Careers · Wildfire

A big fire runs on people with laptops.

Not just people with shovels. Somebody maps the fire, somebody predicts where it goes tomorrow, somebody reads the infrared at 3 a.m., somebody orbits overhead running the airspace, somebody stands in front of a frightened town and tells it the truth, and somebody walks the burn afterwards to work out what the soil will do when it rains. These are jobs. Most of them are computing jobs. Here is what they are and how you get them.

Inside the planning tent at a wildfire incident command post. A man in a ball cap sits at a folding table in a canvas wall tent, typing on one of two open laptops, with a power strip and bundles of orange and yellow network cables running across the table beside him. A printed incident command form, ruled into hand-labelled boxes, lies across the table in the foreground.
This is what “planning” looks like on a fire. Not a war room. A canvas wall tent, folding tables, two laptops, a power strip, and a bundle of network cable someone ran across the dirt. The form on the table is ruled into boxes by hand. Everything else on this page happens in a tent like this one. · Neal Herbert / U.S. National Park Service (NPS), Yellowstone National Park · Public Domain Mark 1.0
01 · the job, in one figure

Run it many times. Count. Divide.

When a fire is going to burn for weeks, someone has to say where it might be next week — past the point where a weather forecast is any use. That job is the Long Term Fire Analyst, and the model they run is FSPro, which “predicts probable fire growth beyond the timeframes of reliable weather forecasts (3 through 5 days) using historical climatological data.”

How does it turn a pile of simulations into a probability? Here are the people who built the model, in the peer-reviewed paper describing the method:

“Probability was estimated simply as the fraction of the total simulations where fire arrives at a particular cell.”Finney, Grenfell, McHugh, Seli, Trethewey, Stratton et al. · A Method for Ensemble Wildland Fire Simulation · Environmental Modeling & Assessment (2011) 16:153–167

Read that as a student. Count the simulations where the fire got there. Divide by the total. That is frequency as an estimate of probability — the first thing a Monte Carlo lesson teaches, and the entire method behind a product people use to decide whether to move a fire camp.

Now the part that should make you sit up. Every one of those scenarios is independent — no run needs to know anything about any other run. There is a name for that shape of problem in computing: embarrassingly parallel. You can run them all at the same time. So that is exactly what the authors did:

“The ensemble simulation system was developed for shared-memory computers and parallelized with multithreading among the independent Monte Carlo scenarios. Computers used for the simulations contained 16 or 32 processors with 32 to 64 GB of shared memory.”Finney et al. (2011)

That is the Supercomputing Challenge's founding premise, in a peer-reviewed journal, describing this job. Not a metaphor and not our sales pitch — the model's own authors reached for parallel hardware because the problem handed it to them. This is why the work wants a supercomputer.

Burn probability, by counting
SIMULATED — a teaching toy, not an operational model
p = 0p = 1
The selected cell

The estimate as runs accumulate
N = 1N = 1
What this computes. A cellular-automaton fire on a 45×45 grid, ignited at the circled cell. Every run samples its own weather — wind direction, wind speed, dryness — then spreads to neighbours with a probability that depends on whether the neighbour is downwind. Runs are independent. Each burned cell adds 1 to its counter, and the colour is that counter divided by the number of runs. Nothing else. The seed is fixed, so this is identical on every machine.

What it is not. It is not FSPro. No terrain, no real fuels, no real weather, no spotting, no suppression, and the fire stops at the edge of the grid. The method — sample, run, count, divide — is the part that is real, and the part worth learning.

Click the ignition point: 1.000. It burned in every run, because that is where the fire started. Click far upwind: near zero. Neither is interesting. The interesting cells are the ones in between, where an analyst has to look an Incident Commander in the eye and say "about half the scenarios burn this place." Learning to say that — not "it will burn," not "it's fine" — is most of the job.

At N = 1 the figure refuses to answer, because with one run the only available numbers are 1.000 and 0.000 and neither is a probability. That refusal is the lesson.

