From Firestorms to Fuel Maps | Fi-Sci’s Matt Robards & Kevin Bonner

Summary

Highlights

• Matt and Kevin from Fi-Sci are revolutionizing wildfire management with machine learning 🚀.
• Their focus is on prevention and mitigation rather than just fire suppression 🚒.
• They emphasize the importance of community awareness and good fire benefits 🤝.
• The conversation touches on the industry's tendency to overlook preventative investments 🔍.
• Fi-Sci's software not only predicts fire behavior but also models preventative outcomes 🌿.

Key Takeaways

• Fi-Sci is innovating with software and machine learning to tackle every stage of wildfire management 🌐.
• Their ultimate goal is to focus on fire mitigation, preventing wildfires before they start 🛡️.
• Prevention of wildfires can save assets and communities, making prevention more impactful than suppression 🔥.
• Community awareness of good fire practices can transform fire management strategies on a macro level 📢.
• Investing in preemptive measures rather than just suppression can change the game in wildfire management 💡.

Overview

Chapters

• 00:00 - 00:30: Introduction The chapter starts with a warm welcome to the show 'fire break' hosted by Steve Wolf. The show is sponsored by team wildfire, who have generously equipped them with good equipment to explore and learn about wildfires globally. They introduce Matt, a guest on the show.
• 00:30 - 02:00: Meet Fi-Sci The chapter introduces Robarts and Kevin Bonner from Fi, a company based in Australia specializing in fire science. They develop software and use machine learning to manage wildfires from their inception to the replanting phase. The discussion provides an overview of their innovative approach to wildfire management.
• 02:00 - 05:00: Fi-Sci's Mission and Vision This chapter introduces Fi-Sci and its mission and vision. In a conversation likened to a 'fire break', the discussion centers around unplanned scenarios (fires) and the hope to prevent them. Hosts Matt and Kevin exchange greetings and discuss their work, their initial ideas, and the current progress of their project. Kevin is poised to explain their work while Matt plans to narrate the genesis of their ideas.
• 05:00 - 10:30: Challenges and Data in Fire Mitigation The chapter 'Challenges and Data in Fire Mitigation' discusses the comprehensive approach to managing wildfires. The mission is to transform the various stages of wildfire management including mitigation, containment, suppression, and overhaul. The focus is on addressing the entire life cycle of a fire, starting with actions required before a fire begins to help mitigate its effects and contain it within its existing footprint.
• 10:30 - 15:00: Role of Smoke in Wildfire Management The chapter discusses a company with a unique vision for wildfire management, focusing primarily on prevention rather than suppression. Instead of trying to control wildfires after they have started, the company aims to mitigate the risk of fires from occurring in the first place, thereby eliminating the need for suppression and containment.
• 15:00 - 23:00: Community Awareness and Challenges in Mitigation The chapter discusses the importance of not needing to revert to previous states when making progress and setting high, potentially unattainable goals. It highlights the use of a software tool designed to offer insights into asset and infrastructure risks. The tool also suggests ways to mitigate these risks, promoting proactive community awareness and challenges in achieving effective mitigation strategies.
• 23:00 - 29:00: Investment in Fire Prevention and Suppression In this chapter, the focus is on investment in fire prevention and suppression strategies. It outlines optimal treatment strategies for managing fire risk both in the short and long term. The strategies discussed include prescribed burns, revegetation, and grazing by cattle or livestock to keep fuel levels low. Additionally, mechanical intervention such as understory thinning is highlighted as part of a comprehensive approach to fire management.
• 29:00 - 36:00: Machine Learning and AI in Firefighting This chapter discusses the integration of machine learning and AI in firefighting strategies. It begins with the analysis of fire history using descriptive analytics, which helps understand past fire incidents in a specific area. Next, predictive analytics comes into play, utilizing bushfire or fire modeling to predict the movement and impact of potential fires. Finally, prescriptive analytics offers actionable measures that can be taken to minimize risk and prevent catastrophic wildfires from damaging essential assets and infrastructure. These advancements highlight how technology is transforming firefighting efforts.
• 36:00 - 43:00: Role of Regulation in Firefighting Technology The chapter discusses the important role of regulation in firefighting technology, focusing on fire mitigation strategies. Fire mitigation is defined as the process of preventing fires from starting and ensuring the safety of people well before any fire begins. This involves taking proactive measures in advance, particularly in remote areas, to reduce the likelihood of fires and their potential impact on communities.
• 43:00 - 51:00: Data-Driven Fire Suppression Strategies The chapter discusses the impact of the Black Summer Fires of 2019-2020 on a town in New South Wales, Australia, known as Mimula. The speaker shares their personal experience, explaining how the fires deeply affected them and their family. As a result of this impactful experience, the speaker decided to join the local fire brigade, expressing a newfound passion for firefighting. The chapter highlights the personal and communal effects of the fires, as well as the motivation and drive they instilled in individuals to engage in fire suppression efforts.
• 51:00 - 59:00: Fi-Sci's Approach and Future Vision The chapter discusses Fi-Sci's approach to firefighting, emphasizing the significance of proactive measures in preventing and mitigating wildfires. The speaker shares their personal experience with firefighting, highlighting the profound impact of addressing fuel loads to prevent fires before they start. Recognizing that wildfires cannot be completely stopped due to natural causes, the focus is on strategies to reduce their severity and impact.
