Fireground Commander

The three findings that matter most

1. The market is split into two camps with a canyon between them. The cheap camp (SimsUshare, roughly $75 to $500 per year; Fire Studio, low hundreds to low thousands) overlays animated smoke and flame on static photos. No physics, no map, no consequences: the fire does what the instructor clicked, not what fire does. The expensive camp (XVR, ADMS, FAAC) sells serious 3D simulation to academies and the military at five to six figures, in generic virtual cities that are not your district. Nobody sells "command a dynamic fire in your own first-due at a price a single department can approve."

2. Real-time fire physics in a browser is feasible: the models are published and proven, the packaging is the work. NIST zone models have computed multi-room structure fire conditions faster than real time since the 1980s (CFAST itself since 1990, seconds per run). GPU cellular-automata wildfire simulators (PyTorchFire, JaxWildfire, 2025 papers) run faster-than-real-time wildfire spread using the same Rothermel spread model FARSITE uses. We do not need to invent fire science; we need to package published, validated models into a game loop. That is a software engineering problem, and a tractable one.

3. Your district is likely already modeled in 3D. Google's Photorealistic 3D Tiles API streams the same 3D imagery you see in Google Earth into web apps (via the open-source CesiumJS globe engine). Load it, fly to any address in your response area, and every rooftop, setback, exposure, and access problem is there. This is the "wow" that no competitor has: officers size up buildings they will actually respond to. One honest asterisk: photoreal coverage is strongest in metro areas and thins out rural, so "does it cover Skyler's district well" is a day-one Phase 0 check (the OSM fallback in the Tech tab covers gaps).

Honest caveats (so you trust the rest of this document)

• Physics credibility is the moat and the risk. "Realistic" must mean "consistent with NIST models and UL FSRI research," not "looks cool." The plan below includes validation gates for exactly this.
• Google tiles cost money per session and cannot be cached offline under their terms. Fine at training-budget prices; the plan includes an offline fallback using free government data.
• Fire departments buy slowly. Volunteer departments especially. The go-to-market leans on grants (AFG), state academies, and your own network of chiefs, not cold sales.
• This is a training aid, not a fire-prediction tool. Liability language and scenario disclaimers matter from day one.

The VR wave, and why it is a different laneNEW

FLAIM Systems (Australia, venture-backed via Breakthrough Victoria) sells headset VR with haptic heat vests and force-feedback hoselines; RiVR builds photogrammetry VR scenes and works with XVR at the UK Fire Service College. These train task skills: nozzle technique, search patterns, extinguisher use. Command training is a different discipline: the IC stands outside, reads smoke, and talks on a radio. Screen plus radio IS the fidelity that matters, and it avoids VR's real adoption killers for departments (headset cost per seat, motion sickness, hygiene between users, one-trainee-at-a-time throughput). When a chief asks "why not VR," the answer is: your ICs do not fight fire in a headset, they command from the street. Academies that own both use screen sims for command and VR for tasks.

The AI window (open, but not forever)NEW

The UK's Fire Service College opened an AI-powered "Immersive Skills Lab" in 2025 (a facility, not a sellable product). US fire media is saturated with AI think-pieces, and instructors already use image generators for smoke-reading slides. But there is no AI-native, physics-driven, your-district command simulator on the US market yet. The ingredients (3D tiles, zone models, GPU fire CA, AI voices) all became practical within the last 24 months, which is exactly why the incumbents have not shipped one. That window is the reason to move at Phase 0 pace now: the first credible product to combine these owns the category narrative.

The core loop (trainee experience)

You are the first-arriving officer. The screen is your district, real 3D imagery, the actual building. Dispatch audio sets the scenario. You do a 360, give your radio report (voice or structured UI), establish command, and start assigning companies as they arrive on your real running order. The fire does not wait for you: it grows on a clock driven by the physics engine. Vent the wrong place, it accelerates. Delay water, flashover. Good decisions are rewarded by the model, not by an instructor's mercy.

