Review of “The Community Fire Behavior Model for coupled fire-atmosphere modeling: Implementation in the Unified Forecast System” by Pedro A. Jiménez y Munñoz, Maria Frediani, Masih Eghdami, Daniel Rosen, Michael Kavulich, and Timothy W. Juliano.

The manuscript presents a fire behavior model that can be coupled with the existing atmospheric models. The study presents results when the community fire behavior model is coupled with the Unified Forecast System for fire spread episode focused during Summer of 2020 in Colorado. The model is run in a standalone model, coupled mode and the results are compared against the WRF-Fire model predictions.

Overall, the manuscript is well written and the fire behavior model is well explained. Figures and model schematic depictions are clear. The study is valuable as the CFBM model presented could be run with a user-selected atmospheric model as long as the required variables are present. The manuscript is suitable for publication after revisions addressing the below comments.

Comments:

1. Expand or describe the wind and slope correction terms mentioned in Equation 1. Without the wind correction term formula, I am having difficulty understanding why the user would have to select a fire wind height, when it is much simpler to use the 10-m winds (which are usually available from many atmospheric model outputs) as the driving force for the fire perimeter. Also, is there an upper limit for the rate of spread in the model to address any unrealistic values or sudden spikes such as the one observed for fire heat/vapor fluxes and emissions shown in Figure 9?
2. Since the WRF model and the UFS model were initialized from HRRR, why is there such a bias in the wind speed predicted by UFS? Did the authors perform any sensitivity analysis to identify the source of this discrepancy? Can we expect UFS to underestimate the wind speed in general? Are there other studies in the literature that pointed this out? Choosing a different height simply to match the WRF wind speed may not fully address the complexities involved, especially when the height is user given. As the authors mentioned, the choice of fire wind height is one of the key input parameters in running the model. Further analysis could help the CFBM users to understand the uncertainties involved and make an educated choice of the fire wind height. If the intention is to simply use the WRF-Fire model as the ground truth or reference, use of different physics parameterizations in UFS compared to WRF would obviously yield different results. Also, as mentioned near Line 105; how will the model perform if (say) 10m winds are interpolated to 2.5m or 5m based on the logarithmic profile.
3. It is very surprising that the model heat flux value differs outside the fire perimeter, especially in the UFS runs (Figure 10). Even more surprising is the presence of negative values in the UFS heat flux differences. If the only difference among the coupled and uncoupled simulations is the feedback from the fire pixels to the atmosphere, why will there be any changes to the surface heat flux far away from the fire perimeter! For the purpose of comparison, in Figure 10, it would be better if the WRF variable also includes the total surface heat flux, as the uncoupled values are subtracted from the coupled, only the fire induced heat flux would remain. 
4. Interesting to see the mean wind speed unaffected even if the temperature increased in both WRF and UFS models as a result of 2-way coupling, which would change the vertical velocity in the model. It would be better to show the change in vertical velocity due to 2-way coupling and it could be used to justify the insignificant effect of 2-way coupling on the fire spread and mean winds shown.
5. It is hard to follow the discussion near Line 385, about the large fluctuations in fluxes and emissions in the UFS-2way run. Why would the WRF-Fire time step 18s be relevant to the fluctuations in UFS output? Also, I thought WRF used a fixed 12s timestep! Why do the large fluctuations in UFS-2way variables start 3 hours after the fire initialization?
6. Line 421: It’s hard to justify this line: “In any case, there is no evidence of a systematic bias in the rate of spread for this case.” From Figures 5 and 8, the fire area from UFS (1-way or 2-way) seems to be larger than the WRF simulated area. This is counterintuitive when one takes into account the mean wind speed differences. For a good portion of the simulated time, the mean wind in UFS is weaker than in the WRF and yet the fire area in UFS is larger than in WRF. Authors could add the timeseries of the difference between the fire area simulated by UFS and WRF for the -P1 and -P2 cases to justify if there is a systematic bias.
7. For the 2way runs, please add details about which layer the smoke emissions area added and that the smoke is being added as a passive tracer, i.e., it does not carry any thermodynamic or chemical properties. Even better would be to show a 3D visualization of the fire progression with smoke tracer and stream lines showing any updrafts over the fire.
8. Discuss any limitations of the CFBM. For example, can it be coupled with very high resolution models (grid size less than 100m or large-eddy scales)?
9. Line 294: Expand on or use a reference for the line “And third, we updated the VEGPARM.TBL to correct for a bug in the table.” What bug?
10. Use consistent references, in the discussion around Figure 5, at places the subplots were referred to as Figure 5a and later in the text, they were referred to as Figure 5 (right/left). 
11. Describe the red perimeter line in Figure 10 in the caption. It would be better if Figures 10, 11 and 12 are shown for the same times as in Figures 4 and 7.
12. As this is similar to and compared with WRF-Fire in the manuscript, it would be useful for the end user to know about any benefits CFBM would have over other existing models such as WRF-Fire in terms of computational requirements.

