Fuel Drivers of Fire Behaviour in Coastal Mallee Shrublands

Round 1

Reviewer 1 Report

(line 68-70) Rephrase – Sentence read awkward, esp. the end part,

(line 87-90) Clarify - Text states that physical fire models were excluded from the review but does not give a specific reason that these types of models were not considered. Suggest adding sentence as to reason thereof.

(line 117-118) Question - How were the two essential fire model elements, fire threshold and rate of spread determined from a likely larger suite of elements of fire models?

(line 197-198) Likely typo – “…North American models studies did not produce a new model, but…”.

(line 229 - 236) Move text – Suggest moving the number and background of the reviewers up to Section 2.2, under the methods.

(Table 2) Clarify – starting Table 2 the results distinguish four fuel classes, specifically, canopy, elevated, near-surface and surface. It would be helpful to include a description of how these classes were defined for the purposes of the survey and associated analyses. As is it leaves the reader second guess, for example, what is the difference between canopy and elevated fuels?

(line 269-270) question – Clarify why not all metrics were presented the same number of times. I my opinion this weakens the comparison of the relative rankings of each metric and could skew this comparison to those metrics (and vice versa) presented more frequently. Not a “fatal flaw” in the methodology, but this aspect needs an explanation in the manuscript.

(line 309-310) reword – use of the word “excluded” in this sentence could be interpreted as canopy height and near surface height not being offered as a metric (see comment above). Since these terms were offered but not selected by panel members suggest reword (something like … were not considered most important…).

(line 327) reword – based on survey experts seem to quiet about the role of vegetation height and cover (just the model review support this statement).

(line 345 ) clarify  - This is the first time in the manuscript the term “bulk density” is used. Although used in the right context in the discussion I recommend introducing this term earlier in the manuscript to the reader.

(line 355) reword or clarify – while the manuscript deals entirely with shrub fuel modeling it seems that out of the blue models for forest fuels are brough into the discussion. This seems confusing.

(line 362) incomplete sentence

(lines 363-367) question - is there any data or reported analyses that discuss the reasonableness of using canopy cover as a proxy for more complex fuel metrics? If so, it might be good to add this to discussion.

(lines 378-381) concern – These sentences point back to my concern regarding not presenting a consistent list of metrics to expert panel members. In my opinion it is important how and why this choice was made in setting up the surveys and picture comparisons.

(line 382) typo – “not” to “no”

(line 412) typo – “remotes sensing” to “remote sensing”

(lines 412-414) expand text? – This is a key finding. One issue is that measuring the more complex fuel metrics (no matter how measured) would be compared and validated against fire models that do not consider these metrics at all, or not as the main drivers. As such, this comparison would be limited. A true improvement of the fire modeling for (mallee) shrublands can only come from developing new models or adjust the parametrization of existing ones. I think the discussion can emphasize this aspect better.

(line 424-427) update – text lists several metrics needed for better characterization of fire thresholds but leaves out the fuel connectivity metrics that were the focus of the expert members. It seems that the importance of the latter is a key takeaway from this analysis yet was left out of the conclusion.

Author Response

Reviewer 1:

Thank you for your comments. We have responded to each one below.

Comment:         (line 68-70) Rephrase – Sentence read awkward, esp. the end part,

Response: Sentence improved for clarity
Change: This discontinuity is believed to play a crucial role in driving fire behaviour in coastal mallee; however, it is unclear which continuity fuel metrics should be measured.

Comment:         (line 87-90) Clarify - Text states that physical fire models were excluded from the review but does not give a specific reason that these types of models were not considered. Suggest adding sentence as to reason thereof.
Response:         Added addition explanation around focus on empirical field studies over theoretical models
Change: Physical models rely on idealised fuel inputs and with the exception of a recent study of gorse shrublands [21] most do not replicate the actual fuel complex of shrublands [22], in particular coastal mallee. For this review, physical models were excluded as they attempt to model fire from the physical and chemical principles of combustion and fluid dynamics rather statistical relationships between weather, fuel and fire behaviour. While these physical models can offer insight into influences of fuel variability on fire behaviour from a theoretical perspective [21-23], these simulations still require evaluation against in situ field experiments [24]. Due to these constraints, this paper focused on empirical fire models for shrublands which have structure and climate analogous to coastal mallee.

Comment:       (line 117-118) Question - How were the two essential fire model elements, fire threshold and rate of spread determined from a likely larger suite of elements of fire models?

Response:        Added sentence with explanation

Change:           Rate of spread and fire sustainability are considered the most useful fire behaviour indicators for assessing conditions for prescribed burning and modelling wildfire in coastal mallee [49]

Comment:        (line 197-198) Likely typo – “…North American models studies did not produce a new model, but…”

Response:       Corrected, thanks

Change:           studies did not produce a new model but did

Comment:        (line 229 - 236) Move text – Suggest moving the number and background of the reviewers up to Section 2.2, under the methods.

Response:       Accepted suggestion
Change:           Moved to line 162

Comment: (line 197-198) Likely typo – “…North American models studies did not produce a new model, but…”.
Response: Corrected, thanks
Change: studies did not produce a new model but did

Comment: (line 229 - 236) Move text – Suggest moving the number and background of the reviewers up to Section 2.2, under the methods.
Response: Accepted suggestion
Change: Moved to line 162

Comment: (Table 2) Clarify – starting Table 2, the results distinguish four fuel classes, specifically, canopy, elevated, near-surface, and surface. It would be helpful to include a description of how these classes were defined for the purposes of the survey and associated analyses. As it is, it leaves the reader second-guessing, for example, what is the difference between canopy and elevated fuels?

Response: We define fuel strata using the Overall Fuel Hazard Guide for South Australia in the methods section. Have added a reference to this guide in the results section too to assist the reader.

Change: Definitions of fuel strata were defined using the Overall Fuel Hazard Guide for South Australia [1].

Comment: (line 269-270) question – Clarify why not all metrics were presented the same number of times. In my opinion, this weakens the comparison of the relative rankings of each metric and could skew this comparison to those metrics (and vice versa) presented more frequently. Not a “fatal flaw” in the methodology, but this aspect needs an explanation in the manuscript.

Response: The number of times each metric was present is not evenly distributed because there were more options for some metrics (e.g., different ways of measuring height) than others. We have clarified this in the text in the results section by distinguishing “metrics” from “categories of metrics”. The limitation of the design was only discovered after the workshops, and we do not believe this has an impact on our conclusions.

Change: Table 3 summarises the rankings of various fuel metrics by broad fuel attribute categories. The results are presented in order of their importance to fire behaviour (both rate of spread and fire sustainability) in coastal mallee shrublands and includes the number of times each metric was presented as an option to workshop participants (not all metric categories were presented the same number of times). Results show cover and connectivity were ranked as most important slightly more frequently than load and density. Table 3 summaries metrics based broadly on the fuel attributes being measured. As there were more potential metrics for some categories of fuel attributes, the number of times each type of fuel metric was presented to participants is not evenly distributed across all categories. Despite height metrics being given as an option more often than all other metrics, height was not the top-ranked metric for any of the responses.

