A grognard exercise, not a scientific model. What that means

Roadmap / 0.1.3.9.2

Potash sites: positions and imagery

Version 0.1.3.9.2, 19 Nov 2026. tag v0.1.3.9.2, commit d3260ad. File docs/RESULTS_v0.1.3.9.2_potash_positions_imagery.md.
Written at this version and not revised since. Later versions may change these numbers; see the roadmap.

Switch: none. The model is unchanged from v0.1.3.8; results are identical and the golden check passes.
Runs: none new. This version fills in the potash site list.
Scope: unchanged. A site record says what a place is, who holds it, where (to about a kilometre), how much, its status and its source. Still out: the inside of any site, how a site could be damaged, rankings by harm, routes, methods and tactics.

Headline

  1. Every potash site that is a place now has a location. 33 of the 44 entries, up from 11. The other 11 are blocks of mineral rights, which are areas and have no single point.
  2. 32 site pages now carry satellite imagery. Public 10-metre Sentinel-2, one clean pass per site between 13 July and 6 November 2026, shown plain: a frame 5 km across and, for 28 of them, a frame 15 km across.
  3. Each position says how it was established. 13 come from a filed document or a government survey point, 17 were found and identified on satellite imagery, and 3 are a published locus for a project that has not been built.
  4. No result changes. Positions and pictures do not enter the simulation. They are what districts will be drawn from.

Where it came from

Roger Wilco Imagery's Canada potash baseline, built on 19 November 2026 from Kevin's potash inventory and published at https://github.com/Prime-Rogue-Inc-Corporate/Canada-Potash-Sites. It covers 34 places: every inventory entry with a physical footprint. tools/import_rwi_potash.py reads the delivery and writes the registry and the page images, so the step can be repeated when a new pass arrives.

The 11 locations the registry already held all agree with the baseline to the two decimals the registry keeps.

What was taken, and what was left

Taken - Each site's position, rounded to two decimals of a degree, and its confidence grade. - Both frames for each site, re-saved at web size (500 and 1,000 pixels across). The full-size originals are in the baseline's own repository. - Each frame's scene and date, and the credit and licence lines.

Left out, under the registry's rules - The separate entry for a shaft at Rocanville. A part of a site is not a site in this registry. Rocanville's own record covers it. - What a precise point marks. Several of the baseline's best positions are quoted from filings to a few metres and name the structure they mark. The registry keeps the point to about a kilometre and records only how it was established. - The transport table. The baseline measures each site's distance to the nearest highway, railway and airport. That is not part of a site record, and the ways in to a site sit on the wrong side of this project's line. The loader now refuses fields named routes, approach or access, and a test checks it. Where potash leaves the country is already in the list as the export terminals. - The contact sheets and the monitoring watchlist. The sheets carry labels; the watchlist belongs to the imagery pipeline.

Whether the transport table belongs in the game at all is Kevin's call. If it does, it should come in as a property of a district (how well connected it is), not of a site.

Positions

Grade A is a filed document or a government survey point. B is found and identified on satellite imagery. C is a published locus only.

ID Site Location Grade Scene date Frames on the page
T1-01 Rocanville 50.47° N, 101.55° W A 2026-10-22 5 km
T1-02 Allan 51.93° N, 106.07° W A 2026-09-28 5 km and 15 km
T1-03 Lanigan 51.86° N, 105.21° W A 2026-09-28 5 km and 15 km
T1-04 Cory 52.09° N, 106.85° W A 2026-07-28 5 km and 15 km
T1-05 Vanscoy 52.01° N, 107.09° W A 2026-07-28 5 km and 15 km
T1-06 Patience Lake 52.10° N, 106.37° W B 2026-07-30 5 km
T1-07 Esterhazy 50.65° N, 101.99° W A 2026-09-27 5 km and 15 km
T1-08 Belle Plaine 50.43° N, 105.20° W A 2026-11-06 5 km and 15 km
T1-09 Colonsay 51.94° N, 105.77° W B 2026-09-28 5 km and 15 km
T1-10 Bethune 50.65° N, 105.38° W B 2026-09-25 5 km and 15 km
T1-11 Harrowby 50.75° N, 101.45° W B 2026-10-04 5 km
T2-01 Esterhazy K1 50.73° N, 101.93° W A 2026-09-27 5 km and 15 km
T2-02 Esterhazy K2 50.66° N, 101.84° W A 2026-09-27 5 km and 15 km
T2-03 Picadilly 45.77° N, 65.37° W B 2026-08-05 5 km and 15 km
T2-04 Penobsquis 45.77° N, 65.37° W B 2026-08-05 5 km and 15 km
T3-01 Jansen Stage 1 51.90° N, 104.71° W A 2026-09-28 5 km and 15 km
T3-02 Jansen Stage 2 51.90° N, 104.71° W A 2026-09-28 5 km and 15 km
T3-03 Bethune Ramp-Up 50.65° N, 105.38° W B 2026-09-25 5 km and 15 km
T3-04 Milestone Phase I 50.19° N, 104.30° W B 2026-10-25 5 km and 15 km
T3-05 Tugaske 50.87° N, 106.29° W C none none
T4-01 Southey 51.07° N, 104.52° W A 2026-07-17 5 km and 15 km
T4-02 Milestone full scale 50.19° N, 104.30° W B 2026-10-25 5 km and 15 km
T4-03 Muskowekwan 51.31° N, 104.10° W C 2026-07-17 5 km and 15 km
T4-04 Wynyard 51.70° N, 104.02° W A 2026-09-25 5 km and 15 km
T5-11 Eyebrow 50.80° N, 106.15° W C 2026-09-25 5 km and 15 km
T7-01 Neptune Bulk Terminals 49.31° N, 123.05° W B 2026-08-07 5 km and 15 km
T7-02 Portland Bulk Terminals 45.63° N, 122.78° W B 2026-07-13 5 km and 15 km
T7-03 Saint John Terminal 45.26° N, 66.06° W B 2026-09-19 5 km and 15 km
T7-04 Thunder Bay Terminals 48.38° N, 89.22° W B 2026-10-17 5 km and 15 km
T7-05 Westshore Terminals 49.02° N, 123.17° W B 2026-08-07 5 km and 15 km
T7-06 Port of Churchill 58.78° N, 94.19° W B 2026-09-04 5 km and 15 km
T7-07 Prince Rupert potash terminal 54.23° N, 130.33° W B 2026-08-21 5 km and 15 km
T7-08 Longview export terminal 46.10° N, 122.94° W B 2026-10-11 5 km

