-
Notifications
You must be signed in to change notification settings - Fork 11
Description
Description
The "CO2 emissions & energy consumption" plot representation, where energy consumption and CO₂ equivalent emissions are shown together using two y-axes on the same boxplot, is not valid in cases where carbon intensity (CI) varies across tasks or processes over time, because CO₂ emissions are computed as energy multiplied by CI, meaning that the CO₂ distribution depends not only on energy but also on the CI at execution time.
As a result, the distribution of CO₂ emissions is not a simple rescaling of the energy distribution. For example, tasks with similar energy usage can result in different CO₂ emissions if they were executed under different CI conditions.
Issue
A single boxplot implicitly represents one underlying data distribution. When CI varies, the CO₂ distribution is not a linear rescaling of the energy distribution. As a result, it is not possible to represent both metrics using two y-axes on the same boxplot, since they do not share the same ordering or statistical properties (e.g. median, quartiles).
Proposed direction
- Clearly separate energy and CO₂ visualizations
- Avoid dual y-axis plots for energy vs. CO₂ distributions
- Explore alternative ways to relate energy, CI, and CO₂ (e.g. separate plots or linked visualizations)
Concerns this plot:
