pdf pdf |
|---|
Kaomapi London Atlas
Curated, analysis-ready London datasets for the Kaomapi project, built from the
London Datastore. One directory per domain under
curated/; every domain ships its parquets plus provenance.json (source
URLs, licences, vintages, raw sha256s), QC_REPORT.md (independently re-derived
counts and validations), DATA_DICTIONARY.md, and the _*.py collector scripts
that reproduce it end-to-end. manifest.parquet lists every shipped file with
sha256.
Geography is keyed to ONS codes (boroughs = E09β¦, with geo_level tags for
region/country aggregates), so domains join on (geo_code, period).
Domains
All curated parquets carry data from 2016 onwards only (project cutoff β
periods starting before 2016 are dropped by src/apply_atlas_cutoff.py; the
raw source files retain full history).
| Domain | Contents | Grain | Coverage | Licence |
|---|---|---|---|---|
housing |
House-price-to-earnings ratios β housing_affordability.parquet (1,385 rows, median + lower-quartile, residence & workplace); council rents β la_average_rents.parquet (252 rows, 2g1k1); housing-association rents β rsl_average_rents.parquet (313 rows, 24rn6); MHCLG live tables on rents/lettings/tenancies β govuk_rents_live_tables.parquet (4,894 rows, 16 tables, 292 measures) |
borough / region / country Γ year | 2016β2025 | OGL v2 / v3 |
waste |
Household recycling rates β recycling_rates.parquet (315 rows); borough Reduction & Recycling Plans: tidy actions table β rrp_actions.parquet (4,163 actions, all 33 boroughs, plan + annual updates + GLA actions index). The 93 source plan documents (DOCX/PDF) ship under unstructured/waste/rrp/ |
borough Γ year; action-level | 2016β2026 | OGL v2 / v3 |
airquality |
NOβ diffusion-tube site results β diffusion_tubes.parquet (22,727 rows, 3,140 sites, all 33 boroughs); the 2024 executive summary PDF ships under unstructured/airquality/ |
monitoring site Γ year | 2021β2024 | not specified at source (see caveat) |
stateoflondon |
The GLA's annual State of London report β 6 PDF editions (2022β2025, 111β153 pp), unstructured reference material only (no parquet, per the processing rules), under unstructured/stateoflondon/ |
London-wide narrative | 2022β2025 | not specified at source (see caveat) |
Layout: curated / unstructured / raw
Three trees, per the Kaomapi processing rules:
curated/β structured tidy parquets (spreadsheet sources, structured workflow) + per-domain provenance/QC/dictionary/collectors.unstructured/β the 100 narrative documents (91 RRP DOCX, 2 RRP PDF, the air-quality executive summary PDF, 6 State of London report PDFs) under human-readable names taken from the catalogue resource titles (e.g.Hackney 2023 to 2026 RRP - Year Two Update.docx).raw/β ALL 129 original source files (structured and unstructured; London Datastore + GOV.UK MHCLG live tables) under their original as-downloaded filenames, byte-identical to what the collectors processed (sha256-verified against provenance and the ingest manifests).
files_index.parquet is the map: one row per source resource (129) with
domain, dataset_id, resource_id, resource_name, file_name, raw_path, unstructured_path (null for structured sources), format, workflow, bytes, sha256, source_url, feeds. resource_id joins into rrp_actions.parquet,
so every action row traces to its source document:
con.sql("""
SELECT a.action_ref, a.action_title, f.file_name, f.unstructured_path, f.raw_path
FROM 'hf://datasets/kaomapi/london-atlas/curated/waste/rrp_actions.parquet' a
JOIN 'hf://datasets/kaomapi/london-atlas/files_index.parquet' f USING (resource_id)
WHERE a.borough = 'Hackney' LIMIT 3
""")
All 4,163 rrp_actions rows join to a source file: the 2,850 plan/update
rows to their DOCX/PDF (both trees), the 1,313 doc_role='actions_index'
rows to the structured GLA index XLSX (raw/ only β unstructured_path
is null for structured sources).
Source datasets
| London Datastore dataset | id |
|---|---|
| Ratio of House Prices to Earnings, Borough | 2z04n |
| Household Waste Recycling Rates, Borough | vd67o |
| Waste Reduction and Recycling Plans (RRPs) | 2r801 + 33 borough datasets |
| Air Quality Monitoring β Diffusion Tube Results | e66gp |
| Local Authority Average Rents | 2g1k1 |
| Registered Social Landlords Average Rents | 24rn6 |
| State of London (annual reports) | 2k5rr |
| MHCLG live tables on rents, lettings & tenancies | GOV.UK (external) |
Full per-file provenance (download timestamps, raw sha256, per-resource URLs)
in each domain's provenance.json.
Caveats
- Time cutoff: curated parquets exclude all periods starting before 2016;
full history remains in
raw/. - Air quality & State of London licences: the
e66gpand2k5rrsource pages do not state a licence. The data is GLA-published on the London Datastore (whose default terms are OGL); treat downstream use accordingly and check the source page before redistribution. - City of London is suppressed at source in most residence-based affordability years (small ASHE resident sample) β absence is suppression, not zero.
- 2013 break in the house-price source series (Land Registry β ONS HPSSA).
- RRP action statuses are self-reported by boroughs; plan-stage actions have
no status column (
status='unknown'). - DOCX/PDF sources ship as unstructured documents (
unstructured/, friendly names; originals inraw/); spreadsheets as structured tidy parquet β per the Kaomapi processing rules.
Use
import duckdb
con = duckdb.connect()
con.sql("""
SELECT geo_name, value
FROM 'hf://datasets/kaomapi/london-atlas/curated/housing/housing_affordability.parquet'
WHERE period='2024' AND basis='residence' AND measure='house_price_earnings_ratio'
ORDER BY value DESC LIMIT 5
""")
Built 2026-06-06 from the opencity_london harvest of the London Datastore.
- Downloads last month
- 294