# Fuel Horizon — public evidence note

Assessed 10 October 2026. This is a dated research snapshot, not a live feed.

## United States and regional stock paths

Product: 0–15 ppm sulfur distillate, a shared diesel/heating-fuel pool. Reported primary stocks are not all immediately dispatchable. Observed 2 October 2026; the four-week changes use 4 September–2 October endpoints.

| Area | Observed stocks, million barrels | Average net draw, barrels/day | Conditional 1 November stocks, million barrels |
|---|---:|---:|---:|
| U.S. | 93.734 | 95,500 | 90.869 |
| New England | 2.451 | 14,000 | 2.031 |
| Midwest | 25.339 | 83,607 | 22.831 |

The 1 November values assume the measured draw persists for 30 days from the 2 October observation. They are not shortage dates. Regions are subsets of the national total. The Midwest's latest reported week built 207,000 barrels despite the four-week draw.

U.S. sensitivities for 1 November: unchanged stocks 93.734m barrels; half the draw 92.302m; continuing draw 90.869m; double draw 88.004m. No probabilities are assigned. Net stock change includes replenishment and withdrawals, and is not daily consumption.

Sources:
- [EIA national series](https://www.eia.gov/dnav/pet/hist/LeafHandler.ashx?f=W&n=PET&s=WD0ST_NUS_1)
- [EIA New England](https://www.eia.gov/dnav/pet/pet_stoc_wstk_dcu_r1x_w.htm)
- [EIA sulfur-grade regional series](https://www.eia.gov/dnav/pet/pet_stoc_wstk_a_EPDXL0_sae_mbbl_w.htm)
- [EIA stock definitions](https://www.eia.gov/dnav/pet/TblDefs/pet_stoc_wstk_tbldef2.asp)

No applicable operational floor or complete forward net-delivery path has been established for these stock populations. No defensible U.S. or global physical run-out date is established in this edition.

## Global context

The June–July 2026 JODI fixed 30-entry country panel (29 countries with positive stocks) drew 16.069 million barrels. Screening reporting changes and quality reduces the draw to 8.423m. These historical partial-country figures cannot be converted into a current world deficit. Seven countries built stocks in the broad panel. [JODI original records](https://www.jodidata.org/oil/database/data-downloads.aspx).

Switzerland reported about 20,000 cubic metres of compulsory diesel released during 8–20 September, followed by refinery restart and recovery of normal supply channels. This is an executed episode, not an ongoing daily deficit. [Original government report](https://www.bwl.admin.ch/de/newnsb/B9PaqlJnD0-G).

## Australian regulatory scenario

Observed eligible MSO automotive-diesel stocks: 2,871 million litres on 6 October 2026. Previous observation: 3,008 ML on 29 September. Displayed reduced comparator: 2,262 ML. Headroom: 609 ML. The dashboard's observation date is distinct from its 9 October update date.

The public calculation is (stocks minus comparator) divided by a sustained net draw. It is anchored to the stock observation date, not the website visit date. At 137 ML / seven days, the comparison is about 31 days after 6 October, around 6 November.

| Alternative net-draw assumption | Approximate comparator date |
|---|---|
| Latest weekly draw persists | 6 November 2026 |
| Half the latest draw | 7 December 2026 |
| 7 July–6 October average (522 ML / 91 days) | 20 January 2027 |
| Stocks stabilize | No crossing |

Alternatives have no assigned probabilities and are not a confidence interval. MSO stocks include crude/unfinished refinery stocks converted to product equivalents, pipelines and fuel in defined waters including the EEZ. Aggregate regulatory headroom is not verified finished fuel available for withdrawal. The threshold is not an operating minimum or service-failure trigger. This is not an Australian, U.S. or global run-out date.

The temporary reduced comparator expires 31 January 2027. The baseline is 2,827 ML. A changed policy comparator does not mean fuel has suddenly disappeared. The site suppresses this countdown after policy expiry and identifies a past scenario date without claiming that it occurred.

Validation: across 114 comparable historical weekly targets, repeating the latest stock change had mean absolute error 176.06 ML, versus 112.96 ML for unchanged stocks and 119.44 ML for the four-week trend. Five of the latest thirteen intervals built stocks. There is no demonstrated predictive advantage for the latest-draw countdown assumption, and no basis for calling it the most likely outcome.

Sources:
- [DCCEEW weekly statistics](https://www.dcceew.gov.au/energy/security/australias-fuel-security/minimum-stockholding-obligation/statistics)
- [Current requirements](https://www.dcceew.gov.au/energy/security/australias-fuel-security/minimum-stockholding-obligation)
- [Official stock definitions](https://www.dcceew.gov.au/energy/security/australias-fuel-security/measures-of-liquid-fuel-stocks)

## Research depth and limits

The reviewed work matched 9,650 original source rows and reproduced 14 physical-evidence records; 42 Canadian continuity comparisons were retained. These establish source-record and arithmetic reproducibility, not independent inspection of physical tanks or calibrated shortage forecasts. A separately tested U.S. stock predictor failed its fixed eligibility tests despite a 1.35% final mean-absolute-error improvement; it is not a validated depletion predictor.

Source links, dates, product definitions, material assumptions and negative validation are public. Proprietary implementation, row-level research ledgers and full reproducibility archives are excluded from this publication. Paid analysis does not imply access to model source code or the full database.

## Two distinct conditional countdowns

These stress cases do not establish U.S. or worldwide operational failure. They are not most-likely forecasts or probability intervals.

**U.S. historical-low stock scenario:** 93.734m barrels observed 2 October, compared with the recorded 22 May 2026 low of 91.239m in the post-2015 EIA comparison window. Continuing the measured four-week net draw of 95,500 barrels/day reaches the benchmark around 28 October. Half draw gives 23 November; double draw 15 October; stable/building stocks do not cross. The low is historical, not an operating floor; stocks subsequently recovered. Source: the EIA national series linked above.

**Global reserve-tranche coverage scenario:** [IEA's 7 October statement](https://www.iea.org/news/statement-by-iea-executive-director-on-meeting-of-iea-member-governments-on-7-october-2026) reported over 200m barrels of public emergency diesel. This scenario selects 200m, not all world reserves, and assumes full timely release and delivery. Source-wide common physical observation date and current releasable quantities are unknown.

The assumed deficit of about 1.4774m barrels/day extrapolates July data from 28 JODI country accounts plus U.S. data using assumed world demand of 30m barrels/day and equal draw intensity in missing countries. Cautionary quality codes, mixed product scopes and persistence into October limit that imputation. It is not an observed current global deficit. Full detailed ledgers and implementation remain private.

Starting on 7 October (the statement date, not a verified common stock observation), selected-tranche coverage is about 135 days, around 19–20 February 2027. Half the assumed deficit gives 4–5 July 2027; double gives 13–14 December 2026; no deficit or matching replenishment yields no net depletion. All are sensitivities without assigned probabilities or demonstrated event-date predictive skill. Local shortages can occur earlier; other stocks and replenishment can sustain supplies after the tranche. Unfulfilled all-petroleum release pledges are not added.

The site uses whole calendar days against these fixed result dates. Approximate calendar dates can vary by about one day because intra-day observation times are unknown. Passing a date is not evidence the modeled event happened.