Methodology

What this site computes

For annual demand D and source share M, generation is D × M. Deaths are generation × deaths/TWh. CO₂, land, and cost use the same weighted-sum structure.

Cost and firming

Lazard 2026 is the only cost methodology used. Oil, hydro, and biomass therefore show no comparable data instead of patched values from another source.

The builder cost panel can include Lazard's illustrative firming-cost adder for wind and solar. The toggle uses grid-operator ELCC and Net CONE frameworks and does not claim to be a full reliability model.

Land

m²/MWh/yr and km²/TWh/yr are numerically identical, so one stored coefficient serves the source tables, the builder, and the model. Wind is shown as a dual figure because total wind-farm area includes turbine spacing, while direct land occupation is much smaller.

Land figures currently come from twosources: coal, oil, gas, biomass, hydro and solar use UNECE 2021; nuclear and wind use van Zalk & Behrens 2018. The two methodologies measure land differently and can disagree by an order of magnitude — UNECE's coal figure (~9 km²/TWh/yr, counting the mining footprint) is roughly ten times van Zalk & Behrens' (~0.85, direct transformation). Every page reads the same stored coefficient, so the site is internally consistent, but cross-source land comparisons carry this extra methodological uncertainty on top of the stated band. Reconciling all eight sources onto a single land methodology is a known open item.

Valuing a death (VSL)

The value-of-a-lifelens multiplies a source's deaths per TWh by a value of a statistical life (VSL) to give a mortality cost per MWh — the same shape as the LCOE figure, so the two are directly comparable. The presets are HHS's 2026 published range ($6.6M / $14.1M / $21.5M, constant 2025 dollars); the slider fills the space between. This is presented as a published range, never as an endorsement of any one figure, and the choice among them is treated as a moral rather than a technical question. Costs are undiscounted — a future death is weighted the same as one today — and a mortality price of zero leaves the underlying death figures unchanged.

Uncertainty

Low values sum with low values, and high values sum with high values. This treats uncertainties as perfectly correlated and widens the band.

Known limitations

• Data vintage differs by source and country.
• Fossil rates derive from a European 2007 baseline generalized beyond its setting.
• AR5 Annex III emissions still require direct hand-verification against the primary table.
• Nuclear figures depend on the contested linear no-threshold model.
• The firming-cost toggle is not a planning model, total system-cost model, or operational reliability model.
• Supply-chain labor conditions, waste-disposal long tails, wartime scenarios, terrorism, and siting-specific land use are outside the calculation.
• Country-level estimates apply global-average rates and will misstate some countries.

Country adjustments

Country estimates differ from the global source table in two deliberate ways. Hydro uses the rate excluding the 1975 Banqiao Dam failure, and fossil death rates are re-anchored to a coarse three-way pollution-controls tier (stringent / moderate / limited) using the geometric mean of the global range as the interior split. This is a site-derived adjustment, not a published per-country figure.

Changelog

2026-07-28 — Added the value-of-a-life lens: HHS VSL range, mortality cost per MWh on the sources and in the builder, and a counted-vs-modeled split on built mixes.

2026-07-26 — Renamed to Level; split hydro ex-Banqiao, added pollution-controls tiers, human-scale anchor, log/linear scale toggle, and dominance warnings.

2026-07-24 — Added firming-cost toggle, no-comparable-cost handling, and wind dual land-use treatment.