• 59:00 - 67:00: Comparison to Other Fire Prediction Models The chapter discusses the impact of wildfires on the built environment and identifies three major factors affecting wildfires: fuel, terrain, and weather. It highlights that fuel is the only factor that can be controlled, though there is ongoing work to influence weather conditions. The difficulty of containing fires, which can spread over hundreds of acres or hectares, is also mentioned.
• 67:00 - 76:00: Current Product Development and Offering This chapter discusses the process of addressing and managing fires, focusing on the different stages of a fire. It outlines the distinctions between burnt areas, threatened assets, and the active combustion line where reactions occur. The speaker emphasizes the importance of controlling factors such as temperature, wind direction, and humidity at the combustion line to potentially stop a fire.
• 76:00 - 81:00: Conclusion and Contact Information The chapter discusses the challenges firefighters face when dealing with fires that create their own weather. Fires can continue to move rapidly even in calm conditions because they generate their own wind, which propels them forward.

From Firestorms to Fuel Maps | Fi-Sci’s Matt Robards & Kevin Bonner Transcription

• 00:00 - 00:30 [Applause] [Music] welcome back to the fire break i'm Steve Wolf as you know team wildfire sponsors us she's very generous of them they paid for all this nice equipment around us so that we could go out and meet some people all around the world who could tell us interesting things about wildfire and we have I've got um Matt
• 00:30 - 01:00 Robarts and Kevin Bonner bonner Bonner Bonner Bonner kevin Bonner i knew that it's the team um uh coming to us live from Australia and they've got a company called Fi which is you know clever take on sci-fi I think uh but it stands for fire science and they're working on software and machine learning to manage every phase of a wildfire from conception through uh replanting oh wait
• 01:00 - 01:30 no no not conception right because we didn't in most cases plan the fire maybe sometimes some someone plans the fire for the most we're we're hoping that doesn't happen um Matt Kevin welcome to the fire break how are you today very well thanks how are you we're we're well here so what are you working on how did this idea come about and where are you in the process kev do you want to explain what we're working on and I'll explain how it came about yeah absolutely
• 01:30 - 02:00 um our vision uh gives us a little bit of a clue and you've already touched upon it Steve but our uh mission is to transform the mitigation the containment suppression and overhaul of wildfire and that uh plays into the you know the full life cycle of a fire uh it beginning or or work that needs to happen before it begins in order to mitigate it to contain it within its existing footprint
• 02:00 - 02:30 to suppress it and put it back out and then the overhaul to put things back to the way they were um before the fire started however we've kind of got a weird already even though we're quite a relatively new company our vision is to replace our vision um we actually want to get it down to just the mitigation is the sole focus of the company if we don't allow the fires to start in the first place then you don't need to contain them you don't need to suppress
• 02:30 - 03:00 them and you don't need to put things back to the way they were previously it's a really lofty crazy we might never achieve it sort of goal but I think you should have something like that when you're looking into weighins in the future so yeah as as you mentioned it's a software tool and it provides our users insights into risk of their assets and infrastructure and it goes about prescribing how you might reduce
• 03:00 - 03:30 that risk using optimal treatment strategies over the short and long term so and when we mention treatment strategies we're talking about prescribed burns uh reveation grazing in fact having cattle or or livestock on land and gra keeping the fuel low and even mechanical intervention so understory thinning and uh things like that so comprehensive there brother yeah yeah exactly it's uh
• 03:30 - 04:00 it it sort of takes all the different analytics that you need to consider so descriptive i.e what was the what's the fire history in your area through predictive we use um bushfire or fire modeling to predict where the fires go and then prescriptive analytics what should you do about it in order to reduce the risk and and hopefully prevent a catastrophic wildfire hitting your your assets and infrastructure wow matt when when Kevin says that the
• 04:00 - 04:30 focus is mitigation what does that mean to you how do you define fire mitigation yeah um how do we prevent the fire starting in the first place and potentially killing people um so it means getting out way in advance of the fire actually you know way out in advance before the fire actually starts um so where I live as you can tell I'm not in the city i'm in a remote area on
• 04:30 - 05:00 the far south coast of New South Wales uh called Mimula is the town I'm in and we were hit quite badly by the fires in 2019 2020 the Black Summer Fires they're now known in Australia and it it was quite impactful for myself and my family as a result I actually went out and signed up for the local fire brigade uh after those fires and yeah decided I wanted to do my part and I love firefighting now i live and
• 05:00 - 05:30 breathe everything firefighting um and yeah I've seen firsthand kind of you know the impact of the fire and also the impact you can have on preventing the fire if you can get out ahead of it and actually treat those fuel loads and prevent it from occurring in the first place so we we obviously we can't stop wildfires right right i mean nature starts a few thousand of them every day but I I guess what what what we're hoping to do is is to mitigate the
• 05:30 - 06:00 impact on the built environment exactly i mean the the factors that affect wildfire are fuel terrain and weather you can only control one of those things and that's the fuel you can't control the terrain i don't know about that Matt that's I'm working on the other part yeah we're trying to we're working on controlling the weather now because you know on on a fire line right it's we talk about you know hundreds of acres or hectars of fire right well that's that's
• 06:00 - 06:30 just referring to the part that's already burnt right and then you have the threatened assets that's the part that haven't burnt yet and then you have that line where the actual combustion reaction is taking place and the only me you know as far as I can tell the only weather that matters is what's going on right at that point where fuels are being converted into oxidated fuels and if right at that point you can control the temperature you can control the wind direction or you can control the humidity you can you can stop a fire that's my
• 06:30 - 07:00 I mean that's exactly what you're doing when you try to put water on the fire um local problem is and that the challenge you have as a firefighter is that's the point where it starts creating its own weather as well so you know I've been at fires where the surrounding environment there is no wind it's you know it's become very calm but still that fire is moving really fast because right at that point that you're referring to it's creating its own wind and it's pushing itself along clever isn't it yeah
