Missing factor keys for waste and biomass in carbon_intensity_per_source.json

The Finnish and Danish bidding zones carry 11–32 % of their generation at factor zero when computed on carbon_intensity_per_source.json, because the table has no key for Waste or Biomass. The direction of the error is known: both types emit at the stack. Its size is not something the table can express.

The numbers. From ENTSO-E A75 for 2025 on your own factor table, unfactored types carried in the denominator at zero:

The Norwegian and Swedish values you ship are barely affected; their coverage fields say so. Finland is affected materially — the one zone where biomass exceeds a tenth of generation.

How the code handles the absence today. In _global_energy_mix_to_emissions_rate (codecarbon/core/emissions.py:353-356) the field name is stripped of _TWh and looked up in the factor table, so biofuel_TWh resolves to biofuel, finds nothing, and is skipped from both the numerator and energy_sum_computed. Most countries never reach that path, since 212 of 213 entries in global_energy_mix.json carry carbon_intensity directly, but it is the path that defines what a missing key means. The nordic_emissions.json path handles the same absence differently: the type stays in the denominator at zero. Neither is a factor, and they do not agree.

What other products do. Electricity Maps carries a biomass key (0 direct, 230 lifecycle) but no waste key. Energinet publishes a Danish waste factor, 329.87 gCO2/kWh under its 125 % CHP allocation. Separately, Eurostat puts Danish waste generation at roughly 58 % non-renewable by energy (2023). Eurostat and Statistics Finland both split waste into renewable and non-renewable as separate codes. ENTSO-E does not.

That split is why we are asking rather than proposing: a single Waste number is not well defined without the fraction, and the fraction moves by country and year — 48.0 % non-renewable for Norway in 2023, 45.1 % Finland, 58.0 % Denmark.

The question: would you want keys for these two types at all, and if so, one value per type, or a scheme pricing only the non-renewable fraction of waste? We have not proposed values, and would rather hear the preferred shape first. Happy to bring the per-zone measurements if useful.

Disclosure: drafted with Anthropic Claude models (Opus 5 via Claude Code, Fable 5.1 in chat) under my direction. The measurements were computed with those tools from ENTSO-E and Eurostat primary data and checked by me; I am responsible for them and can walk through every number.

Start with codecarbon/core/emissions.py:353-356 and compare its missing-key behavior with the nordic_emissions.json path. Review carbon_intensity_per_source.json, global_energy_mix.json, and the cited ENTSO-E and Eurostat measurements. Done requires an agreed representation for Waste and Biomass, approved factor inputs, and consistent handling across both calculation paths.