Comment: (line 309-310) reword – use of the word “excluded” in this sentence could be interpreted as canopy height and near-surface height not being offered as a metric (see comment above). Since these terms were offered but not selected by panel members suggest reword (something like … were not considered most important…).

Response: This has been clarified

Change: Canopy height and near-surface height were not considered the most important metrics with respondents choosing between elevated height and gap between layers as being most important (Figure 6 (e)). Canopy and surface strata were considered least important to measure for estimating bulk density metrics, with elevated, near-surface, and combined total density being most frequently selected (Figure 6 (f)).

Comment: (line 327) reword – based on survey experts seem quiet about the role of vegetation height and cover (just the model review support this statement).

Response: Clarified and added references

Change: The results of the fire behaviour literature indicate that vegetation / fuel height is often used as a fuel structure metric in shrubland fire models, followed by cover. Vegetation height is correlated to fire spread thresholds and rate of spread but is noted as a proxy or surrogate for more complex fuel structure in both literature and comments provided by experts during the workshops [12,13,47].

Comment: (line 345) clarify - This is the first time in the manuscript the term “bulk density” is used. Although used in the right context in the discussion I recommend introducing this term earlier in the manuscript to the reader.

Response: Now defined in methods and results of literature

Change: The primary structure fuel input in Rothermel model is bulk density (fuel weight per volume), which is derived from fuel load and fuel bed height [27] line 261 - Although bulk density (weight of fine fuel per volume, e.g., kg/m^3) only features once in the list of model parameters [54], some studies provided an alternative model which used bulk density and discuss the importance of bulk density in shrublands [13,40,47].

Comment: (line 355) reword or clarify – while the manuscript deals entirely with shrub fuel modeling it seems that out of the blue models for forest fuels are brought into the discussion. This seems confusing.

Response: Comparison to forest fire models adds little to the discussion. Removed

Change: removed 353

Comment: (line 362) incomplete sentence
Response: Fixed, Thanks
Change: Future fire studies in shrublands should consider near-surface fuel continuity based on results of the expert workshop.

Comment: (lines 363-367) question - is there any data or reported analyses that discuss the reasonableness of using canopy cover as a proxy for more complex fuel metrics? If so, it might be good to add this to discussion.

Response: References discuss using canopy cover as a proxy for fuel continuity. Given the statistical correlation found, a certain level of reasonableness can be inferred. Expanding the discussion much more would require data and results which aren’t readily available.

Change: This apparent conflict can possibly be explained by the note by the authors that the canopy cover is a proxy for other more complex fuel metrics [12,47].

Comment: (lines 378-381) concern – These sentences point back to my concern regarding not presenting a consistent list of metrics to expert panel members. In my opinion, it is important how and why this choice was made in setting up the surveys and picture comparisons.

Response: See comments above

Change: This aligns with Cruz, Alexander, and Wakimoto [76] and Cruz, Alexander, and Fernandes [77] who use fuel strata gap as an input for crown fire threshold model albeit in pine forests. There is overlap of model authors and workshop participants, and caution is needed when interpreting these results due to potential double counting of metrics which were considered important in both the literature and workshops results. Although not presented to workshop participants, an alternative metric is the average height of the near-surface and elevated fuel strata weighted by the percent cover of the respective layer [78].

Comment: (line 382) typo – “not” to “no”
Response: Changed
Change: is no clear consensus

Comment: (line 412) typo – “remotes sensing” to “remote sensing”
Response: thanks
Change: This paragraphs has been updated from other comments

Comment: (lines 412-414) expand text? – This is a key finding. One issue is that measuring the more complex fuel metrics (no matter how measured) would be compared and validated against fire models that do not consider these metrics at all, or not as the main drivers. As such, this comparison would be limited. A true improvement of the fire modelling for (mallee) shrublands can only come from developing new models or adjust the parametrization of existing ones. I think the discussion can emphasize this aspect better.
Response: Thank you for this suggestion. We’ve added some discussion to clarify that future studies could derive existing metrics from remote sensing, but more importantly be used for novel metrics which may be better correlated to fire thresholds and spread rates.
Change: Recent advances in the development of remote sensing methods to derive fuel metrics show promise in providing systematic and reliable fuel metrics, particularly for canopy [85-87]. Terrestrial and mobile laser scanners (TLS and MLS) have been used to derive sub canopy fuel metrics which could be used for existing fire spread models [72,86,88-95]. TLS and MLS offer a significant advantage over airborne laser scanning (ALS) because they can differentiate between elevated and near-surface features without being hindered by canopy occlusion. Research and workshop findings highlight the importance of near-surface fuel to fire sustainability. The ability to derive complex fuel metrics such as vertical continuity and bulk density relatively easily and consistently warrants further study and inclusion into new fire spread models. As demonstrated in this study, these more complex fuel metrics are considered to have a key influence on fire behaviour in coastal mallee, but have been considered too difficult to collect reliably using traditional field-based techniques.
The inclusion of traditional fuel metrics (e.g., height and cover) and novel 3D remotely-sensed fuel metrics in shrubland fire experiments offers several opportunities. Measuring existing fuel metrics allows for comparison of new fire sustainability threshold and rate of spread data to be compared to existing fire behaviour models. Traditional field methods could be used for validation of remotely sensed metrics and to improve spatial explicitness of fuel mapping. 3D remote sensing can also be used to derive vegetation continuity metrics suggested by the results of this study [92,96]. To validate the results of expert workshops, future fire studies in shrublands need to not only compare results to existing fire models but develop new models or parameterisation using new metrics.

Comment: (line 424-427) update – text lists several metrics needed for better characterization of fire thresholds but leaves out the fuel connectivity metrics that were the focus of the expert members. It seems that the importance of the latter is a key takeaway from this analysis yet was left out of the conclusion.
Response: This has been clarified in the conclusions now.
Change: Few studies have considered metrics for vertical connectivity of shrubland fuels, which may be an important determinant of crown fires and therefore rapid increases in rate of fire spread and intensity. The results of expert workshops suggest horizontal and vertical connectivity plays a critical role in fire spread thresholds for shrublands, particularly in prescribed burn conditions. Based on the finding of the literature review and workshop, we recommend that vertical connectivity or measurement of gap between fuel strata be included alongside height, and cover metrics in future shrubland fire research.

Reviewer 2 Report

The article makes a clear case for considering coastal mallee shrublands fire modelling parameters separately to those available for sparser semi-arid mallee shrublands.

The authors explore “what are the fuel metrics driving fire spread in coastal mallee shrublands?” and recommend that field work is required to develop measures of horizontal discontinuity of fuels and vertical discontinuity relating to cover and fuel connectivity that can be consistently applied and empirically tested in experimental prescribed fires. Vegetation height is often used as an indicator of bulk density, cover and connectivity. This assumption can also be tested.

The paper analyses the problem by surveying appropriate empirical fire spread models from analogous Mediterranean vegetation types followed by expert workshops.

