Every Warning Was On The Record: The CFB Wainwright Deaths Were Highly Likely Preventable

Kevin J.S. Duska Jr. | Prime Rogue Inc | Calgary, AB | October 21, 2026

The fire had burned for hours, the deadly wind was forecast and then warned of, and the training was routine. Public satellite, weather, fire-model and flight data also contradict key parts of the military’s account, and point to where C Company was when the wind turned.

The Deaths Resulting from the Fire at CFB Wainwright Were Avoidable

In Prime Rogue’s assessment, the deaths of four soldiers at CFB Wainwright on Oct. 17 were highly likely preventable. Every warning sign that mattered was on the public record before the wind turned: a fire burning for hours about a kilometre away, an “Extreme” fire-danger rating, and a forecast, then a warning, of the wind that drove it onto C Company. Whether the failure lay with the chain of command that kept training going, or with those responsible for giving it that information, the outcome was the same. A catastrophic error of judgment led to the deaths of four of Canadian’s finest.

The grass fire that killed four soldiers at CFB Wainwright on Oct. 17 was burning on the range by 12:17 p.m., almost four hours before the time National Defence gave. The wind that drove it onto C Company of the 3rd Battalion, Princess Patricia’s Canadian Light Infantry, was forecast twice that morning and named in a warning about an hour before it arrived.

A Prime Rogue analysis of public satellite imagery, weather records, fire-behaviour models and flight-tracking data finds that if the fire was about a kilometre away when the wind shifted, as Maj.-Gen. Wade Rutland said, C Company was likely in the area around the base’s grenade range and a north–south range road. The fire most likely reached them between about 3:25 and 3:45 p.m. Standard fire models put its speed at 4.4 to 12 km/h, five to 14 minutes to cover that kilometre. The military has said it covered it in 30 to 60 seconds.

The training was routine: grenade throwing and a section attack, for a deployment to Latvia about six months away. In Prime Rogue’s assessment, the decision to continue it about a kilometre from an active fire, under “Extreme” fire danger and a forecast wind shift toward the troops, likely exposed them to a risk out of proportion to its purpose.

National Defence has not explained why training continued. Asked 10 specific questions before publication, it declined to answer them, saying the investigation remains active, at 05:15 AM MSt on October 21, 2026.

This is an analysis of public data, written in calibrated probabilities. It is not a finding of cause or fault; three investigations are under way. Every data file is in Prime Rogue’s public archive.

Key judgments

Prime Rogue uses the probability language of intelligence analysis. Almost certainly means 95% or more; highly likely, 80–95%; likely, 55–80%; roughly even, 45–55%; unlikely, 20–45%; very unlikely, 5–20%; remote, under 5%. Confidence describes the quality of the evidence behind each judgment, not the probability.

1. The fire was almost certainly burning on the range by 12:17 p.m. MDT, nearly four hours before the “about 4 p.m.” time National Defence gave. (High confidence: NASA’s Landsat 8 recorded a 1.9 km fire line; NOAA-20 recorded it again at 2:49 p.m.; Maj.-Gen. Wade Rutland said range control knew of a fire “for several hours.”)
2. The wind shift that drove the fire onto C Company was almost certainly forecast. Environment Canada forecast west winds gusting to 80 km/h that afternoon at 4:30 and 10:48 a.m., and added Wainwright to a wind warning at 2:14 p.m. (High confidence: Environment Canada’s archived bulletins.)
3. If the fire was about a kilometre away when the wind shifted, as Rutland said, C Company was likely about one kilometre downwind of it, in the area around the range that OpenStreetMap labels “Range 2 (Grenade)” and a north–south range road. We put that at 55–70%, and about 80% including an adjacent strip north of Jamieson Lake. (moderate confidence depending on Rutland’s distance and on range names – which remain unverified. Without his distance, the satellite data still rules out Range 3 and the east lobe.)
4. The fire most likely reached the troops between about 3:25 and 3:45 p.m. (High confidence: total fire power measured by NOAA’s GOES-18 satellite over the base rose from 76 MW at 3:10 p.m. to 1,215 MW at 3:30 p.m.)
5. It is almost certain that the fire did not cover “a kilometre in 30 to 60 seconds.” Three standard fire models put its speed at 4.4 to 12 km/h, or five to 14 minutes per kilometre. Rutland’s figure most likely describes warning time, not speed. (High confidence.)
6. The warning signs were on the public record before training continued: an active fire about a kilometre away, a fire-danger rating of “Extreme,” and a forecast west wind that would push the fire toward the troops. Whether range control had them, and what its rules required, is not known. (High confidence on the record; what range control knew is unknown.)
7. In Prime Rogue’s assessment, the deaths were highly likely preventable, and the decision to continue training likely exposed troops to a risk out of proportion to its purpose. The drills were routine, grenade throwing and a section attack, for a deployment about six months away, and could have been paused or moved. The conditions match those under which wildland fire-safety practice calls for withdrawal. Halting or moving the training before the front arrived would have kept C Company out of the fire’s path. Whether the failure lay with the chain of command or with those who should have given it the information, the outcome was the same. A catastrophic failure of good judgment likely led to the unnecessary death of these troops. (High confidence: we have not seen the range standing orders, range control’s logs, or whether the 2:14 p.m. warning reached the range.)
8. There is no sign that gas infrastructure started the fire. No well or pipeline lies within 500 m of where satellites first saw it; the nearest pipeline was about 870 m away. An operating natural gas pipeline does cross the likely area of the overrun. (High confidence on location; nothing links the pipeline to the fire.)
9. The first STARS helicopter on public flight tracking left Edmonton at 4:58 p.m. and most likely landed near Range Control, about 3 km north of the burn and 4–6 km from the likely scene. (Moderate confidence: straight-line projections from ADS-B positions.)

