Time limit for competing teams to extinguish a test fire?

For our proposed prize (click “show” below to see Proposed Prize Design).

Proposed Prize:
[spoiler]We are interested in focusing a prize specifically on wildfire suppression. In light of the increasing risk to lives and assets, the focus is on suppression of Wildland-Urban Interface fires, before these escalate into large fire events that put communities at risk. Here is an initial description of a proposed prize. In the other topics in the community, you’ll find different categories of feedback we would love to get on this design.

We understand that a prize designed in this way may result in a different approach or paradigm for firefighting, and we invite you to explore—for example, with faster detection and response, can fires be extinguished without traditional containment and control strategies?

The Proposed Prize: We are proposing a prize that works something like the following: Teams are invited to test a fully integrated autonomous system that rapidly detects, responds and suppresses wildfires. There will be a competition testing area of X by X acres (NOTE: we are proposing testing in an outdoor environment). At the beginning of the test, competition officials will ignite a fire somewhere within the testing grid. Once Y threshold of spread, temperature, flame height, or another variable is crossed, the competing team will have Z minutes to detect and completely extinguish the fire. The overall cost of the teams’ system must be no more than C dollars. Solutions must not pose an risk to the environment and/or lives.[/spoiler]

What is the right number for the “Z” variable above? ("the competing team will have Z minutes to detect and completely extinguish the fire…") I.e., what is the right length of time to give teams to extinguish a test fire? Why is this the right amount of time?

Share any comments, ideas, thoughts, or examples you might have!

When answering, please also consider the Y variable, ‘Threshold to Respond,’ in the above prize. Again the idea here is how long should teams have–once the fire has crossed a certain threshold in size that indicates it’s a fire that should be suppressed–to fully extinguish the fire.
(Click hereto weigh in on what that threshold should be)

Please focus on why this time limit is the right time limit in your answer.

Ok, I’ll bite!

Z / Y variables: I will refer back to the 97% (readily suppressed fires) / 3% (suppression -challenging fires) figure used in the general forum on this topic.

A (seemingly) valid approach here would be to ‘drill down’ into the actual data that produced the (above) statistic; calculate the AVERAGE time taken to suppress these ‘wildland-urban interface’ fires and use this average to define the upper and lower (range) limits on fire suppression time (for the challenge tests). Or, more specifically – and in the context of this wildland-urban interface – identify those fires which qualify under this descriptor (said ‘interface’) then estimate the average time (given real world demonstrations) to suppress said fire BEFORE serious* / pervasive damage occurred.

This prior averaging also applies to the total time that a fire suppression team will be given to DETECT and suppress a test fire; this time limit should be realistic and based upon real-world data from past fires (that meet the criteria [fire type] set forth in the challenge).

Obviously, we want to discover successful ‘autonomous and integrated’ solutions here…so, in arriving at our average time estimate, we should use (as gauges) examples of successful wildfire suppression which include an average estimation of the TIME taken between DETECTION (meeting the “threshold in size that indicates it’s a fire that should be suppressed.” @DanSelz ) and INITIATION of the suppression effort, and then, additionally, the TIME taken to achieve a successful suppression effort – once detected – to the point of complete containment and/or extinguishing of the fire.

 Note: this 'threshold in size' that warrants a suppression effort needs to be clarified    
 or defined, as any miscalculation (on the part of a challenge team) could result in a 
 'false start' and result in a a time expenditure (penalty, as it were) and/or require a 
 'restart' of the challenge test.
  • ‘serious’ here would mean any fire broaching the ‘interface’ and causing a yet-to-be-calculated amount of damage to homes (number of homes?), property (amount of property in square acreage?), infrastructure (loss of X amount), etc. Thus, a ‘threshold’ or ‘damage limit’ will need to be calculated based upon past similar types of fires (those meeting the yet-to-be-set criteria of this challenge).

I recognize that there is a major wildcard here in calculating these thresholds/variables: the ‘extreme’ weather conditions advocated by many here. Consequently, we must use some averaging formula (which could consist of several variables) to arrive at a ‘best estimate’ range for our putative Z and Y variables.

The time element for suppressing a given fire is a not a set time, rather it is based on the fire behavior (rate of spread and intensity) of the specific fire. Some fires move slowly and present minimal threat which allows for more time to suppress, while others spread rapidly and need immediate action to suppress. Fire agencies currently scale their level of response using predicted fire behavior and damage potential using observations of weather, fuel conditions, location, assets at risk, and availability of fire resources as determinants of the speed, weight, and type of response. Not all fires are treated the same.

The challenge we should address here is the fires that are immediately dangerous and threatening to people and built environment and don’t need to be “large” fires to be “damaging” fires. Every fire starts small, they may be damaging early on or become damaging after they move into an area that is vulnerable. For example, the 1991 Oakland Hills fire (Tunnel fire) actually had firefighters at scene in daytime mopping up a small fire from the previous day, when a wind blown ember from the “thought to be controlled” fire ignited new material and was immediately dangerous…detection, response, and suppression action time was 0 seconds and still the fire killed 20+ people and destroyed hundreds of homes in a less than 5 hours. The 2017 Thomas fire in Ventura county started at night and immediately destroyed built environment and fire manager struggled to know where the actual fire was and what areas downwind were igniting from flying embers that resulted in hundreds of properties destroyed before daybreak.

