Carbon crediting lost trust because the baseline was self-appointed
In most voluntary carbon markets, the registry that issues the credit also decides the counterfactual: what would have happened without the project. That baseline sets the number of tonnes. When the same party defines the baseline, approves the methodology, and earns a fee per tonne issued, the incentive runs one way — toward a generous counterfactual.
The consequence is the story of the last decade: over-crediting, retracted claims, buyers writing off portfolios, and a market where a "tonne" means whatever the issuing standard says it means. The measurement problem was never the physics. It was who got to pick the number.
Karbera does not invent an emission factor. We prove the MWh.
Brazil already has an official grid displacement factor, published by the Ministério da Ciência, Tecnologia e Inovações (MCTI) from its dispatch analysis. Karbera adopts that factor unmodified as the factor of record. We do not set it, adjust it, or compete with it.
What Karbera contributes is the layer underneath: independent, metered evidence that the claimed megawatt-hours were physically delivered to the grid, by which asset, in which hour, and to whose beneficial-ownership account. The emissions arithmetic belongs to the government. The proof of delivery belongs to us.
Operating Margin, Build Margin, Combined Margin
The official factor is built from two halves, following UNFCCC CDM Tool 07:
The plants already running that get turned down when your MWh arrives. On a Brazilian evening peak, that is usually a gas or oil thermal unit.
The plants the system would otherwise have had to build. Brazil builds mostly renewables now, so this half is small.
CM = 0.75 × OM + 0.25 × BM. Weighted three-to-one toward what is displaced today. Base year 2025.
Full derivation, the five-year OM/BM/CM vintage table, and the disclosure notes live on the methodology page — they are not repeated here. See the derivation table
Average versus marginal — and why the choice is not Karbera's to make
MCTI publishes more than one factor. There is an inventory-style average factor for Brazil's grid of roughly 0.02–0.03 tCO₂/MWh, and there is the dispatch-analysis factor of 0.3271 tCO₂/MWh. The gap is roughly ten-fold, and every so often someone points at it as evidence that one of the numbers must be wrong.
Neither is wrong. The average factor spreads all fossil emissions across all generation, including Brazil's enormous hydro base — that is the right way to attribute the footprint of electricity you consumed. The dispatch factor describes the plant that actually backs down when one more MWh shows up — that is the right way to measure a reduction. The ~10× is not an accounting choice. It is physics: hydro-heavy average, thermal-heavy margin.
| MCTI publication | Stated purpose | tCO₂/MWh | Karbera uses it for |
|---|---|---|---|
| Fator Médio | Inventários corporativos — footprint of consumed electricity | ~0.02–0.03 | Not used for crediting |
| Análise de Despacho (CM) | Exclusivamente for estimating emission reductions | 0.3271 | Factor of record for every credited tonne |
We revised the factor downward — 0.476 was never a published number
An earlier version of this site credited at 0.476 tCO₂/MWh. That value was not published by MCTI: it paired a 2021 Operating Margin with a Build Margin the ministry never issued, and it overstated the Combined Margin. When we found it, we replaced it with the published 2025 figure of 0.3271 tCO₂/MWh and logged the correction in the version history.
Every credited tonne on Karbera got roughly a third smaller as a result. We consider that the point. A measurement system that only ever revises upward is a marketing system. Publishing the correction — and the arithmetic that produced it — is what makes the remaining number worth believing.
The factor refreshes annually with each new MCTI dispatch publication, and every figure on this site states its vintage.
A verification layer beneath the government's factor
If the emissions arithmetic is the government's, the value Karbera creates is evidentiary. Each item below is cited and independently checkable.
Hourly generation per plant from ONS verified dispatch (geracao_usina_2_ho / geracao_termica_despacho_2_ho) — the same metered values ANEEL and CCEE use for market settlement.
Jointly held assets are credited only to each operator's ownership fraction, so a JV cannot be claimed twice across partners.
No figure on Karbera originates from the operator being credited. Every input comes from the grid operator, the regulator, or a third party the operator authorises.
Each figure carries its verification tier and vintage, so a reviewer audits the mechanism that produced it — metered, aggregate-reconciled, or modeled — not just the number.
MMGD portfolios are reconciled against EPE PDGD cell aggregates within ±15% tolerance, with anything outside tolerance labelled modeled rather than verified.
Submarket-specific displacement and transmission-loss uplift, so a MWh injected in a constrained node is not treated as identical to one injected at a surplus node.
The geographic and loss refinements are design-stage and are not applied to any figure currently shown on this site. They ship no earlier than Tier 1 rollout, and will be versioned in the methodology history when they do.
The No-Self-Reporting Principle
Conventional carbon credit verification relies on operator-submitted data. Projects file monitoring reports; validation/verification bodies audit what operators claim; registries issue credits based on the audited operator submissions. The operator sits in the middle of the chain of custody for their own credits.
Karbera operates differently. Every Karbera-verified credit is grounded in data already being recorded by parties other than the operator — government-published grid metering (ONS Brazil), regulatory plant registries (ANEEL SIGA), utility-issued documents, and hardware-attested device readings. Karbera reads what's already being recorded; the operator doesn't have to assemble or submit anything. The verification runs on the data, not on the operator.
This is the proof, not the pitch. It is what makes the rest of the engine possible.
This principle is foundational to both Karbera (the verification methodology) and CRIOS (the issuance platform that accepts only methodologies operating under it).
The derivation, the vintage table, and the per-operator arithmetic are all public. Start with the methodology, then see the two quantities side by side.