify the shape of the equation (13). If there is no or limited evidence, how can you justify an equation with 4 unknown parameters and an undefined function. Please provide a derivation of this equation and an explanation for the wind function.

Dear authors,

many thanks for your revised version. I can see that the manuscript has been improved. However, I think that still a range of minor clarifications are required before this manuscript can be published.

My comments:
P3 L5 remove this sentence

Eq 3: The integral still does not seem be correct. I assume that what you are wanting to express is that for any grid cell you calculate the term within integral for each subfraction occupied by a particular land cover type and then sum this up to the gridcell by i) multiplying with the cover fraction and ii) then making a sum? This would be more correctly then represented as a sum over the land cover types. In any case, the integral borders LU=0 and LU=LUtot are not meaningful, because LU is not an index of any of the terms within the integral

Eq 4: This seems to be a fairly complicated way to say that LUn is constrained to solutions between 0 and 1. Consider rewriting to LUn= max(min(fraction_of_Else,1),0)

Eq 11: Please explain how these thresholds were derived, or clearly indicated that these values are based on heuristic reasoning. Also, now that I understand what this does - is this really fragmentation? In my book, fragmentation refer to how homogenous a landscape is, whereas you seem to mostly discount for certain land-use types, irrespective of whether they occupy the gridcell in one block or fragmented into several small pieces.

P11 L8: Please clarify that this constraint is only needed because you do not simulate fuel loading explicitly?

P12 L4. Isn’t it trivial that 100% correspond to a fraction of 1?

P22 L20ff. Please reword (Figure 6 in this paper) compared to model X (Figure X in Author#1) etc. Also it should be added that this comparison is at at best of qualitative nature.

P22 L20 and 21. Check english: strong performance on? not “with respect to”?

P23L13: “This issue has been reported before?” Really? I thought HESFIRE was a new model...?

P24 L 16: No reference to conclusion needed here.

P26 L 13 remove notes on removed paragraphs

P26 L16: Please be more specific in that you briefly mention/respeat the main improvements of HESFIRE compared to other models, as well as the main limitations here

Figure 3: remove “(which happened...)”.

Figure 4: Please reformat Y-axis to match the parameters as they are listed in Table 1 (in particular with respect to subscripts)

Page 38: What is this text at the top of the page referring to?

Figure 5: What are the numbers at the top of each panel. I don’t think it’s necessary to state these values assuming they correspond to those of Table 1

Figure 7: Please provide Y axis for all plots. Please prepare this graphic such that it fits onto the journal pages (break into parts if needed).Please separate the map from the rest of the figure. I am not sure what “significance of the IAV correlation means” Please elaborate. Please clarify in the legend, which number belongs to GFED and HESFIRE

Many thanks for the revisions. However, I still have an issue with equation 3.

It is still not clear what LU actually is (never defined), and how you can integrate over it. My assumption is that LU represents the grid cell fraction of one particular land use type over which you integrate (which leads me to think that this should be a sum rather than an integral), though at some places in the manuscript it seems that you rather refer to the agricultural fraction of the grid cell. If that was the case, why is there an integral at all? Note that there is also some ambiguity in the units of this equation as LU_tot has the unit km2, whereas LU_thres is dimensionless?

In Table 1: Why is there a range for LU_thres, and not one value? How does this vary? Or is this a typo, implying that 0-0.1 is the optimisation range?