And that is not a classroom nicety. NWCG's own course for the strategic planners who work off this model's output says the quiet part out loud:

“The FSPro model outputs can be easily misinterpreted, so it is critical for the SOPL to work closely with the LTAN.”NWCG · S-482, Strategy and Analysis for Complex Incidents

Read that carefully, because it tells you what the job actually is. The analyst is not there to run the model. A computer runs the model. The analyst is there to be the person in the room who knows what the number does and does not say — and who can stop a planner from hearing “less than half” and writing down “no.” That judgement is the thing you are training every time you argue with your own output.

02 · the roles

Who is at the fire

Codes are the mnemonics these jobs are actually ordered by on an incident. If you say “we need a GISS” on a fire, everyone knows what you mean.

LTAN

Long Term Fire Analyst

Planning

The figure above, for real. “Provides probabilistic and deterministic information on long-term fire advancement, fire behavior, and spread direction.” Reports to the Situation Unit Leader or Planning Section Chief; may supervise other LTANs, RAWS technicians, or Technical Specialists.

Your project already does this Monte Carlo ensembles, probability, parallelism. If you ran a model many times with sampled inputs and reported a distribution instead of a single number, you have done this job's core method.
How you qualify S-491 (Intermediate National Fire Danger Rating System) and S-495 (Geospatial Fire Analysis, Interpretation and Application). Supporting training added January 2026: M-410 and N-9035 (RAWS).
FBAN

Fire Behavior Analyst

Planning

The short horizon: what will this fire do today and tomorrow, in time for the morning briefing. The work is “collecting weather, fuels, and other relevant localized data to develop strategic and tactical fire behavior information, predict fire growth, and interpret fire characteristics.” Reports to the Situation Unit Leader or Planning Section Chief, and may provide leadership for Incident Meteorologists, other FBANs, Fire Effects Monitors, Air Resource Advisors and RAWS technicians.

Your project already does this NetLogo fire models and cellular automata — the most-built model in Challenge history. Spread as a local rule applied over a grid, driven by wind and fuel, is genuinely how this class of model reasons. The FBAN's advantage over your project is calibration and consequence, not concept.
How you qualify S-491 (Intermediate National Fire Danger Rating System), required as of January 2025. Your sponsoring Fire Management Officer holds the current course list.
GISS

Geographic Information System Specialist

Planning
A GIS specialist in a grey ball cap, seen from behind, sits at a laptop inside a tent at an incident command post and looks up at a large screen mounted on the tent frame. The screen shows a mapping program: a dark terrain and satellite basemap overlaid with a bright fire perimeter line and labelled division breaks marking sections of the fire.
Neal Herbert / NPS, Yellowstone · Public Domain Mark 1.0

The incident's mapmaker: “responsible for providing timely and accurate spatial information about the incident to be used by all facets of the Incident Management Team and the national coordination system.” Every map in the briefing package, every perimeter, every progression map showing what the fire did overnight — right, and done before the briefing. Reports to the Situation Unit Leader.

Your project already does this The school-bus routing lesson routes across a real directed road graph built from OpenStreetMap: 4,141 ways, 21,609 nodes, one-way streets honoured. And its oneway tag has four states, not two — “yes” (3,717 ways), “no” (110), blank (312 nobody tagged), and “-1” (2 ways that run backwards from how the line is drawn). That gap between what the data says and what the world is is the GISS's daily problem.
How you qualify ICS-100 and IS-700 as prerequisites, then S-341 (GIS Specialist for Incident Management) plus RT-341, an annual refresher. Then a Position Task Book, opened with a sponsoring Fire Management Officer before you work incidents: a qualified GISS watches you and signs off tasks, trainee → qualified, “typically four to five incidents over multiple years.” Then a red card through IQCS (federal) or IQS (states). An agency sponsor is required — federal, state, or a local fire department. Standards are NWCG PMS 936; the software is ArcGIS.
SITL

Situation Unit Leader

Planning

The data-fusion job, and the one the analysts report to. “Responsible for overseeing map production, and for fire behavior, fire weather, and incident status inputs.” Infrared imagery, field observers, weather feeds, GIS layers, model output — many sources, different formats, different reliability, arriving at different times — and by the briefing there is one picture, and it is the true one. Supervises “any configuration of” FBAN, LTAN, FOBS, IRIN and GISS. Reports to the Planning Section Chief.

