Start here · how to navigate this surface

RegenCity interactive analysis

Four analysis surfaces over the RegenCity indicator work, organised into two groups: the parcel study, and the H3 framework. This page explains what lives in each tab, which one to open for a given question, and the interaction rules they share.

Read-only presentation surface
Nothing here calculates an indicator, changes a method, or promotes a sandbox candidate. The Python pipeline and the sandbox scripts remain authoritative; every tab consumes an already-validated snapshot.

Parcel study

Singapore + Tokyo · parcel grain

Parcel explorer

Parcel-level comparison of eight supplied normalized indicators across two cities. District filtering sets the analysed population, district highlighting sets visual emphasis, and a pairwise matrix carries the same selection into correlation space.

  • Single-city or pooled two-city parcel populations
  • Linked X/Y distributions with quartile references
  • Relationship plot as points or square density, with a fitted line
  • District comparison table ranked by the selected indicator
  • Configurable pairwise correlation matrix
  • Source and method audit with per-indicator caveats
  • No map layer — the supplied files carry district names but no geometry
Open the parcel explorer →

H3 framework

Which tab answers which question

Rules the explorer tabs share

The parcel explorer and the cell explorer follow the same interaction contract, so a control learned on one tab behaves the same way on the other.

Filtering decides what is analysed City, district, subzone, and percentile controls remove rows from the analysed population. Every statistic below them is recomputed on what remains.
Highlighting only decides emphasis Highlight controls never remove a retained row. They split the retained population into context and highlighted treatments for rendering.
Percentile filters apply jointly X and Y percentile intervals are applied together, so the distributions and the relationship plot describe the same retained rows.
Statistics use the full retained population Correlations and fitted lines are computed on every retained row. Point sampling is a rendering concession only, and it always preserves highlighted rows.

The two verification tabs are review surfaces rather than filtering surfaces: they present fixed, pre-validated evidence panels, and the built-form tab adds a city selector.

Where the data comes from

Parcel explorerDeterministic snapshot built by src/data/parcel/snapshot.json.py from the two supplied parcel CSVs, harmonized into one browser payload.
Cell explorerDeterministic snapshot built by src/data/snapshot.zip.py, which runs the canonical evaluation preflight and joins the declared sandbox outputs.
Built-form verificationDeterministic snapshot built by src/data/built-form.zip.py from existing sandbox outputs. It validates and packages; it does not recalculate.

Snapshots are cached under src/.observablehq/cache/. The cache, dist/, and node_modules/ are generated and are not versioned. The parcel CSVs are not versioned in this repository; the loader looks for them under REGENCITY_PARCEL_DATA_DIR, then data/parcel_study/, then the sibling Regen_Parcel_Study/ directory. Method definitions, indicator specifications, and evaluation eligibility all remain upstream in the pipeline repository — see observable-framework/README.md for the publication boundary in full.