General Comments:

The paper presents the development and implementation of the framework that allows coupling between the surface fire model and the Unified Forecasting System (UFS). The proposed framework/coupler is presented as designed to enable coupling with various atmospheric models using the Earth System Modeling Framework (ESMF). The fire model itself is generally a copy of the WRF-Fire code packaged so that it can be run in standalone mode or coupled with atmospheric models like the Unified Forecast System (UFS). It includes features such as a dynamic fuel moisture content model, fire propagation based on the level set method driven by terrain and wind conditions, simplified fuel consumption, and emission modules that mimic the existing capability of the fire code in WRF.

The paper describes the model's structure, its coupling with UFS, and the results of simulations of the Cameron Peak Fire for cases initialized from a point ignition and fire perimeter.

At a fundamental level, the paper does not appear to meet the publication standards of GMD. The fire model presented is essentially a refactored version of an existing code, which replicates its shortcomings and issues without introducing substantial novelty. Additionally, the description of the fire model's structure lacks clarity, which would hinder community-based development. It is crucial to clearly separate the fundamental processes involved, such as the representation of the fire front and its propagation, fuel consumption, emissions, fuel moisture, and rate of spread. Unfortunately, these aspects are not adequately addressed in the paper, undermining its potential as a foundation for a community model.

The description of the modeling system is challenging to follow and often appears disorganized. A more structured explanation, beginning with a high-level overview and then detailing specific components, would greatly assist the reader in understanding the overall architecture of the proposed system, including the coupling strategies. Additionally, a fundamental diagram illustrating the relationships between key components such as UFS, ESMF, NUOPC, CFBM, and the data flow in UFS, CFB, UFS-1way, and UFS-2way would be beneficial. The current use of diagrams needs improvement, as they do not effectively clarify the presented developments.

Furthermore, the file system structure is incomplete; for example, it does not specify the location of configuration files. The coupling diagram also fails to indicate the data flow for the experiments presented. It is unclear how the one-way coupled simulations are performed and which components are used in their execution.

The language in the document requires some revisions. There are inconsistencies in the tense used throughout the writing, and many sentences could be rephrased for clarity. Additionally, shifting the focus from "we developed..." to "the module... has been developed..." would improve readability.

At the technical level, the paper faces several significant issues, including a flawed design in its numerical experiments, a lack of scientific rigor in validation and coupling strategies, and insufficient detail for replication. Critical elements, such as domain sizes and placement, are omitted. Additionally, the plots depicting fire simulations are missing scales and axis labels.

Although UFS and WRF are different models, the parameterizations used in UFS are essentially a subset of those available in WRF. Therefore, it is important to focus on creating a comparable setup. Key aspects of the simulation, such as the time step and the number of vertical levels, could have been adjusted in WRF to align more closely with UFS, which would help reduce inconsistencies between the model configurations.

More attention must be paid to the ignition strategy. The issue of integrating fire observations into coupled fire-atmosphere models has been studied, and there are existing solutions that enable both smooth ignition and selective fire activation see for example Kochansky et al. 2023 (https://doi.org/10.3389/ffgc.2023.1203578). The instantaneous ignition method used in this study is shown to negatively impact model stability and although widely used for uncoupled models is not optimal for coupled simulations. Another issue is the fuel moisture. It remains unclear how it was preconditioned and why the fuel moisture model implementation was not evaluated.

