Skip to main content

You are viewing a project that is currently in draft state for the Standards Working Group in the Green Software Foundation. This project should not be considered finished or officially supported in any way by the Green Software Foundation or its members.

EWIF — Average vs. Marginal Water Intensity

EWIF — Average vs. Marginal Water Intensity

In short: there are two different, both legitimate, ways to answer "how thirsty is this electricity" — the average across everything generating power right now, or the specific plant that would ramp up if your workload used more. They can give different numbers for the same situation, and the SSWG hasn't yet settled which one implementers should default to.

Open methodological question

This is an open, unresolved question within the SSWG, not a settled methodology choice. The specification deliberately takes no position on average versus marginal water intensity. This page presents both approaches and the considerations relevant to each — it does not tell you which to use. See Open Methodological Questions for current status.

The two approaches​

Average water intensity uses the water intensity of the overall generation mix supplying a region over a given period — every plant feeding the grid, weighted by its share of generation.

Marginal water intensity uses the water intensity of whichever generation source responds to a change in demand — the plant that would ramp up or down if the software's electricity consumption changed at the margin.

A meaningful concept of marginal water intensity does exist for electricity generation, in the same way it does for carbon intensity, and the two approaches — average and marginal — can produce materially different results for the same workload, in the same region, over the same period.

Why this remains unresolved​

Unlike some other open questions in this guidance, this one is not simply awaiting a dataset. It is a genuine methodological dispute about which question an SWI assessment is trying to answer:

  • An average intensity answers: what water is this workload's fair share of, given how the grid actually generates electricity today?
  • A marginal intensity answers: what water impact would result from a marginal change in this workload's demand?

Both are legitimate questions; they are not interchangeable, and neither is a rounding error of the other. Which one is more appropriate can depend on what the SWI result is being used for — reporting an existing baseline versus evaluating the effect of a proposed change (see Comparing an SWI to a Baseline in the specification).

Status​

A revised EWIF methodology brief addressing this question, among others, has been committed to within the SSWG. Until that work concludes and the group reaches a position, implementers should:

  • Choose whichever approach — average or marginal — best matches the purpose of the assessment.
  • Disclose explicitly which approach was used; do not present an average figure and a marginal figure as if they were the same metric.
  • Expect this guidance to be updated, and this caution notice removed or narrowed, once the SSWG resolves the question.

Please submit any comments you have here.