A carbon footprint reports one number, say 150.8 tCO2e (tonnes of CO2 equivalent). It is still an estimate: meter readings have limited precision, emission factors are averages, and some categories rest on nothing more than spend in euros. Without an estimate of carbon footprint uncertainty, you cannot tell whether a 5% change from one year to the next is real progress or noise.

This article covers where uncertainty comes from, which levels ADEME's Base Carbone publishes for its emission factors, how to combine them with a root sum of squares (the IPCC approach) or a Monte Carlo simulation, and what it means for verification. An illustrative worked example runs the calculation end to end.

Table of contents

  • Where does uncertainty come from?
  • What uncertainty levels do emission factors carry?
  • Combining uncertainties: the root sum of squares
  • Worked example: three categories, one total
  • When to switch to Monte Carlo
  • What the GHG Protocol asks for
  • Uncertainty, verification and assurance
  • How to reduce uncertainty
  • Common mistakes
  • How Kabaun handles uncertainty
  • FAQ
  • Conclusion
  • Sources

Where does carbon footprint uncertainty come from?

Most emissions are calculated the same way: activity data × emission factor. Uncertainty enters in three places.

  • Activity data: litres, kWh, tonnes, kilometres or purchase amounts. A meter reading is precise; estimated kilometres, or a euro amount used in place of a physical quantity, are far less so.
  • The emission factor: an average by country, year or technology that can differ from the actual site or supplier. See the article on emission factors.
  • The model and the boundary: system boundaries, allocation rules, assumptions on product lifetime or use, omitted sources.

The GHG Protocol separates parameter uncertainty (activity data, emission factors), which a company can reasonably quantify, from model uncertainty and scientific uncertainty (which affects, for example, GWP values, that is global warming potential), usually beyond the reach of a corporate inventory.

Two kinds of error coexist. Statistical (random) uncertainty can be detected and estimated through repeated measurements or sampling. Systematic uncertainty is bias: a factor built on a non-representative sample, a forgotten source, a faulty meter. No statistic reveals it; only data quality control does. The IPCC asks that known biases be addressed before any propagation calculation.

What uncertainty levels do emission factors carry?

ADEME's Base Carbone, searchable through the Base Empreinte, publishes an "Incertitude" (uncertainty) field in percent for each factor. A few values taken from the public dataset (version V23.6):

  • Road diesel B7 (id 43013): 3.1 kgCO2e/litre, published uncertainty 10%
  • Electricity, average mix 2024 (id 43642): 0.0519 kgCO2e/kWh, published uncertainty 10%
  • Electricity, heating 2023, seasonal method (id 43276): 0.115 kgCO2e/kWh, published uncertainty 30%
  • Spend-based ratio "Travel agencies, tour operators, bookings", 2023 vintage (id 43515): 136 kgCO2e/k€ excl. VAT, published uncertainty 80%

Across the 11,445 non-archived lines of the database (elements and sub-items combined), the most frequent levels are 30% (3,505 lines), 50% (1,198), 5% (844), 20% (803) and 10% (605). The IPCC describes the uncertainty of its default carbon factors for fossil fuels as "quite low" (5 to 10%), while noting that a default value may be less representative in a given country.

Uncertainty therefore depends on the type of category: a burned fuel is well known, a spend converted to euros much less so. The spend-based method is covered in the article on automating carbon accounting.

Combining uncertainties: the root sum of squares

The 2006 IPCC Guidelines propose two approaches: Approach 1, error propagation with simple formulas that run in a spreadsheet, and Approach 2, a Monte Carlo simulation. The GHG Protocol tool is built on Approach 1.

Uncertainty is expressed as half the 95% confidence interval, divided by the central value (± x%). Two rules are enough.

Multiplication (activity × factor): the relative uncertainty of the product is the square root of the sum of the squared relative uncertainties.

U_product = √(U₁² + U₂²)

Addition (sum of categories): add the squares of the absolute uncertainties, then divide the square root by the total.

U_total = √[(U₁·x₁)² + (U₂·x₂)² + …] ÷ (x₁ + x₂ + …)

The GHG Protocol guide reminds readers that a product is less certain than its least certain component: very precise kWh multiplied by a factor that can be 20% off still gives a result around 20% uncertain.

