Designing Operationally Relevant Daily Large Fire Containment Strategies Using Risk Assessment Results

Round 1

Reviewer 1 Report

Questions have been well revised and clearly addressed. One minor suggestion is the figure. 

1) Figure 2, please make the location map scale bar and north arrow larger, and remove the background for the legend. 

2) Figure 3, please remove the background for the scale bar, north arrow, and legend. 

3) Figure 4, please remove the background for the legend. 

4) line 153, managers’ or manager's?  You used the singular in other sentences (e.g. line 144, 330...)

5) line 570, citation format [47-51],[10,16]->[10,16,47-51]

Author Response

Questions have been well revised and clearly addressed. One minor suggestion is the figure. 

1) Figure 2, please make the location map scale bar and north arrow larger, and remove the background for the legend. 

I increased the size of the scale bar and north arrow. Thanks for the suggestion.

Removing the background of the legend turns out to be difficult (if I understand it correctly). My  understanding is the current ArcGIS Desktop does not support this function. I would have to install ArcGIS Pro and redo all the maps to remove the background. I agree removing the background will make the legend looks cleaner, but this might require a new software and a lot of time for me to redo the maps. Please advise if there is any shortcut to solve this problem, or I didn't correctly understand your comments. Thanks. 

2) Figure 3, please remove the background for the scale bar, north arrow, and legend. 

I cannot find an easy way to do it right now. 

3) Figure 4, please remove the background for the legend. 

I cannot find an easy way to do it right now. 

4) line 153, managers’ or manager's?  You used the singular in other sentences (e.g. line 144, 330...)

I searched through the manuscript and made the change to use manager's. Thanks for catching it.

5) line 570, citation format [47-51],[10,16]->[10,16,47-51]

Corrected it. Thanks. 

Reviewer 2 Report

Thank you for your edits to address the issues I previously identified. I think these figures are much improved and have no further recommendations, aside from a minor edit in the legend of figure 5a, which does not match figure 5b for the last time category (19-21 days vs 19-22 days).

Author Response

Thanks for catching the mismatch. We corrected it. 

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

The authors explained that they obtained fuel, vegetation and topography data from WFDSS to characterize the landscape. However, it is not clear the type and level of detail of the information obtained and what would be required for fire simulations models (for example quantity of fuel biomass, woody/herbaceous and species composition, land elevation, slope, etc.). What is the weight and importance of these variables for fire behavior and risk assessment models? This information can stimulate fire managers to collect and systematize local data in their adscription territories. 

The authors suggested to the fire managers to evaluate more scenarios by extending the use of this model to preseason fire risk analyses or fire management training programs including, for example, different stakeholders’ social, economic and political concerns. I consider this a very important recommendation. What specific variables could be useful to define better fire management policies and actions? How to quantify social and political concerns? Could the authors make some suggestions?

Please revise text citations and the format of reference list (there are several mistakes using criteria for the citation and recommendations from the editor about References format)

Minor comments

Line 35: The term GIS is not mentioned in the text. You will need to include an explanation of its use in the article. In case you decide to delete it, a useful keyword you could add to the list is Operations research (OR) models. 

Line 160: we describe -> We described 

Line 167: In figure 1 please complete description including the meaning of white boxes (or refer it to the text).

Line 198: In the figure 3 is not possible to distinguish the legends of the images, please increase the resolution of the figure. 

Line 239: We select -> We selected

Line 243: We use -> We used

Line 246: We assume -> We assumed

Line 247: We use -> We used

Line 309: (27. Wei et al. 2018) -> Is this reference the number 28: “Wei, Y.; Thompson, M.P.; Haas, J.R.; Dillon, G.K.; O’Connor, C.D. Spatial optimization of operationally relevant large fire confine and point protection strategies: model development and test cases. Can J For Res2018, 48, 480-493.”(line 633)? If it is correct, replace “27. Wei et al. 2018”in the text by 28. 

Line 341: We build -> We built

Line 343: We use -> We used

Line 345: We also explore -> We also explored

Line 357: We assume -> We assumed

Line 382: crew s -> eliminate s

Line 409: discover -> discovered 

Line 432: We use -> We used

Line 451: involvestakeholders -> involve stakeholders

Line 468: In table 5 replace meter by m (under Line length)

Line 525: Include (Riley et al. 2018) in the reference list and replace it by the assigned number. 