Feature pillars

• Anywhere scenarios: type any address in your response area; the sim loads that building/terrain in photorealistic 3D (Google 3D Tiles via CesiumJS). Draw a footprint, tag construction type and occupancy, done.
• Living fire: compartment fire model for structures (stages: incipient, growth, flashover, fully developed, decay; ventilation-driven behavior; suppression vs required fire flow), Rothermel cellular automata for wildland/grass. Smoke and flame visuals keyed to model state (smoke color/volume/velocity telegraph the fire like the real thing).
• Inject engine: seeded-random event deck: trapped occupant reports, mayday, collapse warning signs, wind shift, hydrant failure, EMS surge, second alarm delays. Weighted by scenario and difficulty; instructor can force or veto any inject. Same seed = same drill for grading fairness.
• Real command structure: ICS/NIMS-correct UI: command establishment, division/group assignments, tactical worksheet, PAR checks, radio log with timestamps. Terminology aligned to NIMS and Blue Card conventions so it reinforces, not fights, existing training.
• Multiplayer: instructor as dispatcher and scenario controller, trainee as IC, optional peers as company officers on a shared live map. Browser-based, no install: a volunteer department drills from home on a Tuesday night.
• After-Action Review that sells the product: full timeline replay, decision log vs model state (water on fire at T+6:40 vs flashover curve), benchmark scoring on the tactical priorities every officer recites (life safety, incident stabilization, property conservation), exportable PDF training record per NFPA 1021 job performance requirements. Those records feed department ISO documentation, which is a budget argument, not just a training argument.

Ecosystem integrations: plug into what departments already haveNEW

• First Due (pre-incident planning): thousands of departments keep their pre-plans, building data, and NFPA 291 color-coded hydrants in First Due, importable via CSV/ArcGIS. FireSim scenario authoring should import those pre-plans: "drill the building exactly as your pre-plan describes it" turns their existing data investment into our content library. Long-term, First Due is also the most logical acquirer or distribution partner in the space (and the one to watch competitively: they consolidate everything).
• NERIS (national incident data, sole system since Feb 2026): modern APIs replace the old NFIRS. A department's own incident history can drive the scenario mix ("you ran 61% single-family, 12% garden apartment last year; here is a drill calendar to match"). Evidence-based drill planning is a line no competitor can say today.
• Vector Solutions LMS (TargetSolutions): the dominant fire-service LMS has a Training Records API and an ISO Training Tracker product. FireSim pushes AAR training records straight into the LMS and ISO tracker the department already uses. Training officers adopt tools that reduce paperwork, not add it.
• Blue Card CTC ecosystem: certification = 40-50 hr online course + 3-day simulation lab + continuing education on a 3-year recert cycle, and Train-the-Trainer lets departments run their own Command Training Centers (dedicated sim rooms with 10 workstations exist at Phoenix, South Metro, Oceanside, Mountain View...). FireSim gives that kind of program its practice environment without the dedicated room: any classroom, any laptop. To be precise: cert/recert credit runs on Blue Card's own licensed sims, so we position as the drill sandbox alongside their curriculum, never as a substitute. Alignment first, formal partnership later with traction.

AI voice role-play: the Phase 2/3 differentiatorNEW

The loneliest part of command training is that one IC needs four role-players (dispatch, interior crews, incoming officers, occupants). Priority Dispatch's AI SkillLab and VoiceDispatch Lab already sell AI-voice simulation commercially to 911 dispatch centers: trainees talk, the AI caller responds in real time. The same stack (speech-to-text, LLM role-play constrained by the sim state, text-to-speech) applied to fireground radio means a company officer can run a full command drill alone at the kitchen table: AI dispatch answers, AI engine companies acknowledge and report conditions from inside the physics model, simulated occupant reports raise the stakes. Nobody offers this for fire command today. The sim's structured radio UI ships first (gradeable, simple); AI voices layer on top of the same event system.