• 07:00 - 07:30 so so what's the what's the scope of the of the project i mean because if you're looking at everything from premitigation prevention fuel load management ground conditions you know right on through strategies for suppression and then remediation afterwards that's that's a lot to pack into a software how how long have you been working on it and how long do you think it'll take to get to a point where
• 07:30 - 08:00 you feel like it's moderately functional yeah I think it's really important to have focus so when I talked earlier about um the idea to supplant our vision with just the mitigation side of things um that's kind of where our focus is as a relatively early stage startup we are focusing 100% just now on mitigation and that's not to say that in the future that we might not um put time and effort into the
• 08:00 - 08:30 suppression containment and overhaul but um we're we're seeing others in the industry focus on the suppression so the identification that there is actually a fire happening the containment suppression and overhaul it seems uh to me anyway uh in my opinion which doesn't stand for a lot but my opinion regardless is that there seems to be a um a lot of focus from private companies
• 08:30 - 09:00 going into those areas where we we've seen um you break it up a little bit there could you just repeat that part about what you what you said the focus the focuses that you're seeing are yeah like as in terms of what we're seeing the focus on in terms of private companies they're they are focusing on more like the identification that there's a fire in the first place that there it needs to be suppressed it has to be contained in some way so we've seen that there's a a a gap
• 09:00 - 09:30 around the mitigation in terms of preventing fire starting in the first place so where we have put all of our focus in uh for the for the foreseeable future is is exactly that what can we do what part can we play in order to for our users to give themselves the best chance that a catastrophic wildfire doesn't happen in the first place yeah I'd echo that by saying that um
• 09:30 - 10:00 when it comes to investment dollars as well there's a a lot more investment seems to go into actually uh dealing with a going fire it's it's a much more attractive thing to deal with but the needs and the desires that have been expressed for where we need technology to focus is actually preventative it's the upfront piece that we need to get to where the investment is matching where the need is as opposed to where it is today
• 10:00 - 10:30 which is it's it's much cooler to say you know we do X Y and Zed to put out a fire or to stop a fire and it it it looks great to to do that but we need to get more attention and investment dollars focusing on using tech to actually prevent the fire starting in the first place so I'm curious about just how you define investment do do you define investment as what's being spent by the government towards supporting different types of efforts or are you talking about actual you know venture capital private equity investment
• 10:30 - 11:00 well because I see a lot of the the the VCP side you know going into software projects satellites sensors cameras yeah there is a lot a lot of the satellite work is in detecting hotspots which by nature is a fire that's already going um there's some uh satellite technology now going into monitoring fuel conditions and things like that as well which is much better for upfront preventative measures um but yeah I think there's the the X-P prize wildfire at
• 11:00 - 11:30 the moment for example is a huge prize aimed at detecting and extinguishing a fire in its incipient stage so before it actually gets out of control which is fantastic and we absolutely need to do that but I'd like to get the world to a point where we can't detect the fire because it didn't happen so in when when conditions are hot dry and windy in in a remote in a remote area and there's a lightning strike or something like you're going to get a
• 11:30 - 12:00 fire absolutely yeah yeah i I I'm using hyperbole a little bit when I say get the world to a point where we don't see any fire it it's we don't want to see fires that are impacting civilization in the way they have been recently okay and and where do you put smoke in that kev do you want to talk about smoke a bit yeah smoke's a really interesting one because there's the when we were doing research into smoke um it actually uncovered an
• 12:00 - 12:30 industry that we didn't know um would be impacted by it as much as it was so in terms of the industries that we can service it's really simple it's just anyone who's got assets or infrastructure on fireprone land um and quite often what happens is those industries put their infrastructure on fireprone land because there's benefits to that so I'll give you example
• 12:30 - 13:00 renewables you put renewables like a wind farm up high that's got access to wind and that's where a fire likes to go it likes to go up and the same goes with uh solar panels you want to put solar panels where they're uh close to it it's flat but it's also got access to good sunlight um the one use case in terms of smoke that was interesting was um wineries so
• 13:00 - 13:30 um what you would have thought is that if you've got a lovely I guess in California would be um you've got a lot of wineries over there uh it's no different in Australia we've got lots of wineries right about us and you think just the fire alone would be their primary concern but it's not it's the smoke if smoke gets into the grapes into that year's harvest then it can completely ruin that winery's batch of
• 13:30 - 14:00 grapes uh it's called smoke tain and it's a very well-known thing both in Australia and in the US which we had no idea about at the time but it's it can completely write off your harvest for that entire year if a fire um and the subsequent smoke gets into those grapes so there's been studies that we've looked at um by some universities over here saying that if you do prescribed
• 14:00 - 14:30 burns they will have less of impact and effect on the grapes versus a catastrophic burn because prescribed burn by nature is controlled that the the fire won't be as big therefore the smoke won't be as much um so yeah it's it's definitely smoke's a concern and and that's just wineries i mean what we're really conscious of and what our tools really good at is you may want to do prescribed burns around a community
• 14:30 - 15:00 but you don't want to smoke that community out you don't want to be doing it frequently enough that you know they they suffer as a result due to the smoke and particular if it's near a hospital if it's an age care facility or if it's kindergarten you want to be very conscious of that and um you want to treat that with kid gloves as it were because if you continue to do it's all in the best interest the ve the fuel load vegetation management but at what
• 15:00 - 15:30 cost you don't want to harm the population around you as a result matt what do you add to that um particularly around the prescribed burn uh perspective and the effects and impacts of smoke from prescribed burns is something you have to be very cognizant of it's actually fortunate that you're talking to me when you are because in a few hours time there'll be some smoke wafting over from just behind me there we've got some forestry burns
