Major project pipeline emissions

What is in scope

Two source lists, deduplicated to 128 unique undertakings. From the Canada Investment Summit prospectus, all 104 entries in the seven non-mining chapters. From the Major Projects Office, all 18 referred projects and all 9 transformative strategies. 4 undertakings appear on both lists and are held once, so no project is counted twice.

Every prospectus entry was checked against the text of its own page in the source PDF. An independent re-extraction, run for the audit, matched all 104 entries on their cited page with no name, page, capital-cost or stage errors — the strongest-performing part of the work.

• 31Clean Energy
• 19Advanced Manufacturing
• 16Marine and Port Infrastructure
• 11Conventional Energy
• 11Power and Utilities
• 10Digital Technology
• 6Transportation

Entries in the register

128

Quantifiable

120

With an operating estimate

93

Jurisdictions

13

27 quantifiable entries still carry no operating estimate. They are shown as unassessed on the chart face rather than as zero — an earlier version reported a portfolio total beside a majority of blank entries, which the first audit called out and which is why coverage is now published per entry, in three states: assessed, assessed below the reporting threshold, and not assessed.

What is counted

The headline is net operational emissions at full build-out, in kilotonnes of CO₂ equivalent per year, on the boundary set by Canada's Strategic Assessment of Climate Change. That is the boundary proponents already use in federal impact assessment filings, which keeps published and estimated values comparable.

We depart from it in one place. Avoided emissions and offsets are recorded and excluded from the headline, because they depend on a counterfactual the proponent chooses. They stay visible in each project's detail panel, labelled as such.

Six quantities are held apart and never netted against each other. A filed number is not automatically a headline number: where a proponent's declared scope is narrower or wider than the boundary, the filed figure is kept, visible and quoted, and a comparable estimate is charted beside it.

• 0

  Layer 0 — Filed, different boundary

  A published figure whose declared scope is narrower or wider than layer 1, or gross where layer 1 is net. Never charted, never summed.

  6 records
• 1

  Layer 1 — Net operations

  Direct emissions plus acquired energy, less CO₂ captured or removed. The headline. Removals plot below zero.

  93 records
• 2

  Layer 2 — Upstream

  Production and processing feeding the project.

  4 records
• 3

  Layer 3 — Downstream

  Combustion of the product carried or exported. Not required by the federal boundary, and it dominates for LNG and pipelines.

  9 records
• 4

  Layer 4 — Construction

  One-time build emissions. Held apart because a total is not comparable to an annual rate.

  18 records
• 5

  Layer 5 — Avoided

  Displacement claims. Recorded and excluded from the headline, because they rest on a counterfactual the proponent chooses.

  8 records

Where the numbers come from

Every one of the 138 records carries a tier, and the tier is shown on the chart rather than kept in a methodology appendix. 26 values are filed or stated by the proponent, drawn from 21 cited sources. The audit opened every one of those sources and checked the quoted phrase against it: one record was found to have taken a figure from the wrong column of a filed table and was rebased, and one reported defect did not reproduce and was rejected in writing. The remaining 112 records are our own estimates and say so on the bar.

• A

  Tier A — Regulatory filing

  Impact assessment registries, provincial environmental assessment offices, the Canada Energy Regulator, the CNSC, the AER, BAPE.

• B

  Tier B — Proponent or government statement

  Prospectus capacity and capture figures, Major Projects Office pages, sustainability reports.

• C

  Tier C — FTCG derived estimate

  Capacity multiplied by a benchmark intensity from the published factor table. Arithmetic on a filed number is always flagged: summing filed components keeps the tier, a doubling or a pro-rata split introduces an assumption and drops to C.

• D

  Tier D — Not quantifiable

  Strategies and corridor concepts with no defined physical scope. Listed with the reason, never charted.

Estimates rest on a published factor table — 103 factors across 56 project archetypes, each with its source, date, boundary and confidence rating, of which 43 are high confidence, 24 medium and 17 low. 2 could not be established at all, and the table says so rather than substituting a plausible number.

Ranges are not decoration either. 8 distinct kinds of range appear in the data, and each record states which kind it carries. 21 filed point values are left as points: inventing a band around a number a proponent published without one would manufacture precision in the direction of looking rigorous.

The input that decides whole bars

Most of these projects are electrical loads, so the carbon intensity of the grid they draw from is often worth more than anything about the project itself. Every project carries a commissioning year and draws the grid factor for that year. Federal published reference values stop at 2027 and all but one of the 120 quantifiable projects commission later, so holding them at a single present-day figure would have erased exactly the differences the exercise exists to capture.

The largest single data-centre estimate in the register swings by roughly a third of its own value across a four-year window, with nothing about the facility changed — only the year it plugs in. 28 records are grid-driven and carry a published ±2-year schedule sensitivity for that reason.

How it was checked

The brief given to each auditor instructed it to treat its own checklist as a floor, to spend effort on things the brief did not mention, and to report what it examined beyond what was asked — because the brief was written by the people who built the analysis, which is a conflict.

1. Round 1

   Not fit to publish

   Nineteen findings — ten must-fix, nine should-fix. All closed, each with the recomputed number recorded rather than an assertion that it was handled. The auditor judged the analysis had applied a materially higher standard to the prospectus than to itself. That charge was accepted.

2. Round 2

   Not fit to publish

   Five findings, all closed. Three of them moved the portfolio total by more than the total itself was worth, because closing the coverage gap added more than the corrections removed. The recurring defect was a rule written in a comment and not applied in the next function.

3. Round 3

   Not fit to publish

   Nine material findings, seven accepted and two rejected because they did not reproduce against the files. The pattern had moved: the arithmetic was sound and four of the nine were statements the analysis made about itself that its own published data contradicted. The headline moved 1.5%. Six method problems remain open.

Findings close only with the recomputed number recorded; nothing closes on assertion. Two audit claims were rejected, both because they did not reproduce against the files, and both rejections are written into the log beside the accepted ones. Defects found during remediation get their own identifier — remediation that finds nothing new has usually not looked.

The pipeline rebuilds byte-identically from source into a throwaway directory, so a change in any published figure is a decision someone made rather than drift. A separate check recomputes every load-bearing figure in the methods note from the published data and fails the build if one no longer holds.