The issue of horizontal resolution is significant. A resolution of 3 km is inadequate for testing a coupled fire-atmosphere model. This level of resolution causes a dilution of the heat flux across an atmospheric box of that size, leading to under-resolved buoyancy and compromising the two-way coupling effect. Surprisingly, the authors did not reevaluate their choice of horizontal resolution after observing virtually no impact of fire feedback on the atmosphere. For proper validation of a two-way coupled model, it would be beneficial to use in-situ wind observations or airborne vertical velocity observations and plume top height data. Alternatively, model-to-model comparisons could be utilized, provided that the benchmark model is appropriately configured to resolve the processes involved in fire-atmosphere coupling. To effectively resolve fire-atmosphere interactions, coupled models like WRF-Fire are typically run at resolutions of hundreds of meters. For example, WRF-Fire in COFPS operates at approximately 111 m resolution. Consequently, a 3 km resolution is insufficient. If the UFS (Unified Forecast System) does not support a higher resolution, it becomes essential for model developers to create a coupler that can bridge the resolution gap necessary for a coupled fire-atmosphere model. Regarding observational data for model validation, utilizing datasets from experimental campaigns such as FireFlux, CALFIDE, or FIREX-AQ is recommended.

One of the most alarming practices highlighted in the paper is the decision to manipulate the reference height solely to ensure consistency in the simulations. There is no scientific justification for this adjustment. The authors should consider the potential impact this decision may have on future model results, particularly when the wind overestimation in the Unified Forecast System (UFS) is resolved, or in instances where the UFS model accurately captures wind patterns. In such cases, this adjustment could lead to unrealistically low rates of spread (ROS) for fires.

Additionally, the rationale for interpolating winds from arbitrary levels to the fire wind height is also unconvincing. The Rothermel model was developed based on laboratory data where the prescribed wind was situated close to the fireline. The experimental data relied on handheld devices deployed at the surface, which were relatively close to the fireline -certainly closer than the described model's horizontal resolution of 3 km. This coupling strategy is integral to WRF-Fire, and altering such a critical component of the model without proper scientific justification is unacceptable. Fire-atmosphere interactions lead to low-level jets, resulting from the inflow into the base of the pyroconvective column, driven by the heat of the fire. This phenomenon is documented, for example, in Benik et al. (2023) https://doi.org/10.3390/fire6090332. Raising the reference height above the layer influenced by the fire contradicts the core principle of coupled fire-atmosphere modeling.

Another significant shortcoming is the lack of quantitative assessment in the validation section. The paper contains many unsubstantiated and generic statements, such as “The results show consistency between CFBM, UFS, and WRF-Fire, indicating proper implementation and potential for further development.” However, the values of the heat fluxes in uncoupled simulations differ by as much as 100% (see Figure 9 at the end of the simulation). Additionally, the simulated fire perimeters are not qualitatively analyzed. It would be advisable to use the Sørensen coefficient for this analysis.

The authors assert that "The CFBM is expected to facilitate joint developments and improve the accuracy of wildland fire simulations, ultimately contributing to better fire management and mitigation strategies." However, the results indicate very poor model performance, and using the uncoupled model still requires WRF and WPS, as the processing tools for the fire data have not yet been developed. This dependency complicates the creation of a community model, given the significant effort needed to build the libraries and set up WRF to generate the WPS.

Regrettably, this paper cannot be accepted for publication in its current state. While the effort to develop a coupling framework between the fire model and the Unified Forecasting System (UFS) is commendable, the work suffers from significant flaws that fundamentally undermine its scientific validity and contribution to the field.

The lack of sufficient innovation in the fire modeling component, poorly designed coupling strategy and numerical experiments, inadequate validation, and poorly organized presentation all contribute to a body of work that does not meet the standards required for publication. Despite the potential value of a robust community fire behavior model and coupling framework, this paper does not provide the necessary advancements or rigor to support such an outcome.

This recommendation is not given lightly. Rejecting a paper is always a difficult decision, especially when the effort invested by the authors is evident. However, the issues identified are too substantial to be resolved through even major revisions and require a complete rethinking of the methodology, validation strategies, and presentation. It is hoped that with further refinement, the authors can address these challenges and eventually produce a stronger contribution to the field.

Specific Comments:

L4  “WRF-Fire is a state-of-the-art fire behavior model".  Other models like CAWFE or NesoNH-ForeFire offer a similar fictionality. It is unclear what specifically makes WRF-Fire a leading model in this field as suggested by the authors. There are other models built upon WRF that provide a higher level of coupling, including chemistry, as well as more advanced fire parameterizations.

L76, L82, 85  The absence of perimeter processing functionality and reliance on WPS are significant shortcomings of a model aspiring to become a community fire model.

L105 The current wind interpolation appears to be incorrect. The Rothermel model requires wind speed measurements at mid-flame height. To achieve this, the wind speed should either be interpolated to 20 feet (6.1 meters) and then adjusted using the fuel-specific reduction factors, or it should be directly interpolated to the mid-flame height based on those reduction factors.