Your project already does this Data fusion, visualization, and the judgement about sources. Every team that pulled data from several places and had to decide which one to believe when two disagreed has met the real difficulty of this job. The rest is doing it at 4 a.m. with a briefing at 6.
IRIN

Infrared Interpreter

Planning
An infrared scan image of a wildfire, of the kind interpreted to map a fire overnight. Against a near-black background, ragged clusters of orange, yellow and white mark heat picked up by the infrared sensor, scattered in patches and streaks across the burned area rather than forming one solid flame front.
An infrared scan — the raw material. U.S. Forest Service, Pacific Northwest Region · Public Domain Mark 1.0

Finds the heat. Fires are flown at night, and the imagery comes back to someone who has to say where the fire actually is — “interpreting infrared (IR) imagery from National Infrared Operations (NIROPS) sensors to determine the heat perimeter of fires.” The output is real files people use by morning: “GIS-ready files, GIS layers, geo-referenced PDF maps, KMZs, and logs.”

Look at the scan. NIFC describes the job better than we could: a trained interpreter “can take a thermal image of splotches, wiggles, and dots and locate hotspots, hand lines, dozer lines, and rocks, to say nothing of obvious terrain features like roads, streams, ridges, and buildings, with extreme accuracy.”

The duties list is a computer-vision syllabus. “Identify heat perimeter of fire(s), as well as areas of intense, scattered, and isolated heat.” And then the part every student who has debugged a pipeline will recognise: “Review infrared imagery for issues, including incomplete file transfer, ortho-rectification, false positives, coverage, and image quality.”

The IRIN “works remotely and often independently” but reports to the Situation Unit Leader — or, if none is assigned, the Planning Section Chief.

Your project already does this The laser lab is this job at classroom scale — thresholding, blob detection, tracking a target across frames, and arguing with your own false positives. Those are not algorithms like the ones used on infrared fire imagery. They are the same algorithms.
How you qualify — read this one twice Required training is ICS-100 and IS-700. That is the whole list. Both are free and online. Then a Position Task Book (PMS 311-26) signed off on real incidents. Physical fitness required: none. The work is done remotely.
Note what that means. Of every job on this page, the one that is pure computer vision has the lowest barrier to entry — two free courses and someone willing to sign your task book. S-443, Infrared Interpreter exists, but it is listed as supporting training, not required. It is widely miscited as mandatory. It isn't.
UASD

Unmanned Aircraft Systems, Data Specialist

Air Operations
Three wildland firefighters in blue hard hats and sunglasses stand in dry grass on a rocky, oak-covered hillside, all looking up at the sky at something out of frame. The man in front holds a drone controller with a tablet clipped to it and a thin antenna angled upward; the man beside him holds a red handheld radio.
Rebecca Paterson / DOI · Public Domain Mark 1.0

The drone job — and read NWCG's own description closely, because the word it uses matters: the UASD “manages, synthesizes, and disseminates intelligence collected by agency and contracted unmanned aircraft systems (UAS) for wildland fire incidents,” to “provide tactical and strategic information for incident command teams.” It works alongside a UAS Pilot or for a UAS Manager, and it sits in Air Operations, not Planning — the imagery work on a fire is not all in one place on the org chart.

Your project already does this The same imagery pipeline as the laser lab, from a different sensor at a different altitude. Supporting training for this job includes S-341 — the GIS course — and S-443, the infrared course. The catalog is already telling you these are one skill.
How you qualify ICS-100, IS-700, IS-800, L-180, RT-130, S-130 (Firefighter Training) and S-190 (Introduction to Wildland Fire Behavior), then a Position Task Book (PMS 311-127). Supporting training includes A-100, A-109, A-200, S-341 and S-443.
ATGS

Air Tactical Group Supervisor

Air Operations
The view from inside an air attack plane circling a wildfire. The aircraft's white wing juts across the lower left of the frame, and below it a tall column of tan smoke boils up from fire burning along a steep mountain front, with a green valley and the edge of a town visible beyond and cumulus clouds in the blue sky above.
The office. Rachel Nielsen / BLM · Public Domain Mark 1.0

Orbits above the fire and runs the airspace. The ATGS “coordinates incident airspace, manages incident air traffic, and is the link between ground personnel and incident aircraft” — “an airborne firefighter who coordinates, assigns, and evaluates the use of aerial resources in support of incident objectives.” When they're activated, the ATGS supervises the Leadplane Pilot, the Aerial Supervision Module and the Helicopter Coordinator.