The problem here is to define a time relative to the expected fire behavior for the test case to detect and suppress the test fire, are we simulating a fire igniting inside a community (typically the most dangerous) or one that is some distance away in a forested landscape? What are the fire behavior parameters; "average bad day"or “extreme” conditions.

As an additional component of the challenge should be inclusion of real time fire behavior prediction in the WUI environment (none of the current forest fire models in widespread use model fire behavior in WUI; some even treat it as un-burnable, substantial evidence shows this to be incorrect).

Additionally, since wind driven fires frequently involve night fires where intelligence about the fire’s location and trajectory are practically unknown in real time. The challenge should include fire behavior prediction and close to “real time” location and trajectory that is available to first arriving fire fighters and fire managers

@DTurner first of all allow me to welcome you to the community- we’re very excited to have your expertise add to the conversation.

We definitely appreciate your valuable points regarding the nuances between time, size, and potential to be “damaging” and will keep them in mind (as well as potentially circle back to you) as we are fleshing out the details of the rules and regulations.

Re: your questions- We are going to treat this as something of a proof-of-concept test that will take place in a forested area as though it’s some distance away from a community but will the potential to spread, although we see your point that fires that start in built areas can be the most dangerous. We have discussed lab-scale testing as well that might be able to simulate that. We are hoping for conditions that are as bad as possible but we are trying to work around that given what might be available to us and what will be allowable.

We have also had many conversations about real time fire behavior prediction and modeling, and we could consider incorporating it in some way, either as a bonus “innovation” prize or an additional prize entirely. That will be an ongoing discussion for sure.

@DTurner - you wrote:

" The time element for suppressing a given fire is a not a set time, rather it is based on the fire behavior (rate of spread and intensity) of the specific fire. "

Yes, right. All fires are unique (spread rate and intensity factors have been noted previously), and, their characteristics depend upon unique weather conditions. This is fairly clear to all. However, if one rereads my longer post above (my next to last post before this one) you will see that ‘fire behavior’ is implicit in my mention of calculating (averaging from data) a ‘range’ of ‘fire suppression times’…an observation based upon the fire suppression statistics noted at the opening of the post.

There is no mention of arriving at a “set time” (a range of times, yes). But note that I mention this in the context of judging a fire detection-to-suppression effort by a given team participating in the challenge; there simply has to be a time gauge for determining the scoring/awarding of points to a challenge team. I suggested an re-analysis of the data (from which the 97%/3% figure was derived) as a way of arriving at an average time – for this purpose. One may always ‘weight’ such an average (e.g., awarding more/less points – or more time – depending on specific/specified conditions [rate of spread, etc] under which the fire is fought). This ‘weighting of probabilities’ (averages) is done commonly in probabilistic reasoning methods (like Bayesian inference formulas and Markov chains).

You also wrote: " The challenge we should address here is the fires that are immediately dangerous and threatening to people and built environment and don’t need to be “large” fires to be “damaging” fires. "

Granted. Damaging fires are not necessarily large ones (at first, anyways, although there is a clear quantitative relationship between the total amount of damage from a fire and its size [multiplied by its rate spread]).

Your observation is certainly a valid criterium…But how should we determine/evaluate such a fire in a fire challenge zone (one designed for this challenge and one NOT near homes or critical infrastructure)? We can only do so, if we base our evaluation method on some averaging of prior data about those “immediately dangerous and threatening” fires. Note also that I did call for a clarification of the ‘threshold in (fire) size’ that warrants suppression – a clarification that relates to your point about “damaging fires”.

You also wrote: **" The problem here is to define a time relative to the expected fire behavior for the test case to detect and suppress the test fire, are we simulating a fire igniting inside a community (typically the most dangerous) or one that is some distance away in a forested landscape? What are the fire behavior parameters; "average bad day"or “extreme” conditions. " **

“A time relative to the expected fire behavior” Yes…the key word there is ‘expected’…which in my view must be based upon prior probabilities and averaging(s) of past similar fires (let’s say we classify fires by type*, first, then calculate their ‘damage potential’ under certain conditions, then use these to inform our ‘expected fire behavior’, and arrive at an average time, etc.).

You also wrote: " …since wind driven fires frequently involve night fires where intelligence about the fire’s location and trajectory are practically unknown in real time. The challenge should include fire behavior prediction and close to “real time” location and trajectory that is available to first arriving fire fighters and fire managers."

This is an important observation and you note critical factors (location and trajectory) in real-world fire suppression efforts. I do not know at present how, or if, night-flying drones, for example, operate or perform in this regard (fire detection at night). Intuitively, it would seem that certain aspects of fire detection would be easier at night (no haze or glare to obscure the view; the human eye can see a single photon in a visual field of total darkness). But I wonder if this is at all true (but see my followup post with a link to a very recent NPR story about ‘old school’ fire lookouts).