L153 It is unclear where the 2% comes from. Emission factors available in the literature should be used instead. Additionally, the moisture content in the fuel should be accounted for when the moisture fluxes are computed.

L157 The reorganization of the code should be discussed in more detail. A diagram of the code could be used to support this argument and strengthen this point.

L163  I recommend using more formal language; instead of "facilitate understanding of what is being done," ,"improve code clarity by relating parts of the code to its physical properties and specific functions".

L172-175 it would be helpful to explain how this stand-alone code relates to the original stand-alone version of WRF-Fire .

L175 Change certain to required

L182-183 “testing sensitivities in model parameters and methods” consider rephrasing and providing more specific information. For example “testing model sensitivity with respect to XYZ”

L183-184 It would be beneficial to provide more details about how the model variables are coupled, the re-mapping strategies used, and so on. The fire code in the examples presented operates at a 100 m resolution, which leads to issues with the handling of fuel and slope data. How is it ensured that the fuel properties are preserved when integrating from a 30 m resolution to a 100 m resolution?

L185 “directories with the fire code”, change to directories including the fire code, then modules “used to define”

L189-190. Rephrase and clarify what “extended with WRF-Fire approaches” means

L190 Consider rephrasing to say modules supporting writing standard [...] and reading […].

L193 Consider rephrasing to say "module that supports reading of NetCDF files".

L194 Change to module defining…, initializing, updating etc.

L194 Instead of “There is also”, consider rephrasing to say "the last two modules X and Z are used to decompose the domain into tiles…and …"

L200 Consider changing to “when linked to the atmospheric model using two-way coupling”.

 L215. Are these three-dimensional winds – U,V,W or only horizontal winds U,V on the 3D grid? How are winds projected to the terrain plane?

L217 Capitalize nuopc

Figure 1 just says nuopc which isn't very informative a higher-level diagram showing modeling components should be used here. Details about the file organization should be presented later.

L220 The description of the EXMX is not sufficient. How does it work, is it similar to the WRF Registry mechanism? More explanation is needed here.

L224 How does "cap" defines the WRF-data grid? Is there a namelist/configuration file that defines variable names, grid location and such?

L226 Figure 1 lists ESMF as a part of the CFBM code, while according to the text says “The main difference is that this offline coupling requires ESMF whereas the standalone code described before does not ” This is confusing. Isn’t the standalone code using offline coupling? This part needs some clarification.

L227 Figure 2 does not effectively illustrate the coupling mechanisms. It is important to differentiate the paths and connections between the modules in offline, online, and standalone scenarios to clarify the data flow and coupling mechanism for each case. Additionally, the coupling between ESMF and ESMX should be explained more clearly.

L239 This line is confusing, did the authors mean that only testing doesn’t support MPI, but the rest of the code does support MPI?

Organizing Figured 2 and 3 as a two-panel plot to highlight the similarities and differences between the coupling strategies would be beneficial.

L243. How is the machine environment specified? What information is included there? How would users add other machines?

L253, Where is the fire namelist? Shouldn’t it be listed in Figure 1?

L258-259 Is there any specific reason why the original coupling strategy from WRF with the extinction depth was abandoned? How would the model be kept stable under rapid fire growth and an instantaneous initialization of the fire from perimeters?

L273-274. This sentence needs rephrasing. It is unclear how “WRF-Fire procedures” are related to the comparison between the results from the models.

L274. More justification is needed for selecting this specific fire for comparison. Given the poor performance of WRF-Fire in this fire event, this choice is confusing.

L278. Considering the wide selection of parameterizations in WRF it should be possible to reduce the setup differences to the dynamics core alone. Such a comparison would be much more meaningful.

L280 A wind-driven fire like the one described here is not the best test case for a coupled fire-atmosphere model. Under strong wind conditions, evaluating the role of fire-atmosphere coupling processes is challenging.

L293. This is a significant modification. Are there any observational data supporting the need for that? Do experimental data from experiments like FireFlux and FireFlux II suggest that the wind profile before and after changed significantly due to the roughness change?

How was the roughness change estimated? Was it modified in both models?

Correct the tense “we updated”

Consistency in the tense. The text is hard to follow and should be rephrased.