Structure fires: zone-model approach (NIST heritage)

NIST zone models have simulated multi-compartment structure fires since the 1980s; the current one, CFAST (Consolidated Fire and Smoke Transport, released 1990), runs faster than real time, typically finishing a whole-fire run in seconds. That is the existence proof: physically grounded structure fire modeling is cheap enough for a game loop. Our engine follows the same architecture, simplified where training value permits:

• Building as a compartment graph: rooms are nodes; doors, windows, and breaches are edges (vents). Tagging a real building takes the scenario author minutes.
• Fire growth: industry-standard t-squared growth curves (slow/medium/fast/ultrafast per fuel package), modulated by oxygen availability: the modern ventilation-limited fire behavior UL FSRI has drilled into the fire service for a decade. Open a door, feed the fire; coordinated ventilation with water works, uncoordinated ventilation kills.
• Flashover: upper-layer temperature threshold via the published MQH correlation; the sim's flashover timing stays inside envelopes from published test burns.
• Suppression: applied flow vs required fire flow (National Fire Academy formula, length x width / 3, adjusted for involvement). Hit the required flow, knockdown follows the curves; short of it, steam and stalemate.
• Collapse and mayday: probabilistic timers keyed to construction type (lightweight truss vs legacy), burn time, and load, grounded in the UL/NIST lightweight-construction burn research; collapse-zone standoff doctrine then governs how the sim scores exposure of crews when warning signs appear.

Wildland/grass: Rothermel + cellular automata (GPU-proven)

The Rothermel surface spread model (1972) still powers FARSITE/FlamMap, the federal wildfire tools. Recent open research (PyTorchFire, JaxWildfire, 2025) runs Rothermel-driven cellular automata on GPUs at real-world scale, faster than real time. Our wildland module is a grid CA over the same terrain tiles the map already loads: fuel model, slope, wind per cell. This gives Skyler's "fields" scenario type nearly free, and it is the flashiest demo mode.

System shape

Phase 0: The Demo That Sells The Vision (4-6 weeks, nights/weekends pace)

• CesiumJS + Google 3D Tiles flying to any address in your district.
• One hand-tagged building (pick a real target hazard you know cold).
• Structure fire engine v0: growth stages, one or two vents, smoke/flame particles telegraphing state.
• Size-up drill mode: 360 the building, timed radio report, three command decisions, one inject, simple score.
• Exit test: you show it to two chiefs and they say "when can my officers use this."

Phase 1: MVP a department will pay for (12-16 weeks after Phase 0)

• Scenario authoring: address search, footprint draw, construction/occupancy tagging, hydrant placement.
• Full single-family + commercial strip scenarios; engine validation Gates 1-2 green.
• Inject deck v1 (8-10 event types), difficulty tiers, seeded replays.
• Instructor mode (force/veto injects, pause, voice notes) + trainee IC mode, 2-player over WebSocket.
• AAR v1: timeline replay + benchmark report, PDF export for training records.
• Pilot with your department and 2-3 friendly neighbors, free, in exchange for structured feedback and testimonial rights.

Phase 2: The sellable platform (following 2 quarters)

• Multi-company multiplayer (peers as company officers), PAR/mayday drills, radio discipline scoring.
• NEW First Due pre-plan/hydrant import (CSV/ArcGIS) so scenarios auto-populate from data departments already maintain; Vector LMS Training Records API push so AARs land in the LMS/ISO tracker automatically.
• NEW AI voice role-play v1: AI dispatch + AI interior-crew radio reports driven by live sim state (speech-to-text grading of the trainee's transmissions comes with it). Commercial precedent: Priority Dispatch AI SkillLab in the 911 market.
• Building type library rounding out Blue Card's five types alongside the MVP's residential and commercial: multi-unit residential (garden apartment), strip mall, commercial, big box, plus legacy construction variants (balloon frame), so drills slot into existing curricula.
• Wildland module (Rothermel CA) for grass/interface scenarios: your "fields" case and the western-state sales hook.
• Org accounts, training-records dashboard, ISO documentation exports; AFG-grant-ready quote packets.
• Validation Gate 3 (IC blind review) becomes a public advisory board: names on the website.

