The framework, in depth.

Why eight criteria, not twelve or twenty.

The eight criteria the dashboard holds, each with a threshold the viewer can set, are climate trajectory, water stress, soil organic carbon, forest-cover trajectory, solar-energy potential, conflict proximity, regenerative-network density, and population density. They share the one property that makes them filterable: each reduces to a sourced number in a native unit. That is the quantitative spine of a fifty-to-hundred-year settlement decision, lose any one axis and the project's runway collapses on it.

A dimension The Collective treats as just as load-bearing sits deliberately outside that slider set: legal and economic context, who may buy land, under which ownership regime, at what price. It is documented region by region in the dossiers and the case studies, with its own source list below. But it is not a threshold criterion, because it does not reduce to a single sliderable number. "Non-EU citizens may acquire farmland only through a domestic company" is a categorical fact, not a point on a scale; forcing it onto a slider would be the exact false quantification the framework refuses. So legal and economic context lives in the prose and the dossiers, where it can stay honest, rather than in the filters.

Other criteria considered, cultural fit, food sovereignty, language barrier, biome-specific factors, are real but derivable downstream from the eight: cultural fit follows from regenerative network, population, and legal context; food sovereignty from water, soil, and energy; language barrier sits inside the demographic picture. The pruning principle is to keep as a filter only what reduces to a checkable number, and to hold everything else as honest prose.

Each of the eight is measured in its own native unit, drawn from a named public dataset. The grid below lays out the whole quantitative spine on one page: each criterion, the unit it is actually reported in, and the source that supplies it.

State paired with trajectory.

A current value tells the practitioner what a region is now; a trajectory tells them where it is heading. Both matter, and the framework holds them side by side rather than folding one into the other. The trajectory is a directional reading, stable, improving, declining, or volatile, filled by a credible projection where one exists, by an observed trend where it does not, and left null where direction is meaningless on the timescales that matter. Soil composition, for instance, does not meaningfully "trend" over a single settlement lifetime, so it carries a state and no trajectory.

Climate is the case where state and trajectory fold into one sourced number. The value on the dashboard is not today's weather but the mid-century projection, the 2041–2060 mean annual temperature from the WorldClim CMIP6 downscaled ensemble under SSP2-4.5. Alentejo reads 19.5°C at mid-century, Connemara 15.0°C, Transylvania 11.0°C, each a projected state that already encodes the warming a fifty-year project is planning into. Water works the same way: WRI Aqueduct 4.0 supplies both a baseline stress score and a 2050 projection, and the framework carries both without averaging them.

Forest cover is the pure-trend case, where the number is the trajectory. Hansen Global Forest Change tracks tree-canopy change from 2001 to 2023 at 30-metre resolution, and the framework reads it as percent change per decade. Alentejo runs −2.4% per decade, canopy declining under eucalyptus plantation cycles and summer fire, while Connemara runs +1.1% per decade, slowly recovering under afforestation policy. Same criterion, opposite directions. No single "forest score" could separate those two regions the way the signed trajectory does, and the discipline throughout is to never collapse "fine today" into "will keep being fine."

No composite scoring.

One of The Collective's non-negotiables is that the framework never produces a single composite score across criteria. A composite score requires coefficients, how much rainfall offsets a degree of warming, how much cheap land offsets a thin civil-society stack, and once those coefficients live inside the framework, the framework is making the decision instead of the practitioner. Composites also let regions with one fatal flaw rank highly because compensating scores average it away, which is exactly the failure mode for fifty-year decisions where no flaw is averageable. Alentejo makes the point concretely: its solar resource is top-decile in Europe, around 1,865 kWh/kWp/yr, while its mid-century water stress is extremely high. A composite would let the exceptional sun pull the water problem back toward the middle of the range, yet water is the single axis that actually decides whether a settlement there survives to 2070. Averaging buries the one number that matters. The alternative is filtering: each criterion has a threshold the viewer sets, regions that pass all thresholds remain visible, regions that fail any one threshold dim. The shape of the result is a smaller honest set, never a ranked list.

Native units throughout.