What happened, according to officials

Four soldiers died and 39 people were taken to hospital after a grass fire swept over a training range at CFB Wainwright, Alberta, on Saturday, Oct. 17.

The dead are Sgt. Morgan Wong of Edmonton, of the Loyal Edmonton Regiment, serving with the 3rd Battalion, Princess Patricia’s Canadian Light Infantry; Pte. Shaun Farrell of Prince Albert, Sask.; Pte. Aidan Goulding of North Vancouver, B.C.; and Pte. Hayden Lykkemark of Langley, B.C. Premier Danielle Smith said 19 of those taken to hospital were in critical condition. Acute Care Alberta put the split at 19 critical, eight stable and 12 ambulatory.

At an Oct. 18 briefing, Maj.-Gen. Wade Rutland, commander of 1st Canadian Division, said about 84 soldiers of C Company, 3 PPCLI, were training with live munitions as part of high-intensity training ahead of a deployment to Latvia, reported for April and May. According to reports of the briefing:

• Most of the company was throwing grenades on a hand grenade range; about 10 soldiers were practising a section attack against a simulated enemy.
• Range control had known of the fire “for several hours,” and it was decided training could carry on.
• “The fire was a kilometre away when the wind shifted.” Rutland added: “It is not normal for a fire to move a kilometre in 30 to 60 seconds.”
• Relaying his commander on the ground, he said that “within 30 seconds to a minute they saw no fire, and then there was a fire right on top of them.”
• The flames reached the range’s administrative area. The grenade range was not affected, and the injured had been at the section-attack range, according to Postmedia’s account.
• The fire started within the training area and crossed established fire guards before it was contained, according to CBC’s account.
• He called the wind “uncommon, rapid,” and said winds had been “well below 50” km/h for most of the day.

National Defence and Rutland put the fire at about 4 p.m. Chief of the Defence Staff Gen. Jennie Carignan described a fire “far removed from the training area.” A Canadian Armed Forces memo reported by Global News and The Globe and Mail described the wind change as unexpected and, as the outlets paraphrased it, a fire index that was low. Prime Minister Mark Carney called it a “violent, fast-moving fire.

Officials say the cause is unknown. Rutland said a “multitude of ordnance” was in use that day, and that the cause “could have been any of that, or none of it.” The RCMP and military police, the fire marshal and a board of inquiry are investigating. Defence Minister David McGuinty said Oct. 20 that training procedures had not changed “as of now.”

The afternoon, hour by hour

The fire crept north for at least three hours under a southwest wind. Then a forecast cold front swung the wind west, then northwest, and the fire’s long southeast side became its front. All times are Mountain Daylight Time.