 

 

Reviewer 2 Report

I found this paper to be a compelling read. The concepts are novel and relevant, and content was generally easy to follow, despite some heavy math backing up the science. My main suggestions revolve around providing a bit more minor supporting background content for readers who may be less familiar or unfamiliar with PODs, risk assessments, and spatial fire planning. Also, figure 5 does not work very well in its current form and could benefit from a rework in how it's referenced in the text and how it's displayed. Overall, these are all minor edits.

suggested edits in order of text:

-          Line 71 – insert “landscape” after “PODs are”

-          more description of the landscape surrounding the test-case fire would be beneficial when it is described starting on line 179; this fire and location may not be familiar to most readers. e.g., terrain, vegetation and forest-type, fire regime, remoteness, proximity to towns or high value point resources, etc.

-          why was this fire selected as the test case over others?

-          Figure 2 map should be improved. Inset should be location in the USA, not within the state of Montana, and some readers may not know what I-90 is, which is labeled twice on the map. Also, since you use rivers and ridges as POD delineators, perhaps a hydro layer with a hillshade and DEM could be included. From this map and description in text it is unclear how the study area is defined/what the outer boundary shows, and what the FSPro Fire Starting PODs are.

-          Line 195-197; don’t some PCL boundaries not need fire line construction (e.g. a large river or major road?)

-          Figure 3 and related text

o   All panels will need high-res versions for publication

o   legends and map elements would benefit from larger font

o   Panel 3d figure; study area should be the same size as in other 3 panels. Point feature depicting “Properties” is different than how this is referred to in-text (“Structures” or “Point protection”), which is confusing. Also, based on the caption it’s unclear if this map shows all structures or just some structures that are potentially threatened, and if they’re just some that may be higher risk, it’s unclear how this was decided.

-          Figure 4 could benefit from additional explanation of 6 flame length categories used. Line 114 is the only reference of details of 6 FL categories that I could find in the text, and it is unclear how the FL parameters for the 6 panels are delineated. Labels for the 6 panels could be cleaned up as well to clearly demarcate cut-off values (e.g. panels d and e overlap based on labels) and to clearly show what is being mapped (e.g. panel A specifies FL: but other panels do not).

-          Line 224-253: revisit for consistency of tense (mixture of past and present tense used here)

-          Line 258-259 edit to read “between line construction and fire arrival under the assumption that larger time gaps equate to increased safety for suppression actions.”

-          Line 284: “(e.g. one day)” … is this an example of a management period timeframe or is a management period always one day?

-          Line 291: insert “a” between “above certain”

-          Line 292-3: Cut sentence beginning with “Fire simulation …” It’s unnecessary.

-          Table 1: remove “We will explain later” from Expected total NVC in POD i parameter description.

-          Table 1: change “homes” to “structures” in the point protection benefits multiplier description.

-          Line 315: change “times” to “time” and insert “the” between “along selected”

-          Line 349: insert “fire” before “fighter”

-          Line 350: change “time” to “times”

-          Line 347 and section 3.1: it would be very helpful to have a brief descriptive scenario name added to Table 3 to help the reader keep track of the different letter scenarios while reading in-text

-          Figure 5 is difficult to interpret and supporting text is largely absent. Several suggestions here:

o   Definitely need more in-text reference to describe what this example is showing and how it relates to decision tree discussion

o   An arrow or some kind of visual indicator linking a to d, b to e, and c to f would be helpful. Could also change labels to something like 1a and 1b.

o   In some panels, model parameters are overlaying pod boundaries. In others they are not.

o   Why are structures only shown in some panels? These need to be included on the legend and should be a different color than the model parameters

o   Size gradient lines are too challenging to decipher.

-  a floating "Legend" label is showing in the bottom left corner of several of the panels

o   Is this figure really necessary?

-          Line 430-431 change to “primary fire lines due to multiple factors, such as spotting, rolling logs, and inadequate or incomplete line construction.”

-          Line 435: delete “based on test scenario B”

-          Line 436: change “are” to “were” ; insert “the” between “to fire”

-          Line 439: change “rerun” to “reran”

-          Line 485: FLEP-GEN is currently not available in WFDSS, right? Maybe mention this here.

-      In the Discussion section, I think a few points can be added or expounded upon. I'm interested to hear more discussion about suppression resource use, which is rarely a single resource type working alone in large fire support. If data availability poses challenges for line production rates for single resource types, how can real progress be made to accurately model production for a mixture of aviation and ground resources? Similarly, in the paper you bring up challenges associated with knowing real production rates and resource requirements for point protection, and the need to default to a basic assumption for required resources. Arguably, this is often a big driver in suppression decisions on large fire. Where do you see room for progress related to understanding point protection resource needs and related factors?

