What the signal measures
One question: are the real-world emissions of the facilities a company owns going down, staying flat or going up? The answer is measured from emissions observed independently of the company, not taken from the company’s own report, and it is shown for the company and for each facility behind it.
| Signal | Rule |
|---|---|
| Strengthening | Ownership-weighted emissions across the company’s facilities falling by more than 2% a year on the trailing series. |
| Stable | Change within ±2% a year. |
| Deteriorating | Rising by more than 2% a year. |
| No signal | Fewer than three years of observed data, or every facility sits in a sector where activity is not observed (below). |
It is a signal, not a rating. It does not assess the quality of a company’s climate strategy, the credibility of its targets or its governance. It reports what its operations have done.
Three kinds of emissions data, kept apart
Each kind answers a different question, and none is allowed to stand in for another.
| Kind | Source | Its job |
|---|---|---|
| Observed | Climate TRACE facility-level estimates: power, steel, cement, refining, oil and gas and other large sources worldwide, monthly. | Sets the signal. |
| Verified | EU Emissions Trading System installation data; US EPA Greenhouse Gas Reporting Program; US EPA Clean Air Markets continuous monitoring. | Calibrates and tests the observed data. Never sets the signal on its own. |
| Reported | Company-reported Scope 1 and 2 emissions, from the World Benchmarking Alliance dataset and company reports. | Comparison only. Shown in its own panel; never sets the signal and never calibrates anything. |
How satellite observation enters
For several sectors — power generation and heavy industry in particular — Climate TRACE estimates whether a plant is operating, and how hard, from satellite and remote-sensing observation: visible plumes, thermal signatures and production activity seen from orbit. That activity is converted to emissions using the plant’s fuel and technology characteristics.
We do not process satellite imagery ourselves. We use Climate TRACE’s published estimates, and each facility states the release its figures come from.
Where activity is not observed, there is no signal. In some sectors the estimates are derived mainly from a plant’s capacity and standard emission factors rather than from observed activity, and they barely move from year to year whatever the plant actually did. Those sectors are detected from the data itself — by the share of facilities whose estimate is identical across consecutive years — and are marked activity not observed rather than shown as Stable.
Verified data, and what it calibrates
- EU Emissions Trading System. Annual emissions by installation, checked by an accredited verifier and published in the April after each year.
- US EPA Greenhouse Gas Reporting Program. Annual emissions for every facility above 25,000 tonnes CO2e, with named parent companies and ownership shares. EPA’s most recent published year is 2023; the program’s future reporting is under federal review, and only EPA’s own releases are used.
- US EPA Clean Air Markets Program Data. Continuous stack monitoring at fossil power plants, measured hourly and published quarterly. The live verified series for US power.
Facilities in the verified data are matched to the same plants in the observed data by location and sector. For each sector we publish how the two compare: the typical ratio of observed to verified, the spread around it, and how often the two agree on direction from one year to the next.
Why direction holds where level does not
An observed estimate for a single plant can sit some distance from the verified figure, and how far depends on the sector. So an observed level is never presented as a precise number.
Direction is different. When the same model is applied to the same plant every year, a consistent over- or under-estimate is present in every year alike, and it largely cancels in the change. A plant overstated by the same proportion in two consecutive years still shows the right change between them.
- One release only. Every year in a facility’s series comes from the same release of the observed data. Each release re-estimates history, and mixing releases would create changes that never happened. A new release is adopted whole and the history re-run; it is never spliced onto the old one.
- Direction is tested, not assumed. In every sector with verified data, the share of facility-years where the observed and verified changes agree in sign is measured and published. A sector that fails the test carries no signal.
From facilities to companies
Each facility is linked to its owners and, where there are several, emissions are attributed by ownership share. Owners resolve to the ultimate listed parent through the same issuer graph and LEIs used for determinations. A company’s signal is the ownership-weighted change across its facilities over a trailing period of at least three years, compared with peers in the same sector and region.
A company whose facilities cannot be linked to it is reported as unlinked, not as Stable.
Seeing the year as it happens
Verified figures arrive once a year, a few months after year end; a company’s own report often a year or more after that. The observed data is monthly, so the signal can show the shape of the year — emissions rising in the first half and falling in the second, a plant offline for a quarter, a restart — months before the verified annual total.
In-year readings are marked provisional until the verified figure for that year is available, and are never compared against a verified year as if they were one.
Coverage, region by region
| Where | Level | Direction |
|---|---|---|
| EU; large US facilities | Calibrated against verified data and shown with its error band. | Measured, and tested against verified direction. |
| Rest of world; smaller US facilities | Not shown as a figure: no verified reference. | Measured within each facility on one release; marked uncalibrated. |
| Activity not observed | Not shown. | No signal. |
Facilities without verified data — much of Asia among them — still receive a direction, because it compares each plant with its own past rather than with plants elsewhere. Differences in technology and fuel between regions do not enter it.
Reported against observed
Where a company reports its own emissions, the report is set beside the observed figures in a separate panel. A gap between the two is shown with both series and their sources; it is not a finding, it is not a determination, and it never fires a band. Its use is practical: any metric built on reported figures inherits whatever gap the panel shows.
How it relates to the bands
The transition signal sits beside the bands and never inside them. It cannot move a company into or out of any band, including Leader, which keeps its own published rule. It is not a transition-risk model, an implied temperature rise or a projection: every value describes emissions that have already been observed.
Limitations
The signal covers emissions at facilities a company owns. It does not cover purchased energy or value-chain emissions, and it does not assess targets, plans or governance. Observed estimates are models, and their errors are reported by sector rather than hidden. Facility ownership changes; attribution is point-in-time, as for determinations.
All published output is research. Sustain ESG Lab is not authorised or regulated by the Financial Conduct Authority, and nothing here is investment advice or a personal recommendation. The signal identifies holdings whose observed direction warrants review; the decision stays with the reader.
← Back to the methodology