Why measuring power's climate impacts has become a battleground

There is a little-known concept used in efforts to measure the climate impacts of electricity consumption known as ‘marginal emissions.’ It represents a shift away from hard-headed measurements of the current ‘real’ state of the emissions of your power draw. It’s more akin to claims of ‘offsetting,’ and it has shaky, uncertain scientific foundations. 

Shifting away from measuring reality towards estimating ‘how well you avoided a worse possibility’ could end up with massive underreporting of emissions - and therefore, less impetus to act to reduce them. 

It is tough to imagine a worse time to blur reality, when it comes to tackling the climate impacts of power. Let’s dig into the details.

Avoiding emissions

“Carbon offsets” claim to undo the release of a tonne of greenhouse gas because someone, somewhere else, took an action to avoid it. The logic behind this is their avoidance cancels out your release, and the net effect of your action is zero.

The logic underpinning these ‘avoided emissions’ estimates have begun to spread in many other parts of climate accounting. This takes something good - climate action - and uses it to help justify something bad (climate emissions).

When carbon offsets are blended with measurements of emissions, you end up with an artificially muted picture of a company’s climate impacts. It isn’t just logically incoherent to claim action elsewhere undoes your own harm: carbon offsets are fundamentally rooted in measuring-the-impossible-to-measure, and so have been rife with controversies. It is a clear and present danger to climate action. Fakery can lessen the need to enact real emissions reductions.

Measuring emissions

Efforts to estimate the climate impacts of power consumption date back to the early 2000s, when the Greenhouse Gas Protocol first published guidance on what it called ‘Scope 2’ emissions - that is, the greenhouse gas release from coal, gas and other fossil-fuelled power stations on electrical grids, serving power demand, from the perspective of those doing the demanding. 

The Protocol is up for review and being lobbied in a few different directions - I’ve covered that briefly here. You can also read a great post by EnergyTag 's Killian Daly here.

The flow of electrical energy on a grid is tough to understand. Power grids are the world’s largest machines: a collection of generators all interconnected with thousands of points of demand, synchronised through a common heartbeat. Just as you can’t isolate which lake the water in your tap came from, you can’t know which generator “made” the energy you consume to power your phone.

We have to settle for the best guess: take the average emissions of the power grid at the moment you were connected, and multiply that by the amount of energy you used. If you were to add up the real, measured emissions of the power stations, and this method of estimating the emissions of power demand, you’d get two numbers exactly matched.

This is known as ‘location-based’ reporting, and it has been the default for a long while. 

Sitting on the margins

There has been a shift in the underlying philosophy of reporting on the climate consequences of a company’s decisions, in recent years. “Marginal emissions” try to capture whether a unit of new demand either worsened or avoided greenhouse gas emissions. 

Flick the power switch on your air conditioner, and you are increasing the power demand on your local grid. But what meets your new demand? You can imagine a simple model where a generator (the cheapest still available) spins up to serve that demand, and the emissions of this generator would be the emissions you induce by switching on your appliance. This is known as “marginal emissions.”

Olivier Corradi , founder of Electricity Maps , wrote: “Marginal emissions represent a different paradigm of attributing responsibility for emissions which attempts to capture the consequence of our decisions. As they represent a different paradigm (similar to carbon offsets), they shouldn’t be mixed with the use of average emissions for carbon accounting.”

Mixing the concepts of "attributing emissions" (measurement of history) with the one of consequential accounting (the consequences of our actions) could mean claiming more emissions avoided than emitted. 

As an example, one could purchase certificates from wind turbines on a power grid that’s generating 100% renewable energy while having coal or gas on the margin (which is never or rarely used). Under the logic of consequential accounting, and without strong protections for verification and auditing, you could claim that purchasing these certificates avoids coal or gas emissions from the margin, although they were never or only rarely used.

The reality of power grids is far more complex. Millions of people and many businesses are shifting demand up and down, and hundreds of generators are being managed in real-time to match that as closely as possible. 

When you decide to measure something so inherently fuzzy, you will invariably end up with wildly different numbers coming from different sources. You can see this nicely on the Electricity Maps blog, where they compared different estimates of ‘marginal carbon intensity.’ 

“Marginal emissions factors cannot be measured, and even grid operators are not able to validate them historically. In the absence of any ground truth, it is not possible to confidently assess the accuracy of a marginal signal.”

This brings us back to the logic of carbon offsetting. Over the past few years, there have been a drumbeat of scandals and criticisms directed at the offsetting industry. These centre around the companies claiming significant reductions in emissions, when different methodologies actually return different numbers. 

Part of what these scandals show is that trying to accurately measure the difference between this world, and the world where you took no action, is very hard. Blending those numbers with real-world measurements takes this from ‘hard’ to actively counterproductive.

Instead of simply measuring the direct emissions of certain actions, the ‘marginal emissions’ concept tries to shift towards showing progress against unverifiable counterfactuals, and then mixing up this very different approach with traditional emissions measurements.

The danger is simple: companies could end up wildly underreporting their greenhouse gas emissions, because they’ve assumed their actions had climate benefits that, in reality, are extremely difficult to measure, report and verify. 

The stakes for the big players

The stakes are huge here: many large companies running data centres are seeing their power consumption rise at an unprecedented rate, and that is causing their emissions trajectories to blow out. The chart below shows Meta 's greenhouse gas emissions for its scope 1 (direct) and scope 2 (electricity) emissions, compared to its recently established climate target:

But the picture changes dramatically when you shift to using ‘location based’ data: 

If you accept Meta’s renewable procurement claims, it looks like their scope 2 emissions fell 99% between 2018 and 2023. If you take the 'unadjusted' grid average, it looks like a 315% increase. I hope this gives you a nice view of the stakes here, when it comes to taking measured emissions and then adjusting them for how much 'climate action' you took.

Several companies, including Meta, are pushing for a revision to the Greenhouse Gas Protocol, the basis for emissions accounting for most corporations in the world. This voluntary program sets the standard for reporting how electricity consumption worsens the release of greenhouse gases, and the ‘Emissions First’ partnership proposes embedding this ‘avoided emissions’ thinking into that measurement.

It is very likely to badly overstate the emissions reduction from actions taken by companies (such as purchasing renewable energy certificates). Given the urgency of this problem, recreating carbon offsetting within the power sector is not a useful intervention.

(part 2 of this will explore why large power consumers, such as Amazon and Meta, are pushing for a more problematic approach to accounting for power emissions)