• 15:30 - 16:00 happening today to mitigate the fuel loads around here um and that's one of the constraints that you really have to consider when building a a a burn plan for a region and you you build it out over several years to try and minimize the impact of smoke on the local community you want to treat as much as as as much of the fuel as you can but recognizing that once you get above literally a handful maybe two
• 16:00 - 16:30 burns of a certain size then that's more than that community can handle in a given year so you need to come up with the optimal mosaic over the area to actually not impact the community to utilize the resources you've got in the best way possible and bring those fuel loads and hence the risk down as low as possible yeah so do you think of prescribed burn not only in terms of managing fuel load but um managing the smoke load that that a prescribed burn is going to well it's
• 16:30 - 17:00 it's going to ensure that you're breathing smoke uh but a little bit versus uh in an uncontrolled fire where the the smoke levels are just um not survivable oh absolutely um I remember one day in 2020 the fires I was talking about that actually are now responsible for me being here in this position I think um led me into the whole fire space the sun disappeared at about 3:00 in the afternoon the smoke
• 17:00 - 17:30 from a fire which was I think it was still about 100 km away the smoke from that fire literally blacked out the sun for 24 hours i I don't mean it went darker i mean it turned to night wow yeah um so yes I'll take a prescribed burn over that any day right so you divide the fuel load up over the number of days that you want to burn it rather than have it burn all at once exactly i see uh so so I guess it's one thing to to you know sit at the keyboard
• 17:30 - 18:00 and figure out what you know what what the best way to go about uh fire planning and mitigation looks like um and then you get to the realization that you don't have control over many of these factors uh be because of either political will political climate budget existing policies willingness of the community to participate in these activities uh so at some point the success of all these
• 18:00 - 18:30 plans becomes a marketing challenge right we've got an answer now how do we sell this to the people who have the authority and the budget to implement it how do you go about that so it's a pretty horrific marketing policy that you sometimes have as well because if you sometimes your best marketing policy is an uncontrolled wildfire uh which is pretty grim and horrific to think about really i mean we've got fires just now in South Australia uh and
• 18:30 - 19:00 Tasmania uh more famously over over your side of the pond you've had the fires in LA and and more recently there've been fires further south uh in Texas so that usually brings focus uh and it it means that folks who want to mitigate that risk are out there actively looking for solutions that can bring about um a a better way of managing wildfire
• 19:00 - 19:30 rather than letting it get to what we may have seen over the last couple of months and years so it's it's pretty grim like that is one of the marketing things that that that we see in the wildfire space but generally when we reach out to people in different industries whether it's the uh power transmission and distribution companies whether it's carbon abatement firms uh
• 19:30 - 20:00 or and so on they're very aware of the dangers and the risks to them they just have no way of then taking those fears and doing something about them we've had many conversations where people will say "I've got all this data from different sources there's no way that I am able to consolidate all of this information in one that would in one uh
• 20:00 - 20:30 source that would then allow me to make a decision about what I should then do about um my property or my particular conditions in order to reduce my risk to wildfire so we find that the industries that we talk to which is just about every industry you can think of are very well educated and understand the risk to them but they just need someone or something to give them a helping hand in order to work out
• 20:30 - 21:00 what they should then do about it matt how do you see those challenges yeah so um I think community awareness is a huge part of what we need to target we put out a LinkedIn post just this morning actually around um the benefits of good fire it's I think there's a fear of fire in general and when you have horrific events like Australia's Black Summer or
• 21:00 - 21:30 the recent LA fires it's you know naturally it it's a natural human instinct to be scared of fire and there's there can be a stigma around fire um but raising community awareness around the benefits of good fire that there is actually good fire and bad fire and the benefits of good fire and you know you can imagine the the forestry burns around here that I was talking about a second ago you can imagine the fear that some
• 21:30 - 22:00 of the community will experience if they didn't know to expect that or you know they see the smoke coming in there are a lot of people still suffering from the trauma of previous fires um that smoke will bring things up for those people but this is good fire this is the fire that will actually save them from that happening again and raising that community awareness community awareness ultimately flows into government awareness I think um so yeah we we need to keep pushing out material like this raising people's awareness that good
• 22:00 - 22:30 fire is good so how do you divide uh your efforts between uh writing you know code um and analyzing the situation and softening the beaches for adoption of what it is that you're suggesting yeah this is the challenge of startup there are only so many hours in a day and there's so much to be done um I'm not sure what the good answer is
• 22:30 - 23:00 not sleep maybe uh no you know I I a lot of people have tried that myself included it turns out to be not such a good strategy especially if you then have to go out on the fire ground you you don't want to do that with no sleep right and the same thing about approaching just about everything else is uh a good night's sleep you'll get a lot more return on every hour of work absolutely i think it's a really important point you bring up though Steve because the it's focus it's as I said before like our our vision was
• 23:00 - 23:30 those all of the caveats of a wildfire life cycle but we decided to focus on mitigation it's it's always been very quite cutthroat in terms of what is it that we should be doing this given day this given week this given quarter we need to ensure that the decisions that we take are in the best interests um for the business really uh it does help that I'm my background other than
• 23:30 - 24:00 being a soft a lapsed software engineer is um the project and product management so I'm very much the boring one who does the process and all of that stuff that you know Gant charts and all the sort of stuff that everyone quite rightly hates that's my bread and butter and I I quite enjoy doing it while Matt is the more exciting one the blue sky thinker who goes out and thinks about the how we