The idea to derive metrics on fuel cover and continuity with 3D remote sensing is a good one– say something about the use of LiDAR data to collect such metrics and give some reference of research that has collected similar tree or crop volume metrics from LiDAR data. I think this idea can be introduced in the discussion and then briefly referred to in the conclusion, rather than introducing it for the first time in the conclusion.

Could the authors please complete “Data Availability Statement”?

It might be useful to refer to Moreira et al. 2020 for more discussion on reasons for reticence to implement prescribed fires, in addition to insufficient ability to model or predict fire behaviour.

Francisco Moreira et al. 2020. Wildfire management in Mediterranean-type regions: paradigm change needed. Environ. Res. Lett. 15 011001. DOI 10.1088/1748-9326/ab541e

Author Response

Reviewer 2:

Comment: The article makes a clear case for considering coastal mallee shrublands fire modelling parameters separately to those available for sparser semi-arid mallee shrublands.

The authors explore “what are the fuel metrics driving fire spread in coastal mallee shrublands?” and recommend that field work is required to develop measures of horizontal discontinuity of fuels and vertical discontinuity relating to cover and fuel connectivity that can be consistently applied and empirically tested in experimental prescribed fires. Vegetation height is often used as an indicator of bulk density, cover and connectivity. This assumption can also be tested.

The paper analyses the problem by surveying appropriate empirical fire spread models from analogous Mediterranean vegetation types followed by expert workshops.

The idea to derive metrics on fuel cover and continuity with 3D remote sensing is a good one– say something about the use of LiDAR data to collect such metrics and give some reference of research that has collected similar tree or crop volume metrics from LiDAR data. I think this idea can be introduced in the discussion and then briefly referred to in the conclusion, rather than introducing it for the first time in the conclusion.

Response: Thank you for your positive comments and suggestion of more discussion about Lidar. We have added some addition discussion as suggested.

Change: Recent advances in development of remote sensing methods to derive fuel metrics show promise in providing systematic and reliable fuel metric, particularly for canopy [85-87]. Terrestrial and mobile laser scanners (TLS and MLS) have been used to derive sub canopy fuel metrics which could be used for existing fire spread models [72,86,88-95]. TLS and MLS offer a significant advantage over airborne laser scanning (ALS) because they can differentiate between elevated and near-surface features without being hindered by canopy occlusion. Research and workshop findings highlight the importance of near-surface fuel to fire sustainability. The ability to derive complex fuel metric such as vertical continuity and bulk density relatively easily and consistently warrants further study and inclusion into new fire spread models. As demonstrated in this study, these more complex fuel metrics are considered to have a key influence on fire behaviour in coastal mallee, but have been considered too difficult to collect reliably using traditional field-based techniques. 

The inclusion of traditional fuel metrics (e.g. height and cover) and novel 3D re-motely-sensed fuel metrics in shrubland fire experiments offers several opportunities. Measuring existing fuel metrics allows for comparison of new fire sustainability threshold and rate of spread data to be compared to existing fire behaviour models. Traditional field methods could be used for validation of remotely sensed metrics and to improve spatial explicitness of fuel mapping. 3D remote sensing can also be used to derive vegetation continuity metrics suggested by the results of this study [92,96]. To validate the results of expert workshops, future fire studies in shrublands need to not only compare results to existing fire models, but develop new models or parameterisation using new metrics. 

Comment:   Could the authors please complete “Data Availability Statement”?
Response: Yes, apologies for that oversight

Comment:   It might be useful to refer to Moreira et al. 2020 for more discussion on reasons for reticence to implement prescribed fires, in addition to insufficient ability to model or predict fire behaviour.
Response: Thank you for that useful suggestion. Added a section in introduction about mitigation risk through prescribed burning in Mediterranean landscapes, and the complexities.
Change: In fire-prone Mediterranean landscapes, undertaking prescribed burns adjacent to the values being protected is desirable to mitigate risk of socio-economic losses from unplanned wildfire [15]. However, this brings increased risk of negative impacts should a planned fire exceed the boundaries.

Reviewer 3 Report

Comments to the manuscript “Fuel drivers of fire behaviour in coastal mallee shrublands”

Overall comments

The authors investigate / discuss which fuel metrics are the most important to include in models aiming to predict fire spread in Australian coastal mallee shrublands. Research done on semi-arid mallee resulted in models of fire sustainability and rate of spread for this nature type. However, prescribed burning experiences in coastal mallee reveal that the models developed for semi-arid mallee are not applicable in areas covered by coastal mallee.

To derive the most important fuel parameters for fire spread in coastal mallee the authors examined which metrics were utilised in models developed for shrublands in other areas (e.g., Mediterranean), and arranged a series of workshops with shrubland fire experts to inquire their opinions.

The inclusion criteria for the review process were: 1) shrubland vegetation, 2) empirical modelling of fire sustainability / rate of spread, 3) measurement of fuel attributes. Table 1 lists the 17 models which met the criteria. Thereby, a summary of fuel metrics, based on the models was compiled. The results indicate that Height was the most frequently used metrics, followed by Cover and Load. Some studies discussed that Height is easier to assess than other metrics, and (correctly or not) may be used as a proxy for other parameters, which influence fire behaviour more. Vertical continuity and horizontal continuity have received less attention than their real role in fire spread represents.

Additionally, eight shrubland fire experts with shrubland fire experience participated in workshops where they were asked to compare the importance of different fuel metrics relevant for coastal mallee shrublands, under different assumed weather conditions.

However, NO PHOTOS OR FIGURES ARE INCLUDED IN THE MANUSCRIPT which I received for review, which makes it impossible to follow the workshop process. I have sent two e-Mails to MDPI, on February 9^th and 13^th, without getting any answer. MAKE SURE TO INCLUDE ALL PHOTOS AND FIGURES IN THE SAME FILE AS THE TEXT, SUCH AS THEY SHALL APPEAR WHEN THE ARTICLE IS PUBLISHED.

The research design and questions as listed in Appendix A sound reasonable. To partly compensate for the lack of figures, I have tried to find photos on the Internet and the referenced literature. For example, Cruz et al. (2013), provide in Figure 1 of their article an idealized profile of mallee-heath fuel complex.

Experts express that connectivity metrics are more important than height, especially for prescribed burning.  

I experience it like the article does not distinguish between wildfire modelling and prescribed burning modelling. In wildfires the tree crowns get often involved in the fire. This makes height a relevant metrics, as it also usually scales with other dimensions of the tree. A taller tree is usually also thicker than a lower one, thus containing more biomass. For the Rate of spread in wildfires when tree-crowns are involved, Height may be an important metrics.

In prescribed burning we don`t want to have crown fires. Near surface vegetation continuity is very important for fire spread (and vertical continuity in the unwanted event that fire crowns). In semi-arid malle, near surface continuity is low, while in the coastal mallee, due to more precipitation, low vegetation cover is more continuous. However, these shrubs are not multi-stemmed eucalyptus, but other species. Suggestions for future research are missing in the manuscript. Perhaps it is not too tedious during future research to determine the cover percentage and height of understory shrubs, as well as their vertical continuity to the eucalyptus, in selected plots (flat ones, as well as different slopes and aspects).