As for “fire behavior prediction” I recapitulate my earlier assertion that analysis of fire data (and assigning probabilities, for example, like Markov chains, similar to weather forecasting) will be required for any said prediction (especially if one is presupposing the use of computer modeling/AI/Machine learning in some form) about location and trajectory intel that can be fed to first arriving fire fighters/managers. Otherwise, we are just using a ‘best guess’ methodology.

*Your observation about types of fires is important; these are critical distinctions to be made in terms of real-world fire suppression efforts. However, in this challenge (as implied in the comment by @"DavidPoli "), I doubt we would be permitted to ignite a “test fire” in an actual community where people live.

That said, perhaps the XPrize sponsors would agree to build a quasi- “Potempkin’s village” [an empty fascade] or the equivalent of a ‘Hollywood film set/town’ [a fake town with ‘skeletal’/scaffold-only buildings] that has target infrastructure can be ‘threatened’ (even burned down and restored by a work crew) by a set/test fire. Scoring can be based in part on how much damage is permitted / prevented by a given team, etc. working with the same fire type (weather conditions, fire parameters, etc.).

Further thought: what if the wildfire challenge required detection and suppression of more than one type of fire (but all meeting the criteria that falls under that 3% [of fire, types] datum)…? Although more time consuming and logistically expansive, this would give teams more of a chance to demonstrate their fire fighting methods AND potentially reveal the best / optimal method(s) for fighting multiple types of fires under varying conditions!

This is a great piece by NPR journalist Nathan Rott (that oddly appeared in my browser start up news feed this morning) that offers a ‘Romantic’ view of fire fighting but which is also relevant to our wildfire suppression challenge…and perhaps one might contact these folks to help answer the question of whether nighttime fire detection is easier that day time detection…?

‘A Fire Lookout On What’s Lost In A Transition To Technology’
https://www.npr.org/2019/09/15/749547034/a-fire-lookout-on-whats-lost-in-a-transition-to-technology

I think it is a challenge and perhaps simplification to think that we can describe possible fire scenarios in terms of just the time needed to extinguish the fire and size. @DTurner is right that management decisions are made on every fire based on different objectives. Land management objectives for fires can be much more complicated that public safety objectives, which often include preventing the movement, as much as possible, of the fire towards a populated area. I think that the best path forward is to choose a realistic configuration, such as a rapidly spreading fire from a canyon or wind-driven area to expose a potential community. The goal would be to prevent the fire from reaching that community or igniting anything there. Solutions should include remote sensing and modeling capabilities - otherwise how will they know where to go? Whether that modeling is simply an interpolation or an actual fire model could be up to your team, but detection is key.

In terms of designing an activity I see an issue as fires are rarely repeatable. Winds change, fuel moisture, and you don’t get to burn on every day you desire to. Getting over this will be difficult, however I think the clearest objective would be to protect some area or group of structures (probably fake) from an oncoming fire. The system would be designed to both prevent spot fires that start and spread within a community as well as the larger fire front which is heading to the community, as that could in and of itself ignite structures if it is too close.

Ultimately, the limit then would not be time but the parcel that you are trying to prevent. You could have some easy to ignite targets there and the goal would be to prevent those from igniting while the fire spreads. Getting the fire to spread there every time would be really hard, but that’s a goal.

@mgollner - I agree with your assessment/critique and I have suggested similarly in previous comments – especially the use of fake targets for judging fire suppression success).

Your point that fires are not ‘repeatable’ is well-taken (and it certainly will have an impact on any challenge design plan).

One question: in your statement above, your wrote: "In terms of designing an activity I see an issue as fires are rarely repeatable. Winds change, fuel moisture, and you don’t get to burn on every day you desire to. "

Did you mean to say ‘fuel moisture’, or, did you forget a comma after ‘fuel’ ?

@marz62 fuel moisture would refer to the moisture content in the fuel. There is both live and dead fuel moisture, which further complicates things, and this changes both throughout the day (diurnal variation) and a longer variation with time as weather shifts. The wind is an even bigger factor and can change on the second timescale and dramatically change the outcome of a test, whether in speed or direction. This is a common problem when doing prescribed burns that are used as research validation tests, you never quite know what you’re going to get! The fire may not even go towards your targets.

@mgollner - Good. Thank you, that makes perfect sense (e.g., wood has different moisture content depending upon its state – living or dead – and that of the immediate environment, including the weather)… One of those details (where in the devil dwells) or factors that can have a MAJOR impact on fire spread and suppression.

I was stuck on my recent reading on fire forensics in which there is much research on volatile (liquid) fuel residues that remain post fire (e.g., gasoline).

As for ‘hitting the targets’ (e.g., fake infrastructure, ala a Hollywood film set)…perhaps this can be aided (even ‘directed’) by creating a ‘fuel trail’ from the fire origination site to the target infrastructure to be ‘saved’ – giving teams a MIN/MAX time allotment to detect and suppress the challenge fire. Granted, this is potentially another (fire control) variable in the challenge design, but it is one that could be greatly controlled if the fuel type (and its moisture content) is known before hand (i.e., made for the test).