Market math (NFPA / NVFC / FEMA figures)

• 29,452 fire departments in the US (NFPA profile). 82% all- or mostly-volunteer; those departments protect ~30% of the population and have the least access to live-burn training, which makes simulation MOST valuable exactly where budgets are smallest. Browser-based + grant-fundable is the only model that reaches them.
• Career/mostly-career departments (18%, ~5,300) have training divisions, budgets, and ISO incentives: the early revenue.
• FEMA AFG grants explicitly fund training and training technology; response software and computing devices are eligible categories. "We help you write the AFG line item" is a sales feature.
• Price anchors: SimsUshare normalized departments to paying for sim software (good for us) at $75-$500/yr (bad ceiling for them, fine floor for us); XVR/ADMS anchor "serious sim" at five figures. $1,500-$3,000 per department per year sits in empty space: less than one live-burn day, less than one conference trip.
• Even modest capture is a real business: 500 departments x $2K = $1M ARR. 2% of US departments (~590) = roughly $0.9M-$1.8M ARR across that price band. International (UK, AUS, CA fire services all use English-language ICS variants) is upside.

The ISO/insurance argument: training software that pays for itselfNEW

Verisk's Fire Suppression Rating Schedule scores every US fire department (the PPC class, 1 to 10), and that class feeds directly into the community's property insurance premiums. Training is a scored line item with specific, documented hour requirements: 12 hours/year of officer continuing education for maximum officer credit, 192 hours/year of company training (credited up to 16 hours per member per month), plus driver and recruit requirements. Two consequences for us:

• Every FireSim drill generates a timestamped, exportable training record formatted to support ISO training documentation. Chiefs do not buy "simulation," they buy "documentation that protects our PPC class." Vector Solutions built an entire ISO Training Tracker product on this pain. One diligence item before we print it on the website: confirm with Verisk how the FSRS credits computer-simulation hours vs facility drills (it varies by category), then market exactly what is creditable.
• The pitch to a city council becomes financial: a worse PPC class raises constituents' insurance premiums. $2K/year of software that helps defend the training score is the cheapest line in the budget.

Efficacy evidence (the tender-response ammunition)NEW

Procurement asks "does simulation training work?" The answer file: DHS S&T funded development of an incident commander training simulator and published on it; a performance-based multimedia IC training program documented 600+ firefighter completions; USFA Executive Fire Officer papers study simulation-based command effectiveness; and the fire service's own literature (FAAC, MFRI's dedicated Incident Command Simulation Center, aspiring-officer programs) consistently reports gains in decision-making, communication, and resource allocation, precisely because sims deliver the high-impact/low-frequency incidents officers rarely see live. We cite these, then add our own AAR data as pilots complete.

Go-to-market (in order)

• 1. Skyler's network. Pilot department + neighboring chiefs + your training-officer contacts. Fire service buys on peer trust more than any market I researched.
• 2. Conference demo circuit. FDIC and state chief associations; the demo IS the marketing (officers line up to size up their own town).
• 3. Grant channel. AFG application-season content + templates; regional AFG consultants as a referral channel.
• 4. Academies and Blue Card ecosystem. Position as the practice sandbox for their curriculum, not a competitor. NEW Concrete wedge: departments with Blue Card Train-the-Trainer status run their own Command Training Centers and need fresh sim content for the 3-year recert cycle; physical CTCs (10-workstation rooms at Phoenix, South Metro, Oceanside, Mountain View) prove departments already budget for command sim infrastructure. We are the CTC that fits in a classroom they already have. South Metro Fire Rescue, a Blue Card flagship, is 20 minutes from Grant's house.

Skyler (subject-matter authority)

• Pick the Phase 0 demo building: a target hazard in your first-due you know intimately.
• Write the "truth table" for one bread-and-butter scenario: single-family, kitchen fire, evening. What should happen at T+2, T+5, T+8 with good vs bad decisions? This becomes the engine's first acceptance test.
• List the 10 injects that most separate good ICs from bad ones, and what a correct response looks like for each.
• Name 3-5 experienced ICs who would blind-test builds and eventually form the advisory board.
• Sanity-check the NFPA 1021 benchmark list for the AAR report.

Grant (engineering)

• Stand up the CesiumJS + Google 3D Tiles proof-of-concept over Skyler's district (Google Maps Platform key, metered).
• Build fire engine v0 (compartment graph + t-squared growth + vent logic + seeded injects) with the automated CFAST-comparison test harness from day one.
• Wire smoke/flame particles to engine state; ship the size-up drill loop.
• Then: the two-chiefs demo. Decision point for Phase 1 commitment.