Detailed comments

Line 31, 32: The authors suggest that practitioners are reluctant to perform prescribed burning in areas covered by coastal mallee, among others, because appropriate fire behaviour models for this fuel type are not developed.

I question the postulation that the lack of models discourages prescribed burning. In line 30, it is mentioned that some burns have already been undertaken. Those sessions have provided experiences, so that practitioners, to some extent, know what to expect. However, if reliable models get developed, both Prescribed Burning practitioners and the Fire Service will get increased confidence to operate in the area, during controlled operations or in case of wildfires.

Line 38: Sparse shrubby understory

Line 40: Denser shrubby understory which provides increased fuel continuity for fire spread.

Line 41: Figure 1, (is intended to) show archetypes of open semi-arid mallee and coastal mallee. However, NO FIGURES were included in the manuscript version which I received.

Line 119 – 120: “The results of this compilation are reported separately for fire sustainability models and rate of spread models.” Where are these results?

Line 197 - 198: It should be noted that the North American study did not produce new model, but … (add not)

Line 209 – 210: “Cover and load metrics were each used twice in shrubland fire spread sustainability models.” I have two comments here. First, clarify whether you mean fire sustainability models or rate of spread models. “Fire spread sustainability models” is confusing. Second, in the far-right column of Table 1, Fuel metrics used in fire sustainability models, most indications are Non-Applicable, and only two places we see Cover. Load does not appear.

Line 212: Table 1, please explain (n = 16)*

In the Table the sign * is used two places, as FHS*, which is *Fuel Hazard Score, as noted under Table 1. I cannot relate (n = 16)* to data on Table 1.

Reference [12] is mentioned twice, line 4 and 5 in Table 1.

[59], Pepin and Wooton, use Bulk Density. However, I cannot see in their article that they define it in a comprehensive way. It would be fine if you explain in your article How this metrics is derived.

Line 221 - 223: Although bulk density features only once in the list of model parameters, some studies (WHICH ONES?) provided an alternative model which used bulk density and discuss the importance of bulk density in shrublands. (. appears on line 224)

Lines 305 – 308: Vertical continuity was chosen as having the most influence on fire spread /and crown fire thresholds, compared to other vertical-type measures. This (WHAT?) was followed by the vertical gap distances, noting when the responses of two gap-based metrics are combined they are equal in frequency to the percent vertical continuity metrics. PLEASE EXPLAIN WHAT THAT MEANS.

Line 314: Table 5, indicates 5 questions, but Section 4 presents 4 questions. The questions must be included in the text, before or in Table 5.

Line 352: What is an “eaten out grassland”?

Line 354: Distinguish between fire sustainability and rate of spread.

Lines 356 – 366: The expert group you consulted, held near-surface fuel continuity being the most important predictor of fire spread. Which is true in an early phase, and during Prescribed Burning. Canopy height and cover become important spread drivers when fire has crowned.

Line 360: which used overstory cover

Line 360 – 362: Incomplete sentence

Line 392: Bulk density can be calculated from height, cover and load information. Perhaps you can explain how it is calculated and what it means to fire development if bulk density is low, or high.

Line 424: Remove “models”

Author Response

Reviewer 3:

Overall comments

The authors investigate / discuss which fuel metrics are the most important to include in models aiming to predict fire spread in Australian coastal mallee shrublands. Research done on semi-arid mallee resulted in models of fire sustainability and rate of spread for this nature type. However, prescribed burning experiences in coastal mallee reveal that the models developed for semi-arid mallee are not applicable in areas covered by coastal mallee.

To derive the most important fuel parameters for fire spread in coastal mallee the authors examined which metrics were utilised in models developed for shrublands in other areas (e.g., Mediterranean), and arranged a series of workshops with shrubland fire experts to inquire their opinions.

The inclusion criteria for the review process were: 1) shrubland vegetation, 2) empirical modelling of fire sustainability / rate of spread, 3) measurement of fuel attributes. Table 1 lists the 17 models which met the criteria. Thereby, a summary of fuel metrics, based on the models was compiled. The results indicate that Height was the most frequently used metrics, followed by Cover and Load. Some studies discussed that Height is easier to assess than other metrics, and (correctly or not) may be used as a proxy for other parameters, which influence fire behaviour more. Vertical continuity and horizontal continuity have received less attention than their real role in fire spread represents.

Additionally, eight shrubland fire experts with shrubland fire experience participated in workshops where they were asked to compare the importance of different fuel metrics relevant for coastal mallee shrublands, under different assumed weather conditions.

However, NO PHOTOS OR FIGURES ARE INCLUDED IN THE MANUSCRIPT which I received for review, which makes it impossible to follow the workshop process. I have sent two e-Mails to MDPI, on February 9^th and 13^th, without getting any answer. MAKE SURE TO INCLUDE ALL PHOTOS AND FIGURES IN THE SAME FILE AS THE TEXT, SUCH AS THEY SHALL APPEAR WHEN THE ARTICLE IS PUBLISHED.

Response: Thanks for the comments. It is very unfortunate that the version of the manuscript that you had did not contain the figures and photos. The other reviewers did not find this issue, so we are not sure what happened to your copy. Figures were embedded within the manuscript. Appendix B was also included in the manuscript and contained all the photo pairs used in the workshops, which represent typical coastal mallee vegetation. We think having access to these images would address many of your concerns. Nonetheless, comments are addressed below.

Comment: The research design and questions as listed in Appendix A sound reasonable. To partly compensate for the lack of figures, I have tried to find photos on the Internet and the referenced literature. For example, Cruz et al. (2013), provide in Figure 1 of their article an idealized profile of mallee-heath fuel complex.
Response: As above, the manuscript included figures and photos which would address these concerns. We are unsure why the version you had did not included them. No Change

Comment Experts express that connectivity metrics are more important than height, especially for prescribed burning.  
I experience it like the article does not distinguish between wildfire modelling and prescribed burning modelling. In wildfires the tree crowns get often involved in the fire. This makes height a relevant metrics, as it also usually scales with other dimensions of the tree. A taller tree is usually also thicker than a lower one, thus containing more biomass. For the Rate of spread in wildfires when tree-crowns are involved, Height may be an important metrics.
In prescribed burning we don`t want to have crown fires. Near surface vegetation continuity is very important for fire spread (and vertical continuity in the unwanted event that fire crowns). In semi-arid mallee, near surface continuity is low, while in the coastal mallee, due to more precipitation, low vegetation cover is more continuous. However, these shrubs are not multi-stemmed eucalyptus, but other species. Suggestions for future research are missing in the manuscript. Perhaps it is not too tedious during future research to determine the cover percentage and height of understory shrubs, as well as their vertical continuity to the eucalyptus, in selected plots (flat ones, as well as different slopes and aspects).
Response: Although crown fires are not desirable in certain vegetation types we have found (as part of planned burns we have planned and carried out in our agency role) it is not normally possible to conduct burning in mallee, including the coastal mallee described in this study, without a crown fire. On most occasions a crown fire is desirable in order to regenerate coastal mallee ecosystems, see Taylor [1]. Mechanical treatment is sometimes used to ensure complete combustion of fuels and regeneration. Added discussion about including near surface in future studies
1.      Taylor, D.A. Prescribed burning to increase the richness of long-unburned and fragmented mallee communities. 2019.
Change: Line 406 - The results of the expert user workshop found that surface (or litter) and near-surface cover and continuity metrics were the most important fuel structure metrics for fire spread in shrublands. Despite this strong recommendation by experts, only one of the empirical fire models reviewed used near-surface fuel as a predictor of fire spread [13] and one which notes that near-surface fuel continuity is “key to accurately predicting fire sustainability” but uses overstorey cover as a proxy for fuel continuity [12]. Future fire studies in shrublands should consider near-surface fuel continuity based on results of the expert workshop.