Time	What the record shows	Source
Oct. 12	Last rain at the base	Environment Canada
4:30 a.m.	Forecast for the area: west wind 50 km/h gusting to 80 “this afternoon”	Environment Canada forecast
10:48 a.m.	Same forecast reissued	Environment Canada
1 p.m.	Fire Weather Index at the base 30.8: “Extreme” on Alberta’s scale	Natural Resources Canada (CWFIS)
12:17 p.m.	Landsat 8 records 59 fire detections in a line about 1.9 km long on the range, running southwest to northeast	NASA FIRMS
1–2 p.m.	The front reaches Edmonton International Airport	Environment Canada stations
1:30 p.m.	Dust storm warning upwind, Vegreville to Camrose: 100 km/h gusts, “near zero visibility”	Environment Canada
2 and 3 p.m.	At the base: 23 °C, humidity 16–17%, southwest wind 30–32 km/h, gusts to 46–51	Environment Canada, CFB Wainwright station
2–3 p.m.	The front reaches Camrose and Vegreville	Environment Canada stations
2:14 p.m.	Wainwright added to a wind warning: gusts to 100 km/h, a line of storms	Environment Canada
2:49 p.m.	NOAA-20 records eight fire pixels; the fire has grown about 1 km north since 12:17, one detection inside the area mapped as Range 3	NASA FIRMS
3:05 p.m.	Wind warning kept for Wainwright, ended elsewhere	Environment Canada
3:10 p.m.	GOES-18 flags fire at the north end of the area: 76 MW	NOAA
3:16 p.m.	An 82 km/h gust from the west at the base’s weather station; by 4 p.m. the wind is northwest	Environment Canada
About 3:20–3:35 p.m.	The front reaches the fire, about 13–15 km east-southeast of the station	Prime Rogue estimate
3:30 p.m.	GOES-18: 1,215 MW across three pixels, including one about 4 km south-southeast; 16 times the total of 20 minutes earlier	NOAA
3:40 p.m.	GOES-18: only the southern pixel, 336 MW	NOAA
About 4 p.m.	Fire “reported” (National Defence). At the station: northwest 48–51 km/h, 12.9 °C, humidity 43%	DND; Environment Canada
4:58 p.m.	First STARS air ambulance on public tracking leaves Edmonton	adsb.lol
Oct. 17 evening	Municipality: fire crossed the base fence, burning south of Township Road 44-0	MD of Wainwright
Oct. 18, 2:12 and 2:32 p.m.	Satellites still detect heat in the burn’s east lobe, outside the base	NASA FIRMS
Oct. 19	Sentinel-2 maps a 10.8 km² burn running 6.2 km to the southeast	Copernicus

GOES-18 flagged no fire before 3:10 p.m. Its cloud mask shows the base covered all afternoon, by cloud, smoke or dust, so that absence does not mean no fire. Landsat and NOAA-20 had already seen it.

Where C Company was

Take Rutland’s distance at face value, and the satellite data puts C Company in one place: the area around the grenade range and a north–south range road, about a kilometre southeast of where the fire stood when the wind turned.

National Defence has not said where on the 609 km² base C Company was training. Prime Rogue reconstructed it in four steps:

1. Where the fire was before the shift. The 12:17 p.m. Landsat line and the 2:49 p.m. NOAA-20 detections outline the fire just before 3:16 p.m.: about 2 km², north of Jamieson Lake.
2. Which way the wind pushed it. After the front, the wind blew from 306°, pushing fire toward about 130°. The burn scar’s own long axis runs at 130°.
3. Where “a kilometre away” falls. Every 50 m patch of burned ground lying 0.6 to 1.5 km downwind of the fire at the shift.
4. Testing every alternative. Each candidate area was scored against all the evidence, looking for what rules it out, using the Analysis of Competing Hypotheses method. A separate analyst did the scoring without seeing our first estimate, and reached the same ranking.

The result. About 1 km downwind lies a band of burned ground between the north–south range road near 110.90° W and the north shore of Jamieson Lake, at roughly 52.760–52.775° N. Most of it is around the spot OpenStreetMap labels “Range 2 (Grenade)”. That is where we judge C Company most likely was.

Three things fit the official account closely:

• The grenade range burned only lightly. The mapped Range 2 lies just inside the burn’s edge, with low burn severity, below our threshold for a clear burn. The ground directly south and west of it burned hard. That matches Rutland’s account, as Postmedia reported it, that the grenade range was not affected and the flames reached the range’s administrative area.
• There is access. A road dead end and a building sit at the band’s eastern edge, the sort of place a vehicle-supported admin area would be.
• The numbers suggest a shared admin area. Four dead and 39 hospitalized is 43 people, but only about 10 were on the section-attack range. Many casualties may have been at an admin area shared by both ranges, though Postmedia reports the injured had been at the section-attack range.

What the data rules out.

• Range 3 and the north end. Fire was already there at 2:49 p.m.: a NOAA-20 detection falls inside the mapped Range 3, and the northwest wind then pushed the fire away from it. Nobody there was “a kilometre away.” The Range 3 polygon itself did not burn.
• Where the fire had already burned. The area burning at 12:17 and 2:49 p.m. was the fire, not a kilometre from it.
• The east lobe past the boundary road. It is 3 to 4.5 km downwind. At realistic speeds that gave 15 to 60 minutes of warning, not seconds. Most of it is outside the base, and it was still burning on Oct. 18.

