Network carbon intensity: whoever picks the emission factor picks the number

Both bids on the table carry the same line: network carbon intensity, in kilogrammes of CO2 equivalent per terabyte carried. One declares 12, the other 4. You award the environmental score to the second and close the tender. You have just rewarded the choice of a statistical dataset, not a cleaner network: the two figures come out of the same formula, they measure comparable networks, and neither is false.

One indicator, two measurements and a coefficient

Recommendation ITU-T L.1333, Carbon data intensity for network energy performance monitoring, edition 09/2022, first and only edition, approved by Study Group 5 on 22 September 2022 and listed In force in the catalogue, defines that number: network carbon intensity energy, abbreviated NCIe, freely downloadable in English.

Equation (8-1) gives it in kgCO2e/TB — numerator “Total Carbon Emission of network operation”, denominator “Total Data traffic” — with the TB set by a note at 2⁴⁰ bytes, not 10¹². Equation (8-2) opens the numerator: NCIe is total energy divided by total traffic and multiplied by EF, the emission factor, in kg CO2e/kWh; in disaggregated form, Σ(Ej × EFj) over Σ(data trafficj), site by site. Equation (8-3) splits the energy by the sources listed in clause 6: grid electricity, backup energy (diesel or other types, solar panels included) and energy generated locally that does not serve as backup.

Two of the three inputs you measure yourself: the kilowatt-hours from meters or from equipment monitoring, the terabytes at the aggregation point of the user interface. Both are auditable, and both can be rebuilt from raw data if anybody challenges them. The third one is not measured: it is sourced from somewhere, and the result is proportional to it.

Eleven obligations, seven on the factor

The body carries eleven normative shall; the twelfth sits in the FOREWORD, where the word is only named to explain its meaning. They sit in clauses 8, 8.1, 9 and 9.2, and seven of the eleven concern the emission factor: two in clause 8, five in 8.1. The four should are all together, in the list of requirements in clause 6 — including the line asking the KPI to cover all sites and their supporting plant, which is therefore advisory rather than binding. Not a single may appears in the operative clauses: the eight in the document sit in the FOREWORD, in the intellectual property and copyright notes, in the introduction and in a definition taken from L.1330. The two must are the FOREWORD formula and one sentence in the introduction.

Eleven obligations and no declared option: the choice that moves the number sits inside an obligation.

The datasets are named, the hierarchy is not

Clause 8, immediately after equation (8-2): “In line with [ITU-T L.1450], the EF shall be sourced from recognized public sources and shall consider energy supply chain and distribution losses. Different emissions factors apply for different energy supplies such as grid electricity and local sources such as diesel and solar.” Three obligations in two sentences, and none of them on which dataset.

Clause 8.1 does name institutions, and that matters: “EFj shall be derived from existing national statistical data available in the country where the network is located or from international data such as that available from the IPCC, IEA or others.” Two intergovernmental bodies, plus an open referral.

But the word holding the sentence together is or, and the list ends with or others. The national figure and the international figure for the same country, in the same year, do not coincide; between the two the document sets no preference, nor does it define what makes a source recognised. Neither reading is a deviation: both are compliance, and a tender that cites the Recommendation and stops there has accepted both in advance.

Two mandatory methods, one slot in the formula

Clause 8.1 doubles up. If the network owner buys renewable electricity, “the emission factor shall be referred to as market-based for this share of EF”, with the specific values to be requested from whoever sells the energy. Then: “It is necessary to calculate emissions according to both location-based and market-based methods, in line with the [b-GHG protocol].”

Both. But equation (8-2) holds a single EF, and the indicator that ends up in a bid is a single number: which of the two is the NCIe to be declared, the document does not say. For an owner with multi-year renewable purchase agreements the gap between the two is no rounding error: it is the difference between declaring the country electricity mix and declaring your own contract, and the second one is the number a bidder will pick.

A blunt provision closes the clause: “In the case of renewable energy produced locally on site, the related EF shall be equal to zero considering that losses in power distribution are negligible.” Zero by provision, on a rationale that concerns distribution losses only. The supply chain that clause 8 requires to be considered for the other sources does not come back in that line.

The period is fixed, the boundary is named by subtraction

On time the text leaves no room: clause 9, “The time duration of the measurement shall be one year (365 days).” One month (30 days) and one week (7) are allowed only “for their internal consideration”, for the network owner’s own use rather than for anything published. Definition 3.2.1 speaks instead of “a defined period of normal operation”: the clause is what counts.

On the boundary the document says little: across eighteen pages neither boundary nor reporting period appears. Scope 1, 2 and 3 occur twice each, in the introduction and in definitions 3.1.5-3.1.7 taken from L.1450, but no clause says which of them NCIe falls into; clause 7 delimits by removal: “This Recommendation consider only network energy consumption; other sources of GHG emission such as owner maintenance activities are excluded.” The same knot as the boundary in the data centre count, and it unties the same way: in writing.

What is already enforceable as it stands

Three provisions can be copied into a specification and verified on acceptance.

Clause 8 — the factor comes from recognised public sources and covers supply chain and distribution losses: the supplier is asked for the source, not only for the value.
Clause 8.1 — the factor for each source derives from national statistical data of the country where the network sits, or from international data; for the purchased renewable share it is market-based, with the specific values requested from whoever sells the energy.
Clause 9 — the measurement runs for one year, 365 days, and for fixed and access networks the traffic is read at DSLAM or OLT level.

The four lines that make two numbers comparable

Which emission factor, in kg CO2e/kWh, source by source. From which dataset, named with publisher and year of edition. By which method — location-based or market-based — and, calculating both as clause 8.1 requires, which one makes the declared NCIe. Over which period, with a start date and an end date. Plus the clause that holds the choice still: changing dataset is allowed, with the recalculated prior year attached.

In the specifications and compliance work we deliver, those lines are clauses, verified on acceptance one by one: the declared source exists and is the one named, the value is rebuilt from the electricity and traffic meters, testing and documentation are handed over with the job. On the other axis, meters, invoices, purchase agreements, traffic at the aggregation points and as-built records stop being scattered files and become a single map of the network: on it an AI flags the stalled meter, the factor still on last year’s edition and the drift that takes you outside your declared figure, and a team of ours, no subcontracting, steps in before that figure reaches a bid. Where an AI is needed on that data, it runs inside the client perimeter — on-premises or in a dedicated cloud with data centres in Italy — together with CSIDIA, the other company in the group.

Do you have to declare your network carbon intensity, or compare the figures two suppliers have declared? Talk to an engineer: the site visit is free of charge, and the factor is chosen before the number is written.