• 24:00 - 24:30 improve the product and how we can bring our uh users greater value in what we do so how we delineate in terms of what Matt's responsibilities and mine is quite obvious but um yeah as a startup you find yourself going from financial matters to business development matters to product matters to coding matters in the space of about 15 minutes so it is a challenge and you need to be quite um
• 24:30 - 25:00 tough on the decisions that you take very interesting um Matt from a from an ML standpoint um to what degree do you see a a translation between uh what should be done and who does it so which is to say can we can machine learning teach machines to put out wildfires so we stop putting people in harm's way
• 25:00 - 25:30 fascinating question um I'd love us to get to that point as soon as possible i think one of the big hurdles we have to that is regulation um you know time and time again we hear the fear that governments have around AI in in anything not just not something as dangerous as firefighting but any domain
• 25:30 - 26:00 because there's the the kind of the the fear of the black box we need accountability we need traceability um when it comes to putting out fires then that just becomes so much more scary I guess um I would love to see us get to that point i think u autonomous aircraft in particular is a fascinating area it's one that is entirely hamstrung by regulation at the moment um I I think we can advance a lot a lot more quickly
• 26:00 - 26:30 there if we remove some of those regulatory barriers but I I also think there for the foreseeable future will be the need for boots on the ground um I'm sure not seeing any technology right now that can completely remove that need so So if we were to take the regulatory inhibitors off um how close would we be to being able to do
• 26:30 - 27:00 this i think one of the things if we remove the regulatory barriers to drones for example on the fire ground um there was a fire I was at a year or so ago where we received an emergency call to go to a property that there were people at risk and we got to that property and the driveway was completely engulfed in flames there was no way to get the truck into the property um the e even from one
• 27:00 - 27:30 of the road at the front the the radiant heat from the flames was so like it was searing hot inside the truck it was impossible to get in there and so we made the decision that we couldn't it wouldn't be safe for the crew to go in as it turned out it was very fortunate there was nobody in that property and so there were no lives at risk um and that's a a decision of the crew at that point in time with drone technology we're talking this driveway
• 27:30 - 28:00 is hundreds of meters long it's not a suburban driveway it's a a rural property and so we could have put a drone over and into that property to see if there are people there and lives at risk you could easily see the headlines if it was the other way around and we did decide to take the risk and go into that property you know could have been overrun by a fire could have you could have a truck full of firefighters that are dead for nothing because there's no one in that property um and so you know that's one of the things
• 28:00 - 28:30 I've seen firsthand where removing the regulatory barriers would enable such greater intelligence on the fire ground for us to make it so much safer for firefighters and and just make the decision-m process so much smoother and and presumably the the restriction on aerial drones is is simply so that they inter don't interfere with manned aircraft is that right exactly and it that's a very valid concern don't get me wrong i mean it it does what are the
• 28:30 - 29:00 worst days that wildfires occur on it's hot dry windy weather and those winds could easily sweep a drone into the engines of a manned aircraft which is a very bad thing so I yeah completely understand those the the safety reasons behind those regulations but I think there's got to be a way that we can account for both and enable both to happen well is there a temporal factor there if you if you have no aircraft on the fire are you still restricted from
• 29:00 - 29:30 putting your own drones up well exactly so I if I recall correctly the story I just told you was actually at nighttime um and water bombers don't fly at night at least here in Australia so there were no water bombing aircraft in the air um so it would seem logical that it would have been safe enough to put a drone up oh wait now you're trying to combine logical with government policies apologies yeah I I make that mistake
• 29:30 - 30:00 from time to time right uh so so if you could gather data on a fire you know if you had wind speed location direction topography fuel load fuel moisture humidity you know terrain wind movement etc uh and you could you know get lots of live data on a fire could you then develop an ideal suppression plan based on the resources that you have available and do you think that AI could develop a
• 30:00 - 30:30 better uh optimized suppression plan than people who say I've been doing this longer than you've been alive i would I mean I would always say like what we are in terms of what we are bringing um it isn't to replace it's to augment the decision makers um I think that's really important to to stress that we in no way shape or form think that the final decision should be taken by um certainly our application or
• 30:30 - 31:00 software application it should be done by a person with the experience and authority to do so um that being said the um where we actually originally started as a company was around the containment of a fire before we put all our focus into mitigation we thought that AI would be really useful around optimizing resource management and by resource management I mean the
• 31:00 - 31:30 application the appliances that you need to use in a fire around a fire so in particular heavy plant like um diggers and JCBs and water tankers and things like that having them effectively moved around the fire uh and being at the right place at the right time and doing the right job for the capabilities that that particular appliance might have was something that is
• 31:30 - 32:00 um is is good for is AI can do very very well um we'd seen examples of how it was done today in Australia where it was a physical map and yellow sticky notes to denote where that particular vehicle was at any one time and that was um you know it it it left the opportunity to optimize that uh fairly open and we thought yeah we could do we
• 32:00 - 32:30 could do a better job here and so that was where we originally started so we do believe um and it might be something that we go back and revisit that you know using AI to optimize resource management in containing fires could be uh something incredibly useful i think it's also important to remember it it doesn't matter how many resources you have when you get a situation like the LA fires you're not stopping them
• 32:30 - 33:00 it's it's about putting those resources in places to do property protection and try and prevent the impact of that fire on various properties as it passes over you're not optimizing to stop that fire at this point it's you know that that fire is out of control yeah so so at that point you have to prioritize the assets that are at risk and decide which ones you're going to extend resources to protect and which