A view from the road. Calgary-based independent reporter Sydney Fizzard (@SydFizzard, thefizz.ca) filmed the burn by drone from Township Road 441, east of the base. His flight log, which he shared with Prime Rogue, puts the drone’s takeoff at 52.7707° N, 110.8684° W, about 440 m outside the base boundary and outside the restricted area marked on his flight app. He held the drone in one spot, about 25 m up, panning and zooming.

From there, the building at the burn’s edge is 1.9 km due west, down the east–west track to the boundary road, and the mapped grenade range is 2.6 km away on almost the same line. Jamieson Lake lies 3.8 km to the southwest. His camera was looking straight into the most likely area. Prime Rogue has not published his footage, which is his to license; we encourage readers to support his work and buy his photos at thefizz.ca.

Stress test. We repeated the analysis across 18 combinations of wind direction (270° to 325°) and of how much the fire might have grown between 2:49 and 3:16 p.m.

Fire growth before the shift	Share of the 1 km band around Range 2 and the lake strip	Share in the ruled-out areas
None (as detected)	91–100%	7% at most
400 m	69–86%	0%
800 m	27–53%; the rest shifts south toward the centre of the run	0%

The ruled-out areas stay ruled out in every combination. The estimate’s one weak point is fire growth: 800 m would shift the likely area about 1 km south. The fire models rule that out: even burning freely, the fire’s side could have crept only 200 to 320 m in those 27 minutes.

Our ranking:

Area	Likelihood
Around Range 2 and the north–south range road	Likely, 55–70%
Strip north of Jamieson Lake, next to it	Unlikely but credible, 15–25%
South-central part of the run	Very unlikely, 5–15%
Just outside the burn at its northeast edge (overlaps the Range 2 area)	Very unlikely, 5–10%
East lobe past the boundary road	Remote, under 5%
Range 3 and the north end	Remote, under 5%
Where the fire already was	Remote, under 3%

How fast the fire moved

Three standard grass fire models, fed the base’s own weather readings, agree: after the wind turned, the fire covered a kilometre in five to 14 minutes, not 30 to 60 seconds.

The models: the Canadian Forest Fire Behavior Prediction System, used by Canadian fire agencies, for matted and standing grass; the CSIRO grassland model (Cheney, Gould and Catchpole, 1998), for natural and grazed pasture; and the Cruz et al. (2022) rule that grass fires spread at about 20% of the wind speed. Inputs: Environment Canada’s readings at the base, Prime Rogue’s hourly fine-fuel moisture code (92.3–92.9), computed from Natural Resources Canada’s daily value of 91.8, and grass 90–100% cured.

Period	Wind (km/h)	Head spread (km/h)	Minutes per km	Fireline intensity (kW/m)
Before the shift, 2:49–3:16 p.m.	32, southwest	4.8–7.8 if burning freely	8–12	7,200–15,500
After the shift	48, northwest	4.4–12.0	5–14	6,500–21,000
In the 82 km/h gust	82	7.8–16.4	4–8	11,800–28,000

What that means:

• The soldiers’ account holds; the arithmetic doesn’t. “A kilometre in 30 to 60 seconds” is 60 to 120 km/h; at the top end, faster than the 82 km/h peak gust that drove the fire. The fastest grass fires in the studies we reviewed ran at under 30 km/h.
• It was warning time. When a 2 km side of a fire turns into its front, with smoke and dust blowing ahead of it, a fire a kilometre away can be invisible until it is close. For 30 to 60 seconds of visible warning, the flames were roughly 40 to 270 m away when first seen. The same storm line was raising “near zero visibility” dust upwind.
• The fire had been held. Burning freely in the southwest wind, it could have run 5 to 8 km/h. Satellites show it grew about 1 km in the 2.5 hours from 12:17 to 2:49 p.m., roughly 0.4 km/h. That fits a fire that was monitored, judged contained, and held on its east side by the north–south road.
• It could not be fought. Above about 4,000 kW/m, ground crews cannot attack a fire. This one ran at 6,500 to 21,000 kW/m. Soldiers on foot, in kit, cannot outrun 10 to 12 km/h for long.

The models assume continuous grass; aspen clumps, wetlands, grazing and roads slow real fires, so the true speed may sit at the lower end. Either way, the gap with 30 to 60 seconds is five- to 27-fold.

What was knowable before training continued

By 3 p.m. on Oct. 17, five facts were on the record, and each pointed the same way.

