The Planet Health Score
An educational estimate of Earth-system health, and an honest account of what it does and does not know.
What it is
One number per region and one for the planet, on a 0 to 100 scale where higher is healthier. It is a coverage-weighted average of nine Earth-system domains. Five of them are scored as distance from a published scientific boundary: 100 inside the safe operating space, falling to 0 at the high-risk threshold. The other four are kept as contextual proxies rather than boundary control variables, and the score says which is which.
The boundary framing is grounded in the planetary boundaries literature, cited in the score document as Richardson and others 2023 and Steffen and others 2015.
What it is not
It is an educational estimate and not a scientific index. It is not certified, it is not peer reviewed, and it is not a measurement of current conditions on its own. Every signal carries a confidence label, and the score is built so that missing data lowers confidence rather than quietly moving the number.
Where the number comes from
The globe viewer is keyless. It reads a public score document from object storage and displays it; it never recomputes or alters the document. On any failure, whether the reader is offline or an object is not deployed, each source falls back to a representative sample bundled with the viewer and labels itself as such.
The document bundled in the public repository is one of those representative samples. It carries a representative pattern rather than current conditions, and its own metadata says so. Its numbers are not reproduced on this site for that reason. Its structure is, because the structure is the part that is stable and checkable.
The AI footprint stays outside the score
The application's own AI energy footprint, computed under AIEDS, is attached to the score document as a separate field with an explicit flag saying it is not blended into the score. A planetary health number and a product's own footprint are different claims, and folding one into the other would make both unreadable.
The rest of this section
- What it measures, the nine domains with their weights and their basis.
- Confidence, what the number beside the score means.
- Data sources, every provider and its license.
- Limits, what the score cannot currently tell you, and why.