Other Questions
Last updated
Last updated
The analytical approximation used in propagating the uncertainty assumes no correlation between different entries (see ). With that assumption, more individual data points are less likely to deviate from the mean than one big data point with the same uncertainty distribution. (it is less likely that all data points deviate from the mean in the same direction simultaneously).
If your data split reflects a split into different independent variables, then this makes your uncertainty assessment a little more accurate. If the split creates dependent variables (that would move up or down together), then this split decreases the accuracy of the uncertainty estimate. Overall, this effect is small (1-3 percentage points), and should be considered in the background of the overall accuracy of the uncertainty estimate.
To assess volume uncertainty in full, you need to consider the following parameters
Reliability: is reliable and accurate is your numeric data
Completeness: how likely is it that some part of the data was overlooked or incomplete
Representativeness: are you using data from another time period / location / technology to represent something else? And how closely related are those to the application level?
See the page on picking a for more details and references.
What you are describing is a mismatch in “emission factor representativeness” (Temporal, Geographical or Technological representativeness). This reflects how well the emission factor matches the activity data provided.
Theoretically, this is a part of Emission Factor uncertainty (not volume uncertainty), so it is set at the emission factor level. In the current Carbon+Alt+Delete implementation, a value for this assumed and hard coded in the Emission Factor uncertainty assumptions. (see our page on )
If you encounter this problem, please take the following steps 1) try to find a more suited emission factor, perhaps by looking at the .
2) have a look at the assumed representativeness of the emission factor used to see if it broadly matches your situation.
3) if unsuccessful, try downgrading the volume uncertainty level to reflect this mismatch (if serious enough) to increase the overall uncertainty of this entry.
Custom emission factors will have the default EF uncertainty of Good, as the assumption is that are custom-made, and they closely match the activity data used.
4) that you have encountered the limits of our current uncertainty model, so we can prioritize expanding it.
We plan to expand on this functionality to open up EF uncertainty to be fully customizable. Please that you would be interested in that, and explain how you would approach calculating your own uncertainty.
See the instruction on picking a for when your situation really does not align with the predefined levels. If that still does not scratch the itch, please and make your case for this missing volume uncertainty level.