Rubicon Ridge: a new reconstruction moves the CFB Wainwright fire’s timeline 12 minutes later

Kevin J.S. Duska Jr. | Prime Rogue Inc. | Calgary, AB | November 6, 2026

Weather stations around the base put the wind shift over the fire 12 minutes later than our last model assumed. That moves the fire’s likely arrival at Range 2 and Range 15 to about 3:39 to 3:42 p.m., closer to the Department of National Defence’s “just before 4:00”. This update corrects several figures we published.

How the Fire Grew at CFB Wainwright

The cold front that turned the Oct. 17 fire at CFB Wainwright reached the fire at about 3:28 p.m., according to Prime Rogue’s new reconstruction, working model v5, “Rubicon Ridge”. Our previous model, v4, used 3:16 p.m., the time of the 82 km/h gust at the base’s weather station. That station is about 15 km west-northwest of the fire, and the front took about 12 minutes to cross the gap.

The later front moves the fire’s likely arrival at the two ranges named in the base’s own range orders to about 3:39 p.m. at Range 2 and 3:42 p.m. at Range 15. v4 put the fire at Range 2 at about 3:21. Measured from the moment the front hit, the fire’s run takes about as long as before: roughly 10 to 13 minutes to those points.

That narrows the gap with DND’s account. DND has said the soldiers were overtaken “just before 4:00”, the latest in a series of accounts of the afternoon. In v4, fewer than one in ten model runs reached the road head after 3:45. In v5, about a third do, depending on how fast the fire ran after the front, the biggest unknown left.

This update also corrects five things we published, set out in the next section. v5 is a working model, not a final finding. Its data and code are public. For everything we’ve published on the fire, start with our live updates page

Corrections

Five figures we published were wrong or imprecise, and two model results have changed. We have corrected the archive files, and each change is recorded in our public corrections. v4 stays available exactly as published at the v4-published repository

What we published	Correct figure	Why
The cold front reached the fire at 3:16 p.m.	About 3:28 p.m. (3:27 to 3:29)	3:16 was the gust at the base weather station, not at the fire. A line fitted to nine stations and sensors puts the front over the fire 12 minutes later.
The base weather station is about 12 km (or 13 km, or 15 km west) from the fire	14.7 km west-northwest	Measured from the fire centre (52.7645 N, 110.9052 W) to the station (CZWN). Our files gave three different figures.
GOES satellite pixels are about 2 km across	Several km: about 2.6 by 5 km at the fire	The 2 km figure is nominal. At Wainwright’s steep viewing angle the heat-sensing pixels are much larger.
The burn covers about 11.5 to 11.7 square km	About 1,050 to 1,100 ha	Earlier totals included a separate patch, probably the Oct. 14 fire. The 70 cm outline gives 1,050 ha net of unburned islands inside the image, about 1,075 ha for the whole burn.
DND’s “approximately 4:00 p.m.” release was dated Oct. 17	Dated Oct. 18	A date error in one of our analysis notes.

Model results that changed (updates, not errors):

• Fire arrival. v4 put the fire at its Range 2 point at about 3:21 p.m. and at the road head at about 3:26. v5 puts it at the official Range 2 grid at about 3:39, at Range 15 at about 3:42 and at the road head at about 3:42. Most of the shift comes from the later front.
• Range 2’s location. v4 used the Range 2 position mapped in OpenStreetMap. The base’s 2024 Range Standing Orders place Range 2 about 375 m east of that point. v5 uses the official grids.

One wording fix: Environment Canada’s 2:14 p.m. bulletin did not issue a new warning. It added Wainwright to a wind warning in force for areas to the west since 11:42 a.m.

When the front hit

The front swept across the region from the west-northwest at about 77 km/h and reached the fire at about 3:28 p.m. It arrived at stations to the west and north first and at those to the south and east last.