1. A fire was burning about a kilometre away, and had been for hours. Range control knew, Rutland said. Satellites saw it at 12:17 and 2:49 p.m.
2. Fire danger was Extreme. Natural Resources Canada’s Fire Weather Index for the base’s own station was 30.8 at its daily reading (1 p.m. MDT), in Alberta’s “Extreme” class. Humidity was 16–17% at 2 and 3 p.m. There had been no rain since Oct. 12. Prime Rogue’s hourly calculation puts the index at 53.6 at 2 p.m. and 59.4 at 3 p.m. A military memo reportedly described the fire index as low.
3. The wind shift was forecast. West winds gusting to 80 km/h “this afternoon” were forecast at 4:30 and 10:48 a.m. A west wind would push the fire east, and the northwest wind behind the front southeast, toward where the data puts C Company.
4. A warning was in force. Environment Canada added Wainwright to a wind warning at 2:14 p.m., for gusts to 100 km/h and a line of storms, and kept it for Wainwright at 3:05 p.m.
5. The front was visibly on its way. It hit Edmonton between 1 and 2 p.m. and Camrose and Vegreville between 2 and 3 p.m. At 1:30 p.m. the same storm line was raising “near zero visibility” dust upwind.

Against that, the training was routine. C Company was throwing grenades and practising a section attack, foundational infantry skills, for a deployment to Latvia reported for April and May. The only rationale in Rutland’s reported remarks was: “We train in all types of conditions, because you don’t know where you’re going to operate.”

Wildland fire crews are trained to recognize a short list of “watch out” situations, and to withdraw when they stack up. Three were present that afternoon: the wind forecast to change direction, unburned grass between the troops and the fire, and a main fire that could not be seen once smoke and dust blew ahead of it. Soldiers are not firefighters, but the hazard is the same.

Prime Rogue’s assessment: the warning signs were on the public record before training continued. And it is likely that the decision to continue exposed the soldiers to a risk out of proportion to its purpose. The drills could have been paused or moved for an afternoon at little cost.

What could change that assessment: the base’s range standing orders and fire-danger restrictions; range control’s logs; whether the 2:14 p.m. warning reached the range; what fire guards and suppression were in place; and what escape plan the admin area had. If they show controls we cannot see, the judgment weakens. If they show the warning never reached the range, or orders weren’t followed, it strengthens.

The official account against the record

Read in full, the military’s own account points to the same place the data does. It breaks down only on timing, speed and distance from the training area.

Statement	Who	The record	Holds?
Range control knew of a fire “for several hours” and let training “carry on”	Rutland	Fire on the range from 12:17 p.m., modest at 2:49	Yes
“The fire was a kilometre away when the wind shifted”	Rutland	That kilometre falls around the grenade range in 13 of 18 model runs, and never at Range 3 or past the boundary road	Yes: it is the basis of our estimate
“They saw no fire, and then there was a fire right on top of them”	Commander on the ground, via Rutland	A 2 km side turning into the front, with smoke and dust ahead	Yes
Flames reached the range’s administrative area; the grenade range was not affected	Rutland, as reported by Postmedia	The mapped grenade range barely burned; the ground beside it burned hard	Yes
Fire reported or ignited “about 4 p.m.”	DND release; Rutland	Burning at 12:17 p.m. If 4 p.m. were the overrun, the front was then 4–6 km southeast, 15–60 minutes from anyone a kilometre away	Not as a start time
“A kilometre in 30 to 60 seconds”	Rutland	Models: 5–14 minutes. 30–60 seconds needs 60–120 km/h	No, as a speed; plausibly as warning time
“Far removed from the training area”	Carignan	Fire among the mapped ranges from midday, inside Range 3’s polygon at 2:49	No
“Uncommon, rapid”; unexpected (memo, as reported)	CAF memo; Rutland	Forecast at 4:30 and 10:48 a.m.; warning at 2:14 p.m.; the front crossed Edmonton two hours earlier	No
Winds “well below 50” for most of the day	Rutland	True until 3 p.m. (gusts 46–51 km/h); then 82 km/h	Partly
A fire index that was low	CAF memo, as paraphrased	Fire Weather Index 30.8, “Extreme”	No
Crossed established fire guards	Rutland	The fire crossed the base boundary road to the east	Yes
Not oil and gas: “initial impression is it’s not”	Rutland	Nothing within 500 m of where the fire was first seen	Consistent

Officials disagree with each other on one point: Rutland said the fire started within the training area, according to CBC, while Global News also reported him saying it did not start within the training area, and Carignan called it “far removed” from it.

Gas wells and pipelines

Nothing in the energy records points to a gas ignition: the nearest pipeline was about 870 m, and the nearest well 1 km, from where satellites first saw the fire. But an operating gas line runs through the area where C Company most likely was.

The Wainwright training area is shared land. A 2014 Western Producer report described about 180 buried natural gas wells on the base, with pipelines buried twice the usual depth to withstand armoured vehicles; Global News cited a federal figure of more than 200 active gas wells in the area. Global News reported that unnamed military sources had raised the possibility of a gas line or well being hit. Rutland said: “Initial impression is it’s not, but we will definitely check.”

