The Noon Photograph: What Landsat Shows About Where The CFB Wainwright Fire Started, and The Gas Lines It Crossed

Kevin J.S. Duska Jr. · Prime Rogue Inc. · Calgary, AB · Oct. 24, 2026

Working model v3

At 12:17 p.m. on Oct. 17, a Landsat satellite passed over CFB Wainwright and photographed the fire that would, three hours later, overrun soldiers of C Company. It is the only close-up of the fire while it burned. We obtained the full image and rebuilt our model around it. It puts the start in the grenade launcher range’s impact area, where officials said the fire began. It shows the gas lines cold at noon. And together with every other record we have, it is another sign pointing away from gas.

Landsat 8, 12:17 p.m. MDT, Oct. 17, shortwave infrared. Graphic: Kevin Duska Jr. (Prime Rogue Inc.), CC BY-NC 4.0. Data: USGS, NASA FIRMS, AER, Copernicus Sentinel-2.

The short version

• By noon, the fire had burned a strip about 2 km long, pushed northeast by a southwest wind. Its back end, near where a wind-driven fire starts, lies 1.2 to 1.5 km east of Range 11, the automatic grenade launcher range.
• Our updated model (v3) puts the start in Range 11’s impact area: rounds landing 1.4 to 2.2 km out, between about 3:30 and 7 a.m.
• No gas line or well lies in the grenade launcher’s firing sector within its reach, and the gas lines were cold at noon. The fire reached them around 3:25 p.m., after the wind shift.
• There is no sign of a sustained gas fire or leak: no overnight hot spot, no scorch circle at the lines, an unburned well site, no sign of a large methane leak the next day. A brief release as the fire crossed the lines cannot be ruled out.
• But the gas lines were probably live. The regulator had ordered the wells shut in on Oct. 5; the operator says it was told on the evening of Oct. 17 that none had been.
• Essentially nothing changes for C Company. Our escape findings stand, and we set out below what we still infer about where the company was and what it could see. We correct two arrival times.

1. What the satellite saw at noon

Landsat 8 passed over the base at 12:17 p.m. MDT. Thin, cold cloud covered the area, but fire shows through in the satellite’s shortwave-infrared bands. We found 34 active-fire pixels, each 30 m across; 47 of the 59 detections NASA’s fire service logged from the same pass fall within 45 m of them.

The full bands show more than the fire pixels. The flames ringed a strip of ground that had already burned: about 2.0 to 2.3 km long and 260 to 420 m wide, running southwest to northeast. That is the shape of a fire pushed by a steady wind for hours. Measuring burned ground under cloud is imprecise, so we give a range: 0.26 to 0.84 km² had burned by 12:17 (central estimate 0.44 km²).

A wind-driven grass fire grows mostly downwind, so its start lies near the back of the burned strip. Here that back end is 1.2 to 1.5 km east of Range 11, on a bearing of about 80°, inside the C16 grenade launcher’s reach of about 2.2 km and its likely firing sector. Officials said at the Oct. 22 briefing that the fire started on the automatic grenade launcher range; the base’s standing orders, as reported by Global News, identify that as Range 11.

2. Where and when it started: working model v3

Our working model simulates the grass fire minute by minute on the base station’s hourly weather, with its afternoon pace tuned to what satellites measured. Until now it was fitted to the satellites’ fire detections. Version 3 fits it to the whole burned area Landsat photographed, and lets the fire’s width vary to fit it as closely as it can.

• Where: the best fit is unchanged: a line of starts 1.4 to 2.2 km east of Range 11 along about 80°, as from rounds landing along the line of fire.
• When: between about 3:30 and 7 a.m., depending on how fast the fire ran in the cool morning. In every case, it was burning for hours before noon.

Caveat: this is a model fitted to two satellite snapshots, not an observation of the start. Its fire at 12:17 (0.9 to 1.1 km²) is still wider than Landsat’s central outline; thin cloud probably hides some burned ground. Range 11’s position comes from OpenStreetMap; its firing direction is not public.

3. The gas question

The night of the fire, military sources told Global News the live-fire exercise “was believed to have hit a gas line or gas well on the base, causing an explosion.” Maj.-Gen. Wade Rutland said his “initial impression is it’s not, but we will definitely check.” The claim spread widely; the Canadian Press fact-checked several versions. CFB Wainwright sits on an active gas field, so it is not far-fetched on its face. We tested it.

In the grenade launcher’s line of fire.

We mapped every licensed pipeline and well. None lies within 2.5 km in Range 11’s likely firing sector (bearings 40° to 120°), or even in a generous one (20° to 160°). The nearest gas lines are 350 to 560 m behind and beside the firing point. The gas lines at Range 2 and the road head are 3.5 to 4 km away, beyond the C16’s maximum range of about 2.2 km.

At noon.

Landsat shows no heat near any gas line: within 600 m of every operating line, readings sit at background. The nearest flames were 0.86 km from a gas line. The fire reached the Range 2 and road-head lines around 3:25 p.m., minutes after the wind turned.

In the afternoon and overnight.

The GOES-18 weather satellite saw 76.5 MW of fire at 3:10 p.m., 1,215 MW at 3:30 spread across three pixels along the front, 336 MW at 3:40, then nothing. From about 7:30 p.m. the sky was clear all night, with no hot spot. A ruptured line feeding a fire would have shown.

On the ground.

A gas fire at a break leaves a scorched circle on the line. Sentinel-2 images before and after show none. Within 30 m of the lines the fire burned lighter, not hotter, and the well’s registered site did not burn at all.

In the air.

The Sentinel-5P satellite passed at 2:34 p.m., before the fire reached the lines, with cloud over the base, so it cannot settle the question: burning gas makes little smoke, and a short release would spread out and dilute within hours. But on Oct. 18, with good coverage around the base, it found no sign of a large methane leak at or downwind of it.

4. The paperwork: wells ordered shut, still live

On Oct. 5, the Alberta Energy Regulator suspended Sumerian Oil & Gas Inc., the licensee of the Range 2 lines, and gave the Orphan Well Association 14 days to shut in and secure its wells, facilities and pipelines. The amended order lists the well at Range 2, 1-11-44-7W4, as active. In an Oct. 19 statement, Sumerian said an association contact told it on the evening of Oct. 17 that “none of the Sumerian wells on the range had been shut in.” The regulator said the day after the fire it had received no incident reports about infrastructure on the base. National Observer and Bloomberg have reported on the order.

So the lines the fire crossed were probably still under pressure, with no operator on site. That is why the physical evidence above matters, and why we are asking for the records below.

5. What we still infer about C Company

DND has not said where C Company was or what it could see. What follows comes from public data and our model. It is inference, not a finding, and we will revise it as records come out. Graphics and methods for each point are in our earlier reconstruction.

• The forecast. Environment Canada’s forecast, issued at 4:30 a.m., called for west winds gusting to 80 km/h that afternoon. At 2:14 p.m. it warned of gusts up to 100 km/h. An 82 km/h gust turned the fire at 3:16 p.m.
• Where they were (low to moderate confidence). Maj.-Gen. Rutland said the fire was a kilometre away when the wind shifted. Under all 12 wind directions we tested, the band about a kilometre downwind of the fire’s edge lands around Range 2 and the end of the range road. Read 20 different ways, 16 put most of the likely ground there, give or take about 400 m. Rutland has said the fire overtook a section-attack range’s administrative area; our analysis locates ground, not range names.

• What they could see. A low rise sat between the fire and most of that ground. From 46 to 63% of it, a standing soldier could not see the fire’s edge before the wind turned; lying prone, from about two-thirds. At the two likeliest spots, flames came into view only 3 to 11 minutes before the fire arrived. If the smoke stayed low, our model puts visibility at tens of metres.
• Where the fire was. Before the front it moved slowly, about 0.45 km/h, northeast and away from Range 2. At 3:16 p.m. its edge was roughly 430 to 590 m from the Range 2 point, depending on the method: 566 to 591 m on working model v3, which holds it near where satellites last saw it at 2:49 p.m.; about 425 to 480 m if it kept creeping until the front, as our earlier fire-edge test found.
• The escape window. On our model, about four in five soldiers who left the Range 2 area at the 3:16 gust got clear, heading east across the wind; three minutes later, about two in five; seven minutes later, one in ten. Every model version and starting fire we have tested gives the same pattern. Moving at the gust, two-thirds of the likely ground had better-than-even odds; with the fire 200 m away, none did.

6. What this changes for C Company: essentially nothing

What happened to C Company depends on where the fire’s southeastern edge was when the wind turned at 3:16 p.m. In three of the four version 3 cases, as before, that edge is set by where satellites saw fire at 2:49 p.m.; in the fourth, the modelled fire reaches 25 m closer to Range 2.

	Published (v2.1)	Working model v3
Fire’s edge to the Range 2 point at 3:16 p.m.	586 m (part 6: 430-480 m)	566-591 m
Fire’s edge to the road head at 3:16 p.m.	1,105 m	1,047-1,109 m
Escape odds from the Range 2 area, jogging: leaving at the gust / 3 min later / 7 min later	80% / 43% / 10%	80% / 40-43% / 10%
Same, faster fires (8-12 km/h)	70% / 15% / 0%	70% / 10-15% / 0%
Escape odds from the road head	100% / 83% / 53%	100% / 80-83% / 50-53%
Fire’s arrival after the 3:16 p.m. front: Range 2 point / road head (median)	5.4 / 10.2 min (corrected)	5.4 / 10.0-10.2 min

What we can’t rule out

A brief gas release, seconds to minutes, as the fire crossed the lines around 3:25 p.m. It would fit inside a single GOES scan, below Sentinel-2’s resolution, and after the Sentinel-5P pass. Reports of an explosion could also have other causes: munitions in the fire, or violent fire behaviour in an 82 km/h gust. Satellites can’t tell these apart.

What would settle it

• The result of the army’s check of the oil and gas infrastructure, which Rutland described the day after the fire.
• The Orphan Well Association’s and Sumerian’s records for Oct. 17: pressures, shut-ins, damage.
• What Sumerian’s 7:04 p.m. call to the regulator’s 24-hour line on Oct. 17 (report 460044, per Sumerian) concerned.
• Range 11’s firing records: what was fired, on what bearing, and when range control first saw the fire.

We are putting these questions to DND, the Alberta Energy Regulator, the Orphan Well Association and Sumerian, and will update this article when they reply.

Corrections

• Arrival times. We previously reported that the fire reached the Range 2 point a median 6.4 minutes after the 3:16 p.m. front, and the road head 13.2 minutes after. Measured consistently across all 30 modelled fires, the figures are 5.4 and 10.2 minutes. The earlier numbers came from a different set of runs. Escape odds are unaffected.
• Well location. Earlier work placed “the wellsite” at the mapped west end of gas line 58333-78, about 170 m from the Range 2 point. The regulator’s registered location for well 1-11 is about 360 m north of the road head, outside the burn. Both positions are mapped, not surveyed.

Methods and data

Everything here is reproducible from public data. Raw Landsat files: CFB-Wainwright-Landsat-8-9. Model v3: archive, model_v3. Gas test: archive, part 9. Sentinel-5P: its own repository. Sources: USGS Landsat 8; NASA FIRMS; NOAA GOES-18; Copernicus Sentinel-2 and Sentinel-5P; Environment and Climate Change Canada; Alberta Energy Regulator licence data and orders; Global News; the Canadian Press; Sumerian Oil & Gas. The fire model is a calibrated grass-fire spread model (Canadian Forest Fire Behaviour Prediction System, fuel type O-1a).

This is a working model, not a final finding. Further data will confirm or disconfirm it until, or unless, we build a new one. Updates on our CFB Wainwright Live Updates Page

FAQ

Where did the CFB Wainwright fire start?

In the impact area of Range 11, the automatic grenade launcher range, about 1.4 to 2.2 km east of the range on our model. Landsat’s 12:17 p.m. image puts the back of the burned strip 1.2 to 1.5 km east of Range 11. Officials have said the fire began on the grenade launcher range.

When did it start?

On our model, between about 3:30 and 7 a.m. on Oct. 17. It had burned a strip about 2 km long by 12:17 p.m.

Was a CFB Wainwright gas line hit?

Not by a round from the grenade launcher range: no gas line or well lies in its firing sector within reach. The fire reached gas lines at Range 2 and the road head around 3:25 p.m.; a brief release then is not ruled out.

Were the gas wells shut in?

Probably not. The regulator ordered them shut in on Oct. 5; the operator says it was told on Oct. 17 that none had been.

Does this change what happened to C Company?

Essentially no. The fire’s edge at 3:16 p.m. moves by at most 25 m toward Range 2 across our model cases, and the escape odds are unchanged or within a few points.