Detailed comments

Comment: Line 31, 32: The authors suggest that practitioners are reluctant to perform prescribed burning in areas covered by coastal mallee, among others, because appropriate fire behaviour models for this fuel type are not developed.

I question the postulation that the lack of models discourages prescribed burning. In line 30, it is mentioned that some burns have already been undertaken. Those sessions have provided experiences, so that practitioners, to some extent, know what to expect. However, if reliable models get developed, both Prescribed Burning practitioners and the Fire Service will get increased confidence to operate in the area, during controlled operations or in case of wildfires.
Response: We have added some additional clarification around this point. Here we were not suggesting a reluctance to burn, rather  difficulty in implementing burns due to the unavailability of models to estimate the burn characteristics. As part of a team that regularly plans and carries out such burns in coastal mallee the point was an observation made based on first-hand experience on behalf of the team. While ourselves and other practitioners have been able to conduct some burns, if there is not a robust model to codify the successes (and failures) knowledge is not necessarily gained, documented and passed on to future practitioners. This is the experience of the authors and communicated to us by practitioners (referenced as pers comms lines 68-71). Added “partially owing” to clarify this is only one limiting factor.
Change: limited burns have been undertaken in coastal mallee shrublands outside of large wildfire events partially owing to the lack of a suitable fire behaviour model for this fuel type.

Comment: Line 38: Sparse shrubby understory
Response: No Change

Comment: Line 40: Denser shrubby understory which provides increased fuel continuity for fire spread.
Response: No Change

Comment: Line 41: Figure 1, (is intended to) show archetypes of open semi-arid mallee and coastal mallee. However, NO FIGURES were included in the manuscript version which I received.

Response: Hopefully the revised version of the manuscript has the images. No Change

Comment: Line 119 – 120: “The results of this compilation are reported separately for fire sustainability models and rate of spread models.” Where are these results?
Response: Table 1 in the results section
Change: Line 137 (Table 1)

Comment: Line 197 - 198: It should be noted that the North American study did not produce new model, but … (add not)|
Response: Fixed, thanks
Change: the North American studies did not produce a new model but did validate

Comment: Line 209 – 210: “Cover and load metrics were each used twice in shrubland fire spread sustainability models.” I have two comments here. First, clarify whether you mean fire sustainability models or rate of spread models. “Fire spread sustainability models” is confusing. Second, in the far-right column of Table 1, Fuel metrics used in fire sustainability models, most indications are Non-Applicable, and only two places we see Cover. Load does not appear.
Response: Fire spread sustainability is the term used in the literature to describe whether a fire will sustain propagation or fail to propagate and therefore be extinguished without intervention. Although this term is defined within the reference literature, we have added a definition and reference. Refences to load were removed, thanks for picking that up.
Change: Line 249 Cover metrics were used twice in shrubland fire spread sustainability models.
Line 123 . Throughout the remainder of this paper we use the term fire sustainability model to encompass fire spread probability and crown fire threshold models.

Comment: Line 212: Table 1, please explain (n = 16)*
Response: n = 16 was the number of models fitting the criteria. Have moved the (n=16) to make it clear. Also it should have been 17!
Change: Summary of the fuel metrics used for empirical shrubland fire models (n = 17) as reported in the key literature.

Comment: In the Table the sign * is used two places, as FHS*, which is *Fuel Hazard Score, as noted under Table 1. I cannot relate (n = 16)* to data on Table 1.
Response: An additional study has been added to bring to 17, and n=17 has been relocated to clarify that it refers to the number of studies which met the criteria. Astrix was typo and has been removed
Change: see above

Comment: Reference [12] is mentioned twice, line 4 and 5 in Table 1.
Response: Yes, this paper produced sperate models for different vegetation types (mallee and heath), clarified in the text. Refence is now 13
Change: ** Two separate vegetation types were included in a single study

Comment: [59], Pepin and Wooton, use Bulk Density. However, I cannot see in their article that they define it in a comprehensive way. It would be fine if you explain in your article How this metrics is derived.
 Response: Added a brief definition of bulk density
Change: line 103: l fuel types [23-26]. The primary structure fuel input in Rothermel model is bulk density (fuel weight per volume), which is derived from fuel load and fuel bed height [27].  
Line 261: Although bulk density (weight of fine fuel per volume, e.g. kg/m3)

Comment: Line 221 - 223: Although bulk density features only once in the list of model parameters, some studies (WHICH ONES?) provided an alternative model which used bulk density and discuss the importance of bulk density in shrublands. (. appears on line 224)
Response: Thanks, references were missing, fixed
Change: Although bulk density (weight of fine fuel per volume, e.g. kg/m3) only features once in the list of model parameters [54], some studies provided an alternative model which used bulk density and discuss the importance of bulk density in shrublands [13,40,47].

Comment: Lines 305 – 308: Vertical continuity was chosen as having the most influence on fire spread /and crown fire thresholds, compared to other vertical-type measures. This (WHAT?) was followed by the vertical gap distances, noting when the responses of two gap-based metrics are combined they are equal in frequency to the percent vertical continuity metrics. PLEASE EXPLAIN WHAT THAT MEANS.
Response: Clarified in text
Change: Vertical continuity was chosen as having the most influence on fire spread (and crown fire) thresholds compared to other vertical-type metrics (Figure 6 (d)).  Vertical gap distances were ranked next most important of the vertical structure metrics, however if the two gap metrics (near-surface to elevated gap and elevated to canopy gap) are combined they are equal in importance to vertical connectivity

Comment: Line 314: Table 5, indicates 5 questions, but Section 4 presents 4 questions. The questions must be included in the text, before or in Table 5.
Response: A question was missing from Appendix A, replaced Q1-5 with text to help table stand alone without need for Appendix.
Change: Updated Appendix A and Table 5

Comment: Line 352: What is an “eaten out grassland”?
Response: This is defined in the reference, but have change to sparse (“eaten-out) for clarity. Agree it is an unusual term, but it is what is used in the source reference (Cheney et al 1998).
Change: This is logically true of other fuel types including sparse (“eaten-out”) grasslands [74] and forest fuels [75], but is more pronounced in shrublands.

Comment: Line 354: Distinguish between fire sustainability and rate of spread.
Response: This line was removed based on other reviewers feedback

Comment: Lines 356 – 366: The expert group you consulted, held near-surface fuel continuity being the most important predictor of fire spread. Which is true in an early phase, and during Prescribed Burning. Canopy height and cover become important spread drivers when fire has crowned.
Response: The semi arid mallee models suggest that these models also have an effect on surface fire sustainability and spread. This section has been reworked to clarify what other literature suggests.
Change: The results of the expert user workshop found that surface (or litter) and near-surface cover and continuity metrics were the most important fuel structure metrics for fire spread in shrublands. Despite this strong recommendation by experts, only one of the empirical fire models reviewed used near-surface fuel as a predictor of fire spread [13] and one which notes that near-surface fuel continuity is “key to accurately predicting fire sustainability” but uses overstorey cover as fuel parameter [12]. Future fire studies in shrublands should consider near-surface fuel continuity based on results of the expert workshop.
Canopy metrics are ranked low by experts during workshop, and this contrasts with the importance of canopy height and cover as used in models such as the semi-arid mallee and generic shrubland models. Canopy cover has a negative effect on fire be-haviour possibly due to the effect on wind speed in subcanopy fires and influence on litter, and lower shrub fuels [12,13]. This apparent conflict of importance of canopy metrics between workshop results and literature can possibly be explained by the note by the authors that the canopy cover is a surrogate for other more complex fuel metrics [12,13,47].

Comment: Line 360: which used overstory cover
Response: Reworded sentence

Change: See above

Comment: Line 360 – 362: Incomplete sentence
Response: Reworded sentence
Change: See above

Comment: Line 392: Bulk density can be calculated from height, cover and load information. Perhaps you can explain how it is calculated and what it means to fire development if bulk density is low, or high.
Response: Added some discussion about bulk density correlation to fire behaviour and references to alternative models in the literature..
Change: Bulk density can be calculated from height, cover and load information and is a logical way of combining these three metrics into one metric. Increasing bulk density has been found to decrease rate of spread and is used in alternative models for several studies [40,47,54,79], however given the correlation to fuel load on cover it is can increase the likelihood of sustained fire spread and crown fire. Bulk density is also the primary fuel structure variable used in the Rothermel model [22] and is usually derived from fuel loading (weight per unit area) and fuel bed height [27].

Comment: Line 424: Remove “models”
Response: Thanks, typo fixed
Change: of those models developed for Mallee systems[12-14], so collecting

Reviewer 4 Report

Brief Summary

This paper reviews a selection of existing empirical shrubland fire behaviour models to understand which fuel metrics are commonly used in existing empirical models. This is compared with interview feedback from 8 practitioners and researchers, who, via a structured workshop process, gave their opinions on the importance of various fuel metrics on fire behaviour and the appropriateness of some existing fuel assessment methods. Fuel height was identified as one of the most frequently used fuel metrics in the chosen existing empirical models although it is noted that this may also reflect its ease of measurement and it’s usefulness as a proxy for other fuel structure parameters. The expert panel, regularly highlighted the perceived importance of near-surface and surface strata despite a greater focus on canopy metrics in many of the reviewed models. The authors overall recommendation is that vertical continuity or fuel strata gap size should be considered in future research efforts alongside other fuel metrics.

Major Comments

Overall: There seems to be some cross-over between workshop participants and the authors of the reviewed empirical models. This may require consideration when attempting to compare expert opinions with the common approaches in existing models, and therefore should be highlighted in the main manuscript. There are a couple of specific places where I believe this is worth highlighting further and these are listed under minor comments.

Methods: For the literature review, can you discuss in this section, any limitations of this approach given that this is considering studies based on statistical analysis in which any physical mechanisms responsible for any fuel driver are not explored/identified. Especially where multiple attributes are included in the model. Perhaps this can be addressed by commenting/summarising further the actual performance of the reviewed models and providing a summary of the key stats info e.g. R² value?

In the methods, could you also provide some additional discussion of how the photos shown in workshops were chosen? Including a description of the range of variability in the vegetation structure between photos and how this relates to any existing quantifications of typical ranges across these natural environments (if this exists).

Minor Comments

Line 37: Typo – ‘Mallee shrublands [are]…’

Lines 39-41: ‘Coastal mallee shrublands share the short mallee form eucalypt overstorey as 39 open semi-arid mallee shrublands but have higher canopy cover and denser shrubby 40 understorey which provides increased fuel continuity for fire spread.’

Can a reference be cited to support this, as was done for the preceding statements.

Figure 1: Can you include location and/or co-ordinates? Are these pairs of photos from one site each or two? Was the connectivity quantified at all in these areas?

Line 66: Is further development of physics-based models required to help support the limitations inherent in empirical model approaches?

Lines 135-136: Can you clarify here, did all 8 participants participate in all 3 workshops? Or was this 8 over the course of the workshops?

Lines 217-219: Might be worth mentioning the strong influence of height in the Rothemel model..

Line 222: Can you add a reference(s) for the studies mentioned?

Lines 267-270: When discussing 'importance to fire behaviour' in this specific section, is this in relation to sustainability or spread rate (or something else) or a combination of both?

Table 5: Can you explain why it was decided that the litter density and arrangement should be combined here (in Question 1)? These are potentially quite different factors for example, if the fuel is very sparsely arranged.

Table 5: In Question 1, Is it worth discussing that some of these metrics are essentially a combination of each other? i.e. the litter depth is determined by the litter density and fuel load for a given area?

Table 5: For Question 5, could you clarify what is meant by visual fuel hazard method and data hazard method? These categorisations are for me a little unclear.

Lines 327-329: As per earlier general comment, I think it should perhaps be  acknowledged here that the workshop attendees included authors of some of the reviewed literature.

Lines 360-362: Not sure if there is something missing here but maybe revise for clarity.

Lines 376-378: As per earlier comments, I think here you probably need to highlight that some of these model authors were also workshop participants. This is particularly important in a situation like this where you are making the point that there is alignment between workshop participant

Line 416: ‘I think this is the first time fuel age is mentioned. Is it worth discussing this in more detail earlier e.g. in the intro? Was age incorporated into the workshops discussion e.g. in selection of site images etc. If so could this be discussed as well?’

Lines 420-422: ‘Maybe the conclusion should be more like ' are considered to have a key influence on fire behaviour' as this study is focused on this expert opinion rather than determining/quantifying the fuel effect.’

Lines 430-432: As above, should this be 'are considered to play..'?

Author Response

Reviewer 4:

This paper reviews a selection of existing empirical shrubland fire behaviour models to understand which fuel metrics are commonly used in existing empirical models. This is compared with interview feedback from 8 practitioners and researchers, who, via a structured workshop process, gave their opinions on the importance of various fuel metrics on fire behaviour and the appropriateness of some existing fuel assessment methods. Fuel height was identified as one of the most frequently used fuel metrics in the chosen existing empirical models although it is noted that this may also reflect its ease of measurement and it’s usefulness as a proxy for other fuel structure parameters. The expert panel, regularly highlighted the perceived importance of near-surface and surface strata despite a greater focus on canopy metrics in many of the reviewed models. The authors overall recommendation is that vertical continuity or fuel strata gap size should be considered in future research efforts alongside other fuel metrics.