Water stays in mm/year. Temperature in degrees Celsius. Soil organic carbon in g/kg. Land cost in euros per hectare. The framework refuses to reshape every variable onto one normalised grid because reshaping costs information. The modifiable areal unit problem (MAUP), familiar to anyone who has worked with spatial data, has two parts: the scale effect, where statistics change as the unit of aggregation changes, and the zoning effect, where they change as the boundaries shift. Water stress on a one-kilometre grid loses the catchment that actually matters. Cost-per-hectare averaged across a NUTS-II region disguises the difference between marginal upland and irrigated valley. Keeping each variable in its native unit, at the analytical scale where it makes sense, is the discipline the framework chose over visual neatness.

In practice that means each criterion is read at the resolution its source actually resolves, never one invented for tidiness: climate on the roughly 18 km (10 arc-minute) WorldClim CMIP6 raster, water on HydroBASINS level-6 catchment boundaries, soil carbon on the 250 m SoilGrids grid, forest change on the 30 m Hansen grid. Reshaping all four onto one universal grid, an H3 hexagon say, would either blur the 30 m forest signal up to 18 km or invent 30 m detail for climate that the model never produced, and either way manufacture a precision the data does not hold. The zoning effect bites just as hard: the Alentejo interior that reads a single moderate stress average across a whole NUTS-II region resolves, at HydroBASINS catchment scale, into specific sub-basins already flagged "extremely high." The aggregation scale is not a cosmetic choice; it changes the number the practitioner reads.

Source, vintage, and license for every value.

Every value the framework presents carries three things. A source, a named public dataset. A vintage, the year or scenario the value represents, "2041–2060 SSP2-4.5" for a climate projection, "2050 BAU" for water stress, "2001–2023" for a forest trend. And a license, the terms under which the data may be reused and redistributed. The public backbone is eight datasets: WorldClim CMIP6 (SSP2-4.5) for climate, WRI Aqueduct 4.0 for water, SoilGrids 2.0 for soil carbon, Hansen Global Forest Change for forest trajectory, the Global Solar Atlas for solar potential, UCDP GED for conflict, JRC GHSL for population, and the Global Ecovillage Network directory for regenerative sites. Most carry open CC-BY or CC-BY-SA terms, and the license field records which, because a value nobody is allowed to reuse is not a value a community can build on.

On a fifty-to-hundred-year decision, vintage and license are not bureaucratic footnotes, they are the point. A climate projection is a model run that will be re-run: newer ensembles, revised SSP pathways and better downscaling will move the numbers, and a value stamped "2041–2060 SSP2-4.5, WorldClim CMIP6" tells a community exactly which run to re-check it against years from now. The per-region dossier files in /data/research-dossier/<region>/<dimension>.md are the audit trail: no value without a source, no source without a vintage, no claim that cannot be re-checked. Where two sources disagree the dossier notes it; where the value is uncertain the dossier says so. Transparency is what lets a place that will outlive this dataset verify it for itself.

The seventeen map layers, and the gaps not shipped.

The interactive map carries seventeen toggleable layers, grouped into six themes. Climate & water: precipitation, water stress 2050, water depletion 2050. Land & soil: forest loss, soil organic carbon, land cover. Energy: solar-PV potential. Hazards: coastal flood and sea-level rise, seismic hazard. People & access: population density, travel time to cities, conflict density, ecovillage sites. Terrain & imagery: relief hillshade, topographic map, recent satellite, night lights. Every data endpoint is a verified public tile or WMS service read live, ISRIC SoilGrids, the Global Solar Atlas via WRI Resource Watch, WRI Aqueduct floods, GEM's global seismic-hazard map, the Malaria Atlas Project travel-time surface, Global Forest Watch, SEDAC population, Sentinel-2 cloudless imagery, NASA's Black Marble night lights, each carrying its own attribution and none behind a login.

What the map does not ship is as honest as what it does. There is no clean, openly-served public tile service for several layers the framework would want: air-temperature climatology, an aridity or Köppen-zone surface, riverine (as distinct from coastal) flood, a real SPEI drought index, or ESA WorldCover at its full 10 m. Rather than fake those with a proxy dressed up as the real thing, the map leaves them out and says so. The overlays you can toggle are the ones that resolve to an actual verified service today; the rest wait for a later build.