Reports to the Incident Commander on initial attack; to the Operations Section Chief or Air Operations Branch Director on bigger incidents.

Your project already does this Real-time allocation under hard constraints. Several aircraft, one piece of sky, limited fuel, changing priorities on the ground — assign the right resource to the right task, right now, and don't let two of them want the same airspace. That is the assignment problem with a deconfliction constraint and no undo.
How you qualify Crew Resource Management training, S-378 (Aerial Supervision), RT-130, and RT-378, a refresher every three years. You also need to have already performed as a Division/Group Supervisor or an Incident Commander Type 3, plus the ATGS Position Task Book (PMS 311-18). This one is not an entry point — it is where a fire career arrives after years on the ground.
FEMO

Fire Effects Monitor

Planning
A fire effects monitor in a yellow Nomex shirt and yellow hard hat sits on a high ridge, writing on a clipboard with a pen. Below and in front of her, a broad column of white smoke rises out of forested ridges from a prescribed burn, filling most of the sky.
NPS · Public Domain Mark 1.0

Ground truth, collected by a human being standing in it. The FEMO collects “incident status information from personal observations at the incident” — “fire perimeter location, onsite weather, fire behavior, fuel conditions, smoke, and fire effects information needed to assess firefighter safety and whether the fire is achieving established incident objectives.”

Your project already does this Measurement, and the discipline of it. Every model on this page is eventually checked against what a FEMO wrote down. A model nobody validates is a rumour with a colour ramp.
How you qualify RT-130 and S-290 (Intermediate Wildland Fire Behavior), then performance as a Firefighter Type 2 plus the FEMO Position Task Book (PMS 311-30). Fitness level: moderate — unlike the analyst seats, this job is on the fireline.
PIOT

Public Information Officer, Technician

Command
A public information officer in a yellow Nomex shirt and sunglasses stands in front of a long row of large canvas tents at a wildfire incident command post, talking and gesturing to a group of children and adults who have gathered on the dry grass in front of her to listen.
Ashley National Forest, USFS · Public Domain Mark 1.0

The person who tells the public the truth while it is still happening. A PIO “participates in the collection and dissemination of incident information and may be the first public point of contact” — information centres, community engagement, media. On a large incident the PIOT reports to the lead Public Information Officer as part of the Command Staff.

Your project already does this Explaining a model to people who will act on it. Every Challenge team presents to judges and has to answer "so what does that actually mean?" without hiding behind the maths. That is this job. If you write well and stay calm when people are frightened, this is a real career and almost nobody tells students it exists.
How you qualify — the open door ICS-100, ICS-200, IS-700, RT-130 and S-110 (Basic Wildland Fire Orientation). Required experience: none. NWCG says it plainly: “satisfactory completion of the required training meets the position qualification requirements.” No task book.
The ordering codes are PIOT, PIO3 and PIOC (Complex).
BAER

Burned Area Emergency Response

Post-fire · a program, not an ICS code
A soil scientist in a yellow Nomex shirt, green trousers and a white hard hat kneels on blackened ground in a burned pine forest, reaching out with a small hand tool to examine ash and a charred branch. A red pack and a tan gear bag rest on a boulder beside her, and scorched tree trunks and dark rocks cover the slope behind.
Ground-truthing the satellite. USFS · Public Domain Mark 1.0

The job most people don't know exists. When the fire is out, the hazard isn't: a burned watershed sheds water instead of absorbing it, and the next storm can put a debris flow through a town that the fire never touched. BAER teams “rapidly evaluate the burned area and prescribe emergency stabilization treatments” — to “minimize threats to life or property resulting from the effects of a fire.”

They assess “presence and extent of water-repellent soils,” “debris-prone areas and downstream effects,” “potential for increased erosion or sedimentation,” and how much of each watershed burned at high severity. The assessment “usually begins before the wildfire has been fully contained,” and emergency stabilization happens within one year of containment.