These formulas assume normally distributed, unbiased, independent errors, with individual uncertainties below 60% for the GHG Protocol (a standard deviation below 30% of the mean for the IPCC). Beyond that they give an order of magnitude, not a reliable result.

Worked example: three categories, one total

The activity data are fictional and serve only to illustrate the calculation. Factors and uncertainties come from the previous list, read as 95% half-intervals as an illustrative assumption.

  • Fuel: 20,000 L (±3%) × 3.1 kgCO2e/L (±10%) = 62.0 t; uncertainty √(3² + 10²) ≈ ±10.4%, i.e. ±6.5 t
  • Electricity: 400,000 kWh (±1%) × 0.0519 kgCO2e/kWh (±10%) = 20.8 t; uncertainty √(1² + 10²) ≈ ±10.0%, i.e. ±2.1 t
  • Travel agency purchases: €500k (±5%) × 136 kgCO2e/k€ (±80%) = 68.0 t; uncertainty √(5² + 80²) ≈ ±80.2%, i.e. ±54.5 t

Total: 150.8 tCO2e. Absolute uncertainty: √(6.5² + 2.1² + 54.5²) ≈ 54.9 t, or about ±36% of the total (roughly 96 to 206 tCO2e). Three takeaways:

  • Purchases make up 45% of emissions but about 98.5% of the variance. Fuel and electricity, well measured, barely matter.
  • Replacing that line with physical data at ±30% (a textbook assumption, same 68 t) brings the total down to about ±14%.
  • That line exceeds 60%, so the root-sum-of-squares assumption no longer holds and the ±36% remains an indication.

When should you switch to a Monte Carlo simulation?

The simulation randomly draws a value for each parameter from its distribution thousands of times, recalculates the footprint each time and reads the 95% interval from the results. According to the GHG Protocol, it accepts any distribution, range and correlation structure, provided they are properly quantified. It requires software.

It becomes preferable in three cases:

  • individual uncertainties above 60%, such as spend-based ratios;
  • asymmetric distributions: emissions cannot be negative, and a ±80% interval is not well represented by a normal distribution;
  • correlated parameters: one factor reused across several lines does not make those lines vary independently, and the IPCC notes that positive correlations widen the uncertainty of the result.

What the GHG Protocol asks for

The GHG Protocol addresses uncertainty in the "Managing Inventory Quality" chapter of its Corporate Standard and in a dedicated guide with a spreadsheet tool. The guide states that uncertainty estimates are themselves subjective, that they do not measure inventory quality on their own, and that the causes of uncertainty should always be documented qualitatively. Its quality scale (±5%, ±15%, ±30%) is described by the guide itself as arbitrary.

For Scope 3, the dedicated standard recommends describing the level of uncertainty, qualitatively or quantitatively, and the efforts made to reduce it when it is high. See the Scope 3 guide.

None of these GHG Protocol documents sets a maximum value: they ask you to assess, document and reduce.

Uncertainty, verification and assurance

A verifier does not check that the figure is exact: they assess whether the statement is free of material misstatement. The GHG Protocol Scope 3 Standard cites ISO 14064-3 as an example of an assurance standard and states that uncertainty is distinct from materiality: it does not describe a known error, but indicates how well the data represent the processes in the inventory.

In practice:

  • the GHG Protocol documents reviewed set no uncertainty level above which a footprint could no longer be verified; what they expect is that it be assessed and documented. The requirements of the assurance standard used (for example ISO 14064-3) should be confirmed with the verifier;
  • the verifier asks for evidence: data sources, assumptions, justification of factors;
  • uncertainty assumptions (ranges, distributions, method) are written down and kept like any other calculation assumption.

How to reduce the uncertainty of a carbon footprint

For the IPCC, an uncertainty analysis is first a way to prioritise improvement efforts. Four actions:

  • Compute each category's contribution to the variance, as in the example. One line almost always dominates.
  • Replace spend data with physical data on the dominant lines: quantities purchased, tonne-kilometres, kWh, tonnes of waste.
  • Ask key suppliers for primary data rather than applying an average factor. Their allocation brings its own uncertainty, which must be documented.
  • Pick a specific factor (country, technology, year) over a generic one.

Effort goes to the lines that move the total, not to a category worth 1% of the variance.

Common mistakes

  • Confusing display precision with accuracy: 150.8 tCO2e does not mean known to within 0.1 t.
  • Applying the root sum of squares outside its domain: uncertainties above 60%, correlated parameters. Flag the caveat or move to Monte Carlo.
  • Adding percentages instead of squaring them: a linear sum overstates the uncertainty of independent categories.
  • Treating bias as uncertainty: an incomplete boundary is not fixed with a wider interval.
  • Ranking companies on uncertainty: except in very restricted cases (similar facilities, identical methods), the GHG Protocol guide rules out reading it as an objective quality measure.

How Kabaun handles uncertainty

Kabaun provides uncertainty analysis (CBC-005a): calculation and display of the uncertainty intervals attached to footprint results, in line with GHG Protocol recommendations. Several other features help reduce uncertainty:

  • each data point is linked to a traceable source (invoice, meter reading, supplier declaration), with supporting documents attached;
  • custom emission factors can be added, with administrator validation and traceability;
  • Scope 3 data requests go out to suppliers, with scheduled reminders;
  • a timestamped audit trail keeps data, changes and calculations, which helps a verifier.

FAQ

What is carbon footprint uncertainty?
It is the range in which the true value of emissions lies, at a given confidence level (most often 95%), expressed as ± percent. It is attached to a figure, not to a company.

How do you calculate the uncertainty of an emissions category?
For a category calculated as activity data × factor, the relative uncertainty is the square root of the sum of the squares of the two relative uncertainties. With activity at ±3% and a factor at ±10%, you get about ±10.4%.

How do you aggregate uncertainty across several categories?
Square the absolute uncertainty of each category (emissions × relative uncertainty), add them up, take the square root, then divide by the total. Highly uncertain categories dominate: in the example, a 68 t line at ±80% explains about 98.5% of the variance of a 150.8 t total.

What uncertainty do Base Carbone emission factors carry?
Each factor has its own value. In version V23.6, road diesel B7 is at 10%, the 2024 average electricity mix at 10%, seasonal heating electricity at 30% and a sector spend-based ratio at 80%. The most frequent levels across the database are 30%, 50%, 5%, 20% and 10%.

When should you use a Monte Carlo simulation?
When the root-sum-of-squares assumptions no longer hold: uncertainties above 60%, asymmetric distributions or correlated parameters. It requires software.

Can an uncertain carbon footprint be verified?
The GHG Protocol documents reviewed set no maximum uncertainty level for verification; the requirements of the assurance standard used should be confirmed with the verifier. The verifier assesses the absence of material misstatement and the quality of evidence. Uncertainty, a concept distinct from materiality, must be assessed and documented.

Conclusion

Estimating carbon footprint uncertainty means assigning a range to each activity data point and each factor, combining them with a root sum of squares (or Monte Carlo when its assumptions fail), then spotting the categories that dominate. A single spend-based line can explain almost all the uncertainty of a total, and that is where data collection should focus.

Do this now: list your five largest categories, look up each factor's uncertainty in the Base Carbone, and apply the two IPCC formulas in a spreadsheet.

Kabaun helps you document and reduce the uncertainty of your carbon footprint → www.kabaun.com/en/contact

Sources

  • IPCC, 2006 Guidelines for National Greenhouse Gas Inventories, Volume 1, Chapter 3 "Uncertainties": Equations 3.1 and 3.2, Approaches 1 and 2, 95% confidence interval. Accessed 2026-10-02. ipcc-nggip.iges.or.jp
  • GHG Protocol, "Guidance on uncertainty assessment in GHG inventories and calculating statistical parameter uncertainty". Accessed 2026-10-02. ghgprotocol.org
  • GHG Protocol, Corporate Accounting and Reporting Standard, Chapter 7 "Managing Inventory Quality". Accessed 2026-10-02. ghgprotocol.org
  • GHG Protocol, Corporate Value Chain (Scope 3) Accounting and Reporting Standard: Appendix B (uncertainty) and the assurance chapter. ghgprotocol.org
  • ADEME, Base Carbone® version V23.6, public dataset ("Incertitude" field), values read on 2026-10-02. data.ademe.fr
  • ADEME, Base Empreinte®. base-empreinte.ademe.fr