Station or sensor	Distance and direction from the fire	Front arrival (MDT)	Peak gust (time)	Temperature fall
Albert Hall	43 km NNW	about 3:02	82 km/h (3:07)	22.9 to 12.7 °C
Bellshill	43 km WSW	about 3:02	81 km/h (3:25)	23.0 to 12.7 °C
Castor (air sensor)	91 km SW	3:07	—	—
CFB Wainwright (CZWN)	14.7 km WNW	3:16	82 km/h (3:16)	23.3 to 12.9 °C
Wainwright town (air sensor)	8 km NNE	3:28	—	—
The fire	—	about 3:28 (3:27 to 3:29)	—	—
Gilt Edge North	39 km NNE	about 3:30	83 km/h (3:28)	23.5 to 14.1 °C
Coronation	85 km SSW	3:31	94.6 km/h (3:33)	22.1 to 13.7 °C
Hughenden (nearest station)	22 km SSE	about 3:39	77 km/h (3:43)	23.1 to 13.8 °C
Lloydminster	82 km NE	about 3:45 (3:41 to 3:48)	98 km/h (3:48)	22.8 to 15.1 °C

Front arrival is when the air mass changed at each site: – the wind shift at stations with minute-by-minute data (Environment and Climate Change Canada); – the jump in humidity at hourly stations (Alberta Climate Information Service); – the dust spike at the two air sensors (PurpleAir).

The fire’s time comes from a moving line fitted to these nine sites (average miss 3 minutes). No instrument stood at the fire itself. Bellshill is the worst fit: it turned about 7 minutes earlier than the line predicts.

The peak gust at a station can come well after the front, as it did at Bellshill. Peak gusts are 2 m or 10 m readings as reported, and gusts are not the sustained wind that drives a fire.

The front’s arrival at the base station at 3:16 is real. It is the 3:16 we and others reported. It just happened about 15 km upwind of the fire.

When the fire reached Range 2 and Range 15

On v5, the fire reached the official Range 2 grid at about 3:39 p.m. and Range 15 at about 3:42, roughly 10 to 13 minutes after the front hit. How fast it ran after the front is the biggest unknown, and it moves those times by up to 15 minutes.

v5 runs 2,160 versions of the fire. They vary the front’s timing, the wind after the front, the grass type, the spread rate, and how much the April 2026 planned burn slowed it. Each is weighted by how well its final shape matches the actual burn scar. The full results are in the v5 model folder.

Point	v4 (published Oct. 25)	v5 median	Likely range (10 to 90%)	Share reached 3:45 to 4:00
Range 2 (official grid)	not modelled (v4 used a point 375 m west: 3:21)	3:39 p.m.	3:34 to 3:51	25%
Range 15 (official grid)	not modelled	3:42 p.m.	3:36 to 3:58	31% (plus 7% after 4:00)
Road head (entrance to Range 15)	3:26 p.m.	3:42 p.m.	3:36 to 3:56	30% (plus 5% after 4:00)

Spread rate decides the clock. The burn scar cannot tell a fast fire from a slow one: both leave the same shape. So v5 brackets it:

Spread rate after the front	Range 2	Range 15	Road head
Slow, 3.5 to 5 km/h	3:48	3:54	3:53
Standard, 7 to 10 km/h	3:38	3:41	3:40
Fast, 11 to 13 km/h (near the 12.7 km/h an independent wildfire expert said to expect)	3:35	3:37	3:37

Two other findings: – The wind after the front. The wind that best reproduces the scar blew from the north-west (310 to 315 degrees), as in v4. – The April 2026 planned burn. The ground burned in it did not clearly slow the fire: giving it a lower spread rate makes the fit slightly worse.

What this means for “just before 4:00”. v4 found DND’s time hard to square with the model: fewer than one in ten runs reached the road head after 3:45. v5 narrows that gap. A slow-running fire reaches Range 15 at about 3:54, and about a third of weighted runs reach it between 3:45 and 4:00. DND’s time is now within the model’s range, at its slow end

What “1 km in 30 to 60 seconds” measures

The Canadian Armed Forces has said the fire was about a kilometre away when the wind shifted and covered that distance in 30 to 60 seconds. That is the speed of the wind front, not of the flames.

• The front: at 77 km/h, the wind shift crossed a kilometre in about 47 seconds. Our first fit, at 87 km/h, gives about 41 seconds. Either sits inside the CAF’s 30 to 60 seconds.
• The fire: grass fires driven by winds like these spread at roughly 3.5 to 13 km/h in our model, so a kilometre takes about 5 to 17 minutes. One kilometre in 60 seconds would mean 60 km/h; in 30 seconds, 120 km/h. Both are far beyond the grass-fire spread rates measured in field studies, and beyond the head-fire speeds the Canadian Fire Behavior Prediction System gives for grass in these conditions.

A fire can seem to arrive in seconds without its front moving that fast, in three ways: – smoke and dust hide the flames until they are close; – embers start spot fires ahead of the main front; – a long flank turns into a head fire all at once when the wind veers, the scenario our Oct. 19 analysis of satellite imagery explored.