Response: Thank you for the comments. In particular the comment about the overlap between workshop participants and model developers. While we don’t think it changes the outcomes and conclusions of this paper, it was an oversight not to discuss it. We have attempted to address this. Other comments below.

Major Comments

Comment: Overall: There seems to be some cross-over between workshop participants and the authors of the reviewed empirical models. This may require consideration when attempting to compare expert opinions with the common approaches in existing models, and therefore should be highlighted in the main manuscript. There are a couple of specific places where I believe this is worth highlighting further and these are listed under minor comments.
Response: Clarification of the overlap between workshop participant and model authors has been included in the methods and discussion.
Changes:
line 169 - Three of the four researchers published models which were included in the literature review which may have influenced some of the results of the workshops (bias towards existing methods) or result in double counting of metrics in literature and workshops. While the workshop was design to minimise the influence of anchoring or confirmation biases, they cannot be completely ruled out. We do not believe these significantly effected the outcome of the workshops.  .
Line 383 - Workshop results indicating height as an important predictor may be due to overlap between model authors and workshop participants
Line 434: There is overlap of model authors and workshop participants and caution is needed when interpreting these results due to potential double counting of metrics which were considered important in both the literature and workshops results.

Comment: Methods: For the literature review, can you discuss in this section, any limitations of this approach given that this is considering studies based on statistical analysis in which any physical mechanisms responsible for any fuel driver are not explored/identified. Especially where multiple attributes are included in the model. Perhaps this can be addressed by commenting/summarising further the actual performance of the reviewed models and providing a summary of the key stats info e.g. R² value?.
Response: Added a discussion of limitations of the models reviewed. We originally attempted to compared models and fuel parameters between studies. Unfortunately there is not a consistent method of reporting correlation of fuel independent variables in the reviewed literature (even models with the same authors), so it is difficult to compare the performance of each model and the influence of fuel. In more recent literature particularly from W. Anderson, correlation coefficients, Mean absolute error and Mean absolute percentage error appear as standard methods of statistical analysis. However, many older studies do not use these and we believe it would be confusing or misleading to have partial information.
Changes: line 144 - There are some limitations to inferring influence of fuel from empirically derived statistical models. The reliance on finding a statistical relationship may be limited by the lack of variation within the experimental sites [14,43,50]. Cruz, et al. [51] suggest a di-minishing influence of fuel on fire spread as fire danger conditions increase. Addition-ally, the physical mechanism for fire propagation cannot always be inferred simply through correlation of fuel metrics to fire behaviour. Despite these limitations, analysis of trends in peer-reviewed literature provides a basis to understand potential influence of fire behaviour in coastal mallee shrublands based on analogous vegetation types globally. 

Comment: In the methods, could you also provide some additional discussion of how the photos shown in workshops were chosen? Including a description of the range of variability in the vegetation structure between photos and how this relates to any existing quantifications of typical ranges across these natural environments (if this exists).
Response: Thank you for this suggested clarification. Added more detailed about why photo pairs were selected. We deliberately avoided comparisons to existing quantifications to avoid biasing responses from participants.
Changes: Line 200 - Figure 3 presents a photo pair of planned burn sites, illustrating burns executed under similar weather conditions yet resulting in distinct fire behaviour. All corresponding photo pairs are documented in Appendix B. Photo pairs represent typical range of coastal mallee fuel structure on Kangaroo Island and Eyre Peninsula. Considering that all other factors influencing fire behaviour were controlled across each site pair, it is assumed that the observed variation in fire behaviour between sites within the photo pairs is primarily due to differences in fuel characteristics. Photo pairs were presented to workshop participants without additional commentary on fuel characteristics to avoid biasing responses.

Minor Comments

Comment:        Line 37: Typo – ‘Mallee shrublands [are]…’|
Response: Fixed, thanks
Change: Mallee shrublands are characterised

Comment: Lines 39-41: ‘Coastal mallee shrublands share the short mallee form eucalypt overstorey as 39 open semi-arid mallee shrublands but have higher canopy cover and denser shrubby understorey which provides increased fuel continuity for fire spread.’ Can a reference be cited to support this, as was done for the preceding statements.
 Response: Yes, added
Change: included Berkinshaw, T. Native Vegetatin of the Eyre Pensisula, South Australia; Finsbury Green Printers: Adelaide, South Australia, 2010; p. 234.

Comment: Figure 1: Can you include location and/or co-ordinates? Are these pairs of photos from one site each or two? Was the connectivity quantified at all in these areas?
Response: Added in figure label
Change: Line 51 - Figure 1. Photographs of coastal mallee on Kangaroo Island, South Australia (a and b) and semi-arid mallee shrublands in northern Eyre Peninsula, South Australia (c and d) contrasting vertical and horizontal connectivity of from eye level and oblique aerial perspective.

Comment: Line 66: Is further development of physics-based models required to help support the limitations inherent in empirical model approaches?
Response: Possibly, although we have focused on empirical modelling we have expanded this section and added some discussion towards the end on uses of physical modelling and further development.
Changes: Lines 144 - There are some limitations to inferring influence of fuel from empirically derived statistical models. The reliance on finding a statistical relationship may be limited by the lack of variation within the experimental sites [14,43,50]. Cruz, et al. [51] suggest a di-minishing influence of fuel on fire spread as fire danger conditions increase. Addition-ally, the physical mechanism for fire propagation cannot always be inferred simply through correlation of fuel metrics to fire behaviour. Despite these limitations, analysis of trends in peer-reviewed literature provides a basis to understand potential influence of fire behaviour in coastal mallee shrublands based on analogous vegetation types globally
Line 462 - Physical fire models have been used to study the influence of fuel on fire spread[17,80,81]. Physical fire models continue improve through inclusion of addition physical processes and become cheaper and faster to run [21,82-84]. However, field experiments and direct or indirect measurement of fuels will continue to play a major role in validation, whether for empirical or physical model.

Comment: Lines 135-136: Can you clarify here, did all 8 participants participate in all 3 workshops? Or was this 8 over the course of the workshops?
Response: Yes, Clarified in text
Change:  Line 162 – one of three

Comment : Lines 217-219: Might be worth mentioning the strong influence of height in the Rothemel model.
Response: Height is used to derived bulk density in Rothermel, so have added a note for the readers in the results and methods about bulk density being important to Rothermel, but also related to height
Change:  Line 103 - The primary structure fuel input in Rothermel model is bulk density (fuel weight per volume), which is derived from fuel load and fuel bed height [27].
Line 259 - Height is also commonly used together with fuel load in implementations of the Rothermel model particularly using the BEHAVE system [23,24,31,36,70].

Comment: Line 222: Can you add a reference(s) for the studies mentioned?
Response: Yes, done
Change:  Line 264 - features once in the list of model parameters [54], some studies provided an alternative model which used bulk density and discuss the importance of bulk density in shrublands [13,40,47].