• 33:00 - 33:30 ones are on exactly and and I guess I I you you spoke about um having all the data inputs and things like that one area where we could potentially apply that data to make really good decisions is which assets are defendable and which are not is it a waste of time sending resources to this asset to defend it because all those variables you mentioned the fuel loads the fuel moisture content the the terrain and
• 33:30 - 34:00 topography is it such that actually that asset just can't be saved um so definitely you could apply that data and make decisions there but that's also a a very difficult uh in many ways ruthless decision that nobody wants to have to make and and do we want machines making that decision on our behalf right i I I think they couldn't um and and hopefully the software would have a place for user input there so that you could order the the values at risk there
• 34:00 - 34:30 um but in all likelihood I would imagine that the resources that would we'd want to protect you know were in more hardened areas exactly optimal you know I would say if you have a fire coming through probably one of the the the key thing you might protect would be like a a nuclear power plant because the the consequences of that burning could be catastrophic yeah so this conversation has actually naturally kind of gone through the cycle
• 34:30 - 35:00 of thinking that's led us to where we are as a company it's come back around now to actually how do you harden before the fire starts and and this kind of it's that line of thinking that led us to build the software that we've built it's going through all of this all right how do you respond to the fire how do you make these decisions and then actually wait let's make a world where we don't have to make those decisions because we've prevented the fire in the first place and so you've got the nuclear power plant how do you harden it where will the worst fires come from how
• 35:00 - 35:30 do you ensure that you've mitigated against those in the first place yeah so Yes it was it was very much as you you probably can see Matt's passion for this particular problem um his passion around fire that made it a very easy decision back in July 2024 for me to jump on board and and try and do something about it um to to found eyesight such as we
• 35:30 - 36:00 did um even though it was something I know I knew absolutely nothing about and I don't claim to know a huge amount more other than what Matt has been able to um help me understand the passion does rub off however even even if I'm not a world expert on these things the passion to um you know do something about a very relevant problem means
• 36:00 - 36:30 that the 10 months that we've been doing this for has been really really hard work um it's been the hardest that I think both of us have ever tried at anything we've ever done before but it's been so enjoyable um in seeing the progress that we're making and we're really proud that when we demo things that we demo our tool to to potential customers or we're just showing it off just because we like to
• 36:30 - 37:00 show off the tool that that we get the feedback that that that we do so that makes us incredibly proud of of what we've achieved even so far um the big big challenge is that there's always more that you want to do there's always functionality features big goals that you want to add on to make life even easier for users to make you know the quality of the insights that we provide and the treatments that we then prescribe um to be um as legitimate and as good as
• 37:00 - 37:30 possible so it's been yeah it's been a joy the the first 10 months uh and long may it continue um interesting yes you Matt you said something that I've heard many times since the LA fires well you know by the officials there and other people echoed around the world you know no amount of resources
• 37:30 - 38:00 could have made a difference once the fires started we were going to lose it etc right and you know and that that's how it played out is that that just is to me that sounds like the people who said you know we'll never put a man on the moon and things that weigh more than air are never going to be able to fly and you know we're just kind of accepting um the the physics of the situation based on our current technologies and then we're extrapolating that into the future as if things will always be the way they are now and I know that you know the reason
• 38:00 - 38:30 that you're working on this project is because you believe things could be different if the technology were different so um you know you mentioned the X- prize earlier and the X-P prize says is to try to get you thinking outside the box and to say okay well this is impossible but if it were possible how would you go about it and then you test that and then lo and behold you start chipping away at the realm of impossible and making more and more of that possible so
• 38:30 - 39:00 uh if if we were to say you know well these things aren't impossible because they're they're physics from chemistry problems um although in fact they're they're social problems as well because even if you have a good solution you have to get adoption um but you know let me ask you just how would you go about if if we were to say okay actually this fire could have been stopped but here's what the resources would have to look like yeah that's a very good point you make
• 39:00 - 39:30 um I guess again it's it's fire size perspective with where we are where we are in our product life cycle right now that it could have been stopped before the event um that's what we try to do i think the the X-P prize perspective and I would agree with that as well is that there was a point even after it started in the very incipient phase where it could have been stopped had it been detected in that very early incipient
• 39:30 - 40:00 phase and attacked by autonomous aircraft or something like that definitely it it could have been stopped i guess there's the nuance when when people say there was nothing you could do about it it's once it's a raging inferno and as you say yeah with existing resourcing constraints um correct I can't think of any technologies that yeah I I I can't foresee even in into the future kind of right now off the top of my head any
• 40:00 - 40:30 technologies that will stop it at that point in time but you're right it the thought of landing on the moon would have been ludicrous a century ago so right or people who said you know during the black plague there's just there's nothing they can be done to stop it exactly because they hadn't they hadn't heard he heard of penicellin yet yes so it's not that there was nothing it's just that the thing hadn't been created yet exactly yeah very very good point so to me that's encouraging because it means there's there's room for answers for these things where if you simply say
• 40:30 - 41:00 no it can't be done then I guess then you're right i guess you don't want to have to stop the inferno either um you want to have stopped it beforehand one of the really interesting things about how that fire propagated was it was spreading from house to house as if they were bushes trees you know it's the the way kind of wildland fire typically propagates through the vegetation the houses were behaving like that