Prime Rogue mapped every Alberta Energy Regulator (AER) well licence and pipeline segment around the burn, along with NRCan’s CanVec wells and pipelines.

Check	Result
AER well licences within 5 km of the burn	109: 38 abandoned, 26 issued, 19 suspended, 17 reclaimed, 9 other
Wells inside the burn	2, both in the east lobe: one abandoned, one reclaimed
Nearest well to the 12:17 p.m. fire line	1.0 km, Sumerian Oil & Gas, licence issued
Nearest pipeline to the 12:17 p.m. fire line	870 m, Sumerian Oil & Gas natural gas line, 89 mm, operating
Wells, pipelines, buildings or power lines within 500 m of where the fire was first seen	None
Operating pipeline in and within 150 m of the likely area	1.24 km of Sumerian Oil & Gas natural gas pipeline
Well licences in the likely area	1, Sumerian Oil & Gas, licence issued
Share of the burn’s edge that follows a pipeline	2% (roads and tracks: 12%)

What it means:

• No sign of an infrastructure ignition. The fire was first seen about 870 m or more from any pipeline and 1 km or more from any well. That is consistent with Rutland’s initial impression.
• A gas line crossed the fire’s path. About 1.2 km of operating natural gas pipeline runs through the area that burned near the grenade range, the likely scene of the overrun. Nothing in the data links it to the fire or the casualties. It raises a separate question: has the line been inspected since the fire?
• Pipelines didn’t shape the burn. Only 2% of the scar’s edge follows a pipeline right-of-way.

AER licence status shows only whether a licence is active, not whether a well produces. A well licensed for sour gas, with a low hydrogen sulphide content, lies about 300 m east of the likely area, near the east–west track to the boundary road.

Air ambulances and the medical response

The first STARS helicopter on public flight tracking left Edmonton at 4:58 p.m., about an hour and a half after the fire most likely overran C Company, and appears to have landed near Range Control rather than at the scene.

Alberta officials said 30 emergency units went to Wainwright on Oct. 17: 22 ground ambulances and eight aircraft, six fixed-wing and two helicopters. Patients went first to the Wainwright Health Care Centre, then to hospitals in Edmonton, Calgary and Red Deer. Health agencies held their first coordinating call by 6 p.m. Code Orange was declared at several hospitals and lifted the next afternoon. Saskatchewan sent air ambulances.

Prime Rogue searched the adsb.lol open archive, which records aircraft position broadcasts, for every aircraft heard within about 60 km of the base from 10 a.m. Oct. 17 to 8 a.m. Oct. 18: 151 aircraft.

• STAR03, a STARS Airbus EC145 (C-FLCI), left Edmonton International at 4:58 p.m. It was last heard at 5:32 p.m., descending through 4,675 ft, 13 km west-northwest of the burn, heading 112°. Low flight there falls below receiver coverage.
• It was heard again leaving at 9:40 p.m., heading 295° toward Edmonton, and made a second round trip from about 12:36 to 2:10 a.m.
• All four of its lines, both arrivals and both departures, pass within about 1 km of Range Control headquarters (52.814° N, 110.901° W). They pass 4–6 km from the likely scene of the overrun. That is consistent with casualties being brought to a collection point near Range Control.
• Later aircraft: fixed-wing aircraft came within 30 km of the burn from 5:26 p.m., when King Air C-FSUG arrived, followed by CWA921 at about 6 p.m., Saskatchewan Air Ambulance’s SLG2 at about 6:20 p.m., C-GWRS at about 7 p.m. and SLG3 at about 9 p.m. More STARS helicopters, STAR11 and STAR01, came after 9:40 p.m. Operators and purposes of the King Airs are not confirmed.
• No aircraft was recorded low or slow within 30 km of the burn. Military helicopters such as the CH-146 Griffon do not normally broadcast positions, so their absence from the record means nothing.

The gap between about 3:30 p.m. and the first STARS departure at 4:58 p.m. is a question for National Defence and Emergency Health Services Alberta: when was air support requested, and how were the wounded moved before it arrived?

These are straight-line projections from the last and first positions heard; a helicopter can turn out of coverage. Data: adsb.lol (Open Database License).

The fire leaves the base

The same northwest wind carried the fire across the base’s eastern boundary road and onto farmland, where it was still burning the next afternoon.

