Colchuck lake, through sparse trees, with bare rock mountains behind; Apline Lakes Wilderness, Enchantments area, Washington.
Colchuck lake, way out of scope.

Organizations are being encouraged to track their carbon emissions via public pressure and perceived competitive advantage, but what exactly are they tracking and how are those boundaries defined? Do the scopes even make sense for the purpose to which they are being put? Let’s dig in and see what’s what.

The Greenhouse Gas (GHG) Protocol is a widely used international accounting tool that defines three scopes for greenhouse gas emissions. Already we can see some challenges, this isn’t carbon accounting, and isn’t even a full GHG accounting framework. This is emissions. Active contributions in the day to day steady state. While we may find individual items throughout that do not fit the scope framework, one thing that jumps out immediately is the agency in embodied carbon. The entity authorizing and paying for the work is not the entity emitting the greenhouse gas, and the closest we could get there would be an aggressively broad interpretation of Scope 3 emissions, but before we get there let’s see how the scopes have been defined.

The scopes track the emissions of seven greenhouse gases of interest within the Kyoto Protocol:

1. Carbon dioxide (CO2)

2. Methane (CH4)

3. Nitrous oxide (N2O)

4. Hydroflourocarbons (HFCs)

5. Perflourocarbons (PFCs)

6. Sulfur hexafluoride (SF6)

7. Nitrogen trifluoride (NF3)

Folks love to point out that most of the air isn’t made up of these things and most of the heat retention comes through water vapor in the atmosphere. Both are true but suggest a dramatic misunderstanding of how impact needn’t be proportional to volume or mass, like expecting the empty land in a typical color-coded voting map to overwhelm the population centers where humans actually live. Small changes in heat retention increase temperature, increase water carrying capacity, increased water content, further increases heat retention, further increasing water carrying capacity. Sure the water is doing most of the work, but the water is the dependent variable here and the GHGs are in the drivers seat.

Graphic with the 7 greenhouse gasses and illustrations showing upstream, direct, and downstream examples of contributions to emissions scope.
Graphic from

Scope 1 — All direct GHG emissions

· Fuel combustion

· Company vehicles

· Fugitive emissions

Scope 2 — Indirect GHG emissions from purchased electricity, heat, or steam.

· Purchased electricity, heat, and steam

Scope 3 — Other indirect emissions, for instance supply chain or outsourced functions.

· Purchased goods and services

· Business travel

· Employee commuting

· Waste disposal

· Use of sold products

· Transportation and distribution (up and downstream)

· Investments

· Leased assets and franchises

Out of Scope — Nothing, or everything?

The scopes are set up with an old school heavy industry focus, as such some things like purchasing heat and steam are specifically called out while the purpose for the ML algorithm running in the cloud is left ambiguous in impact: did that workload generate a number that the customer needed, or did that workload double the extractive capacity of the Permian Basin? This specific example absolves or implicates big tech in the fossil fuel extractive responsibility, and we can all guess which way this breaks when companies are free to perform their own attributions — all hyperscale cloud providers make a big deal about operating on clean power and being (or aspiring to be) net zero, including primary extraction enabled by their technologies and performed by their customers would blow it all apart.

Of the scope definitions, Scope 2 is the least problematic. How much power did your entity or corporation (switching to corporation now) purchase and how clean was its source? There is a bill, an ongoing agreement with another company or utility, and a transfer of energy. In Scope 1 it is easy to quibble on attribution and in Scope 3 we have to grapple with IP and non-physical services.

In Scope 1, some attribution is straightforward. If I operate container ships, their operations is in scope. But, in a world where employees are expected to bring their own devices and when we’ve extended devices into things like vehicles, can we attribute that fuel consumption to the individual instead of the corporation? To take that to an extreme, as a car rental company am I responsible for the vehicle miles travelled by my customers in scope 1? If I’m Uber, is VMT my Scope 1 or am I classing everyone as a contractor and it counts against their Scope 1 emissions?