L302 Consider changing to “Consistency between simulations”

L305- The ignition procedure requires a more detailed description, especially regarding the coupled simulations. Sudden ignition can disrupt the interaction between fire and atmospheric processes, leading to poorer model results. Greater care must be taken to ensure that the integration of fire perimeters is conducted in a manner that keeps both the fire and atmospheric model components in sync at the start of the simulation.

L320 It is puzzling why the WRF time step and the number of levels have not been simply adjusted to match those in UFS. The study would benefit from a more careful experimental design.

L327. Why 20 minutes? This creates inconsistencies between the uncoupled WRF simulation and the new implementation. Are the wind variables averaged over the 20 minutes or are instantaneous values used?

L330- What is the scientific basis to believe that at 3km a coupled fire-atmosphere model could successfully resolve fire-atmosphere interactions? It is over an order of magnitude coarser mesh than COFPS, or other fire studies using coupled models.

L335 It must be explained what the purpose of running a coupled model at such low resolution is, and what exactly the authors expected to achieve by running simulations in such a configuration.

L338 It is unclear based on what the authors made this decision. Rothermel internally uses wind speed at the mid-flame height. The midflame winds are specified using reduction factors that are used to convert winds from 6.1 (20ft winds) to the midflame height. Such changes invalidate the simulations.

L350. That is not true. There is no quantitative analysis so this claim is subjective. Also, the fact that the authors had to adjust the fire wind height from 5 to 2.5 meters to get there is alarming.

L351. The term "underestimate" is not appropriate in this context. The simulation is completely unrealistic, likely due to a configuration error in the WRF-Fire model, which then affects the proposed model. Additionally, without proper validation of the basic weather variables, it is impossible to determine whether the issue lies with the fire or the weather aspects of the model. Overall, it is evident that both models have failed, but this does not provide any conclusive evidence regarding the implementation of the model.

L358 This kind of discrepancy should be minimized through careful planning of the model setup. Running a standalone Fire code with UFS forcing would better serve the purpose of this experiment.

The mean wind speed varies by as much as 1 m/s, which represents a significant difference of 20-25%.

On a “hot, dry, and windy day,” as mentioned in L281, why are the wind speeds only around 4 m/s, which is less than 9 mph? Previously, gusts were noted to reach 71 mph.

The difference between the CFB and UFS-1-way models is alarmingly high, especially considering their resolution.

L361 The "consistency" should be quantified. The plot suggests something opposite. There is no consistency – there is no systematic bias. The WRF winds are higher or lower than the CFB or UFS-1-way depending on the time.

L365-366. “As has been already mentioned, to obtain consistency in the simulated winds from WRF and UFS we had to use a different height for the winds driven the fire in the models”. This is unacceptable, especially because after the adjustment between 8/14/ 03 to 8/14/13 UFS showed lower values than WRF. Change "driven" to "driving"

Figure 5. The actual wind conditions at the same altitude should be presented. Additionally, validation against observed data is necessary. If the Unified Forecast System (UFS) is unable to realistically simulate winds and requires such corrections, its suitability as a platform for a community fire model is highly questionable.

Figure 6. Explain how the fire front is defined here, based on the heat flux threshold, fuel fraction, air temperature?

L372. The consistency hasn’t been quantified or analyzed yet. At that point, we only know that after creating an artificial adjustment to the wind height, the results are comparable. For example, we don’t know if the fuel moisture is consistent, if the fluxes at the fire mesh are comparable, whether fuel consumption is comparable, etc.

L374. The claim “Again, we see an underestimation of the observed burned area with consistency between the simulated perimeters using WRF and UFS” is not supported.

There is no quantitative analysis to support this claim. Interestingly, the backfire rate of spread (ROS) differs significantly between the models, with the WRF model showing a more rapid backfire propagation compared to the UFS. Since the backfire ROS is essentially derived from the no-fire, no-wind ROS from the Rothermel model, one would expect them to match. These discrepancies should not be affected by the modeled winds.

L375-376 “This is the first evidence of a relative small impact of the heat and moisture fluxes from the fire for these experiments that use 3 km grid spacing.” Unfortunately, this is also evidence that the modeling experiment wasn’t carefully planned and is unsuitable for validating coupled fire-atmosphere models.

It should also say here and in a few other places “relatively” not “relative”.

