SIGNAL No.16 — ESA's Biomass satellite is putting forest carbon on a public ledger

Today's Signals

Hosted by Oli.

01 — Biodiversity's baseline is being rewritten from orbit and from the horizon

• ESA marks the European Space Agency's Biomass satellite as having finished commissioning and now opening its global forest archive to anyone — the mission carries a P-band synthetic aperture radar, a wavelength long enough to see past leaves and branches and read the woody trunks underneath, narrowing error bars on tropical forest carbon stocks that have been wide enough to swallow most climate policy assumptions, and the data is open so any researcher anywhere can pull it
• Trends in Ecology & Evolution publishes the 2026 horizon scan in which a panel of conservation researchers across continents flags fifteen emerging issues most likely to reshape biodiversity in the coming years, and the frame the authors keep repeating is refreshed baselines — the reference states many of today's baselines describe are decades old, drawn from ecosystems that no longer exist in the form they were measured, so without updated machine-readable baselines new pressures will only be caught after the damage is irreversible
• IUCN carries the new IPBES Business and Biodiversity assessment, whose finding is direct — the Kunming-Montreal Global Biodiversity Framework cannot be met by governments alone, and what has stalled progress is not ambition but measurement, with no shared language for what counts as a nature impact across an industry, a portfolio, or a supply chain, so the frontier work is now mundane: the data schema, the line items, the disclosure standard

02 — Craftsmanship is being treated as a design problem rather than a relic

• SAGE / IOS Press formalises hybrid human–AI systems for safeguarding cultural heritage and craft, stepping past both the anxious framing that automation will hollow out tradition and the evangelical one that AI will save it — the machine takes pattern capture, search across large catalogues and comparative analysis, the human keeps embodied judgement that a potter or weaver makes with their hands and cannot fully verbalise, and the contribution is to draw the line cleanly enough that you stop defending the craft against the technology and start integrating them
• arXiv publishes Hands-Free Heritage, which couples a coordinated robot manipulator with high-resolution three-dimensional scanning to digitise museum objects, naming the real bottleneck as the expert hand-work — the careful turning and re-positioning that can take a conservator hours for a single fragile vessel — and shifting that hand-work to a robot under supervision, so what used to be a heroic one-off for celebrated pieces becomes a routine pipeline that can address whole collections, and works especially well for the awkward geometries that have stalled digitisation programmes for years
• Met Museum stages The Infinite Artistry of Japanese Ceramics as a re-hang of works already in the collection, an arrangement designed to make a different argument from the old layout — that clay, glaze and fire are physical processes whose decisions a contemporary potter is still making about the same three variables a tenth-century potter was, and reading the show alongside the 3D-scanning paper makes the contrast sharp: one project digitises the object, the other shows why the unrecorded knowledge in the maker's hands still matters

03 — Memory and naming are being rebuilt with consent and taste at the core

• Prism Reports profiles the community-led model emerging for using AI to preserve endangered languages, anchored on Te Hiku Media's Kaitiakitanga licence — kaitiakitanga is a Māori concept best translated as guardianship — under which the community retains control of any model trained on its language data, and on Michael Running Wolf's training pipeline for North American Indigenous languages under similar terms; consent is built into the system at design time rather than asked for at the end as a courtesy, which is what AI ethics looks like when it is structural rather than performative
• UNESCO reframes its Memory of the World programme with a clean distinction — digital heritage is not the work of scanning records and putting them online (that is digitisation), it is giving historically marginalised communities the agency to shape, narrate and share their own past, to decide what enters the record and on what terms; the shift in subject is the point, from institution-as-author to community-as-author, mirroring the Indigenous-language work and making explicit that the old default about who holds the pen on collective memory was the wrong one
• Focus Lab reads 2026 brand naming as two pressures arriving together — trademark saturation, where short and pronounceable names in most consumer categories have already been registered, and the wide availability of large language models as naming assistants — concluding that LLMs sharpen the funnel by expanding the search space and surfacing prior art, while the final call on which name carries the right weight and cultural read still belongs to human taste
• How Brands Are Built runs an extended hands-on test of AI-assisted naming and lands on a precise verdict — the models are excellent at the first twenty ideas because word play, conceptual mashups and semantic clusters around an attribute are exactly what they are well shaped for, but the next round, where you distinguish a name that carries weight from one that is merely available and judge whether it is distinctive enough to defend over years, remains a human call; read with the Focus Lab piece, a steady picture emerges across creative industries where the work the model can do is moving inward fast and the work that stays human is moving up the value stack toward judgement