Comment: Lines 267-270: When discussing 'importance to fire behaviour' in this specific section, is this in relation to sustainability or spread rate (or something else) or a combination of both?
Response: clarified
Change:  Line 305 - Table 3 summarises the rankings of various fuel metrics by broad fuel attribute categories. The results are presented in order of their importance to fire behaviour (both rate of spread and fire sustainability) in coastal mallee shrublands

Comment: Table 5: Can you explain why it was decided that the litter density and arrangement should be combined here (in Question 1)? These are potentially quite different factors for example, if the fuel is very sparsely arranged.
Response: Arrangement here referred to vertical arrangement, which is closely tied to density. You are correct, the term is ambiguous, so have removed arrangement, as density is the best term here
Change:  Change litter density/arrangement to litter density

Comment: Table 5: In Question 1, Is it worth discussing that some of these metrics are essentially a combination of each other? i.e. the litter depth is determined by the litter density and fuel load for a given area?
Response: Added a comment and noted that participant familiarity may also influence choices, rather than just fire behaviour
Change:  Line 267 - It should be noted that many of the metrics (cover, height, density, load) are often highly correlated. Familiarity of definitions of terms or methods may have an influence over workshop participant preferences

Comment: Table 5: For Question 5, could you clarify what is meant by visual fuel hazard method and data hazard method? These categorisations are for me a little unclear.
Response: Added extra definition and clarification at end of paragraph
Change:  Data derived hazard scores refer to hazard scores which turn physical measurements of fuel into a single unitless score or rating [72]. Hazard scores derived from physical data-based methods received lower rankings compared to visual techniques, but a more detailed examination of the feedback is necessary to comprehend the underlying reasons.

Comment: Lines 327-329: As per earlier general comment, I think it should perhaps be acknowledged here that the workshop attendees included authors of some of the reviewed literature.
Response: Additional comment about the overlap between researchers and participants has been added.
Change: 

Comment: Lines 360-362: Not sure if there is something missing here but maybe revise for clarity.
Response: Reworded the sentence for clarity
Change:  Near-surface fuel metrics are considered difficult to assess in situ[12] , however considering the results of the workshops and literature combined, future fire studies in mallee should still attempt to quantify the influence of near-surface fuel on fire behaviour. 

Comment: Lines 376-378: As per earlier comments, I think here you probably need to highlight that some of these model authors were also workshop participants. This is particularly important in a situation like this where you are making the point that there is alignment between workshop participant
Response: Additional comment about the overlap between researchers and participants has been added.
Change:  There is overlap of model authors and workshop participants and caution is needed when interpreting these results due to potential double counting of metrics which were considered important in both the literature and workshops results.

Comment: Line 416: ‘I think this is the first time fuel age is mentioned. Is it worth discussing this in more detail earlier e.g. in the intro? Was age incorporated into the workshops discussion e.g. in selection of site images etc. If so could this be discussed as well?’
Response: Age shouldn’t have been in the list, removed
Change:  Fuel height, fuel connectivity, cover and bulk density are considered to play …

Comment: Lines 420-422: ‘Maybe the conclusion should be more like ' are considered to have a key influence on fire behaviour' as this study is focused on this expert opinion rather than determining/quantifying the fuel effect.’
Response: Accepted, thanks
Change:  are considered to play a role in determining coastal mallee shrubland fire behaviour

Comment: Lines 430-432: As above, should this be 'are considered to play..'?
Response: Accepted, thanks
Change:  The results of expert workshops suggest horizontal and vertical connectivity are considered to play a…

Round 2

Reviewer 3 Report

Comments on version 2

Photos and figures were included this time, it was important to see them. I liked the concept of presenting to the experts the photos taken right before a prescribed burn was initiated, so that you knew how the vegetation fire behaved. I suppose that the experts could not see the result, whether it was “No go” or “Crown” upon ignition, as presented on Figure 3? Remove the text?

What do you mean by “unofficial”, regarding the 7 pairs of photos? Will they not appear openly in the published article? I believe that the photos provide insights to the readers. However, if you have good reasons for not publishing them, it is OK with me.

I assume that Figure 5 is an attempt to present the raw data from the experts` workshop, comparing photo-pairs. The importance score is from 1 (not important) to 7 (highly important). Should the information on Figure 5 correspond with / be connected to the information on Table 2? If YES, how?

Thank you for an interesting read.

Author Response

Thank you for your constructive comments. We are glad that the figures worked this time and helped clarify the intent of this paper.

In response to your comments, the labels (e.g. No-go, crown) were presented to the participants at the workshop. This was to help them distinguish which fuel characteristics may have contributed to different fire behaviour. We will leave the photos as presented.

The UNOFFICIAL text was generated by the software used to export the images, and was not meant to be there. It has now been removed, thank you for noting that. 

Lastly, We have added an additional clarification about the relationship to figure 5 and table 2, which you rightly point out are presenting similar information in a different way.

Thank you again for your time in providing detailed feedback. We believe this has helped refine the manuscript and trust that other readers will find it interesting and useful for future studies of wildfire in shrublands. 

Reviewer 4 Report

Overall Comment

I appreciate the receptiveness of the authors to the previous review comments and their effort and consideration in addressing and responding to these comments. I agree with the authors that the overlap between workshop participants and model developers do not necessarily change the findings etc. but I’m pleased they agree that this is worth clearly highlighting and explicitly discussing. I believe that the changes highlighted clearly explain this overlap, mitigation steps and possible effects (and their likelihood).

In terms of my previous comments regarding the Methods section, I understand and accept that comparison between existing models is complicated by reporting inconsistencies in previous studies. I appreciate the inclusion of the discussion around the limitations of statistical approaches. The additional discussion added for the photo pairs is also useful. I would still be interested in understanding the actual difference in ley fuel properties between paired photos even if this was not made available to the workshop participants (however I understand if this information is just not available).

I am pleased that the minor comments have also been considered in a detailed and thorough manner. The inclusion of a discussion of the possible influence of participant familiarity with terms is I think a very important point and good to see it included here. The additional clarification provided in several places is also appreciated and I hope the authors will feel that the clarity for outside readers has been further polished as a result. Finally, I think the slight change in the conclusions just to acknowledge the subjective judgment of the participants is merited and is then supported by the presented study.

Minor Comments

There seems to be some issues with the formatting for a couple of the references which seem to have been corrupted in the reference list.

Author Response

Thank you again for your time and constructive comments. We agree that clarifying some of the text as suggested by all four reviewers has indeed polished the manuscript.

In response to your comments, we have chosen not to add our own interpretation of the differences in photo pairs, but instead let the responses of the participants speak for themselves. While some pairs have obvious differences, others are more subtle, which was a deliberate design. We would like to not influence the interpretation.

We have added an additional sentence to the conclusion, as you suggested, as we agree that it is worth pointing out the limitations of the methods. While these results are somewhat subjective due to the small number of participants and subjective nature of the assessment of fuel drivers of fire spread, we believe they warrant further investigation. 

Finally, we have fixed the formatting errors in the reference list, thank you for pointing that out!