• The burn. Copernicus Sentinel-2 imagery from Oct. 4 and Oct. 19 maps a 10.8 km² main burn, 6.2 km long and about 2.4 km wide, running southeast at 130° from where the fire stood before the shift. About 1 km² burned east of the boundary road, outside the base.
• The municipality. The Municipal District of Wainwright said the fire crossed the base fence and was burning south of Township Road 44-0. It said Oct. 19 the fire was contained within its boundary.
• Still burning. Satellites detected heat in that eastern lobe at 2:12 and 2:32 p.m. on Oct. 18, so part of it may have burned after the incident.
• A separate, earlier burn. A patch of about 0.6 km² southwest of Jamieson Lake matches a NOAA-21 fire detection on Oct. 14, three days earlier. Its cause is unknown.
• History. NASA’s satellites have recorded fire on this part of the range repeatedly since 2012, including in October. Fire on the range is routine; this one turned deadly because of where people were when the wind changed.

An earlier scar south of the lakes predates Oct. 4 and is not part of this fire.

What we don’t know

The data shows where the fire went and when. It cannot show where any soldier stood, what caused the fire, or what range control knew and did.

Limits of this analysis

• No public record marks the casualty site. Our location rests on Rutland’s “a kilometre,” the fire’s position 30 minutes before the shift, and range names from a single OpenStreetMap edit.
• The wind readings come from a station about 13–15 km from the fire; wind at the fire may have differed.
• GOES pixels are 2–4 km across and uncorrected for terrain; we use them for timing, not position.
• The burn scar shows where the fire ended up, not when each part burned.
• The fire models assume continuous grass.
• Soldiers may have moved hundreds of metres while fleeing, most likely further southeast.

What would change our assessment

Record	What it would settle
National Defence’s grid reference for the admin area and section-attack range, and its range numbering	Where C Company was; whether OpenStreetMap’s labels are right
Range control and radio logs, Oct. 17	When the fire was first logged; when training stopped; whether the 2:14 p.m. warning was received
CFB Wainwright range standing orders and fire-danger restrictions	Whether continuing training complied with the base’s own rules
The briefing video and a full transcript	Rutland’s exact words on distance, timing and the grenade range
Air ambulance request and landing records	When help was called, and where the wounded were collected
High-resolution post-fire imagery	Vehicles, equipment or a staging area in the burn

If any of these places C Company more than about 2 km from the north–south range road near the grenade range, our location estimate is wrong and we will say so.

Questions put to National Defence

Prime Rogue sent National Defence 10 questions at about 5:20 a.m. ET on Oct. 21, with a 7:30 a.m. ET deadline. They are condensed below. National Defence did not answer any of them. In an emailed reply before publication, Andrée-Anne Poulin of National Defence media relations wrote that the investigation “remains active, and commenting on specific details at this stage would be inappropriate. The Office of the Canadian Forces Provost Marshal will provide further updates on the investigation’s status and outcome when appropriate.” We will add any further response.

1. Our analysis places C Company’s section-attack administration area in the area around Range 2, along the north–south range road. Is that correct? If not, where was it?
2. Maj.-Gen. Rutland said the grenade range was not affected, while imagery shows the area around it burned. What did he mean?
3. Which range or training area was C Company using when the fire reached it?
4. Satellites recorded the fire on the range at 12:17 p.m. When did range control first log it, and why did National Defence describe the fire as reported or ignited at about 4 p.m.?
5. Did range control or the unit receive Environment Canada’s 2:14 p.m. wind warning? Why did training continue after it?
6. How far was the fire from C Company when the wind shifted, and at what time? Was “30 to 60 seconds” the warning the soldiers had, rather than the fire’s speed?
7. A CAF memo reportedly described the fire index as low. Natural Resources Canada’s index for the base was 30.8, “Extreme.” Which index did range control use?
8. Does Gen. Carignan stand by her statement that the fire was “far removed from the training area”?
9. When was air ambulance support requested, how were casualties moved before it arrived, and were CAF aircraft used?
10. An operating natural gas pipeline crosses the burned area near Range 2. Has it been inspected since the fire?

Methods and data

Every data file behind this analysis is public, unaltered, hashed and timestamped on the Bitcoin blockchain in Prime Rogue’s CFB Wainwright Fire Archive, so anyone can check the work.

