Self-evaluation #1 — SITREP 005 against Prime Rogue's Wainwright modelling
Bottom line
SITREP 005 grade: C. The satellite collection was strong and independently reproduced Prime Rogue's figures to the decimal. The analysis, however, trailed work Prime Rogue had already published and got two things wrong.
- Miss 1 — the house's own reporting was not read first. Prime Rogue had published five Wainwright pieces between 21 and 24 Oct. A web search did not surface them, and SITREP 005 went out without citing them.
- Miss 2 — the troop-contact time was wrong. SITREP 005 placed contact at ~15:55 from Maj.-Gen. Rutland's "just before 4 p.m.", calling the 1,215 MW GOES peak a run "25–30 minutes before" it reached the soldiers. Station data (an 82 km/h gust at 15:16) and Prime Rogue's spread model put the fire at the Range 2 area around 15:19–15:26. The 15:30 peak is most likely the run that reached them.
- Miss 3 — the origin was paired with the wrong range. SITREP 005 tied the origin to the OpenStreetMap "Range 2 (Grenade)" and said this fit the automatic-grenade-launcher account. Prime Rogue identifies the AGL range as Range 11, with Range 2 as C Company's area. SITREP 005 also did not address the gas-line question.
What SITREP 005 added that the benchmark did not:
- the full VIIRS overpass record, including four passes with no fire before 14:50 and a near-overhead NOAA-21 pass at 13:54;
- a separate fire in Permanent Danger Area 2 on 14 Oct;
- about 0.9 km² burned outside the base boundary, with heat still there on 18 Oct.
Status: all corrections below were issued in SITREP 005 revision 1 on 25 Oct 2026.
The benchmark: Prime Rogue's ten most recent posts
Five of the ten are a Wainwright series that sets the standard for this kind of work. The other five are essays or indices with no overlap with RWI's imagery work.
| Date | Post | Kind | How it is built | Overlap with RWI |
|---|---|---|---|---|
| 24 Oct | The Noon Photograph | Model + imagery | Landsat 8 12:17 SWIR, 34 FIRMS 30 m pixels, FBP O-1a spread model v3 fitted to the burn, AER pipeline data, Sentinel-2/-5P, GOES-18 | Direct: ignition 1.4–2.2 km east of Range 11, ~03:30–07:00; 0.44 km² burned by noon |
| 24 Oct | Another sign pointing away from gas | Hypothesis test | Six tests: location vs firing sector, timing, GOES persistence, burn severity at the lines, Sentinel-5P methane/aerosol, regulator record | Direct: RWI did not test gas |
| 23 Oct | The Pot Calls the Kettle Black | Documentary chronology | ~45 sources, Confirmed vs Not-proven lists | None |
| 22 Oct | What Happened to C Company | Reconstruction | Forecast/warning timeline, four-station front fitting, spread model, five candidate positions, escape-odds model with stated confidence | Direct: arrival ~15:21 at Range 2; burn 10.8 km² |
| 21 Oct | How we found the 82 km/h gust | Method guide | ECCC SWOB XML for station CZWN; gust 82 km/h from 269° at 15:16 MDT; file hashed and timestamped | Direct: RWI had no station data |
| 21 Oct | Alberta medical examiner's investigation | Jurisdiction reporting | Statute, precedent (Moher 2016), one background source, questions to three offices | Context |
| 1 Jul | Intelligence Index No. 001: Ireland | Graded index | Eight-dimension letter-grade rubric; FATF and commission records | None now; possible consumer of RWI data later |
| 24 Jun | Canada–Switzerland Round of 32 scenarios | Scenario tree | Tiebreaker arithmetic, qualitative odds | None |
| 22 Jun | Cold War tech still running | Explainer | Narrative survey of eight systems | None |
| 8 Jun | Qatar's small-state playbook | Essay | Open reporting and secondary literature | Background for Al Udeid |
The Wainwright series already does three things RWI was built to add: versioned models with dated corrections, explicit confidence levels, and hashed, timestamped evidence.
Wainwright, side by side
On shared numbers the two agree exactly. Where they differ, the benchmark is better supported on ground truth; SITREP 005 is stronger only on the satellite pass record.
| Question | Prime Rogue (21–24 Oct) | SITREP 005 (25 Oct, first issue) | Verdict |
|---|---|---|---|
| GOES-18 fire power | 76.5 MW 15:10, 1,215 MW 15:30, 336 MW 15:40, then none | Same values; GOES-19 never detected it | Agree — independent replication |
| Burned area | ~10.8 km²; ~9% outside PDA 2 | ~10.7 km²; ~0.9 km² outside the base boundary | Agree; 005 adds the off-base part |
| Origin | 1.4–2.2 km east of Range 11 (the AGL range) | Near OSM "Range 2 (Grenade)" | 005 misleading — corrected in rev. 1 |
| Fire at noon | Landsat 12:17: strip 2.0–2.3 km × 260–420 m, ~0.44 km² | Called "small" until ~14:30 | 005 weaker — low-intensity, not small |
| Ignition time | ~03:30–07:00 from the spread model | Not estimated | Benchmark ahead |
| Before 14:50 | Not covered | Four VIIRS passes with no fire; GOES clear and blank 12:00–14:40 | 005 adds |
| Front arrival | 82 km/h gust from 269° at 15:16 (CZWN); SW→NW shift; 23.3→12.9 °C | Cloud edge at 14:50 as a proxy | Benchmark measured; 005 inferred |
| Contact with troops | ~15:21 at Range 2 (15:19–15:26) | ~15:55 | 005 wrong — corrected in rev. 1 |
| Gas lines | Six tests; points away from gas; brief release not ruled out | Not addressed | 005 gap — note added in rev. 1 |
| Earlier fires | Not covered | Separate fire in PDA 2 on 14 Oct (~0.9 km²) | 005 adds |
| After the fire | No overnight hot spot | Off-base heat 18 Oct 14:06–14:31 MDT | 005 adds |
The VIIRS "no fire" passes at 13:54 and 14:31 do not contradict a 0.44 km² burned strip at noon. A slow, low-intensity grass front (~0.45 km/h in the benchmark model) can sit below VIIRS's detection threshold. After revision 1, SITREP 005 also found the 14:50 fire pixels 2.1–4.4 km east-north-east of Range 11, which independently supports the benchmark's origin model.
Why it went wrong
The errors came from process, not data.
- House reporting not checked first. A web search returned none of the five posts; reading the blog directly would have taken one request.
- A quote treated as a timestamp. "Just before 4 p.m." is an approximation. Measured station weather existed and was public.
- A map label read as identification. The nearest grenade-named range was treated as the AGL range.
- A reported question skipped. Military sources had told Global News a gas line or well was hit; the claim was never tested.
- No surface weather in the pipeline. RWI had satellites, thermal detections, AIS and ADS-B, but no ground stations.
- Landsat gap accepted. The archive lags about two weeks; the benchmark used the FIRMS Landsat fire product instead.
- Speed over review. The report was issued about an hour after the data pull, with no independent read.
Changes adopted
- House check first. Every SITREP and incident report begins by reading primerogueinc.com, cites matching posts, and states what it adds.
- Live-questions list. Open questions in the reporting are listed before writing, then answered or explicitly parked.
- Surface weather. ECCC station data (wind, gust, direction, temperature, humidity) to be added for Canadian sites.
- Landsat fire product. A FIRMS map key to be added for same-day Landsat 30 m fire detections.
- Measured over quoted times. An approximate official time is never used as an anchor when a measured time exists.
- Independent read before publishing incident reports, by a reviewer that did not draft them.
Scorecard
| Dimension | SITREP 005 | Prime Rogue Wainwright series | Note |
|---|---|---|---|
| Data collection | A− | A | 005: 1,440 GOES scans, 136 VIIRS overpasses, viewing geometry. Benchmark adds Landsat, station weather, Sentinel-5P, regulator data |
| Accuracy of shared numbers | A | A | GOES values and burned area match |
| Analytic judgement | C− | A− | 005: wrong contact time, misleading range pairing |
| Coverage of live questions | D | A | 005 skipped gas, ignition time, escape window |
| Original contribution | B | A | 005: pass record, 14 Oct fire, off-base burn. Benchmark: ignition model, gust, gas tests |
| Sourcing and credit | D | B+ | 005 missed the house's own posts; benchmark's jurisdiction piece rests on one anonymous source |
| Uncertainty and confidence | B | A | Benchmark gives numeric confidence and model versions |
| Corrections and provenance | B | A | Benchmark already hashes, timestamps and logs corrections |
| Overall | C | A− |
The benchmark, graded
The Wainwright series earns A−; the ten-post set overall a B, pulled down by thinly sourced essays.
Why it is the model
- Falsifiable claims. The gas analysis states six tests that could have pointed toward gas, reports that none did, and names what it cannot rule out.
- Measured times over quoted times. The 82 km/h gust is taken from the raw station file, cross-checked against the 44-knot remark, and hashed.
- Versioned models, corrected in the open. The model went from v2.1 to v3. Arrival times were corrected by 5.4 and 10.2 minutes, and the well location moved 410 m.
- Numeric confidence. For example, 80–90% on C Company's area and 70–75% on the attribution.
- Named records requests. Questions went to the OCME, the Justice Minister, DND, the Orphan Well Association, the operator and the regulator.
Where it is weaker
- The attribution outruns the evidence. The 70–75% judgement that the failure sits in range awareness or communication rests on modelled positions DND has not confirmed.
- The 03:30–07:00 ignition window is model-only. It is back-extrapolated from two snapshots. Five of six VIIRS night passes on 17 Oct (02:24–05:00 MDT) saw nothing, and the sixth was unreadable. That is consistent with a small fire but adds no support.
- Range 11 as the AGL range is the load-bearing assumption of the origin model, and its sourcing should be confirmed.
- The jurisdiction piece rests Alberta's legal position on one anonymous official.
- The non-Wainwright essays (Qatar, Cold War technology) are thinly sourced against this standard.
What RWI adopts from it: versioned models with a corrections log, numeric confidence, a tests-that-could-fail section, and named records requests in every incident report.