The data cannot say which, if any, happened here. What it does show is that the wind front, with its dust and the sudden gusts, covered a kilometre in under a minute.

Dust, not smoke, and no visible plume

The air-quality spike recorded in Wainwright at about 3:28 p.m. was wind-blown dust carried by the front, not smoke from the fire.

• Dust. The town’s PurpleAir sensor shows coarse particles jumping for about four minutes as the front passed: PM1 6.5, PM2.5 16.9 and PM10 28.0 µg/m³ in the reading ending 3:29.

• The share of the finest particles fell, the opposite of a smoke signature.
• Matching spikes hit Castor and Lloydminster as the front reached them.
• The town was upwind of the fire once the front passed, so its sensors would not be expected to catch the smoke.

• No plume from space. NOAA’s GOES-18 and GOES-19 satellites show no identifiable smoke plume between 2:30 and 5:30 p.m. A band of frontal cloud crossed the fire as it ran, and post-front cloud covered the area from about 3:50. The fire shows only as a heat signal between about 3:20 and 3:50.
• No close-up satellite pass. No polar-orbiting satellite imaged the fire during its intense phase. (Our week of satellite data lists every pass and detection around the base.) The nearest pass, NOAA-21 at 3:35 p.m., missed it by about 25 km at the edge of its swath. An Aqua MODIS pass at 3:18 p.m., before the front, is the one image worth checking for a pre-front plume.

None of this rules out smoke on the ground. It means the instruments that might have recorded it were in the wrong place, under cloud, or between passes.

The burn

The fire burned about 1,050 to 1,100 ha, stopped just short of both official range grids, and crossed the base’s eastern boundary near its south-east corner. These measurements come from Prime Rogue’s 70 cm satellite image of Oct. 20, published here.

• Area. 1,050 ha net of unburned islands inside the image, about 1,075 ha for the whole burn.
• The ranges. Using the grids in the base’s 2024 Range Standing Orders, the burn stops 16 m from the official Range 2 grid, 12 m from Range 15 and 6 m from the road head at Range 15’s entrance. Ground north-east of the road head, between the Range 2 access road and the base boundary, did not burn.
• The tracks. The fire crossed the north-south track about 90 m south of the road-head junction. North of there, it stopped at a 20 to 25 m unburned strip along the track.
• The base boundary. North of about 52.752 N, the burn stopped at a bare strip along the boundary road. South of that, it burned on both sides of the road, into land beyond the base.
• No burn-out lines. The image shows no sign of crews lighting backfires or cutting new firebreaks.

The model does not know about tracks or bare strips, so it burns the pocket by the road head and too little of the south-east corner. That is v5’s main misfit, and it points to local features such as tracks, fuel breaks or suppression that the model does not include.

What v5 cannot tell us

v5 narrows when the fire reached the ranges. It cannot say where the soldiers were or exactly when they were overtaken.

• Speed. The burn scar fixes the fire’s shape, not its speed. The model’s implied spread after the front ranges from 3.5 to 11.6 km/h, and that range is mostly our own assumption.
• Location. Range 2 and Range 15 are the ranges the base’s orders place at the road head. DND has not said which range the soldiers were using or given a grid reference, and there may be no GPS record to settle it
• Front time. The 3:28 estimate comes from stations 15 to 85 km away. No instrument stood at the fire.
• The burn edge at the ranges. The 70 cm image puts both range grids just outside the burn. Other satellite edges, which v6 will combine, put them on the edge within measurement error. Whether the flames reached the grid points themselves is still open.
• Fit. The best runs match about 54% of the burn’s shape. The model has no smoke, embers, spot fires, roads, firebreaks or firefighting.
• Wind. Wind is applied evenly across the area from one station or model point at a time. Local gusts and eddies near the ground are not modelled.
• Gas infrastructure. The model does not include the oil and gas wells and pipelines on the base (see below).

v5 is a working model. New data, or DND’s own records, may confirm or overturn it.

DND’s response

On Oct. 27, Prime Rogue sent DND these questions, on top of the access-to-information requests we’ve filed – which range was the pairs range; – when the fire reached it, according to DND’s records; – whether tracks were used as fireguards; – when the base fire hall was dispatched.

DND has not responded to eight requests for comment.

DND’s account so far comes from its Oct. 22 briefing (our archived transcript): the fire began on the automatic grenade launcher range, was monitored inside the danger area, and overtook soldiers on a pairs range “just before 4:00” after the wind shifted. That account came days after DND first said the fire “ignited” at about 4 p.m., although satellites had detected it hours earlier. Several investigations are under way, military and provincial.

The gas wells on the base

Sumerian Oil & Gas Inc., formerly Cleo Energy Corp., holds about two-thirds of the roughly 260 licensed well sites on CFB Wainwright, Bloomberg reported. Its wells and gas gathering lines are the nearest to the ranges.

• The order. On Oct. 5, the Alberta Energy Regulator (order amendment, Oct. 9) ordered the Orphan Well Association to shut in and secure Sumerian’s wells, facilities and pipelines within 14 days. That window was still open on Oct. 17. Canada’s National Observer has reported on the status of the wells on the base.
• The company’s statement. In response to Prime Rogue’s request for comment, a Sumerian spokesperson said on Nov. 4 that the company is not aware whether its infrastructure was affected by the fire. The company said it has not been able to physically inspect its wells on the base, despite repeated requests to the Alberta Energy Regulator and the Orphan Well Association.
• Our mapping. v5 does not model gas infrastructure. Our Oct. 19 check of the well records and our later mapping of Sumerian’s wells and lines found no Sumerian well inside the burn. One Sumerian gas line crosses it for about 50 m.
• Sumerian’s own account. The company sets out its dispute with the regulator on its website.

Prime Rogue has asked the Alberta Energy Regulator and the Orphan Well Association: – which of Sumerian’s wells, lines and facilities on the base had been shut in and secured by Oct. 17; – whether any have been inspected since the fire; – why the company has not been given access.

We will add their answers when they come.

What comes next

Rubicon Ridge is the first of a series. Each part tests one piece of what happened on Oct. 17, and each is a working model that later data can confirm or overturn.

• v5.1, Rubicon Cloud: the air response. How the air ambulances and the search-and-rescue aircraft reached the base after the fire, from public flight data.
• v5.2: what the soldiers could see. The fire’s behaviour where it reached the ranges: how wide and how intense the front was, and how long before it arrived it came into view.
• v5.3, Rubicon Casualties. Tests the v5 fire against the reported toll: what warning time and what conditions on the ground fit four dead and 39 hurt.
• v6: the fire’s timing from new satellite data.

• New satellite data: radar coherence, thermal passes and further burn edges from Sentinel-2 and Landsat. These will pin down where the fire’s edge lay at the ranges and when it got there.
• Whether the gas wells and lines on the base played any part, using the regulator’s and the Orphan Well Association’s records once we have them.
• After v6: the air and medical response in detail (v6.1), the chain of decisions behind the training that day, from access-to-information records (v7), and a reconciliation with the official findings when they come (v8).

Method, data and sources

v5 is a 40 m grid fire-spread model using the Canadian Forest Fire Behaviour Prediction System for grass, run 2,160 times and weighted by fit to the burn scar. Full method, limits and code are in the archive.

• Model write-up: Rubicon Ridge, working model v5
• Model outputs and the v4 comparison: model v5 folder
• Weather station data and front-passage table: Rubicon Ridge raw data; Environment and Climate Change Canada historical data; Alberta Climate Information Service
• Air sensors: PurpleAir; our particle analysis
• Satellite: GOES visible-imagery analysis; polar-orbiter pass list; NOAA GOES on AWS
• Burn measurements from the Oct. 20 image: Satellogic-derived vectors
• Range locations: 2024 Range Standing Orders analysis
• Gas wells and pipelines near the site: Sumerian Oil & Gas mapping
• Working model v4, as published: v4-published
• DND briefing, Oct. 22: archived machine transcript
• Earlier coverage: CFB Wainwright fire live updates; the air response, Rubicon Cloud (v5.1)

Prime Rogue’s data, analysis and figures are CC BY-NC 4.0. Weather data: Environment and Climate Change Canada and Alberta Agriculture and Irrigation (Open Government Licence). Satellite data: NOAA.

More on the CFB Wainwright fire

• Live updates: CFB Wainwright fire
• The flank became the front
• Every warning light was on: what most likely happened at CFB Wainwright
• The storm line, then the fire
• CFB Wainwright fire deaths were highly likely preventable
• No GPS trail at CFB Wainwright: fire probe rides on range logs
• DND said the fire “ignited” at 4 p.m. The military now says it was known for hours
• Satellites saw fire on CFB Wainwright range hours before DND says it ignited
• Gas wells played no part in the CFB Wainwright fire, records show
• CFB Wainwright deaths: who is investigating
• ATIP request tracker