-          Line 541: change “helps rate” to “help identify”

-          Line 568: delete “earlier”

Reviewer 3 Report

This paper strings together several emerging fire suppression modeling efforts in a demonstration of a potential OR decision support system, thereby developing a novel approach. While this makes a valuable contribution especially with the decision tree approach, the complexity of the system and its lack of application tomore than one fire location limits the potency of the paper. Completely missing is the aerial firefighting component, present in most large fires.  Helicopters and fixed-wing could likely be described well in sections discussing HRLs and as a delay mechanism when arrival time is close to line construction completion. Some of the foundational concepts such as how firefighter safety is handled/ proxied could be clarified. Much of what is presented in the results section is a continuation of the methods. The paper would also be easier to digest if the supporting figures and tables were presented closer to their references in the text. The descriptions of the objective functions are worded awkwardly but this could be cleaned up easily. With some adjustments this type of system does seem quite valuable as future decision support.  Some suggestions about who could support this modeling and consult the fire managers about various scenarios during the response and how to test and refine the system to ideally include aviation support should be included in the discussion.

Line 15. Are fire risk assessments used widely by fire managers in the western US? I would change that word.

Line 22. Do you calculate eNVC using common weather forecast data? Rewrite so that weather doesn't sound like the only inputs you are using.

Lines 49-56. You are haphazardly mixing up tactics (direct, indirect, contingency), objectives or tasks and resources.  For example, dropping water and retardant are a means to an end, they should not be in the same task or objective list as protecting infrastructure, conducting burn out operations. Add delay fire spread to this list. Clean this section up to describe the various resources. How sometimes they go direct, sometimes they go indirect and sometimes they are doing contingency operations, then provide a more complete list of the operations objectives. You might want to add improving and patrolling as the meaning of holding line.

Line 57. Following what logic?

Line 59-60. Clean up this statement. Are the cost from responding to wildfire or the potential hazard to wildfire to be balanced with the positive and negative wildfire effects or are they the effects?

Line 73. Within or across PODS? Both apply.I think you mainly summarize across (e.g., total line length, number of pods, etc.) in your results.

Line 83. RMAT more than one dozen, but not dozens. Maybe just provide the actual number. Mention this effort later in the discussion as a possible support staff to pilot this tool.

Line 93. predict time-based probability contours or polygons.

Line 100. each grid cell

Line 106. not modeling instead of insensitive

Line 107. Should you be using the full range of weather conditions across the season, would you weight the weather more towards conditions when fire most often occur or forecast? Are the eNVC estimates you project biased low if weather associated with high probability of ignition days isn't weighted heavier?

Line 121. Your response strategy of line construction and point protection seems to neglect the import of reducing fire intensity with hose lays and aircraft and delaying fire spread with aerial liquids.

Line 128. would readily available be better than common for the weather here?

Line 130. Have you validated the fire arrival time modeling at all?

Line 166. Can you explain firefighter exposure and suppression cost in Figure 1 more? Are these combined? How?

Line 198. If you changed your classification in Figure 3(b) to single days the map would look more like a predictive version of a classic fire progression map. Can you explain the high and low eNVC scale in Figure 3 (c)?

Line 208. Figure 4, what do these meter measurements anchor into, Hauling Chart isoclines?

Line 225. Forest roads are used all the time as containment line. Why don't you use them in your POD delineation. They are certainly as good as handline unless severely overgrown. Could you use other data you mention like SDI or PCL results in addition to rivers and ridge lines?

Line 227. Are the flame lengths max predicted, average? Max seems like a better choice.

Line 230-235. This is where helicopters and retardant aircraft are glaringly missing. They spend a lot of time, effort and money reducing flame lengths.

Line 247. Have you consulted any hotshot supervisors about this topic? Line placement is easy with narrow line, much more saw work with wider lines. What about wet line versus dry line? Did you consider Broyles et al for fireline production rates?

Line 253. Its my experience that there are far more than 2 firefighters at each residence for point protection, at least during wrapping and setting up sprinklers. This is often done by engine crews, since residences almost always have roads that access them. Do you allow crew to move with the firefront through time or are they committed all days?

Line 282. What is fire is out in that POD already?

Line 284. Why must all lines be completed in a single shift? This seems unnecessary and unrealistic. Night shift and multiday lines are very common.

Line 290. Could you factor aircraft support in here. They are often cooling the fire to prevent it from being unsafe or ineffective and cleaning up slopovers. Basically, they can help convert HRLs to non-HRLs.

Line 299. Why did you assume all lines built would hold? How inaccurate is this assumption? What alternatives did you consider?

Table 1. What is M? It's role is not explained well.

Line 328. Can you change this to minimize the temporal gap? Or are you saying it is trying to make all of the gaps as long as possible. Its not written in a way that is easy to read or decipher with confidence.

Line 340- 377. Move to methods.

Line 364. What is the averaged minimum?

Line 369. Does this mean you did not do any scenarios where response options were limited by crew resources? Why not? Is there are reason you don't do this?

Line 396. Delete can between rPODS and vary. Please rewrite if the intent is to say that the fire size was 12 times as large and line constructed was 7 times as long.