• 41:00 - 41:30 vegetation and because the houses are more flammable than the vegetation around them so yeah so it's I mean it it's not something that we we haven't seen propagation of fire that deep into urban environments in Australia as far as I'm aware in history um there was a fire in Canberra a couple of decades ago that did go a little way into the urban environment but the the LA propagation through the urban environment was just I
• 41:30 - 42:00 hate to use this word these days but unprecedented these two 200 kph winds don't help much either no that's exactly right um and so yeah envisaging the technology to prevent that once it's already happening uh at the moment is yeah you've got to have a pretty good mind's eye to picture that I think well one one thing you do is you stop building homes out of fuel well that's Yep we're back into preventative measures then right so you
• 42:00 - 42:30 you you you talked about the will you said there's a a strong will for you know suppression technology because of the glamour supposedly of that um you know and less for the the prevention side um and to me I guess that reflects the will of the people right is that people uh they want to spend money on putting a fires out and they don't want to spend money on prevention at least that's you know how I see the the US psychology you know no money for prevention endless
• 42:30 - 43:00 deep pockets for suppression Is is that the case uh in Australia as well yeah I mean there's definitely more of a a slant towards suppression for sure yeah one thing I would say though like in our conversations with whether it's uh potential customers um communities or whether it's partners when we're able to
• 43:00 - 43:30 show and explain why mitigation is incredibly important everyone to a person gets it and understands it so it's not something that's completely foreign to them that they are like I don't know why I'd be interested in this once you've got the visual visualization take them on a journey explain to them why prevention is better than cure then um yeah they they get on board really quickly so uh I think that you know with
• 43:30 - 44:00 the education piece as Matt mentioned earlier on uh that's really key in order to um get uh a greater appetite into mitigating against the risk of fire yeah that's a really good point um we've had people say to us things like your tool uses the data and machine learning to tell the story uh of why I need to do mitigation work so it's yeah great point
• 44:00 - 44:30 that Kev raises whilst the I think because it looks cool the public focus is typically on suppression [Music] um our efforts are around using data to say this is why you need to prevent it in the first place presenting those numbers presenting the predictions of what could happen if you don't mitigate um and actually using the modeling to to show people that risk it's Yeah people
• 44:30 - 45:00 get it they see the like two fires on your computer screen exactly one in a mitigated area one in unmitigated area and show the the the difference in the results yeah in the exact same place so we can do a here's what here's how the fire will propagate if you do no treatment and then we can put the treatment plan in prescribed burn for example and show in the exact same location the difference in behavior of that fire as a result and that's really powerful i mean I I'd say I'm probably in the majority of the camp here that
• 45:00 - 45:30 getting told something is less effective than v getting the visualization and showing the actual benefits of doing something like that so I think again that's possibly why when we get the chance to get in front of people and show what you know the the outputs of our tool um that gets them excited and bought in and we'll also project that out into the future for the users so we'll show you
• 45:30 - 46:00 you can load in an entire bushfire management plan for the area and then project out in the future and we can say on the 14th of May 2028 this is what the fuel loads will look like this is how the fire will behave under the plan that you've put together so a prediction of how it will behave under that plan and presumably how it'll predict under no plan exactly and compare the two yeah exactly right and and part of the reason So would I be able to see a model I'm
• 46:00 - 46:30 sorry sorry go ahead uh if if I'm the fire chief of a district uh based on the way conditions are right now uh you know mitigated partially mitigated unmitigated whatever they might be um you could show me how fire is going to behave to each day right each day I show up and I say okay if a fire start here this is what's going to do and then based on the probability of that I may pre-position assets yes uh
• 46:30 - 47:00 right so so that could be useful even even without mitigation just to have some good fire modeling yeah the the really cool thing about what we can do is as well is all we need an address we already have the data in terms of the topography the vegetation uh the weather so as long as similar to what you do in Google Maps if you give us your address to the location or property that is of interest to you we've had a number of times now that our
• 47:00 - 47:30 first interaction with a customer um is we we walk into the first meeting and demo our product to them of their property so where before Matt and I have worked in places where you need to have month month-long negotiations and get uh non-disclosure agreements and confidentiality agreements going and then they'll hand over the data that's prevalent to them and then we'll do a
• 47:30 - 48:00 demo um we have been able to uh demo specifically their asset back to them and that really helps sell the story because they it's not something that's etheral out there it's they're getting to visualize their property that they potentially might be sitting in when we demo it to them um so that's a really nice powerful way of um of uh building a story for them that they can appreciate and yes um I'm curious Matt um have you
• 48:00 - 48:30 ever taken the software or I don't know what what what stage of the development error but could you envision if you haven't done it already that you're on an active fire and everyone says "Come on right over here we're going to we're going to put a fire break in here and then you could look at your model and say we could trench out 10 meters there but the ember cast is going to just push on." Let me show you something here and you draw a line and you draw a line on the screen and you said but if we put the fire break here now watch what happens you know yes we lose another of
• 48:30 - 49:00 this but we save all of that so that could be used right for decision support on the ground about where you're going to put in fire brakes yeah absolutely I think at this point in time I might get laughed off the truck if I try to bring my laptop with me out on the fire ground but um absolutely we we have actually not not for active fires that I've been involved in but you just for validation of our models and and ensuring that it's working well just monitor the news and
