Roadmap / 0.1.3.9.1
Potash baseline
v0.1.3.9.1, commit fd2d06d. File docs/RESULTS_v0.1.3.9.1_potash_baseline.md.
Written at this version and not revised since. Later versions may change these numbers; see the roadmap.
Switch: potash, with and without potash.embargo. The model is unchanged from v0.1.3.9.
Runs: 3,000 paired runs per variant, three variants, three scenarios (27,000 runs). Seed 20261105.
Scope: district and province totals. No site is switched on or off in these runs.
Headline
- This is the line the district work will be measured against. Before any district's output is taken away, the United States already loses a large share of its potash in every scenario: at the worst month, 38% of US potash use in a full invasion, 44% in the Alberta scenario and 51% with no invasion. The shortfall then eases to between 11% and 28% over two years as other suppliers fill in.
- In a full invasion that shortfall changes almost nothing. With potash switched off, not one of 3,000 wars ends differently and 7 end in a different month. In the Alberta scenario 75 wars end differently, and with no invasion 344 do. The module's worth is unchanged from v0.1.3.8: 0.0, +2.5 and +11.5 points on "the United States gives up".
- A deliberate Canadian embargo is worth less than the potash that is lost anyway. It lifts the worst month to 73% of US use where Canada still holds the mines, and adds +2.8 points in the Alberta scenario and +4.7 with no invasion. It adds nothing in a full invasion. It costs Canada US$0.9 to 3.3 billion in sales.
- By district, the south-east matters most to the US. If US-bound potash comes from every mine in proportion to its capability, south-east Saskatchewan supplies 28% of all US potash use, the Saskatoon district 22%, Regina and Moose Jaw 13%, and east-central 9%.
The baseline
Share of US potash use that goes unmet. "Worst month" is the peak in each run; the range is the 10th to 90th percentile of runs.
| Scenario | Worst month, average | Range | Runs with a worst month over 50% | Potash's worth, points | Wars that end differently without potash |
|---|---|---|---|---|---|
| Full invasion | 38.0% | 30.0% to 46.1% | 2.2% | 0.0 (0.0 to 0.0) | 0 of 3,000 |
| Alberta only | 43.7% | 35.5% to 51.7% | 15.4% | +2.5 (+1.9 to +3.1) | 75 of 3,000 |
| No invasion | 51.0% | 42.7% to 58.9% | 58.1% | +11.5 (+10.3 to +12.6) | 344 of 3,000 |
How the shortfall runs over time
Average share of US potash use unmet, among wars still going in that month.
| War month | Full invasion | Alberta only | No invasion |
|---|---|---|---|
| 2 | 37.2% | 41.0% | 49.0% |
| 6 | 31.1% | 31.1% | 41.0% |
| 12 | 22.4% | 19.5% | 30.5% |
| 24 | 17.8% | 11.3% | 27.9% |
| 36 | 17.2% | 10.8% | 27.8% |
The fall has two causes in the model. Other suppliers replace up to about 45% of the lost potash over about a year. And where US forces hold the mines, they run them at about 60% of normal output, which is why the two invasion scenarios fall further than the one without an invasion.
If Canada cuts the potash off itself
The same runs with a Canadian embargo on US-bound potash from the first month.
| Scenario | Worst month | Month 12 | Month 24 | Added to "the US gives up" | Canadian sales lost |
|---|---|---|---|---|---|
| Full invasion | 48.7% | 32.4% | 29.9% | 0.0 (−0.1 to +0.1) | US$0.9 billion |
| Alberta only | 72.7% | 43.4% | 39.7% | +2.8 (+2.2 to +3.4) | US$3.3 billion |
| No invasion | 72.7% | 43.6% | 39.8% | +4.7 (+3.9 to +5.4) | US$2.7 billion |
72.7% is the ceiling: Canada supplies 79% of US potash imports and the US imports 92% of what it uses.
What each district supplies
Operating capability from the site list (v0.1.3.9), and what that would be as a share of US potash use.
| District | Capability, Mt KCl a year | Share of Canada | Share of US potash use |
|---|---|---|---|
| South-east Saskatchewan | 10.8 | 39.1% | 28.4% |
| Saskatoon district | 8.2 | 29.7% | 21.6% |
| Regina and Moose Jaw district | 5.0 | 18.1% | 13.2% |
| East-central Saskatchewan | 3.5 | 12.7% | 9.2% |
| Western Manitoba | 0.1 | 0.4% | 0.3% |
This rests on one assumption: that potash bound for the US is drawn from every mine in proportion to its capability. The inventory does not say which mine ships where, and Canpotex's offshore sales are not spread evenly across mines.
What it means
- The model is far more sensitive to potash when there is no shooting. In a full invasion the US loses over a third of its potash and the war goes the same way. That may be right: a war with thousands of dead is not decided by fertilizer. It may also be a weakness. The shortfall reaches US politics through two numbers set by hand, one for lost output and one for lost support, and neither knows anything about planting seasons beyond a single multiplier, about yields, or about food prices.
- That is the case for the agriculture work. Before a district is switched off and the difference read, the path from a potash shortfall to US farms and food needs to be more than two weights.
Next
- Output lost by district, on a schedule. Scenario variables that take a district's sites out of action at month X, or cut their output by a percentage from month X, for a set time. Measured against this baseline.
- US food supply and agriculture. Fertilizer application, yields with a lag, food prices and farm income, fed back into the potash module.
- The next site lists: Quebec electricity and Alberta oil, treated the same way.
Limits
- Districts do not yet exist in the simulation. The district table is arithmetic on the site list, not a model result.
- "Wars still going" shrinks with each month, so later months describe the longer wars.
- The embargo here starts in month 1 and never lifts. Partial or timed embargoes are not run.
Sources
- USGS, Mineral Commodity Summaries 2026, Potash: US net import reliance 92%; Canada 79% of US imports.
- Prime Rogue Inc., Canadian Potash Sites Inventory, November 2026 (Kevin Duska), for capability by mine.
- Model parameters
potash.substitution_max,potash.substitution_months,potash.occupied_mine_output,potash.political_weightandpotash.gdp_weightindata/parameters/model.yaml. All five are the modeller's estimates. - Results in
results/v0.1.3.9.1_potash_baseline/; scripttools/potash_baseline.py.