When the tool converts between energy units, the answer is exact. When it tells you the energy or emissions of an actual fuel, the answer is a sourced estimate. This is not sloppiness — it reflects the fact that fuels are real materials whose properties genuinely vary.
Three properties that vary
Every fuel calculation rests on one or more material properties, none of which is a natural constant:
- Density (for volume ↔ mass). Refinery blend, additives, seasonal grade and especially temperature move a fuel's density by a few percent — density falls as temperature rises, so a "kg per litre" figure is only right near the reference temperature the source used.
- Calorific value (for mass or volume → energy). This depends on composition and moisture, and it comes in two bases (HHV vs LHV) that themselves differ by several percent.
- Emission factors (for fuel → CO₂ / CO₂e). These depend on the fuel, the energy basis, the scope, the region and the year.
Where the spread comes from
- Diesel and gasoline: blend, additives, seasonal grade and temperature shift density and heating value.
- Natural gas: composition varies by field, network and time, spreading the calorific value across a band — the reason m³ → kWh is never exact.
- Coal: "coal" spans anthracite, hard coal, sub-bituminous and lignite, whose heating values differ by a factor of two or more; the tool treats named grades as separate fuels rather than one "coal".
- Wood and pellets: energy content is dominated by moisture — oven-dry wood carries substantially more usable energy per kilogram than freshly felled "green" wood, because the water must be evaporated first.
What the tool gives you instead
For these fuels the tool provides a well-sourced representative value with its assumptions on display — not a measurement of the specific fuel in your tank, pipe or pile. Concretely, that means it prefers ranges over false precision (showing ~A–B where a property genuinely spans a range), caps significant figures so an estimate never looks more certain than it is, marks estimates with a leading ~, and never silently averages two sources that disagree — it shows the chosen source's value with provenance, or the range across sources, with the divergence visible. Every non-exact factor resolves to a citation on the sources panel. The point is honesty about spread, not the illusion of a single perfect number.