Data	What we used it for	Archive folder
NASA FIRMS fire detections (Landsat 8, NOAA-20, NOAA-21, Suomi NPP, MODIS)	Where and when the fire burned; 2012–2026 history	data/firms/
NOAA GOES-18 and GOES-19 fire detection and imagery, every 10 minutes	Fire power and timing, 3:10–3:50 p.m.	data/goes/
Copernicus Sentinel-2, Oct. 4 and Oct. 19	Burn scar, by change in Normalized Burn Ratio (dNBR ≥ 0.27)	data/copernicus/
Environment Canada forecasts, warnings and station observations	Wind, temperature, humidity, the front’s timing	data/eccc/, data/reference/
Natural Resources Canada fire-weather index	Fire danger; inputs to fire models	data/nrcan_cwfis/
Alberta Energy Regulator wells and pipelines; NRCan CanVec	Gas infrastructure, roads, buildings	data/infrastructure/, data/reference/
OpenStreetMap	Range names and boundaries (unverified)	data/infrastructure/
adsb.lol aircraft positions	Air ambulance tracks	data/adsb/

Location estimate. The fire’s position before the shift is the outline of the 12:17 and 2:49 p.m. detections, clipped to the burn. For each 50 m patch of burned ground, we measured the distance upwind to that outline along the post-front wind. The “kilometre away” band is 0.6 to 1.5 km. Seven candidate areas were scored with the Analysis of Competing Hypotheses method (Heuer), by a Prime Rogue analyst working independently of our first estimate, and stress-tested across 18 combinations of wind direction and fire growth.

Fire models. Canadian FBP System fuel types O-1a and O-1b (Forestry Canada, 1992; Wotton et al., 2009); CSIRO grassland model (Cheney, Gould and Catchpole, 1998), natural and grazed; Cruz, Alexander and Kilinc (2022). Hourly fine-fuel moisture by Van Wagner’s (1977) method from Natural Resources Canada’s daily value.

Air ambulance tracks. Positions broadcast by aircraft and collected by volunteer receivers. Arrival and departure lines are straight-line projections from the last and first positions heard.

Probability language. Estimative terms follow the scale in our key judgments. Confidence reflects the quality of the evidence.

Licences. Prime Rogue’s analysis and graphics: CC BY-NC 4.0. Code: MIT. Copernicus: contains modified Copernicus Sentinel data 2026. NASA and NOAA data: public. Environment Canada and NRCan: Open Government Licence – Canada. OpenStreetMap and adsb.lol: Open Database License.

Have something? Send documents or information through Signal (Rubicon49.50) or Proton email via kevinjr.duska (at) proton (dot) me. Do not use GitHub for tips; its history is public and permanent.

Our assessment

These deaths were highly likely preventable.

The fire did not appear from nowhere. It had been burning on the range since at least 12:17 p.m., and range control knew of it for hours. The wind that drove it onto C Company was forecast before dawn, forecast again before 11 a.m., and named in a warning at 2:14 p.m., about an hour before it arrived. Fire danger at the base was rated “Extreme.” The training was routine, and could have been stopped or moved for an afternoon.

We cannot yet say who failed. If range control and the chain of command had this information and kept training, the failure was one of judgment. If the information never reached them, the failure was in the system meant to deliver it. Either way, the outcome was the same: four soldiers dead and 39 people in hospital, from a hazard that was visible in public data before it struck.

The military’s own account, read closely, supports this judgment. A fire it had monitored for hours, a kilometre away, turned with a forecast wind and reached a range admin area. What it does not explain is why the training was still there when it did.

Three investigations are under way. Their findings should answer what this analysis cannot: what range control knew, what its rules required, and why C Company was still downwind when the wind turned. Prime Rogue will publish every record it obtains.

This assessment is based on public data and is stated in calibrated probability. It is not a finding of criminal or civil fault against any individual.

Sources

Official statements and reporting

• Global News: Everything confirmed so far about CFB Wainwright fire
• Global News: Fire at CFB Wainwright wasn’t a concern — until a rapid ‘uncommon’ shift in wind
• CBC News: What we know so far about fast-moving fire that killed 4
• CBC News: Findings from fatal CFB Wainwright fire investigation will be made public
• CP24: What happened at CFB Wainwright, the confirmed facts
• CP24: No immediate changes to military training, defence minister says
• CochraneNow (Postmedia): Military names four soldiers killed, details training exercise
• The Globe and Mail: Blaze began as grass fire
• The Globe and Mail: Live updates, the four military members killed

Data

• NASA FIRMS
• NOAA GOES on the AWS Open Data Registry
• Copernicus Browser
• Environment Canada MSC Datamart
• Canadian Wildland Fire Information System
• Alberta Energy Regulator ST37 well list
• adsb.lol
• OpenStreetMap

Science

• Cruz, Alexander and Kilinc (2022), Fire 5(2):55
• Cheney, Gould and Catchpole (1998), International Journal of Wildland Fire 8(1)
• Forestry Canada (1992), Development and Structure of the Canadian Forest Fire Behavior Prediction System, ST-X-3
• Wildland firefighter “watch out” situations, NWCG Incident Response Pocket Guide
• The Western Producer (2014), on gas wells on the base

Prime Rogue

• CFB Wainwright Fire Archive on GitHub
• CFB Wainwright fire: live updates