In Scope 3, everything is a mess. Did I lease this building from a holding company established to manage the land and construct the facilities? In building the facilities do I own the embodied carbon of the whole construction effort, or just the operating costs? If my product is software, am I accountable to what customers do with that software? If my employees need to use a cell phone, but I expect them to use their own cell phone for work, should that embodied carbon count against the corporate Scope 3? Much of this is fringe errors, except for the attribution of intent across corporate boundaries. When Microsoft rebuilds its campus and carves out a narrative around net-zero operations but downplays the 85,000 tons of embodied carbon in the 5,000 underground parking stalls that went in when the mid-life buildings of the former campus were torn down. Perhaps we just think that’s embodied vs. emissions and because we’re counting emissions here we don’t care about any level of embodied carbon on the road to lowered emissions — damn the payback cycle (if payback is even possible).

At some level, an accounting firm taking credit for ExxonMobil’s oil and gas production is a bit silly. The capability provided by the accounting shop is easily replaced by any other accountancy company, which is an ok bar on one end, and ExxonMobil ought to be taking credit for this in their own accounting (literally drawing the line at direct customers instead of lifecycle analysis may let them out of Scope 3 by passing through intermediaries before getting to the point of consumption). But for companies materially impacting the productivity of their customers they must take ownership of the consequences they’ve helped unleash. So yes cloud and software vendors, support services, patent attorneys, etc. should add increased oil and gas productivity to their own carbon tally. Lifecycle analysis matters, and anytime one corporation puts a boundary on another there is a real chance from the other players won’t line up neatly and something will slip through. The entire transformation of Drax from coal to wood pellet burning comes from an accounting alignment — a massive deal.

So, how to handle the attribution equitably when the transfers may be of very different kinds? How about, since this is an accounting game anyway, we attribute a carbon rating by company and decorate revenues and expenses with a record of the green-ness of that transaction. If you do $100M annually with an oil major, you pick up a fraction of their annual GHG load proportional to their expenses (indexing to revenues would be susceptible to dilution as energy prices are often divorced from production costs). The alternative is the madness of every company carving out their own GHG boundaries because they are likely to get away with leaving important things out because the fine details are not often shared outside the walls.

Scope 3 was meant to hold corporations to account for where and how they make their money. In particular, the “Use of sold products” line suggests the consequences of sales, even of the sale of software and services, matters to one’s scope three emissions accounting. But we know this isn’t the case today. Just look at green media darling Microsoft, their recent carbon blog and aspiration to zero out their historic carbon footprint:

Chart showing lifetime carbon dioxide emissions for Microsoft, showing a sharp drawdown to reach net zero by 2050 (including significant negative emissions in the future).
Chart showing lifetime carbon dioxide emissions for Microsoft, showing a sharp drawdown to reach net zero by 2050 (including significant negative emissions in the future).
From the Microsoft Carbon Blogpost, showing lifetime emissions and anticipated future trend.

The problem becomes clear when we read through a couple announcements from last year around activities in the Permian Basin. Just from the press announcement we can see a carbon impact on the order of 50,000 barrels per day, or around 18M tons of CO2 per year — with no mention of impacts on natural gas production or methane. These contributions are explicitly called out as a direct consequence of the investment in Microsoft Azure solutions — this capacity would not be realized without Microsoft getting in the game. So it stands that these contributions must accrue to Microsoft under the “Use of Sold Products” portion of the Scope 3 definition. Look at that chart from the Carbon blog again. The peak, in 2018, was about 12M tons of CO2 a year, the contribution from these two deals, and only a portion of those two deals, is 18M tons of CO2 per year. It isn’t in the chart because a full accounting would dominate the chart. When we let corporations define their own boundaries they are going to include or exclude with intent, and the intent here is clearly to take the money from Exxon, Chevron, and Schlumberger but not to take credit for the carbon that money has produced.

Corporations already play loose with the exchange between embodied and emitted carbon, but they are also doing this with traditional emissions attributions and distancing the revenue from the consequences of taking that revenue. At the end of the day the environment doesn’t care how the carbon in the atmosphere is counted and attributed; the concentration will either lead global temperatures higher or lower based on the concentration itself. The science doesn’t care about your marketing and publicity efforts, so it is sad that we’re supporting a clear double standard in the name of near term revenue.

Former thermoelectrics and fuel cell scientist; current software product manager. He/Him.