In Toovalu, the uncertainty calculation method is based on version 9 of the Bilan Carbone® methodology and combines two approaches: a qualitative rating of the data, and an automatic quantitative calculation that produces a 95% confidence interval.
Key points
What are the two approaches used to determine uncertainty?
The qualitative approach assesses activity data and emission factors against several criteria, each rated on a quality scale.
The quantitative approach uses these ratings to calculate a 95% confidence interval.
How are the results presented?
The results come in three different but related forms:1. a qualitative uncertainty on GHG emissions, ranging from “Very high” to “Very low”
2. a 95% confidence interval [lower bound; upper bound], within which the result has a 95% chance of falling
3. a percentage indicating how much higher the result could be than its value, using the following formula.
How does it work in the tool?
Qualitative approach
In the collection screen, each line of your project has a “Data Quality” column. To fill it in, open the drop-down menu on the relevant line and select a level from “Very poor” to “Very good”.
On the Admin side, each line of the template has an “EF Quality” column. This rating is based on the uncertainty stated in the emission factor’s source, but you can adjust it to fit the context of your project, downwards only.
Example: the EF for Spanish oranges, taken from a very high-quality LCA, is rated “Very good”. If you use it for oranges from another country, it makes sense to lower its rating.
As explained further down in this article, the quality of a piece of data or an emission factor can be assessed by considering its technological, geographical and temporal representativeness, as well as its completeness and reliability.
Quantitative approach
The tool uses the qualitative ratings to calculate the quantitative uncertainty automatically.
How do ratings become numbers?
For each activity data (AD) and emission factor (EF), each rating corresponds to a coefficient, according to the table below. The poorer the quality, the further the coefficient is from 1.00.
Quality | Coefficient (geometric deviation GSD²) |
Very good | 1.00 |
Good | 1.12 |
Average | 1.26 |
Poor | 1.64 |
Very poor | 2.52 |
* Coefficients set by the ABC (Association pour la transition Bas Carbone).
How do we go from coefficients to the confidence interval?
For each emission source, the activity data coefficient and the emission factor coefficient are combined.
This GSD gives the source’s 95% confidence interval, with a lower bound and an upper bound.
The lowest value of this interval is called the “Lower bound” and the highest the “Upper bound”. There is therefore a 95% chance that the actual value lies within this confidence interval.
Overall consolidation
The sensitivity of an emission source is the share it represents in the total footprint. For an emission source with value E1 and a footprint with total value T, the sensitivity S1 is calculated as follows:
Now consider two emission sources with values E1 and E2, respective sensitivities S1 and S2, and respective geometric standard deviations GSD(E1) and GSD(E2). The geometric standard deviation of their sum E1 + E2 is obtained with the following formula:
The 95% confidence interval is then derived using the same method as described above.
Where can I find the result in the tool?
In the tool, this interval is displayed next to your total emissions.
If you filter your data in the Collection, the uncertainty shown is recalculated based on that filter.
Going further: methodology
For activity data and EFs, the organization must give a qualitative rating to each of the five characteristics: very good, good, average, poor or very poor.
Requirements by maturity level
What does the method require for each target level? At every level, uncertainties are both qualified and quantified, and the quantification is calculated automatically from the qualification. What changes is the scope and the level of detail of the rating.
In Toovalu, the rating covers the overall quality of the EF or the data, as allowed at the Initial level (M1). Adjusting the EF rating meets the recommendation of the Standard level (M2). The Advanced level (M3), which requires five ratings per data point, will call for the detailed rating planned for a later stage.