Composite scores.

The first tension is whether a regenerative-settlement framework should ever produce a single composite "livability" or "suitability" score across criteria.

Weighting authority.

The second tension is about who decides how much a given criterion matters.

Both tensions are deliberately deferred to V2 design. This site, the prototype and the upcoming V1 alike, neither scores nor weights; it shows the data and lets the viewer apply their own criteria. The viewer who needs the framework to do more than that should treat it as a sketch and wait for the next layer, or read the dossier files directly.

Localism. Decisions about land are best made primarily by people whose lives are continuous with the place. A tool that helps a stranger imagine a place cannot substitute for the slow work of becoming someone the place knows.

Bioregioning. Thinking in terms of watersheds, ecotones and cultural geographies rather than national borders and real-estate markets. A region can contain three watersheds and three different settlement traditions, and the bioregional scale is the scale that actually matters for land that is meant to last a century.

Reciprocity. Reciprocity is inseparable from bioregioning: you cannot live within the grain of a place while displacing the people already in it. Much of this land — Mi'kma'ki in Nova Scotia, the unceded Sinixt, Ktunaxa and Syilx territories of the Kootenays, the Zapotec, Mixtec and Chatino highlands of Oaxaca — is held in Indigenous or communal title, and arriving means entering obligation with the people, lifeways and nations already rooted there, not acquiring something empty. Where arriving in a place would harm the community already there, the honest answer is not to go. This tool exists to help people belong to and regenerate a place — never to land on top of one.

Healthy integration. Whether a place welcomes new arrivals, and on what terms, is a matter of trust built across decades. It is not a checkbox a framework can verify. The work of integration sits with the arriving person, in relationship with the people already there.

Land stewardship. Stewardship as relationship over time, not transaction. Buying land is the easy part; becoming responsible to it over decades is the work the framework cannot represent in a data layer.

The framework is one input among many, most of which live not in datasets but in conversations with the people who already know a place. These values are how the project keeps that conversation honest: the tool exists to help a community arrive in reciprocity and give back to a place over generations — and where it cannot be used that way, it should not be used.

The dossier work behind this page now covers twenty regions, ten in Europe and ten in North America, all live on the site as of May 2026. The European set came first: Alentejo, West Cork / Connemara, and Transylvania (explored in depth above) anchored the cut, with Galicia, Pembrokeshire, the Cévennes, South Tirol, Asturias, Saxony-Anhalt and rural Estonia joining as their dossiers were ratified. North America extends the framework to a second continent, Cascadia, Vermont, the Southern Appalachians, the Driftless Area, the Ozarks, Northern New Mexico, Nova Scotia, the BC Kootenays, Québec's Eastern Townships, and the Oaxaca highlands, spanning the arid Southwest to the maritime Pacific Northwest and a far wider catalogue of land-tenure regimes: US fee-simple with agricultural-use zoning, Canadian agricultural-land reserves, Québec civil law and CPTAQ farmland protection, and Mexican ejido communal title. The North American values are an honest first pass, every cell is sourced and dated, but several lean on national datasets and triangulation rather than a single pixel read, and that confidence is recorded alongside the data. The slate itself remains a proposal pending The Collective's sign-off.

A V2 data-ingestion build is in early planning. The current site is a hand-built prototype, values extracted by hand from PDF source documents, no live pipeline, no parcel-level granularity, and the planned V1 release polishes that prototype rather than replacing its engine. A serious V2 would ingest WorldClim CMIP6, WRI Aqueduct, SoilGrids and the regional cadastres on a scheduled cadence; surface per-cell source plus vintage in the interface; and let practitioners overlay their own thresholds against live data rather than a static snapshot. That build is months of work for a small team, not a single-person sprint. It hasn't started.

The Collective's invitation status: subscribers receive a link to the invitation PDF with context about the project, the contributor list, and how to join. The full framework documents remain private until everyone involved agrees on public attribution. The site itself stays open; The Collective is opt-in.

This page exists because the modal promised more. If you signed up expecting a deeper layer, this is the first instalment. Further regional case studies, a fourth, a fifth, will be drafted as time allows, and the page will grow with the work.