Methodology

Power-plant repeal — coal-plant health scenario ⚠ CONDITIONAL SCENARIO

How this scenario is built. The comparison throughout is a world where the rule stays in force versus one where it is repealed — the rule is final but not yet published in the Federal Register, so every figure describes a hypothetical outcome.

How the estimate is built

Repealing the CO2 standard → coal-fired power plants run more hours than they would have under the rule → those extra run-hours emit additional SO2 and NOx → SO2 and NOx convert to secondary sulfate and nitrate PM2.5 downwind → modelled excess deaths (GEMM concentration-response), reported as steady-state annual counts. Health is modelled entirely through criteria pollutants — the regulated pollutants that go on to form fine particles. CO2 itself is not one of the five species in the source-receptor matrix and carries no modelled health effect here.

The two anchors and the midpoint

The emissions change is bounded by two independently sourced anchors, not a single point estimate with error bars. The gap between them is the actual uncertainty in how big a repeal's emissions effect would be — collapsing it to one number would hide that. The third card is the midpoint the scenario toggle exposes: an interpolation shown for the toggle's middle position, and the only one of the three that is not evidenced.

Constants and checks behind every figure

Acceptance tests — what was checked. Before this prototype's numbers were trusted, checks were run against the pipeline. The pipeline's arithmetic checks pass on the real Bay-Area source-receptor data, and the Ohio-Valley run has since been performed on the real v02 matrix (four independent data checks, all passing — see the Impact page). What remains unstaged is the full CONUS coal-fleet run. A passing check shows the arithmetic is sound — it says nothing about whether the repeal happens.

TestWhat it checksResult

Core parameters — the constants behind every figure.

Concentration-response function
GEMM (Burnett et al. 2018), used verbatim — no re-fit, no re-parameterization. GEMM is an upper-bound function: it returns more deaths per µg/m³ of PM2.5 than the log-linear functions EPA uses, so any estimate built on it sits at the high end of the published range.
Baseline death rate
0.01182 deaths per person per year — the USA-national rate for non-communicable disease plus lower-respiratory infection, ages 25+, applied to the whole population. A national rate, not a Bay-Area or Ohio-Valley one.
Value of a statistical life
Loading… Illustrative sensitivity range: $7M–$13M around the $11M central value.
Climate TRACE release
Coal-fleet denominators
Meteorology
2005 annual-average, steady state — the same InMAP SR-2005-annual-v02.0 vintage used across every demo in this gallery.

Which numbers bind live, and which are authored. The evidence-anchor figures, value of a statistical life, Climate TRACE release and coal-fleet denominators bind live from policy_payload.json; the concentration-response function, baseline death rate, meteorology and acceptance-test results are authored constants, each traced in the provenance drawer.

Assumptions, caveats and how figures are worded

Honesty ledger — which way each assumption pushes.

AssumptionEffect on the estimate

▲ pushes the estimate up   ▼ pushes it down   ◆ a scope or process choice, with no directional effect
“Cluster” = the group of Ohio-Valley coal plants this scenario models; “cell” = one grid square of the transport model.

Short caveats shown beside each figure.

How every figure here is worded. Figures are modelled estimates under a rule-in-force-versus-repeal comparison, and they are written that way: "would be associated with," "would affect," "we estimate." Each one describes what the model says would change if the repeal took effect and the change were sustained.