Skip to content

ROADMAP.md

Latest commit

 

History

History
145 lines (118 loc) · 8.53 KB

ROADMAP.md

File metadata and controls

145 lines (118 loc) · 8.53 KB
Raw

Roadmap

Single source of truth for status, open work, and methodology decisions. Updated 2026-07-31 (manuscript in editorial review; score + Atlas built and soft-launched).

Current status. The manuscript (paper/latex/main.tex) is drafted, compiles clean, and is under a structural editorial pass before submission. The NEPI score and Atlas are built and soft-launched under noindex (dissemination/launch_checklist.md); full launch is gated on acceptance. Remaining before submission: Zenodo DOIs, front-matter, and the editorial revisions.

The publication sequence

Three self-contained stages; each stands alone.

Phase 0 — submit

The paper stays pure: the empirical finding (two axes, the rate, lock-in), no index proposal, no tool promises.

  1. Apply the editorial revisions; final prose pass.
  2. Mint DOIs: Zenodo code release + deposit of the built per-OA artefacts.
  3. Front-matter: co-author details, acknowledgements, cover letter (kept local).
  4. Submit.

Phase 1 — NEPI score + Atlas v1 (done; full launch on acceptance)

Built 2026-07-29 and deployed 2026-07-31. The score is measured, not modelled: the compositional regression is for inference, the score is descriptive of the place as lived. Specs: dissemination/score_spec.md, dissemination/atlas_architecture.md, inventory at dissemination/README.md.

  • Score (stats/nepi_score.py): per-OA A–G on the rate, bands frozen at the 2021 household-weighted distribution (dissemination/nepi_bands_2021.json).
  • Current vs potential label: each OA scored as-lived and under full technology deployment (deterministic, from the scenarios.py levers); the gap between the two labels is the lock-in result made visible per neighbourhood.
  • Atlas v1 (stats/atlas_export.py → site/): map + label card + the three deterministic technology levers; no model anywhere. Live under noindex on Cloudflare R2; de-noindex and custom domain on acceptance.

Phase 2 — the NEPI paper (after acceptance)

  • Paper 2: the score design, banding rationale + sensitivity, score stability across scales, the equity distribution of letters, current-vs-potential label, the Atlas as artefact; cites paper 1 as evidence base.
  • XGBoost is dropped from the plan (decided 2026-07-29). Nothing in the score or the Atlas needs a model, and modelled form-change what-ifs (densification) would undermine the score's measured-not-modelled identity. The old code stays in git history; any future densification study is a separate project that picks its own method.

Scope decisions (consumption audit)

The rebuild targets only what the two-axis analysis consumes:

  • KEEP (load-bearing): Census 2021, DESNZ postcode metered energy, EPC (build year + dwelling floor area + best-fabric intensity), OS Greenspace/UPRN/ Code-Point, OS Open Roads (network access via cityseer), FSA (food + grocery), NaPTAN, GIAS, NHS, Census workplace jobs, IoD 2025, DVLA vehicles (bev_share), NTS9904 mileage, ONS 2021 RUC.
  • Removed (nothing consumed them): LiDAR/momepy morphology (cityseer centrality not revived — accessibility only) + OS Open Map Local footprints + OS Built-Up-Areas/Boundary-Line. In git history.
  • Removed from the tree (in git history): the summed three-surface / A–G code (scorecard, bands, empirical access-penalty model, three-surface figures) and the old Atlas (XGBoost planning models + static site), taken out in the two-axis migration. The scoring has since been rebuilt on the two-axis frame (stats/nepi_score.py); XGBoost is dropped (Phase 2 note above). Plus the earlier archive: Census 2011, DESNZ LSOA energy, MSOA OD flows, BRES+GVA scaling, NESO projections, the basket index.