The landscape, honestly stated

• Program status caveat: SBIR/STTR went through a five-month congressional lapse with roughly $4B frozen; agencies are restarting on varying timelines through summer 2026. This is a "watch and pounce" situation, not a guaranteed check.
• NSF SBIR reopened with $250M for deep-tech startups. First step is a short Project Pitch (a few pages, no full proposal): first deadline July 27, 2026 (re-verify on nsf.gov before writing; post-lapse dates have moved). Fundability caveat: SBIR wants genuine R&D risk, and "we integrate published models" is a classic rejection rationale. The pitch must lead with the actual research problem (real-time, validated, reduced-order fire modeling coupled to trainee decisions and graded against full-fidelity models), which happens to be true. Submitting costs only the writing time.
• DHS SBIR is the thematic bullseye (they literally have a published incident command training simulator project) and restarts through summer 2026. We watch their topic releases and respond when a first-responder training topic drops.
• Requirements to have ready: a US-based small business entity (the LLC we already need), majority US-citizen ownership, and a registered SAM.gov / SBIR.gov profile. All cheap to set up now, slow to set up under deadline.
• Customer-side grants stack with this: AFG for departments buying the product, and USDA's Community Wildfire Defense Grants for WUI preparedness (relevant once the wildland module ships, especially in Colorado, which just adopted the 2025 Wildfire Resiliency Code).

Market and competitors

• SimsUshare · pricing · CTC
• Digital Combustion / Fire Studio 7
• ETC Simulation ADMS
• FAAC fire simulators
• Blue Card command training
• freefiresimulator.com
• Fire Apparatus Magazine: fireground simulators overview
• SafetyCulture: fire training software roundup (2026)

Fire physics

• NIST CFAST technical guide (SP 1026)
• NIST Advanced Fire Modeling program
• PyTorchFire: GPU-accelerated differentiable CA wildfire simulator
• JaxWildfire: GPU wildfire simulator for RL
• Forest fire spread via cellular automata (Natural Hazards)
• FirefighterNation: using fire models to understand fire behavior

Mapping platform

• Google Photorealistic 3D Tiles docs · Map Tiles API policies
• Cesium + Google 3D Tiles guide · Cesium ion ToS
• Google FAQ on 3D Tiles

Standards, market size, funding

• NFPA US Fire Department Profile (29,452 departments)
• NVFC Volunteer Fire Service fact sheet (Mar 2024 PDF, since taken down by nvfc.org)
• FEMA Assistance to Firefighters Grants (AFG)
• NFPA 1410 training standard · NFPA 1021 Fire Officer (now part of consolidated NFPA 1020) · NFPA 1041 Instructor (also since consolidated)

v2 additions (7/12)NEW

• Verisk ISO PPC training criteria · Vector Solutions guide to ISO training requirements (12 hr officer / 192 hr company) · Vector ISO Training Tracker
• UL FSRI Coordinated Fire Attack study · FSRI ventilation tactical considerations · three new coordination tactics
• Google Map Tiles API usage and billing ($6/1,000 root tiles, 1,000 free/mo, ~3 hr per session)
• DHS SBIR program · DHS-funded Incident Command Training Simulator publication · SBIR topic listings · SBIR 2026 lapse/restart status
• First Due pre-incident planning · First Due hydrants (NFPA 291, ArcGIS import)
• USFA: NFIRS sunset · NERIS FAQ (sole national system as of 2026, modern APIs)
• Vector LMS Training Records API
• Blue Card program · Oceanside CTC · Mountain View CTC · Blue Card at South Metro (EFO paper)
• FLAIM Trainer · UK Fire Service College AI Immersive Skills Lab
• Priority Dispatch AI SkillLab · VoiceDispatch Lab
• USDA Community Wildfire Defense Grants · Colorado 2025 Wildfire Resiliency Code · USFA WUI training
• MFRI Incident Command Simulation Center · USFA EFO: incident command effectiveness

v1 compiled 2026-07-05; v2 deep-research pass 2026-07-12. Prices, deadlines, and program rules should be re-verified before any purchase or filing decisions (SBIR timelines especially are in flux after the 2025-26 lapse).