A BAER assessment team is a science team. Its named roles include BAER Hydrologist, Engineer, Soil Scientist, Geologist, Botanist, Biologist, Forester — and BAER GIS Specialist.

Your project already does this Remote sensing plus hydrology. Teams work from a BARC — a Burned Area Reflectance Classification, “a satellite-derived data layer of post-fire vegetation condition” with four classes: high, moderate, low, unburned. Then comes the part that is pure science literacy: BARC is an input, not the answer. It measures what happened to the vegetation, and the team goes to the ground to turn it into a soil burn severity map. Knowing the difference between what your sensor measured and what you actually want to know is the whole job — and it is the same lesson as the one-way street that OpenStreetMap never tagged.
How you get on a team There is no ICS mnemonic for this one — BAER is a Forest Service and Interior program. You get on a team by being the scientist: “completion of FS or Interagency BAER team member course and/or prior experience as a BAER assessment team member.” If the fire is still uncontained you also need fireline basics — S-130, S-190, the annual safety refresher, and a light fitness rating.
03 · read this one twice

“Infrared Interpreter” is a fossil

Stephen Guerin · who has worked these incidents

“Infrared Interpreter is so limiting and legacy. Imagery Intel Analyst is what that should morph to.”

He is pointing at something structural. Look at how the catalog names these jobs: Infrared Interpreter. UAS Data Specialist. The names are sensors. That taxonomy is a fossil from when the only imagery on a fire was an infrared line-scan from a fixed-wing aircraft at night — the NIROPS aircraft are a Citation and a King Air that has been flying since the mid-1980s — so naming the job after the sensor told you everything you needed to know.

It doesn't any more. Now it's drones, satellites, multispectral, real-time video. And the analytic work is identical across all of them:

That is imagery intelligence. The sensor is an implementation detail.

And here is the thing that settles it — the catalog has already half-admitted this in its own words. Read the two descriptions side by side. The drone job “manages, synthesizes, and disseminates intelligence.” The infrared job interprets. Same verbs, same product, one of them written thirty years later. The newer job got the modern word because it was named in the drone era; the older one is still carrying the name it was given when it was the only game in town.

Named for the sensor

IRIN · Infrared Interpreter — Planning
UASD · Unmanned Aircraft Systems, Data Specialist — Air Operations

Different codes, different functional areas, different task books, different courses — for the same analytic work on different hardware. One of them is even called intelligence already.

Named for the skill

Imagery Intel Analyst

Detect, classify, geolocate, track, fuse. Infrared, drone, satellite, multispectral, video — because the algorithms don't care where the pixels came from.

There's a second tell, buried in the course lists. The drone job's supporting training includes S-341 — the GIS course the GISS takes — and S-443, the infrared course. The catalog is already routing one person through all three sensors. It just hasn't given that person a name yet.

Don't train for infrared. Train for imagery analysis. The skill transfers across sensors, and the job titles will consolidate.

Which is the most useful thing on this page for a student, because it tells you what to actually learn. You built this at the Institute this week. The laser lab — thresholding to find the bright thing, blob detection to decide what counts as one object, tracking it across frames — is imagery intelligence at classroom scale. Point that at an infrared frame instead of a webcam and the code barely changes. That is the whole point.

04 · beyond fire

The same section, a different hazard

ICS is all-hazard — “it is used for all kinds of incidents” — so this same structure stands up on a flood or a chemical release. But there is no flood equivalent of FBAN or LTAN with its own code and course list. Instead, ICS reaches for a Technical Specialist, and FEMA is blunt about what that takes: “No specific qualifications are prescribed, as technical specialists normally perform the same duties during an incident that they perform in their everyday jobs.”

So a hydrologist gets on an incident by already being a hydrologist. There is no shortcut course, which means the road starts where you are: become someone who models water, or air, or contaminants — and then be the person in the region who can do it when the river comes up.

© 2026 RedfishGroup · New Mexico Supercomputing Challenge · Careers · position codes, titles, reporting lines and course numbers from NWCG and FEMA; the ensemble method from Finney et al. (2011) — working notes and sources in the bead. The burn-probability figure is a teaching simulation, not an operational fire model.