Done

  • National OA dataset — assembled in the stats layer.
  • Network access (stats/oa_network_access.py): cityseer over OS Open Roads, national network built once and queried per catchment band, each OA at its own NTS car-trip catchment, the built-once counts matching a literal per-OA computation to ~2% (~12 min); the rate is 3.9× access per kWh (access advantage × energy saving, access_profile.py).
  • Two-axis analysis (paper/summary.md): NTS-anchored car-travel energy, lock-in (per dwelling 2.12× → 1.51×; at equal family size 1.71× → 1.18×), network access rate (3.9× access per kWh) + on-foot gap (~27×), heat-vs-size decomposition, all on the shared stats/oa_data.py core.
  • Decarbonisation scenarios (stats/scenarios.py + paper/figures/fig8_scenarios.png): fabric, heat pumps and EVs as separate levers over the energy axis, at CCC Seventh Carbon Budget Balanced Pathway 2040 uptakes (heat pumps 50%, EVs 75%) and full deployment. Fabric closes ~20% of the log gap, heat pumps ~2% wider (a delivered-energy fuel switch that unmasks travel), EVs ~18%; full rollout leaves 69% surviving, access unchanged in every scenario. lock_in.py is retained as the fabric+EV bound (1.51×).
  • MAUP scale check (stats/maup_scale.py): the compositional energy gap re-fit household-weighted at OA/LSOA/MSOA — total survives re-zoning (2.12/1.88/1.72×; support-respecting dominant-type median 1.74/1.59/1.47×); the heat sub-component's MSOA reversal is an out-of-support extrapolation.
  • Two-axis migration cleanup: stripped the retired three-surface / A–G code and the old Atlas; unified the EPC→OA aggregation (data/aggregate_epc_oa.py); lean orchestrator (urban_energy.pipeline, acquire-only); REPRODUCTION.md.
  • Accessibility bands settled on the minute-clean ladder (400/800/1600/4800/ 9600 m ≈ 5/10/20/60/120 min at ~80 m/min) — kept as-is.
  • NEPI score + Atlas + site (2026-07-29 → 2026-07-31): stats/nepi_score.py (bands frozen), stats/atlas_export.py (aggregates, PMTiles, postcode shards), static site in site/, deployed to Cloudflare R2 under noindex.

Open work — by layer

Data (changing requires re-acquiring)

  • None outstanding. Climate (HadUK-Grid heating-degree-days) is acquired (oa_hdd.parquet) and held as a confound in the energy ladder.

Analyse (computed in stats/; cheap to revise — minutes)

These are the contestable scientific choices; none gate acquisition.

  • Per-household vs per-capita unit. Resolved in the manuscript: per dwelling is canonical, with household size and floor area as freely estimated controls (elasticities ~0.5 and <1, so fixed denominators are rejected). The paper reports the as-lived and equal-household-size views side by side; per-person and per-m² appear only in the Discussion as the units under which the gap narrows. The NEPI score inherits the single as-lived per-dwelling mode (dissemination/score_spec.md §Units).
  • Lock-in end-state. Resolved. scenarios.py reports the full ladder (each lever alone at 100%, the CCC Balanced Pathway 2040 mix, and full deployment) rather than a single ceiling; lock_in.py stays as the fabric+EV bound (1.51×).
  • Rate circularity. Travel energy is partly the cost of low access, so the rate contains the inverse of its own numerator; consider rating access against heat + an idealised/electrified travel cost (see summary.md §7). A circularity-robust rate (2.33×) is already reported alongside the headline in access_profile.py.
  • Under-recording robustness. Addressed. form_size_decomposition.py §7 reports the gas-meter coverage and electricity-meter denominator checks (heat gap 1.61× at coverage ≥0.9, 1.55× at elec-meters ≈ households) plus winsorisation; the gradient is not driven by under-recording.
  • Spatial autocorrelation & MAUP. LAD-clustered SEs are the delivered primary inference; the MAUP scale check is done (maup_scale.py). What remains open is an optional spatial error / lag model as a cross-check.

Forward work (out of current scope)

  • Bettencourt scaling analysis (BRES + GVA) — source archived; revive if pursued.
  • Morphology features (LiDAR/momepy network centrality) — only if a future Atlas needs them. (The network-distance access measure is done — stats/oa_network_access.py, summary.md §3.)

Paper / repo

  • Finalise paper/references.bib (data-source entries; NEED/EHS added).
  • Pytest suite at 49 tests (compositional model, travel constraint, inference, aggregation, EPC bands, postcode, NEPI score/bands); scenario-ladder and access-ratio coverage still pending.