Line 398. Delete By at start of sentence.

Line 399. Can you replace imposing with setting. Imposing implies constraints.

Line 401. Can you change requires to results in, and then add, as shown by ...with reference to one of your results.

Line 410. Please tell us how the benefits and the costs changed, don't just tell us the net result and guess at why.

Line 414. Why?

Figure 5. Move this up to just below when it is first discussed.

Line 429-444. These are methods.

Line 437. What is the probability? Please share.

Line 454-455. This sentence does not make sense. These seem like two ideas for how to increase scenarios, there are so many, but these two don't seem like natural ways to go about it.

Table 4 and 5. Can you portray firefighter safety some way here that accounts for flame length and time?

Line 477. First sentence is stating the blatantly obvious. Can you rewrite to actual list some of the benefits? Do you need a few steps in the development of this model before you argue for its adoption. Should it be run by RMAT? What fires should they target? What is the successful path to get a system this complex up and running and actually integrated into the planning and operations sections?

Line 481, can you rewrite, maybe eliminate integrating and change would be to could become a relatively easy external extension to add.

Line 483. Can you change could to occasionally.

Line 484. Which analysts? FBANs LTANs, RMAT? What skills do they need, do they all have these skills? What is the feasibility of doing this in a yurt? Do you need internet access?

Line 498 and 499. It seems like you casually throw in that this model can help estimate KPIs. If this is something you want to include I suggest you more carefully state which elements from your results would be used.What does providing a benchmark to compare to real world operations mean? You've just been talking about adopting this for real world operations. Perhaps restate.

Line 501-502, restate...improving upon previous spatial response optimization research by considering firefighter safety, contingency planning.... Firefighter safety is not coming through as an obvious result from your model. Risk management seems to be a more apt description. I think you are modeling it, you just need to clarify how you are assessing it, which is mainly time spent building line in various flame lengths. However, your optimization (2) is making sure that firefighters are not even present before the fire arrives. Basically it forces all line building to be indirect, at least if your fire arrival times are accurate, right? Doesn't not constraining crew resources make this artificially possible.

Line 519. What's overall effectiveness? How would you test it?

Line 524. It looks like you are using Plucinski's recent article to show what influences line construction rates. Can you add in crew productivity. Not all crews are Type 1. Can you revisit that assumption of yours here and address the impact of the actual mix of crews on most large fires.

Line 525. Why is Riley et al written out unlike other citations?

Line 535. by forest do you mean land management planners across all type of agencies, federal and state, etc?

Line 537. I think you mean potential control lines, not probability. Can you provide citations for both of the atlas layers?

Line 541, Which models?

Line 545. safety, probability of success at protecting values at risk? If not this, success at what?

Line 551. The Step (2) optimization  to be consistent with Step (1) on line 549.

Line 556-558. You end the paper with a statement about contingency planning. But you have not been very clear about why you picked 5 breach points after you set up the original assumption of 100% line effectiveness. I would state that this part of your demonstration is one of the many refinements that would need to be explored and resolved to make this tool a real workable extension to the WFDSS system.

Reviewer 4 Report

Major comments

1. The authors addressed that raster layers are used in their simulation.  However, I am very confused with the input raster layers. Does the spatial resolution of the raster data influence the fire simulation?  In a fire simulation, should one pixel be entirely burned or it can be partially burned?  What is the spatial resolution of the raster layers used in this study?

2. There are some tense and in text citation errors, please revise these errors.

 

Minor comments

Line 35: The term GIS is only mentioned in the Keywords. If you would like to use GIS to represent the raster layers used, please add more description for the raster data and use the term “GIS”. Otherwise, delete it.

Line 92: Here, fire weather is daily fire weather or fire weather for an entire fire event? If daily, please add “daily”.

Line 126: It should be better to add the units for the flame lengths (meter) and arrival time (day). Or these information can be presented in the method part.

Line 171: We use -> We used

Line 176: The three data sets (fuel, vegetation, and topography) used to characterize the landscape are not clear. Does the fuel mean fuel type or available quantity of fuel? Does the vegetation information require species information or only biome cover type? For the topography data, please list the factors included (e.g., slope, elevation). Because these information required for different fire simulation models are changed so much.

Line 198: The figure 3 is not clear, please use high-resolution images for this figure.

Line 236-Line 253: The tense is confusing. Please use the same tense for this paragraph. (e.g., Line 239: we select -> we selected)

Line 309: (27. Wei et al. 2018) -> In text citation format is wrong.

Line 402: We conducted -> We conduct

Line 430: During large fire suppression-> During a large fire suppression

Line 525: (Riley et al. 2018) -> In text citation format is wrong.

Line 547:  Indentation for the first line