Positions to treat as unsettled

  • Picadilly and Penobsquis (T2-03, T2-04). Two works stand about 2 km apart near Sussex, New Brunswick. The baseline found both and could not establish which name belongs to which. Both records carry the same point, and both frames show both works.
  • Harrowby (T1-11). Identified on imagery at the edge of what 10 m can show. Provisional.
  • Saint John terminal (T7-03). A bulk-loading jetty with covered storage, identified on imagery. Not confirmed as the potash berth.
  • Tugaske, Muskowekwan, Eyebrow (T3-05, T4-03, T5-11). Not built. The point is a village or reserve centroid. Eyebrow had no location in the registry before; it now has this one, at grade C.
  • Southey and Wynyard (T4-01, T4-04). Not built. The point is the province's survey point for the permit area.

Imagery

  • What it is. Copernicus Sentinel-2, true colour, 10 m to a pixel: the lowest-cloud pass over each site between 1 May and 17 November 2026. Contrast is stretched frame by frame, so brightness cannot be compared between sites.
  • How it is shown. As it came. Nothing is marked, labelled or outlined on any frame.
  • Credit. "Imagery: Roger Wilco Imagery. Contains modified Copernicus Sentinel data 2026, processed by Roger Wilco Imagery."
  • Licence. The frames carry the Canada Potash Sites licence: free to use with credit for non-commercial purposes, and not for government use. Each page and the Sites index say so and link to the baseline.

Frames to re-pull

Six frames fall at the edge of a satellite tile and are partly empty. They were not used. Their sites show the frame that is whole; Tugaske has no whole frame and shows none.

ID Site Frame Empty
T1-01 Rocanville 15 km 11%
T1-06 Patience Lake 15 km 25%
T1-11 Harrowby 15 km 8%
T3-05 Tugaske 5 km 44%
T3-05 Tugaske 15 km 48%
T7-08 Longview 15 km 16%

The fix is on the imagery side: build each frame from the two neighbouring tiles.

Two points from the baseline's own notes, for Kevin

  • The licence is not a Creative Commons licence. Adding "no government use" to CC BY-NC makes it a licence of its own. The baseline says so, and this site names it "Canada Potash Sites licence" for that reason.
  • The transport figures are built on OpenStreetMap, whose licence may require share-alike for a derived database. The baseline flags that as a question for counsel. It does not arise here, because the transport table was not imported.

What changed in the code

  • data/sites/potash.yaml: 22 new locations; a loc_grade on all 33; notes on the unsettled ones; a new source key, rwi.
  • forceonforce/sites.py: LOC_GRADES; the loader checks grades and refuses routes, approach and access.
  • tools/import_rwi_potash.py: the import.
  • tools/site/sites/: 60 frames and imagery.yaml.
  • tools/build_site.py: two frames per site page with date, credit and licence; how each point was placed; an imagery line on the Sites index.
  • Tests: 36 (was 35).

Ahead

Unchanged from v0.1.3.9.1: districts in the simulation, then output lost by district on a schedule, then US food supply and agriculture. The positions in this version are what the first of those will be drawn from.