L376 Rephrase “In this direction, we further show the evolution of the burnt area and the wind speeds calculated with the 1way and 2way experiments using WRF and UFS (Fig. 8)”

L380. “The consistency is also evident in the time series of the fire heat and moisture fluxes and the smoke emissions (Fig. 9).” The plots actually show something different. The fluxes from on-way simulations differ by as much as 50% on 08/14-03, or 08/14-05. At the end of the simulation, the WRF-1way heat fluxes are twice as high as the ones from the UFS-1way run. In the two-way simulations, we see identical heat fluxes with UFS-1way and UFS-2way, which is surprising considering that on 08-14 12 the couped winds are visibly weaker than uncoupled. A careful explanation of these issues is needed.

If the time step was an issue, the data should be time aggregated to the same intervals. However, I’m not convinced that WRF heat fluxes are averaged in time. More explanation is needed about why the instability appears at the ignition time. See also the comment about the ignition procedure. At what rate is the ignition implemented? The ignition needs a more thorough explanation and more careful planning.

Figures 10 and 11. The header should be changed. It is unclear what “WRF-Fire x UFS-Fire” means. It suggests a multiplication between the two, which is misleading.

L386. If the timestep is to blame why didn’t the authors run WRF with the same time step? This is again the problem of inconsistency between the model configuration that could have been avoided if the numerical experiments were carefully planned.

L393. “The WRF differences are positive because WRF has a dedicated variable for the fire fluxes in the atmospheric grid and the fact that the fluxes are zero in the one-way experiment.” Actually, WRF also has an integrated variable GRNHFX that should be used here to keep it consistent with UFS since it provides only aggregated fluxes. Alternatively, a new diagnostic variable for fire heat fluxes could be added to UFS.

L395. This claim is not supported by the presented results. The UFS simulation, particularly in the bottom right of Figure 10, shows a hot plume downwind from the fire, which is absent in the WRF simulation. This discrepancy should be investigated further. Additionally, a comparison of the overall heat fluxes from WRF and UFS displayed as the difference (WRF - UFS) would provide more informative insights. The same applies to temperature comparisons.

What is the size of the domains used for testing? Why are the data shown on a coarse 21x17 mesh? Is it the native atmospheric mesh? 

L405 The change in nomenclature is confusing. The switch from UFS-1 (one-way coupled) and two-way coupled UFS and CFB makes the plots difficult to analyze. Additionally, the lack of axes and labels in Figures 7 and 10-13 is unacceptable, especially since a figure depicting the domain configuration is missing.

L406 The sentence, “This spread is simulated accurately in the northern and eastern parts of the fire, but results in a clear overestimation towards the northeastern portion of the perimeter,” needs rephrasing as it is self-contradictory. It suggests that the northeastern progression is both accurate and overestimated. The reality is that the progression does not appear to be particularly good. This is particularly concerning in light of findings from the ignition point analysis, which suggested a significant underestimation of fire growth, while now we see a notable overestimation. I recommend that the authors review their approach to representing fuel moisture. The fact that the model overpredicts the rate of spread (ROS) at 10:00 in the morning but performs considerably better at 16:30, when fuel moisture reaches its minimum, indicates that fluctuations in fuel moisture are not represented accurately. Proper conditioning and modeling of fuel moisture are critical for analyzing diurnal fire activity.

L407 The issue of representing fire activity based on infrared perimeters has been previously studied. A potential solution to this problem is discussed in: https://doi.org/10.3389/ffgc.2023.1203578. The authors should consider employing such a method to enhance the realism of their simulation.

L411 The differences between UFS and CFB are significant, and it is crucial to investigate why uncoupled simulations show faster progression. Generally, fire-induced winds tend to accelerate the fire, so we would expect the opposite effect. If the variations in the frequency of inter-model communication result in such significant differences, the proposed system cannot be considered robust.

L420 The simulated wind should be validated to provide more insight into these differences. Wind speed alone is not sufficient; wind direction is also critical. The authors should consider investigating this aspect as part of their analysis.

L421-423 The plots indicate the opposite: CFB systematically overestimates fire expansion compared to UFS, which also consistently overestimates fire growth along the active parts of the fire. The runs are only consistent regarding backfire propagation and propagation along the northern and northwestern flanks.

L439 Refer to comment L4.

L446  The work presented contradicts this statement, as it highlights significant inconsistencies between WRF and CFBM. There is underestimation when ignited from a point and overestimation when ignited from a perimeter.

L448 Unfortunately, this statement lacks substantiation. Fundamentally, it is impossible to evaluate coupled models using numerical experiments that are inadequate to resolve the processes that these models intend to represent.

L454 The fire community would greatly benefit from a more modular construction of fire models, and greater attention should be given to this aspect.