• 49:00 - 49:30 the various fire websites for any active fires that are just starting in areas where it's you know bad weather it's looking like it could get bad load it into the software and see what it says is going to happen and then watch and yeah that's that's how we've validated in many ways that the model is actually accurate in what it's saying so you compare what the model says would have happened to what actually happened exactly you know how close you are
• 49:30 - 50:00 yeah sorry i was just going to say we've got a really good example of what Matt just said there there was actually um just a couple of hundred meters um uh down the road from where I live um there was an escaped hazard reduction burn so the the RFS were performing a hazard reduction burn and it actually jumped containment lines and uh threatened the local community and there was uh helicopters going AC over my property
• 50:00 - 50:30 and there was an evacuation order and uh my friends who was away at the time I had to go and rescue their dog and uh bring them back to here um however afterwards we were able to perform a retrospective where we knew the outline the behavior of the fire um we were able to plug in the weather conditions uh we knew what the uh topography was and then we uh fit um ran our fire spread model
• 50:30 - 51:00 and it was uncanny how the fire spread model followed the path of what the fire actually did 2 days before um that felt like a really big you know put your fist in the air moment and and celebrate because it really validated our our tool or or one aspect of our tool the modeling part of it that that we were on to something and it was legitimate
• 51:00 - 51:30 and how do you compare your approach to how other people who are working on a similar problem go about it so you know right now you've got you know uh fire prediction modeling from Technosila or from you know Ina or from Aurora or you know many people working on this project and thank thank goodness a lot of people are working on it because there's enough fire for everyone um but but is there a way that you distinguish um your strategy from those that other
• 51:30 - 52:00 people are using yeah it's [Music] um hard right now for us to compare to those we're primarily um monitoring Australian fires at the moment and Australian fire behavior models and fire spread is quite different to in the US totally different fuel types um and yeah different models typically used for those our predictions are from
• 52:00 - 52:30 physics-based models at the moment um building off kind of the state-of-the-art published models coming out of Australia for Australian vegetation our ambition is to very soon kind of do a a combined approach of the physical model and machine learning um we firmly believe that neither of those approaches are perfectly suited to solving this problem it's a physical model misses a lot of data that a
• 52:30 - 53:00 machine learning model could incorporate machine learning model is not perfect for it either because there are only a limited number of fires and they all have totally different conditions and it's not like you can just go out and light thousands of fires to get more training data and we know how hungry machine learning models can be for data so I don't think either of those approaches are perfect our goal is to augment them into a model that's better than any one of them um but yes at the
• 53:00 - 53:30 moment primarily focused on Australian vegetation types and using the Australian state-of-the-art so it's kind of hard to compare to the models that are more for the the US vegetation types okay so you could adapt to current model from the US if you were to change the the fuel properties absolutely the uh the US the the standard model that is used physical model that's used in the
• 53:30 - 54:00 US is easy to load into our software um but as I say it's then layering that machine learning on top of it is you know we'll we'll want to be very specific with the training data that we use for that um to ensure it accounts for the types of fires that you're seeing it predicting over and how close do you feel like you are to um you know commercially viable product
• 54:00 - 54:30 kev do you want to yeah we um so at this moment in time our our long-term goal is to offer this as a SAS product software as a service so that we provide our customers login details and they go and draw their own insights um the data is already loaded in in terms of vegetation fuel load etc or they can bring their own uh data and um they can go and yeah build a 5year 10
• 54:30 - 55:00 year treatment strategy as they see fit or they can go and build one for the next seven days it's really up to them um so we we're building it towards that end that it's really intuitive uh relatively simple to use so that a wide range of stakeholders can uh gain value from it for now um we offer it as a managed service where Matt and I will uh build the
• 55:00 - 55:30 projects and the insights and then present them back to customers and we actually have customers uh signed today that that are are gaining insights from our tool which we're really proud of um and uh yeah it's very much at this stage just building out so that in the future we can offer that SAS solution to them well it feels like you're in the right climate for that i I'd say on a per capita basis there's more wildfire or
• 55:30 - 56:00 bushfire innovation going on in Australia than most places yeah and and the really um happy circumstance from doing the manage platform is you're embedded with the customer understanding their processes understanding their wants and needs so it's actually really useful to build out a road map for the future as well like to validate there's no point in building something great that no one wants um so you want to
• 56:00 - 56:30 build something that people see value in and have a desire for and the fact that we get to hold our customers hands for now anyway really allows us to build out uh something that's going to be incredibly useful to them and to others in the future yeah fingers crossed matt do you want to add anything to that i It was coming through a bit choppy i'm sorry it was a bit hard to hear what Kev was saying
• 56:30 - 57:00 well I'm hoping that it when it uploads it'll all be beautiful know what you said sorry i want to thank you both so much for giving up an hour to to share with me and our viewers and listeners on the project that you're working on if folks want to get in touch with you Matt what's the best way uh our website uh hit contact us or um on LinkedIn or Alternatively
• 57:00 - 57:30 uh fi-s.ai that's fi dash sci aai wonderful well thank you both so much i'm really excited about what you're working on um I I hope you'll share some screenshots with me as the development moves forward and and and I suspect that at some point you'll you'll be working in partnership with some other people who are working on similar things so that you all have the the
• 57:30 - 58:00 efficiency of scale there and grow so best to both of you and if I find myself in your neighborhood I'll definitely stop by absolutely take care thank you Steve thanks so much kevin Bonner and Matt Roberts from fi fi-sci.ai check them out on the web you've been watching or listening to the fire break i'm Steve Wolf thanks so much for joining us right here a lot of
• 58:00 - 58:30 innovation out there if you have a technical project or some fire related thing that you'd like to see featured on the fire break please get in touch with me and we'll see you at the next one take care